CTC TT] VITAL and HEALTH STATISTICS ~ DATA EVALUATION AND METHODS RESEARCH Ne i 2 xX Pd NCHS ES S } ¢ 2 I" H TSA Language and TS Scales for the Thematic Apperception Test for Youths 12-17 Years U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration Library of Congress Cataloging in Publication Data Neman, Ronald S. Language and adjustment scales for the thematic apperception test for youths 12-17 years. (Data evaluation and methods research, series 2, no. 62) (DHEW publication no. (HRA) 75-1336) Includes bibliographical references. Supt. of Docs. no.: HE 20.6209: 2/62. 1. Thematic apperception test. 2. Adolescent psychology—Cases, clinical reports, statis- tics. I. Neman, Janice F., joint author. II. Sells, Saul B., joint author. IIL. Title. IV. Series: United States. National Center for Health Statistics. Vital and health statistics. Series 2: Data evaluation and methods research, no. 62. V. Series: United States. Dept. of Health, Education, and Welfare. DHEW publication no. (HRA) 75-1336. [DNLM: 1. Language. 2. Social adjustment. 3. Thematic apperception test—In adolescence. WM145 N433L 1974] RA409.U45 no. 62 [BF698.8.T5] 155.2’844 73-20183 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 - Price $1.50 DATA EVALUATION AND METHODS RESEARCH Series 2 Number 62 Language and Adjustment Scales for the Thematic Apperception Test for Youths 12-17 Years A report on the development and standardization of objective scoring procedures for five cards of the TAT used in the Health Examination Survey of youths 12-17 years of age. DHEW Publication No. (HRA) 75-1336 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration National Center for Health Statistics Rockville, Md. October 1974 NATIONAL CENTER FOR HEALTH STATISTICS EDWARD B. PERRIN, Ph.D., Director PHILIP S. LAWRENCE, Sc.D., Deputy Director JACOB J. FELDMAN, Ph.D., Acting Associate Director for Analysis GAIL F. FISHER, Associate Director for the Cooperative Health Statistics System ELIJAH L. WHITE, Associate Director for Data Systems IWAO M. MORIYAMA, Ph.D., Associate Director for International Statistics EDWARD E. MINTY, Associate Director for Management ROBERT A. ISRAEL, Associate Director for Operations QUENTIN R. REMEIN, Associate Director for Program Development PAUL E. LEAVERTON, Acting Associate Director for Research ALICE HAYWOOD, Information Officer DIVISION OF HEALTH EXAMINATION STATISTICS ARTHUR J. McDOWELL, Director HENRY W. MILLER, Chief, Operations and Quality Control Branch JEAN ROBERTS, Chief, Medical Statistics Branch LINCOLN I. OLIVER, Chief, Psychological Statistics Branch HAROLD J. DUPUY, Ph.D., Psychological Advisor COOPERATION OF THE BUREAU OF THE CENSUS In accordance with specifications established by the National Center for Health Statistics, the Bureau of the Census, under a contractual agreement, participated in the design and selection of the sample, and carried out the first stage of the field interviewing and certain parts of the statistical processing for the Health Exam- ination Survey. Vital and Health Statistics-Series 2-No. 62 DHEW Publication No. (HRA) 75-1336 Library of Congress Catalog Card Number 73-20183 FOREWORD This is the second and final report summarizing research carried out under a research contract with the National Center for Health Statistics by the Institute of Behavioral Research, Texas Christian University, on the development of objectively scored cognitive and affective scales for the Thematic Apperception Test (TAT). The data for the study were obtained from story protocols given in response to the five-card, orally administered and tape-recorded version of the TAT used in the Health Examination Survey of youths 12-17 years old completed in 1970. A previous report inthis series summarized similar research using data from the Health Examination Survey of children 6-11 years. In keeping with the survey's focus on characteristics as- sociated with growth and development, the TAT research was directed toward the construction of an objective scoring system and the formula- tion of scales useful in the assessment of psychological development and normal behavior. The objectives and procedures of the present study stand in sharp contrast to the usual clinical utilization of the TAT. In typical clinical assessment practice, the TAT is administered in order to confirm hypotheses about maladjustment and personality pathology which the clinician has inferred from his knowledge of an individual's life history and from the individual's responses to other instruments, both objective and projective. In that type of use, standard scoring procedures are of little interest, and protocols are usually recorded by the clinician him- self. Each clinician may use his own idiosyncratic set of notes and symbols, and his diagnosis or decision is largely a matter of subjective interpretation, With regard to the content of the TAT scales, the approach followed in this study was similar to that developedin the study of the children's TAT stories. The research was designed to explore various areas of psychological development, cognitive as well as emotional, which the TAT protocols might illuminate. The cognitive-verbal scales identified in the current study were similar to those in the study of children's TAT stories, but they did not reflect developing ability to the same degree as the younger age group. One important aspect of the extension of the study to ages beyond 11 years was the finding that development of language as assessed by the TAT scales is essentially completed by early adolescence. Also, the TAT affective scales were again not related to the available adjustment criteria. In assessing the ccatribution of the TAT research presented here and in the previous report, the criticism might be leveled that the ob- jective scales are merely another measure of verbal ability. In actuality, the TAT language scales represent innovative measures of oral speech based on controlled samples of spontaneously produced speech and represent an important original contribution, That the TAT scales pro- vide a basis for scoring verbal factors from actual samples of speech should be of considerable interest to linguistic scientists as well as to psychologists. A great many people took part in this research over a period of several years, Special recognition should be given to those who were involved in the production of this report: Steve Angle, Vicky Breed, Joan McGhee, Jane McMahon, Frances Neil, and Gwen Sorsby, for their exceptional enthusiasm and dedication in scoring the TAT protocols; Jan Fox, for her prompt and resourceful attention to the many adminis- trative details associated with this project; and Mary Hostvedt, for the incredible skill, patience, and good nature she maintained while typing the report. All of these persons are or were at the Institute of Behavioral Research where the study was carried out. On the staff of our division the project officer for this contract was Glenn Pinder, Research Psy- chologist, whose contributions to the successful completion of the work were many and substantial and merit recognition. Arthur J, McDowell, Director Division of Health Examination Statistics National Center for Health Statistics CONTENTS Page Foreword ------==-mmmmmmmmm mmm mmm meme mmm mm mae m moo om —-—— eo iii Objectives and Background-------==-====-=--mmmmmmmmm mmm mmm mmm mmm o 1 Health Examination SUurvey---------==---c--mmmomommomomooooooommmm mmo 1 PRycholOZICALl TEL BaLTEIY «wm mmm mmm mmm or sh smo i wt i ie ————————- 1 Review of Previous Studies -~«-=-mwwmm emma mem mm or tm em mm 2 Plan of the Cycle Ill TAT Study----=-==-e=meemmemeem meme meee ————— 5 Description of the Cycle III Subsample----===----=-=-cc---c--oommmmoonon 5 Revision of Scoring Manual----------=--mmmmmmmmmommom moomoo mmm mmo 6 Scoring Procedure---=----=-----c-mmommmmmome mmo oomoeooo ooo moomoo 6 Selection of Criterion Data--==-=-==---=--coccommmmmmmmmmmo moomoo 7 Analysis of the Criterion Data---=---==ee--ommmmmem meme emcee em em 8 Medical History of Youth—Parent's Questionnaire--------------------- 8 Health Habits and History— Youth's Questionnaire-------=-==--=---------- 11 Supplemental J6iormation Prom SCHOOL me = mmm mmm mm mm wm no i mmo 14 Health Behavior-— Youth's Questionnaire---------====--oecee-ocoooomoo- 14 Test Scores and Demographic Data-----=--==-=m-cmomeooo-mmommmm ooo 18 Relationships Among Primary Criterion Factors----------------------- 18 Analysis of TAT Predictor Data--------===------------co----mo-oomoomooo- 21 Defining Variables for Cycle Ill Factors----=--------------------=----- 21 Comparison of the Cycle II and Cycle III TAT Factors------------------ 23 Validation of TAT Scales-=-===-======-===-c-cmmmmmomomommmmmmmm moomoo oo 25 Correlations of Criterion Measures With Age, Sex, and Race------------ 26 Correlations of TAT Composites With Age, Sex, and Race--------------- 27 Correlations Between TAT and Criterion Composites-=------=-=--=-------- 30 National Norms for TAT Scales----------=---ccommmmmmmcm momo mm mmo 31 References —----====m=mmm mmm oo meee ome mmo o-------- 32 List of Supplementary Tables---------=-mewmemeeoe mmm mmm mmm m mmm mm meee 33 Appendix I. Scoring Manual for Cycle Ill TAT Data--------------=------ 57 Appendix II, Questions From Cycle III Health Examination Survey Forms Used in This Study-----=--=-===-----=c--mmmcmommoommmomooommmo mmm m mmm 67 Appendix III, Conversion Tables for Raw Scores on 29 TAT Variables and Percentile Equivalents for TAT Composite Scores--------------------=-= 83 vi SYMBOLS Data not available--- reese I Category not applicable -------eeoeemeemeeeeeee QUANTILY ZErO---=----sesssmmmrmmmmes emma cee ecnene Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision. --------eeeeeeeeeeeees LANGUAGE AND ADJUSTMENT SCALES FOR THE THEMATIC APPERCEPTION TEST FOR YOUTHS 12-17 YEARS Ronald S. Neman, Janice F. Neman, and S. B. Sells Institute of Behavioral Research, Texas Christian University OBJECTIVES AND BACKGROUND This study represents an extension of pre- vious research involving the development of scales for the five-card, orally administered and tape-recorded version of the Thematic Apper- ception Test (TAT) used in both Cycles II and III of the Health Examination Survey (HES) by the National Center for Health Statistics (NCHS). This report, developed from TAT protocols from Cycle III of the HES conducted in 1966-70, doc- uments scale development and presents national norms for these scales based on a national prob- ability sample of 1,398 youths in the age range 12-17 years. A previous NCHS report,! devel- oped from TAT protocols from Cycle II of the HES, presented scales and norms based on a national probability sample of children in the 6-11 year age range. Health Examination Survey The background of the HES is provided in an NCHS report? that describes the developments leading to the enactment of the National Health Survey and the policies, program, and operations of the survey. Procedures used in the Cycle III survey, which provided data for the present report, were similar to those adopted for use in the Cycle II survey, The sample design for the survey was a probability sample of persons in geographically defined segments of the U.S. population. Of those youths selected, 6,768 were examined at 40 lo- cations over a 4-year period, Examination in- cluded measures of visual and auditory acuity, skeletal and dental development, metabolic func- tioning, and psychological development, All sam- ple youths were tested individually in specially designed vans, by qualified professional ex- aminers, A previously published report> presents a summary of the plan and operation of the Cy- cle III survey. Psychological Test Battery The Cycle III test battery included all of the tests used with the sample of younger children in the Cycle II survey. These were the Vocabu- lary and Block Design subtests of the Wechsler Intelligence Scale for Children (WISC), the Read- ing and Arithmetic subtests of the Wide Range Achievement Test (WRAT), and a modified ver- sion of the Goodenough-Harris Drawing Test, as well as the TAT. A reading and writing literacy test, not employed in Cycle II, was added to the Cycle III battery. Whereas the other tests were chosen to measure intellectual-cognitive functioning, the TAT was chosen to measure both language de- velopment and personality. Sells* pointed out, in particular to the Cycle II age range, that neither the TAT nor any other personality test could be recommended for survey use without criticism, The TAT was included in the survey battery because of its widespread acceptance as an indi- 8 Therefore, it should be concluded that not only are symptom data required to describe primary care accu- rately, but that the researcher who labels each episode of illness only with a diagnosis is not viewing the total ambulatory picture realis- tically. Once the importance of the patient’s motiva- tion to visit the physician’s office is established, the question arises: Should the information derived from the visit be expressed in the patient’s own words or should it be translated by the health professional? McFarlane and Norman® examined this dilemma by considering the two methods of describing a symptom or complaint—the use of lay terms by the patient on the one hand and the use of professional terminology on the other. For purposes of mon- itoring illness, McFarlane and Norman advocate the latter. They argue that it reduces the many irrelavancies that occur when lay terms are employed. However, based on data gathered in the out- patient clinic of the Toronto General Hospital, Bain and Spaulding? advocated the use of lay terms. In their study of diagnostic probability based on symptom frequencies, they recognized that most symptoms are subjective and are often expressed by the patient when under stress. Yet how else can it be determined what motivates patients to visit physicians? Murnaghan3 sums up the situation by pointing out that the motiva- tion to visit a physician’s office can best be described by the patients themselves, not by the physician. There is little doubt that different patients not only perceive similar complaints and symptoms in different ways, but also express them verbally using multifarious terms. Can one expect the average person to describe in stand- ardized medical terms—which allows the physician instant insight into the problem—the physical, mental, or social phenomenon that motivates his seeking medical aid? Certainly not. Therefore, if one agrees that data should be collected on the need for primary care from the patient’s point of view, then one must be ready to accept the entire array of complaint vernac- ular as valid and develop some logical system for its classification. For years, nosologists and researchers have advocated the need for uniform standard medi- cal terminology. The NCHS has been in the fore- front in developing coding procedures for mor- bidity and mortality based on classifying standard acceptable medical diagnoses. The re- quirement to develop a logical statistical coding system for the classification of idiosyncratically perceived and communicated symptoms and vague complaint expressions becomes a nosol- ogist’s nightmare. The authors are fully aware of the inherent imperfections and deficiencies that exist in a scheme to classify imperfect and defi- cient terminology. REVIEW OF SYMPTOM CLASSIFICATIONS During the initial search for a realistic symptom classification, consideration was given to the International Classification of Diseases, Adapted (ICDA),!0 with the idea of possibly expanding the sections dealing with symptoms. The ICDA symptom section (780-796) is clearly a catchall for undiagnosed conditions. It is not a complete symptom list and was never intended to serve as an independent symptom classi- fication. Therefore, it was decided that the ICDA would be applicable for coding the physi- cians’ diagnoses recorded in item 9 of the Pa- tient Record Form, but that another classifica- tion would be employed for coding the symptoms and complaints presented in item 5. Further investigation revealed a paucity of contributions toward symptom classification. The symptom coding systems that are available had to be viewed in the light of their original application. Bain and Spaulding? in 1966 devel- oped a very effective coding system from the study of the adult population in the outpatient department of the Toronto General Hospital. Their simple and concise classification has been adapted for use by others and has served as a guide in our deliberations. Morrell, Gage, and Robinson’s! 1 classification system was devel- oped after recording presenting symptoms in general practice over a 1-year period. It is note- worthy that their scheme provides for coding administrative requests in which there was no symptom or complaint present. The results of their study yielded documented frequency data of symptoms encountered in an ambulatory medical care setting. Hurtado and Greenlick!2 at the Kaiser Foundation Hospitals in Portland, Oregon, de- veloped an extensive symptom classification for their medical care utilization study. Renner’s! 6 symptom classification, which he devised for the Department of Family Medicine at the Uni- versity of Wisconsin, proved instrumental to our efforts. This list, which corresponds somewhat to the Kaiser Foundation list, provided a ra- tional framework from which to begin. An addi- 4 tional source of data was given by the Commis- sion on Professional and Hospital Activities in their second edition of Hospital Adaptation of ICDA.'3 This edition contains added categories consisting of signs, symptoms, abnormal labora- tory findings, factors relating to medical history, and the socioeconomic factors affecting health. All these classifications were reviewed in de- tail. Considerable knowledge and an apprecia- tion of the varied but logical approaches to unique problems were gained. The adoption of any one system in its entirety for a statistical classification was prohibited by one or more of the following factors: (1) the system was limited to symptoms observed in a selected population, such as all adults seen in an internal medicine practice; (2) the system was too extensive to be useful as a statistical classification in that the scheme lent itself to indexing and retrieval rather than to statistical grouping; and (3) the system did not classify the vague problems ex- pressed by patients, but rather coded medical terminology physicians use in describing patients’ symptoms. The latter reason proved to be the most challenging problem in developing the NAMCS classification. Unlike coding schemes needed for the evaluation of care processes and decisionmaking for resource allo- cation, the NAMCS classification was generated primarily for its application in this survey. RATIONALE The primary determinant in the development of this classification is its function as a process- ing instrument for the NAMCS. Except for the limited experience gained from the pilot studies, there was no previous experience to call upon in developing a method which would “custom fit” this survey situation. Therefore, three major symptom classification criteria were established: The classification must be simple, flexible, and functional. The following discussion of these ob- jectives will help to explain the structure of the scheme and justify many of the individual inclusions. 1. Simplicity—The decisions regarding the number and character of the individual rubrics were influenced by the requirement that the re- sulting classification be not only simple and easy to use, but relatively short. The NAMCS classi- fication contains 197 rubrics. It is expected that these rubrics will be broad enough to encompass all symptoms, and yet exclusive enough to facil- itate the medical coders’ decisions as to proper code assignment. Also, as part of the classifica- tion system, an alphabetical index of symptom and complaint terms was developed (appendix II). The importance and use of the index are discussed in a later section. 2. Flexibility—Plans call for future revisions of this initial scheme. The relatively simple con- struction and use of a four-digit numeric system will allow this expansion. The numeric assign- ments found in this scheme permit the needed flexibility within classes, and the adaptability of a more detailed subdivision of many separate categories. An example of the expansion possi- bilities may be seen in Class IV, “Cardiovascular and Lymphatic Systems.” Code 201.0, Heart murmur, is followed by code 205.0, Abnormally high blood pressure, leaving the numeric assign- ments 202.0, 203.0, and 204.0 for future ex- pansion. This same technique was employed throughout each class to insure flexibility of the system. One will note that each symptom code has four digits, the fourth digit being “0.” This is not only to assist coders in coding consistently with the four-digit coding of item 9 (see Patient Record Form, figure 1), but also to permit future expansion of individual categories. De- pending on the frequency or interest, code 056.0, Headache, may later be subdivided into 056.1, Headache, tension; 056.2, Headache, migraine; and 056.3, Headache, nonspecific. In this manner, the fourth digit “0” could be used as a summary code. Flexibility is necessary in any classification when grouping of data may be required for presentation. In developing this system, not only do the classes represent major groupings but within certain classes, rubrics can be combined to form logical groups. For example, codes 500.0 through 503.0 may be combined to form Symptoms referable to upper digestive system, and 651.0 through 653.0 could be grouped tcgether as Symptoms and com- plaints regarding menstruation. 3. Functional-The development of a scheme that would realistically fit the NAMCS required that the scheme serve as a statistical classifica- tion. This requirement prevented the accom- modation of all possible entries with individual codes, which would have been an impossible task. There can be scant argument for assigning individual codes to such terms as “tired,” “pooped,” “rundown,” and ‘“‘exhausted’ unless one can distinguish between the relative degree of each term. It is also for this reason that no attempt was made to force coding of slang, vague, or lay expressions of illness into accept- able or standard terminology. In this light, the grouping within this classification is a purposeful effort to accept the varied ways in which pa- tients might express their symptoms. Description of the National Ambulatory Medical Care Survey Symptom Classification The NAMCS symptom classification consists of two major divisions—the tabular list of in- clusive terms and the alphabetical index of terms. The tabular list consists of 13 classes that are further divided into individual categories or rubrics. The inclusions are the symptomatic terms contained within the rubrics. The inclusive terms are not meant to be a complete listing for each category, but they are used rather to give a general idea of the symptoms contained within each category. The primary axis of the classification struc- ture was selected to conform to that of the ICDA, i.e., by principal anatomical site. How- ever, observing that a relatively large number of symptoms could not be classified rationally to any one anatomical site,! I some deviations from the primary axis proved necessary. Basically, the classification is divided into the following 13 classes: I. General (Nonspecific) 000.0-049.0 II. Nervous 050.0-099.0 III. Skin, Nails, and Hair 100.0-199.0 IV. Cardiovascular and Lymphatic Systems 200.0-299.0 V. Respiratory 300.0-399.0 VI. Musculoskeletal 400.0-499.0 VII. Digestive 500.0-599.0 VIII. Urinary 600.0-629.0 IX. Male Reproductive 630.0-649.0 X. Female Reproductive 650.0-699.0 XI. Eyes and Ears 700.0-799.0 XII. Mental Health 800.0-899.0 XIII. Nonsymptomatic 900.0-999.0 5 Class I, General Symptoms, contains complaints of a vague, nonspecific nature (e.g., code 013.0, Aches all over) and terms for which no specific body site can be identified (e.g., code 002.0, Fever). Each of the next 11 classes, which refer to selected systems or specified body sites, follows similar procedures for gen- erating rubrics. Several previous studies provided a partial foundation and listing of the more common symptoms.?»11.12,14 In addition, the NAMCS pilot study symptom data were com- piled and reviewed in the light of frequency and estimated importance of individual terms. Once it was determined that a term merited inclusion as a separate category, synonymous terms and phrases that might be used to express the same symptom were compiled. Consider, for example, the symptom, “nasal congestion.” How might a patient express this problem? It could be “post- nasal drip,” ‘“drippy nose,” “runny nose,” “sniffles,” “stuffy nose,” “stopped up nose,” or a number of related phrases. All these expres- sions would be coded to the same rubric—301.0 Nasal Congestion. Besides grouping synonymous expressions under the same category, several other decision processes were used in selecting the rubrics. First, in some categories all symptoms relating to a selected region or site were grouped to- gether—e.g., code 505.0, Symptoms referable to lips, which includes “abnormal color,” ‘“bleed- ing,” “cracked,” “dry,” “pain,” and “swelling.” This grouping technique was applied to many of the body sites for which no specific symptom could be isolated as having any statistical value either in terms of frequency or interest. Another method of selecting categories was to group together under a single rubric symptoms that are varying degrees of a similar complaint. For example, “pressure in/on chest,” “chest tightness,” and “discomfort in chest” may be considered various levels of severity of chest pain. All these terms are included under the same code—322.0, Pain in chest. The last technique in selecting rubrics was to identify different expressions with an underlying element of similarity. For example, grouped under rubric 062.0, Disturbances of sleep, are “sleepwalking,” “sleepiness,” “drowsiness,” “in- somnia,” “hypersomnia,” “nightmares,” and “time-zone syndrome.” Although these terms express different problems, they may all be classified as disturbances of sleep. For all classes, and in some instances for specific body systems, residual or catchall cate- gories are provided. These categories usually fol- low one or more detailed symptomatic codes and serve to accommodate entries that are un- classifiable elsewhere in the scheme. Residual categories will be used as a last resort only, and their frequency will become a partial measure of the utility of the classification. A high frequency may point to the need for more specificity or possibly to poor response to item 5—i.e., the entries in item 5 may not be in the patient’s own words. Class XIII represents an interesting variation from the anatomical axis presented in the other 12 classes. A large percentage of patients arrive at the physician’s office with no stated com- plaint, but they do have a reason for the visit.! 5 In an effort to adequately accommodate all major reasons why patients say they seek medi- cal care, these reasons as expressed in the pa- tients’ own words were provided for in the scheme. Examples of the reasons are such blunt, yet truthful, statements as ‘“‘the doctor told me to come in,” or “I want to talk to the doctor” or may be the need for an annual checkup, pre- natal visit, or family-planning counseling. Within this context, categories relating to followup care or progress visits are acceptable. And as Morrell stated: Only a proportion of consultations in general practice are provoked by a new illness experience. Many are under- taken for the followup of established disease. In develop- ing a system of symptom recording it is essential to separate these two types of consultation. ! The NAMCS has addressed this particular area not only in its symptom classification but also in items 7 and 8 which attempt to determine which visits can be considered followup visits (see figure 1). In determining the symptom categories, no attempt was made to classify inappropriate answers in response to item 5. It may be an- ticipated that the physician, in answering this item, will not record the problem in the pa- tient’s own words, but will translate lay expres- sions into complex medical terms. If such terms do not represent a problem, symptom, or reason for visit, the response will be coded to a residual catchall code. Bronchitis, diabetes, arthritis, and hemorrhoids are but a few of the diagnoses that a patient can reasonably name as his complaint. Such responses will be coded to the residual categories of the specified body site according to the stated diagnosis. In this manner, bronchitis would be coded to 330.0, diabetes to 990.0, arthritis to 430.0, and hemorrhoids to 560.0. Though this procedure may appear contra- dictory to the original purpose of the classifica- tion—to accommodate the majority of entries—it may be justified by the presumption that if these patients actually have the stated illnesses, their true reason for the visit would be for some manifestation of their conditions or for follow- up care. Because item 9 of the Patient Record Form requires the physician’s diagnostic inter- pretation of the symptoms presented in item 5a, the actual presence of such conditions would be recorded and there would be no loss of vital data. The second part of the classification is the alphabetical index of terms that constitutes an expanded list of symptoms and reasons for visits with the appropriate code number. The coding procedures provide for the addition of new terms to the index as they are encountered. This particular aspect, along with the plans for re- vision of the tabular list, emphasizes the dy- namics and research potential of the scheme within the first year of use. CODING PROCESS AND EVALUATION During the first year of the survey, the serv- ices of the American Medical Record Asso- ciation were secured for the coding of the medi- cal data presented in items 5 and 9. The symp- tom and complaint coding process will consist of three related functions: (1) assignment of appro- priate codes; (2) verification; and (3) evaluation of the coding instrument. Initially, these experienced medical coders will be introduced to the NAMCS to get an understanding of the uniqueness of the data. Procedure manuals and teaching techniques will be developed. Instructions in the use of the classification will emphasize consistency and accuracy. The alphabetical symptom coding index will simplify the job of locating the correct code for the recorded expression. As stated before, the classification scheme itself does not contain all the numerous terms, jargon, and slang expres- sions that are found in this index. "nclsive terms listed under the rubrics are examples of other symptoms that could fall under the same codes; these terms do not represent an exclusive group. The symptoms and complaints listed in the alphabetical index are also cross-indexed when appropriate. Therefore, “pain in neck’ can be located under “pain, neck’ and also “neck, pain.” In determining the appropriate code, the alphabetical index will always be referred to first. If a specific term is not indexed, synon- ymous expressions will be considered. If no applicable term can be located, the coding super- visor will be consulted. He will then decide whether to code the entry to a selected rubric, to a catchall category, or to code 990.0, Problems, complaints, symptoms, or reasons for visit, NEC. Coders will be required to maintain a frequency listing of all terms codable to the catchall categories. As new terms are introduced, they will be added to the index with their proper code assignments. In this way, coders will develop a thesaurus of symptoms, complaints, and reasons for seeking primary care as expressed by patients. For consistency in the coding process, all coders will receive up-to-date lists of new terms and expressions. All coding problems encountered will be evaluated and documented as to their source and solution. During the first year, however, no new rubrics will be created. A two-way independent verifica- tion of the coding process is planned with a tolerated overall coding error rate of 5 percent. Yearly frequency tabulations of each rubric will receive in-depth analysis. Results of both this study and the coding supervisor’s documented appraisal will provide guidelines for revisions of the NAMCS classification. SUMMARY There has been an increasing demand from researchers and medical personnel for reliable national data on the largest segment of ambula- tory medical care, that portion provided in the physician’s office. The National Ambulatory Medical Care Survey (NAMCS), conducted by the National Center for Health Statistics, has undertaken to answer this data need. The NAMCS describes utilization of ambula- tory services in terms of the volume of ambula- tory care visits, the demographic characteristics of the patients, the nature and prevalence of their medical problems, and the nature of the treatment and services they receive. Of particu- lar interest is the NAMCS collection of patients’ reasons for the ambulatory visit as expressed by the patients themselves. The decision to use this item as verbalized by the patient resulted in the need to develop a logical system for classifying such complaint vernacular. Presented here is the resulting classification specifically designed to code patients’ expres- sions of their symptoms, complaints, and prob- lems as collected by the National Ambulatory Medical Care Survey of the National Center for Health Statistics. Based on the experience of others, this classification was developed to serve in the processing and presentation of statistical data. The NAMCS symptom coding scheme con- sists of 197 rubrics grouped into 13 classes around an anatomical axis. Following the first year of experience, the coding scheme will be evaluated for its utility and modified as required. REFERENCES I National Center for Health Statistics: National ambulatory medical care statistics: background and methodology, United States. Vital and Health Statistics. Series 2, No. 61. DHEW Pub. No. (HRA) 74-1335. Health Resources Administration. Washing- ton. U.S. Government Printing Office, Mar. 1974. 2National Center for Health Statistics: Current estimates from the Health Interview Survey, United States, 1970. Vital and Health Statistics. Series 10, No. 72. DHEW Pub. No. (HSM) 72-1054. Health Services and Mental Health Administration. Washington. U.S. Government Printing Office, May 1972. 3Murnaghan, J. H.: Review of conference proceedings. Med.Care 11(No. 2, Suppl.): 13-34, 1973. 4White, K. L.: Improved medical care statistics and the health services system. Public Health Rep. 82(No. 10):847-854, 1967. 5Weed, L. L.: Medical records that guide and teach. N.Eng.].Med. 278(No. 13):593-600, and 652-657, 1968. 6 Renner, J. H., and Piernot, R. W.: A revised symptom code list for ambulatory medical record data. Paper prepared for Family Practice Program. University of Wisconsin, Madison, Wis., 1972. TWestbury, R. C., and Tanant, M.: Classification of disease in general practice. Canad. Med. Ass.]. 101(No. 10):82-87, 1969. 8McFarlane, A. H., and Norman, G. R.: Methods for classifying symptoms, complaints, and conditions. Med. Care 11(No. 2, Suppl.):101-108, 1973. 9 Bain, S. T., and Spaulding, W. B.: The importance of coding presenting symptoms. Can.Med.Assoc.]. 97(No. 16):953-959, 1967. 10National Center for Health Statistics: Eighth Revision International Classification of Diseases, Adapted for Use in the United States. PHS Pub. No. 1693. Public Health Service. Wash- ington. U.S. Government Printing Office, 1967. Morrell, D. C. Gage, H. C., and Robinson, N. A.: Symptoms in general practice. J.Roy.Coll. Gen.Pract. 21:32-43, 1971. 12Hurtado, A. V., and Greenlick, M. R.: A disease classifica- tion system for analysis of medical care utilization with a note on symptom classification. Health Serv.Res. 6(No. 3):235-250, 1971. 13 Commission on Professional and Hospital Activities: H-ICDA, 2d ed. Vol. 1, Tabular List, Based on International Classification of Diseases (1965 Revision, WHO) and Eighth Revision International Classification of Diseases, Adapted for Use in the United States (PHS Publication No. 1693). Ann Arbor. Sept. 1973. 14McFarlane, A. H., and O’Connell, B. P.: Morbidity in wenlly practice. Canad.Med.Ass.]. 101 (No. 6):259-263, 1969. 1 Unpublished pilot study data. APPENDIX | TABULARLIST List of classes Classification code XI. XII. X11. . General Symptoms LL... Nervous System . . . . i . Skin, Nails, and Hair . .............. i nnn.. . Cardiovascular and Lymphatic Systems . .............. 0... ... Respiratory System... . LL... . Musculoskeletal System . . ............. VII. VIII. Digestive System... . LL... eee Urinary System... . ee . Male Reproductive System . . . ........... 0... Female Reproductive System, Including Breast ................. Eyesand Ears . ......... Mental Health . ....... 000.0-049.0 050.0-099.0 100.0-199.0 200.0-299.0 300.0-399.0 400.0-499.0 500.0-599.0 600.0-629.0 630.0-649.0 650.0-699.0 700.0-799.0 800.0-899.0 900.0-999.0 General Symptoms (000.0-049.0) 001.0 Chills 002.0 Fever Includes: Fever High temperature 004.0 Fatigue Includes: Exhausted General weakness Pooped Rundown Tired Worn out 005.0 General ill-feeling Includes: Not feeling well 007.0 Fluid imbalance Includes: Dehydration Excessive sweating Excessive thirst Retention of fluid 009.0 Lack of physiological development Includes: Lack of growth 010.0 Weight gain Includes: Obesity 011.0 Weight loss Includes: Recent weight loss Underweight 013.0 Pain, generalized, site unspecified Includes: Ache Aches all over Cramp Hurt Stiffness 015.0 Swelling or mass, site unspecified Includes: Bulge Bump Knot Lump 020.0 General symptoms of infants and chil- dren, NEC Includes: Crying too much Fidgety 10 020.0 General symptoms of infants and chil- dren, NEC—Con. Fussy Hyperactive Irritable Underactive Symptoms Referable to the Nervous System (050.0-099.0) 050.0 Abnormal involuntary movements Includes: Shaking Tic Tremor Twitch Excludes: Eyes 052.0 Coma and stupor 053.0 Confusion 054.0 Convulsions Includes: Fits Seizures Spells Excludes: Fainting 056.0 Headache Includes: Migraine Nervous Tension 058.0 Memory, disturbances of Includes: Amnesia Lack or loss of memory Temporary loss of memory 059.0 Other disturbances of sensation Includes: Anesthesia Burning Hyperesthesia Loss of sense of: Smell Taste Touch Prickly feeling Tingling 062.0 Disturbances of sleep Includes: Drowsiness Hypersomnia 062.0 Disturbances of sleep—Con. Insomnia Trouble sleeping Can’t sleep Nightmares Sleepiness Sleep walking Time-zone syndrome 065.0 Senility—Old Age 067.0 Stammering or stuttering 069.0 Vertigo-dizziness Includes: Falling sensation Giddiness Lightheadedness Loss of sense of equilibrium or balance 070.0 Other symptoms referable to the nervous system, NEC Symptoms Referable to Skin, Nails, and Hair (100.0-199.0) 100.0 Acne or pimples Includes: Bad complexion Blackheads Blemishes Breaking out Whiteheads 104.0 Discoloration or pigmentation Includes: Blushing Color, change in Flushing Freckles Red Spots Excludes: Discoloration of extremities Jaundice Infectious disorders Includes: Athlete’s foot Boils Ringworm Calluses or corns Skin moles Wrinkles Warts 106.0 108.0 109.0 110.0 111.0 112.0 Allergic skin reactions Includes: Hives Photosensitivity Poison ivy, poison oak, etc. Rash, diaper 113.0 Skin irritations, NEC Includes: Inflammation Itching Pain Ulcer Sore 115.0 Swelling or mass of skin Includes: Bumps Lesion Nodules Welts, except hives 116.0 Wounds of skin Includes: Bites Blisters, non-allergic Bruises Burns (chemical, sun, wind, and stearn) Cuts Scratches 120.0 Other specific symptoms referable to skin Includes: Dryness Oiliness Peeling Scaliness Texture, change in 122.0 Symptoms referable to nails Includes: Breaking Brittle Color, change in Cracked Ingrown Ridges Splitting 124.0 Symptoms referable to hair Includes: Baldness Brittle Dryness Falling out 1 124.0 Symptoms referable to hair—Con. Oiliness Receding hair line 126.0 Symptoms of umbilicus Includes: Discharge Draining Not healing Painful Red 130.0 Other symptoms of skin, nails and hair, NEC Symptoms Referable to Cardiovascular and Lymphatic Systems (200.0-299.0) 200.0 Irregular pulsations and palpitations Includes: Decreased Fluttering Increased Pulse too fast Pulse too slow Other irregular heart beats Rapid heart Skipped beat Unequal 201.0 Heart murmur 205.0 Abnormally high blood pressure Includes: Elevated B/P High B/P Hypertension 206.0 Abnormally low blood pressure Includes: Decreased B/P Hypotension Low B/P 210.0 Symptoms referable to blood Includes: Poor Thin Tired Weak 212.0 Pallor and cyanosis Includes: Ashen color Blueness of fingers-toes Paleness 12 214.0 Syncope or collapse Includes: Blacked out Fainting Passed out Spells 216.0 Other symptoms of the heart, NEC Includes: Bad heart Poor heart Weak heart 220.0 Other symptoms referable to cardio- vascular system, NEC 231.0 Edema and dropsy 232.0 Swollen lymph glands Includes: Enlarged lymph nodes Sore glands Swollen glands 240.0 Other symptoms referable to lymphatic system, NEC Symptoms Referable to Respiratory System (300.0-399.0) 300.0 Nose bleed Includes: Bleeding from nose Hemorrhage from nose Excludes: Injury 301.0 Nasal congestion Includes: Drippy nose Postnasal drip Red nose Runny nose Sniffles Stuffy nose 304.0 Sinus problems Includes: Congestion Impacted Infected Lightness Pain Pressure 306.0 Shortness of breath Includes: Breathlessness Dyspnea 306.0 Shortness of breath—Con. 400.0 Pain, swelling, injury of lower extrem- Sensation of suffocation ity-Cog, : ontracture Trouble breathing C ramp 307.0 Other disorders of respiratory rhythm Hot-cold and sound Hurt Includes: Pulled muscle Abnormal breathing sounds Soreness Hyperventilation Spasm Rales Stiffness Rapid breathing Strain Sighing respiration of Wheezing Ankle 310.0 Sneezing Yaoi 311.0 Cough He 312.0 Cold Leg or thigh 313.0 Flu Lower extremity, Includes: part unspecified Grip Toe Influenza 405.0 Pain, swelling, injury of upper extremity 314.0 Croup Includes: 320.0 Sputum or phlegm Ache Includes: Contracture Bloody Cramp Excessive Hot-cold Purulent Hurt 321.0 Congestion in chest Pulled muscle Includes: Soreness Lung congestion Spasm 322.0 Pain in chest Stiffness Includes: Strain Burning sensation of Chest tightness Arm Pain in lung Elbow Pain over heart Fingers Pain: respiratory, rib, retro- Forearm sternal, sternal, side of chest Hand Pressure in/on chest Shoulder 325.0 Disorders of voice Thumb Includes: Upper arm Hoarseness Upper extremity, Hypernasality } part unspecified 330.0 Other symptoms referable to the respira- Wrist tory system, NEC 410.0 Pain, swelling, injury of face and neck Symptoms Referable to the Musculoskeletal region . System (400.0-499.0) bngtager 400.0 Pain, swelling, injury of lower extremity Contracture Includes: Cramp Ache Hurt Charleyhorse Pulled muscle 13 410.0 Pain, swelling, injury of face and neck region—Con. Soreness Spasm Stiffness Strain of Back of head Cervical spine Face Jaw Neck Upper spine 415.0 Pain, swelling, injury of back region Includes: Ache Contracture Cramp Hurt Pulled muscle Soreness Spasm Stiffness Strain of Back, NOS Back, upper, lower Lumbar Lumbosacral Sacroiliac Spine, NOS Thoracic spine 420.0 Atrophy or wasting of extremities Includes: Numbness Paralysis, partial or complete Weakness 421.0 Difficulty in walking, abnormality of gait Includes: Clumsiness Falling Inability to stand or walk Limping Staggering 422.0 Other specific symptoms referable to limb and joint Includes: Foot drop Posture problems Wrist drop 14 425.0 Other and multiple symptoms referable to unspecified limbs, joints, or muscles Includes: Ache Contracture Cramp Hurt Pulled muscle Soreness Spasm Stiffness Strain 429.0 Bunion 430.0 Other symptoms referable to the mus- culoskeletal system, NEC Symptoms Referable to Digestive System (500.0-599.0) 500.0 Chewing difficulty 501.0 Bleeding, gums (gingival) 502.0 Halitosis Includes: Bad breath 505.0 Symptoms referable to lips Includes: Abnormal color Bleeding Cracked Dry Pain Swelling 510.0 Symptoms referable to mouth Includes: Bad taste Burn Dryness Inflammation Pain Swelling Ulcer 511.0 Saliva, excess Includes: Drooling 512.0 Toothache 515.0 Other symptoms referable to mouth, NEC 520.0 Throat soreness Includes: Painful throat 520.0 Throat soreness—Con. Scratchy throat Sore throat 525.0 Symptoms referable to tongue Includes: Abnormal color Bleeding Blisters Burned Pain Ridges Smooth Swelling or mass Ulcer 527.0 Symptoms referable to tonsils Includes: Bleeding (postoperative) Discharge Inflammation Swelling 528.0 Difficulty in swallowing Includes: Choking 530.0 Other symptoms referable to throat, NEC 540.0 Abdominal pain Includes: Colic, intestinal infants) Pain in the following sites: Epigastrium Iliac Inguinal (groin) Right (left), upper (lower) quadrant Stomach (includes cramps) Umbilical region 541.0 Colic, infantile 542.0 Abdominal swelling or mass Includes: Distention or fullness 543.0 Flatulence Includes: Bloating, gas Distention due to gas Gas, excessive 545.0 Appetite, abnormal Includes: Decrease Excessive Loss (except 546.0 550.0 554.0 555.0 556.0 560.0 570.0 572.0 574.0 575.0 579.0 580.0 590.0 Feeding problems Bleeding, gastrointestinal Includes: Blood in stools (melena) Hematemesis Hemorrhage, cause unknown Vomiting blood Constipation Diarrhea Includes: Loose stools Other symptoms or changes in bowel function Includes: Bulky stools Dark stools Fatty stools Mucous stools Pus stools Unusual color Unusual odor Symptoms referable to anus-rectum Includes: Bleeding Itching Mass Rectal pain Swelling Heartburn or upset stomach Includes: Belching Indigestion Nausea and vomiting Includes: Retching Sick to stomach Throwing up Regurgitation or spitting-up Hiccough Jaundice Other symptoms of liver, gallbladder, and biliary tract Includes: Biliary colic Gallstones Other symptoms referable to the diges- tive system 15 Symptoms Referable to the Urinary Tract (600.0-629.0) 600.0 Urine abnormalities and abnormal constituents Includes: Blood Pus Unusual color Unusual odor Excludes: Passing stones 601.0 Frequency and nocturia Includes: Bed wetting Night discharge 602.0 Incontinence of urine Includes: Dribbling Involuntary urination Retention of urine Includes: Cannot empty bladder Inability to urinate 604.0 Painful urination Includes: Burning 610.0 Other urinary dysfunction Includes: Hesitancy Large volume Slowing of stream 620.0 Other symptoms referable to the urinary tract, NEC Includes: Bladder trouble Passed stones 603.0 Symptoms Referable to the Male Reproductive System (630.0-649.0) 630.0 Infertility—Male Includes: Low sperm count Sterility 631.0 Pain, swelling, or mass of male genital system Includes: Pair, swelling, or mass of the following sites: Penis 16 631.0 Pain, swelling, or mass of male genital system—Con. Scrotum Testes Painful erection 640.0 Other symptoms referable to the male reproductive system Excludes: Psychosexual problems Symptoms Referable to the Female Reproductive System, Including Breast (650.0-699.0) 650.0 Menopause symptoms Includes: Hot flashes 651.0 Premenstrual tension 652.0 Menstrual cramps 653.0 Menstrual disorders Includes: Absence (amenorrhea) Atypical material Blood clots Excessive (hypermenorrhea, menorrhagia, meno- metrorrhagia) Frequent Infrequent Irregular (metrorrhagia) Large flow Onset delayed Scanty Small flow 654.0 Ovulation pain 660.0 Pelvic symptoms Includes: Pain Pressure or dropping sensation Swelling or mass 661.0 Vaginal disorders Includes: Pain Swelling or mass 662.0 Vaginal discharge Includes: Atypical (leukorrhea) Bloody Brown 663.0 Vulvar disorders Includes: Itching Pain 663.0 Vulvar disorders—Con. Perineal swelling or mass Ulcer 665.0 Infertility—female Includes: Sterility 667.0 Problems of pregnancy, NEC Includes: Leaking amniotic fluid Possible labor Products of conception passed Spotting 670.0 Other symptoms referable to the female reproductive system, NEC Breast Symptoms 680.0 Lump or mass of breast Includes: Bump Hard spot Knot Local swelling Nodule 681.0 Pain or soreness of breast Includes: Redness Swelling, generalized Tenderness 683.0 Symptoms of nipple Includes: Bleeding Change in color Discharge Inflammation Inversion 684.0 Postpartum problems of breast Includes: Abnormal secretion Absence of milk Difficulty or inability in nursing Engorgement Excessive milk Improper lactation 690.0 Other symptoms referable to breast, NEC Includes: Sagging 690.0 Other symptoms referable to breast, NEC—Con. Too large Too small Symptoms Referable to the Eyes and Ears (700.0-799.0) 700.0 Blindness Includes: Complete 701.0 Other vision dysfunction Includes: Blurred Cloudy vision Diminished Dull vision Floaters Half vision Hazy vision Photophobia Spots 704.0 Discharge from eye Includes: Blood Excessive tearing Pus Watering 705.0 Eye pain and irritation Includes: Burning Inflamed Irritation Itching Swelling or mass Excludes: Eyelids 708.0 Abnormal eye movements Includes: Abnormal retraction Cross-eyed Pupil unequal Spasms Squinting Twitching 710.0 Symptoms of eyelids Includes: Drooping Inflammation Itching Swelling or mass 17 711.0 Sty 712.0 Pink-eye Includes: Conjunctivitis 715.0 Foreign body in eye 716.0 Eye injuries Includes: Black eye Burns Scratches Excludes: Foreign body 717.0 Abnormal appearance of eyes Includes: Abnormal eye protrusion Bloodshot Cloudy Dull Hazy 720.0 Other symptoms referable to the eye, NEC Ears 730.0 Deafness 731.0 Other hearing dysfunctions Includes: Acute hearing Diminished hearing Extraneous noises Ringing in ears Trouble hearing 734.0 Discharge from ear Includes: Bleeding 735.0 Earache Includes: Pain in ear 737.0 Plugged feeling in ear Includes: Blocked Cracking Popping Stopped up 738.0 Excess wax in ear 739.0 Abnormal ear size 740.0 Other symptoms referable to the ears, NEC Includes: Foreign body in ear Itching Swelling or mass 18 Symptoms and Problems Relating to Mental Health (800.0-899.0) 800.0 Anxiety Includes: Apprehension 801.0 Fears and phobias 805.0 Restlessness Includes: Hyperactivity Overactivity 806.0 Loneliness 807.0 Depression Includes: Bitterness Crying excessively Dejected Discontented Feeling lost Feeling low Feeling rejected Hopelessness Unhappy Worrying 810.0 Nervousness Includes: “Butterflies” Nerves Tension Upset 815.0 Behavorial disturbance Includes: Antisocial behavior Behavorial problems Irritability Quarrelsome Temper tantrums 820.0 Excessive smoking : 821.0 Alcohol related problems Includes: Alcoholism Drinks too much 822.0 Abnormal drug usage Includes: Excessive use of stimulants or depressants Misuse of medications or drugs 824.0 Delusions or hallucinations 826.0 Bad habits Includes: Chewing on hair Nail biting Thumb sucking 827.0 Obsessions or compulsions 828.0 Psychosexual disorders Includes: Frigidity Homosexuality Impotence 830.0 Other symptoms or problems relating to mental health, NEC Nonsymptomatic Visits According To Patient's Purpose (900.0-999.0) Visit for Examination 900.0 General medical examination Includes: Annual examination Checkup General examination Office examination Physical examination Regular examination Routine examination Excludes: Well-baby examination 901.0 Physical examinations Includes: Disability evaluation Driver’s license physical Examination for social secu- rity or insurance forms High school, college, or camp physical Industrial examination Military eligibility Preemployment physical Required company physical Return to work checkup General psychiatric examination Radiological examination Gynecologic examination Includes: Marital examination Pap smear Routine gynecologic examination Pregnancy examination, routine Includes: Postnatal Pregnancy confirmed Pregnancy unconfirmed 902.0 903.0 904.0 905.0 905.0 Pregnancy examination, routine—Con. Prenatal Routine check 906.0 Well-baby checkup or examination 909.0 Other examination, NEC Includes: Routine ophthalmologic examination Excludes: Followup visit Progress visit Visit for therapy 910.0 For medication Includes: Allergy shots Immunizations Injections—vitamins hormones New prescription Renewal of prescription Routine inoculations 911.0 For other therapy Includes: Physical therapy Radiation therapy Excludes: Family planning services and Visit for testing 920.0 Laboratory test Includes: EKG Throat culture To have blood work done Excludes: Pap smear—See gynecologic examination 921.0 Other testing, NEC Includes: Pulmonary function test Excludes: X-ray test Visit for family planning services 930.0 Counseling regarding: Abortion Contraceptives 19 930.0 Counseling regarding: —Con. Infertility Sterilization 931.0 Medication Includes: Instructions in the use of medications Request for medicine 932.0 Services Includes: Abortion performed Insertion (removal) of IUD Vasectomy 935.0 Other family planning services, NEC Visit for advice and instructions 940.0 Physician discusses or instructs patient regarding: Diet change or control Exercise Fitting of prosthesis Home—self-care Medication—injection (use) Use of crutches or cane 941.0 Patient seeks advice regarding situational problems Includes: Family problems Job dissatisfaction, problem Legal problems Marital problems School problems 942.0 Patient requests advice regarding non- specified reason Includes: Has a problem Wants to talk to doctor 950.0 Preoperative visit Includes: Discuss possibility of plastic surgery Instruction and counseling regarding imminent surgery Preoperative check To have surgery 960.0 Visit for minor surgery Includes: Ears pierced 20 960.0 Visit for minor surgery—Con. Joint manipulation Nonspecific lesion to be removed Excludes: Abortion Vasectomy 970.0 Referred visit (referred from another physician or agency) Followup care—progress visit 980.0 Progress visit—specified condition Includes: Examination or checkup for purpose of evaluating the progress of a specified condition: Check—cancer of mouth Check following thyroid- ectomy Check for hypertension Check—postflu Examination following strep throat Old duodenal ulcer—healed Postop—cancer of breast Status arthritis—check hands Surgical aftercare—cataract extraction 985.0 Progress visit—unspecified condition Includes: Examination of checkup for purpose of evaluating the progress of a nonspecified condition: Abnormal laboratory findings Blood pressure check Change or removal of cast Draining wound Healing wound Heart check Postop. visit Suture removal surgery following 990.0 Problems, complaints, symptoms, or reasons for visit, NEC Includes: Diabetes 997.0 Entry of “None” 998.0 Noncodable entry Includes: Out of scope 999.0 Illegible entry APPENDIX II ALPHABETIC INDEX OF TERMS Abdominal distension 542.0 fullness 542.0 pain 540.0 rigidity 542.0 swelling 542.0 Abnormal Abnormality breathing sounds 307.0 drug usage 822.0 ear size 739.0 eye appearanee 717.0 gait 421.0 hair 124.0 heart sounds 200.0 high blood pressure 205.0 lip color 505.0 low blood pressure 206.0 periods 653.0 protrusion (eye) 717.0 retraction (eye) 708.0 secretion (postpartum, breast) 684.0 sounds (respiratory) 307.0 stools 556.0 tongue color 525.0 Abnormal involuntary movements 050.0 eyes 708.0 muscles (see also Twitching) 050.0 Abortion counseling 930.0 performed 932.0 request 930.0 Absence (see also Lack of) appetite 545.0 feeling 059.0 hair 124.0 milk (postpartum) 684.0 Ache all over 013.0 ankle 400.0 Ache—Con. arm 405.0 back 415.0 back of head 410.0 cervical spine 410.0 elbow 405.0 face 410.0 fingers 405.0 foot 400.0 forearm 405.0 generalized 013.0 hand 405.0 hip 400.0 jaw 410.0 joints, not specified 425.0 knee 400.0 leg 400.0 limbs, not specified 425.0 lower back 415.0 lower extremity, part unspecified 400.0 lumbar 415.0 lumbosacral 415.0 neck 410.0 sacroiliac 415.0 shoulder 405.0 site unspecified 013.0 spine 415.0 thigh 400.0 thoracic spine 415.0 thumb 405.0 toe 400.0 upper back 415.0 upper extremity, part unspecified 405.0 upper spine 415.0 wrist 405.0 Acne 100.0 Activity over 805.0 21 Activity—Con. over (infants) 020.0 under (infants) 020.0 Acute hearing 731.0 Alcohol-related disturbances 821.0 Allergic skin reactions 112.0 Allergy shots 910.0 Amnesia 058.0 Anesthesia 059.0 Ankle ache 400.0 broken 400.0 cold 400.0 contracture 400.0 cramp 400.0 hot 400.0 hurt 400.0 injury 400.0 limited motion 400.0 pain 400.0 pulled muscle 400.0 soreness 400.0 spasm 400.0 stiffness 400.0 strain 400.0 swelling 400.0 Annual checkup 900.0 Antisocial behavior 815.0 Anus, symptoms referable to 560.0 Anxiety 800.0 Appetite abnormal 545.0 decreased 545.0 excessive 545.0 loss of 545.0 Apprehension 800.0 Arm ache 405.0 broken 405.0 cold 405.0 contracture 405.0 cramp 405.0 hot 405.0 hurt 405.0 injury 405.0 limited motion 405.0 pain 405.0 pulled muscle 405.0 soreness 405.0 spasm 405.0 stiffness 405.0 22 Arm—Con. strain 405.0 swelling 405.0 Ashen color 212.0 Athlete’s foot 106.0 Atrophy of extremities 420.0 Back, lower upper ache 415.0 contracture 415.0 cramp 415.0 hurt 415.0 injury 415.0 limited motion 415.0 pain 415.0 pulled muscle 415.0 soreness 415.0 spasm 415.0 stiffness 415.0 strain 415.0 swelling 415.0 Bad breath 502.0 complexion 100.0 habits 826.0 heart 216.0 taste 510.0 Balance, loss of sense of 069.0 Baldness 124.0 Bedwetting 601.0 Behavioral disturbances 815.0 Belching 570.0 Biliary colic 580.0 symptoms of 580.0 Bites 116.0 Bitterness 807.0 Black- eye 716.0 heads 100.0 out 214.0 Bladder problems (see 600.0-606.0) Bleed, Bleeding ear 734.0 eye 704.0 gastrointestinal 550.0 gingival 501.0 gums 501.0 lips 505.0 nose 300.0 from rectum 550.0 Bleed, Bleeding—Con. of rectum 560.0 tongue 525.0 tonsils 527.0 Blemishes 100.0 Blindness, partial or complete 700.0 Blind spots 701.0 Blisters nonallergic 116.0 tongue 525.0 Bloating, gas 543.0 Blocked feeling in ears 737.0 Blood in stools 550.0 poor 210.0 tired 210.0 vomiting 550.0 weak 210.0 Blood pressure abnormal 205.0 decreased 206.0 elevated 205.0 high 205.0 low 206.0 Bloodshot eyes 717.0 Blueness fingers 212.0 toes 212.0 Blurred vision 701.0 Blushing abnormal 104.0 excessive 104.0 Boils 106.0 Bowel, Bowels change in 556.0 dysfunction 556.0 Breaking nails 122.0 Breaking out 100.0 Breast bump 680.0 deformity 690.0 hard spot 680.0 knot 680.0 lump 680.0 mass 680.0 nodule 680.0 pain 681.0 redness 681.0 sagging 690.0 soreness 681.0 swelling 680.0 Breast, swelling—Con. generalized 681.0 local 680.0 tender 681.0 too large 690.0 too small 690.0 Breath, breathing bad 502.0 problem 307.0 shortness of 306.0 sounds, abnormal 307.0 Breathlessness 306.0 Brittle hair 124.0 nails 122.0 Bruises 116.0 Bulge (see Swelling and particular site) Bump (see Swelling and particular site) - Bunion 429.0 Burning eye 705.0 sensation (in chest) 322.0 skin 113.0 tongue 525.0 urination 604.0 Burns chemical 116.0 mouth 050.0 steam 116.0 sun 116.0 wind 116.0 Butterflies 810.0 Buzzing in ear 731.0 Calluses 108.0 Change in bowels 556.0 voice 325.0 Charleyhorse 400.0 Chest congestion in 321.0 pain in 322.0 pressure in 322.0 tightness 322.0 Chewing difficulties 500.0 on hair 826.0 Chills 001.0 Choking 528.0 Clammy skin 120.0 Cloudy eye appearance 717.0 vision 701.0 23 Clumsiness 421.0 Coated tongue 525.0 Coitus, painful 661.0 Cold 312.0 Cold ankle 400.0 arm 405.0 elbow 405.0 fingers 405.0 foot 400.0 forearm 405.0 hand 405.0 hip 400.0 knee 400.0 leg 400.0 lower extremity, part unspecified 400.0 shoulder 405.0 skin 120.0 thigh 400.0 thumb 405.0 toe 400.0 upper arm 405.0 upper extremity, part unspecified 405.0 wrist 405.0 Colic biliary 580.0 infantile 541.0 intestinal 540.0 NOS 540.0 Collapse 214.0 Color ashen 212.0 change in nail 122.0 change in nipple 683.0 change in skin 104.0 Coma 052.0 Compulsion 827.0 Conflict job 941.0 marital 941.0 Confusion 053.0 Congestion chest 321.0 nasal 301.0 sinus 304.0 Conjunctivitis 712.0 Constipation 554.0 Contraceptive counseling 930.0 Contracture (see Ache and particular site) Convulsions 054.0 Corns 108.0 24 Cough, coughing 311.0 phlegm 320.0 sputum 320.0 Cracked lips 505.0 nails 122.0 skin 120.0 Cramps (see also Ache and particular site) menstrual 652.0 stomach 540.0 Cross-eyed 708.0 Croup 314.0 Crying 807.0 infants 020.0 Cuts 116.0 Cyst site unspecified 015.0 skin 115.0 Dark urine 600.0 Deafness 730.0 Decreased appetite 545.0 blood pressure 206.0 pulse 200.0 Deformity breast 690.0 ears 757.0 Dehydration 007.0 Dejected 807.0 Delusion 824.0 Depression 807.0 Diaper rash 112.0 Diarrhea, functional 555.0 Diet control change 940.0 counseling 940.0 Difficulty breathing 306.0 chewing 500.0 nursing 684.0 swallowing 528.0 walking 421.0 Diminished hearing 731.0 vision 701.0 Discharge ear 734.0 eye 704.0 nipple 683.0 tonsils 527.0 Discharge—Con. umbilicus 126.0 vaginal 662.0 Discoloration nails 122.0 skin, 104.0 Discontented 807.0 Disorders (see also Disturbance) respiratory rhythm 307.0 respiratory sound 307.0 urinary 610.0 voice 325.0 Dissatisfaction, job 941.0 Distention abdominal 542.0 bladder 603.0 gas 543.0 Disturbance (see also Disorder) hearing 731.0 memory 058.0 sensation 059.0 sleep 062.0 smell 059.0 taste 059.0 touch 059.0 vision 701.0 Divorce proceedings 941.0 Dizziness 069.0 Draining, umbilicus 126.0 Dribbling 602.0 Drinking problem 821.0 Drip, postnasal 301.0 Drippy nose 301.0 Drooling, excessive 511.0 Drooping eyelid 710.0 Drop, dropping foot 422.0 sensation of pelvic floor 660.0 wrist 422.0 Dropsy 231.0 Dryness eye 705.0 hair 124.0 lips 505.0 mouth 510.0 nose 330.0 skin 120.0 Dull eye appearance 717.0 vision 701.0 Dysfunction (see Disorders, Disturbance) Ear abnormal size 739.0 blocked feeling 737.0 buzzing in 731.0 discharge 734.0 extraneous noises 731.0 pain 735.0 pierced 960.0 plugged feeling 737.0 pressure 737.0 ringing 731.0 unusual sounds 731.0 wax, excessive 738.0 Earache 735.0 Edema 231.0 EKG 920.0 Elbow (see Arm and particular condition) Elevated blood pressure 205.0 Empty bladder, inability to 603.0 Engorged nipple 683.0 Enlarged heart 220.0 liver 580.0 lymph nodes 232.0 spleen 240.0 Epigastrium pain 540.0 Epitaxis 300.0 Equilibrium, loss of sense of 069.0 Erection, painful 621.0 Excessive appetite 545.0 crying 807.0 crying (infantile) 020.0 drinking (alcohol) 821.0 drooling 511.0 hair 124.0 menstrual flow 653.0 milk secretion 684.0 phlegm 320.0 smoking 820.0 sputum 320.0 sweating 007.0 thirst 007.0 use of stimulants or depressants 822.0 wax in ear 738.0 Exhausted 004.0 Extremities atrophy 420.0 numbness 420.0 paralysis 420.0 25 Extremities—Con. wasting, 420.0 weakness, 420.0 Eye discharge 704.0 dryness 705.0 examination 909.0 inflamed 705.0 injuries 716.0 itching 705.0 pain 705.0 protrusion 717.0 red 717.0 swelling 705.0 tearing 704.0 watering 704.0 Eyelid closed 710.0 drooping 710.0 dropping 710.0 itching 710.0 red 710.0 swollen 710.0 symptoms of 710.0 Face ache 410.0 contracture 410.0 cramp 410.0 hurt 410.0 injury 410.0 limited motion 410.0 pain 410.0 pulled muscle 410.0 soreness 410.0 spasm 410.0 stiffness 410.0 strain 410.0 swelling 410.0 Fainting 214.0 Falling (out) of hair 124.0 (out) of nails 122.0 sensation 069.0 sensation of pelvic region 660.0 Family planning 930.0 problems 941.0 Fast breathing 307.0 heartbeat 200.0 pulse 200.0 Fatigue 004.0 26 Fears 801.0 Feeding problem 546.0 Feeling bad 005.0 blue 807.0 lost 807.0 low 807.0 numb 059.0 rejected 807.0 Fever 002.0 blister 505.0 Fidgety 805.0 infants 020.0 Fingers (see also Arm and particular condition) blueness 212.0 Fit 054.0 Flashes hot 650.0 light 701.0 Flatulence 543.0 Floaters 701.0 Flu 313.0 Fluid imbalance 007.0 retention 007.0 Flushing 104.0 Fluttering heart 200.0 Followup visit specified condition 980.0 unspecified condition 985.0 Foot (see also Ankle and particular condition) drop 422.0 Forearm (see Arm and particular condition) Foreign body (see also Injury) ear 740.0 eye 715.0 Freckles 104.0 Frequent menstruation 653.0 urination 601.0 Frigidity 828.0 Fullness 542.0 bladder 603.0 Functioning, Functional bowels 556.0 diarrhea 555.0 Fussy, infants 020.0 Gain, gaining weight 010.0 Gait, abnormal 421.0 Gallbladder, symptoms of 580.0 gallstones 580.0 Gas bloating 543.0 distention 543.0 excessive 543.0 Gastrointestinal bleeding 550.0 General, generalized ill-feeling 005.0 pain 013.0 symptoms of infants 020.0 weakness 004.0 Giddiness 069.0 Gingival bleeding 501.0 Glands, swollen 232.0 Grip 313.0 Groin, pain 540.0 Growth, lack of 009.0 Gums, bleeding 501.0 Gynecologic examination 904.0 Hair abnormal 124.0 dryness 124.0 excessive 124.0 loss of 124.0 symptoms of 124.0 Half-vision 701.0 Halitosis 502.0 Hallucinations 824.0 Hand (see Arm and particular condition) Hard spot (see Swelling and particular site) Hazy eye appearance 717.0 vision 701.0 Head (back of) (see Face and particular condition) Headaches 056.0 Hearing disturbance of 701.0 noises (nonpsychiatric) 701.0 Heart beats, irregular 200.0 burn 570.0 flutter 200.0 murmur 201.0 pain over 322.0 rapid 200.0 sounds, abnormal, increased 200.0 weak 216.0 Hemorrhage gastrointestinal 550.0 nose 300.0 vaginal 662.0 Hesitancy of urination 610.0 Hiccough 575.0 High blood pressure 205.0 temperature 002.0 Hip (see Ankle and particular condition) Hives 112.0 Hoarseness 325.0 Homosexuality 828.0 Hopelessness 807.0 Hot (see Cold and particular site) Hurt (see Ache and particular site) Hyperactivity 805.0 infants 020.0 Hyperesthesia 059.0 Hypersomnia 062.0 Hypertension 205.0 Hyperventilation 307.0 Hypotension 206.0 Iliac pain 540.0 Illegible item 999.0 Imbalance, fluid 007.0 Impacted sinuses 304.0 Impending litigation 941.0 Impotence 828.0 Improper lactation 684.0 Inability to nurse 684.0 to stand 421.0 to urinate 603.0 to walk 421.0 Incontinence of urine 602.0 Increased blood pressure 205.0 pulse 200.0 Indigestion 570.0 Infantile colic 541.0 Infected sinuses 304.0 Infectious disorders 106.0 Infertility counseling 930.0 female 665.0 male 620.0 Inflamed, Inflammation eye 705.0 eyelid 710.0 mouth 510.0 nipple 683.0 skin 113.0 throat 520.0 tonsils 527.0 27 Influenza 313.0 Ingrown nail 122.0 Inguinal pain 540.0 Injections of vitamins or hormones 910.0 Injury (see also Foreign body and particular site) eye 716.0 nose 116.0 skin 116.0 Inoculations 910.0 Insertion of IUD 932.0 Insomnia 062.0 Instruction for diet change or control 940.0 exercise 940.0 regarding imminent surgery 950.0 use of contraception 931.0 use of crutches or cane 940.0 Intestinal colic 540.0 Inversion of nipple 683.0 Involuntary movements 050.0 movements of eyes 708.0 urination 602.0 Irregular heartbeats 200.0 menstruation 653.0 pulsations 200.0 Irritability 815.0 infants 020.0 Irritation ear 735.0 eye 705.0 skin 113.0 Itching ear 740.0 eye 705.0 eyelid 710.0 rectum, anus 560.0 skin 113.0 vulva 663.0 Jaundice 579.0 Jaw (see Face and particular condition) Job dissatisfaction 941.0 Joint manipulation 960.0 Joints (see particular site) Joints, not specified ache 425.0 contracture 425.0 cramp 425.0 hurt 425.0 pain 425.0 28 Joints, not specified—Con. pulled muscle 425.0 soreness 425.0 spasm 425.0 stiffness 425.0 strain 425.0 swelling 425.0 Knee (see Ankle and particular condition) Knot (see Swelling and particular site) Labor, possible 667.0 Laboratory test 920.0 Lack of (see also Absence) growth 009.0 memory 058.0 physiological development 009.0 Large menstrual flow 653.0 Leaking amniotic fluid 667.0 Left quadrant pain 540.0 Leg (see Ankle and particular condition) Legal problems 941.0 Light, flashes 701.0 Lightheadedness 069.0 Lightness, sinus 304.0 Limbs, not specified (see Joints, not specified and particular condition) Limited motion (see Ache and particular site) Limping 421.0 Lips abnormal color 505.0 bleeding 505.0 cracked 505.0 dry 505.0 splitting 505.0 symptoms of 505.0 Litigation, impending 941.0 Liver, symptoms of 580.0 Loneliness 806.0 Loose stools 555.0 Loss of appetite 545.0 family member 941.0 hair 124.0 memory 058.0 sense of equilibrium (balance) 069.0 sense of smell 059.0 sense of taste 059.0 sense of touch 059.0 weight 011.0 Lost feeling 807.0 Low blood pressure 206.0 sperm count 620.0 Lower extremity, part unspecified (see Ankle and particular condition) Lower quadrant pain 540.0 Lumbar (see Back and particular condition) Lumbosacral (see Back and particular condition) Lump (see Swelling and particular site) Lymph nodes, swollen 232.0 Maladjustment, social 815.0 Marital conflict 941.0 Marital examination 904.0 Mass (see Swelling and particular site) Medical examination 900.0 Medication visit 910.0 Member of family, recent loss 941.0 Memory, disturbance of 058.0 Menopause symptoms 650.0 Menstrual cramps 652.0 disorders 653.0 tension 651.0 Migraine, headache 056.0 Milk absence of 684.0 excessive 684.0 Misuse of medication or prescription drugs 822.0 Mole 109.0 Movements, abnormal (involuntary) 050.0 bladder 602.0 bowel 556.0 eye 708.0 Murmur, heart 201.0 Muscles (see particular site) Muscles, unspecified (see Joints, not specified and particular condition) Nails biting 826.0 brittle 122.0 discoloration 122.0 falling out 122.0 splitting 122.0 spots 122.0 stained 122.0 Nasal bleeding 300.0 congestion 301.0 Nausea 572.0 Neck (see Face and particular condition) Nerves, Nervous, Nervousness 810.0 headache 056.0 Night discharge 601.0 Nightmares 062.0 Nipple discharge 683.0 inflammation 683.0 inversion 683.0 other symptoms 683.0 Nodule (see Swelling and particular site) Noises, heard (nonpsychiatric) 731.0 Noncodable entry 998.0 Nonspecific pain 013.0 Nose bleed 300.0 drippy 301.0 hemorrhage 300.0 injury 410.0 red 301.0 runny 301.0 stuffy 301.0 Not feeling well 005.0 Numbness of extremities 420.0 Obesity 010.0 Obsession 827.0 Oily hair 124.0 skin 120.0 Old age 065.0 Overactivity adult 805.0 infant 020.0 Oversize breast 690.0 ears 757.0 Ovulation pain 654.0 Pain (see also Ache and particular site) abdominal 540.0 breast 681.0 chest 322.0 ear 735.0 epigastrium 540.0 eye 705.0 face 402.0 generalized 013.0 groin 540.0 head 056.0 iliac 540.0 inguinal 540.0 knee 400.0 29 Pain—Con. left quadrant 540.0 lips 505.0 lower quadrant 540.0 mouth 510.0 nonspecific 013.0 over heart 322.0 pelvic 660.0 penis 631.0 rectal 560.0 respiratory 322.0 retrosternal 322.0 rib 322.0 right quadrant 540.0 scrotum 621.0 side of chest 322.0 sinus 304.0 sternal 322.0 testicle 621.0 throat 520.0 upper quadrant 540.0 urinary 604.0 vaginal 661.0 vulva 663.0 Painful coitus 661.0 erection 621.0 tongue 525.0 umbilicus 126.0 urination 604.0 Paleness 212.0 Pallor 212.0 Palpitation 200.0 Panic 800.0 Pap smear 904.0 Paralysis of extremities, partial or complete 420.0 Passed out 214.0 Passed stones 620.0 Peeling skin 120.0 Pelvis pelvic relaxed 660.0 sensation of dropping 660.0 symptoms of 660.0 Penis pain 631.0 swelling 631.0 Phlegm bloody 320.0 coughing up 320.0 30 Phlegm—Con. excessive 320.0 purulent 320.0 Phobias 801.0 Photo- phobia 701.0 sensitivity 112.0 Physiological development, lack of 009.0 Physical therapy 911.0 Pigmentation nails 122.0 skin 104.0 Pimples 100.0 Pink-eye 712.0 Plugged feeling in ear 937.0 Poison ivy, oak, sumac 112.0 Pooped 004.0 Poor blood 210.0 heart 216.0 Popping in ear 737.0 Possible labor 667.0 Postnasal drip 301.0 Postnatal examination 905.0 Postoperative visit specified condition 980.0 unspecified condition 985.0 Postpartum breast problems 690.0 Posture problems 422.0 Pregnancy examination 905.0 Prenatal examination 905.0 Preoperative visit 950.0 Pressure chest 322.0 ear 737.0 pelvis 660.0 sinus 304.0 Prickly feeling 059.0 Problem (see also Trouble) breathing 306.0 drinking (alcohol) 821.0 economic 941.0 family 941.0 female 670.0 legal 941.0 male 640.0 NOS 942.0 personal 942.0 posture 422.0 pregnancy 667.0 Problem—Con. school 941.0 sexual 828.0 Proceedings, divorce 941.0 Products of conception passed 667.0 Progress visit (see also Visit, followup) specified condition 980.0 unspecified condition 985.0 Psychiatric examination 902.0 Psy chosexual disorders 828.0 Pulled muscle (see Ache and particular site) unspecified site 425.0 Pulsations, Pulse decreased 200.0 increased 200.0 irregular 200.0 skipped beat 200.0 too fast 200.0 too slow 200.0 unequal 200.0 Pupils unequal 708.0 Purulent sputum 320.0 Pus eye 704.0 stools 556.0 Quarrelsome 815.0 Radiological examination 903.0 Rales 307.0 Rapid breathing 307.0 heart 200.0 Rash 112.0 diaper 112.0 Receding hairline 124.0 Rectal Rectum bleeding 560.0 itching 560.0 mass 560.0 pain 560.0 swelling 560.0 symptoms of 560.0 Red, Redness eye 717.0 breast 681.0 nose 301.0 skin 104.0 umbilicus 126.0 Referral from another physician or agency 970.0 Regurgitation 574.0 Relaxed pelvic floor 660.0 Removal of IUD 932.0 sutures 985.0 Renewal of prescription 910.0 Respiratory insufficiency 306.0 pain 322.0 rhythm disorders 307.0 sighing 307.0 sound disorders 307.0 Restlessness 805.0 Retching 572.0 Retention of fluid 007.0 urine 603.0 Retrosternal pain 322.0 Rib pain 322.0 Ridges, tongue 525.0 Right quadrant pain 322.0 Rigidity, abdominal 540.0 Ringing in ear 731.0 Rings on skin 104.0 Ringworm 106.0 Rough skin 120.0 Routine inoculations 910.0 Rundown 004.0 Runny nose 301.0 Sacroiliac (see Back and particular condition) Saliva, excessive 511.0 Scales 120.0 School problems 941.0 Scratches eye 716.0 skin 116.0 Scratchy throat 520.0 Scrotum, pain 631.0 Seizure 054.0 Senility 065.0 Sensation burning 059.0 burning (in chest) 322.0 falling 069.0 of suffocation 306.0 pelvis floor, dropping 660.0 smell (unusual) 059.0 taste (unusual) 059.0 31 Sexual problem 828.0 Shaking 050.0 Shortness of breath 306.0 Shots allergy 910.0 injections 910.0 Shoulder (see Arm and particular condition) Sick feeling 005.0 head 056.0 stomach 572.0 Side of chest, pain 322.0 Sighing respiration 307.0 Sinus infection 304.0 pain 304.0 problem 304.0 Skin bulge 115.0 burning 059.0 change in color 104.0 clammy 120.0 cold 120.0 inflammation 113.0 irritation 113.0 mass 115.0 moles 109.0 rash 112.0 red 104.0 rings 104.0 rough 120.0 sores 113.0 thickened 120.0 warts 111.0 waxy 120.0 wrinkles 110.0 Skipped beat 200.0 Sleep disturbances of 062.0 inability to 062.0 sleep walking 062.0 Slow pulse 200.0 Slowing of stream 610.0 Smell disturbance of 059.0 loss of sense of 059.0 unusual sensations of 059.0 Smoking, excessive 820.0 Smooth tongue 525.0 Sneezing 310.0 Sniffles 301.0 32 Social maladjustments 815.0 Sore glands 232.0 skin 113.0 throat 520.0 Soreness (see Ache and particular site) Sounds breathing 307.0 respiratory, abnormal 307.0 unusual, in ear 731.0 Spasm (see Ache and particular site) eye 708.0 eyelid 710.0 Spells 054.0 Spine, thoracic spine (see Back and particular condition) Spine, cervical, upper spine (see Face and partic- ular condition) Spitting up 574.0 Splitting lips 505.0 nails 124.0 Spots nails 124.0 skin 104.0 vision 701.0 Sprain (see Ache and particular site) Sputum coughing up 320.0 excessive 320.0 purulent 320.0 Squinting 708.0 Staggering 421.0 Stammering 067.0 Stand, inability to 421.0 Sterility female 665.0 male 630.0 Sternal pain 322.0 Stiffness (see Ache and particular site) Stomach cramps 540.0 pain 540.0 upset 570.0 Stones, passed 620.0 Stools abnormal 556.0 bloody 550.0 bulky 556.0 dark 556.0 fatty 556.0 Stools—Con. loose 555.0 pus in 556.0 unusual color 556.0 unusual odor 556.0 Stopped up ears 737.0 nose 301.0 sinuses 304.0 Strain (see Ache and particular site) Stream, slowing of 610.0 Stuffy nose 301.0 Stupor 052.0 Stuttering 067.0 Sty 711.0 Suffocation, sensation 306.0 Surgery, (minor) visit 950.0 Surgical aftercare specified condition 680.0 unspecified condition 685.0 Swallowing difficulties 528.0 Sweating Sweats excessive 007.0 night 007.0 Swelling abdominal 542.0 ankle 400.0 arm 405.0 back 415.0 back of head 410.0 breast, generalized 681.0 breast, local 680.0 cervical spine 410.0 ear 740.0 elbow 405.0 eye 705.0 eyelid 710.0 face 410.0 fingers 405.0 foot 400.0 forearm 405.0 generalized 015.0 hand 405.0 hip 400.0 jaw 410.0 joints, not specified 425.0 joints specified (see site) knee 400.0 leg 400.0 limbs, not specified 425.0 lower back 415.0 Swelling—Con. lower extremity, part unspecified 400.0 lumbar 415.0 lumbosacral 415.0 neck 410.0 pelvis 660.0 penis 631.0 sacroiliac 415.0 scrotum 631.0 shoulder 405.0 site unspecified 015.0 skin 115.0 testicle 631.0 tongue 525.0 tonsils 527.0 upper extremity, part unspecified 405.0 vagina 661.0 vulva 663.0 Swollen ankles 400.0 glands 232.0 Syncope 214.0 Taste disturbance of 059.0 loss of sense of 059.0 unusual sensation 059.0 Tearing of eye 704.0 Teeth, symptoms of 515.0 Temperature, high 002.0 Temper tantrums 815.0 Temporary loss of memory 058.0 Tender breast 681.0 skin 113.0 Tension headache 056.0 nervous 810.0 premenstrual 651.0 Test, laboratory 920.0 Testicle pain 631.0 swelling 631.0 Texture, change in skin 120.0 Thickened skin 120.0 Thigh (see Ankle and particular condition) Thin blood 210.0 Thirst, excessive 007.0 Throat culture 920.0 pain 520.0 33 Throat—Con. scratchy 520.0 soreness 520.0 Throwing up 572.0 Thumb (see Arm and particular condition) sucking 826.0 Tic 050.0 Tightness of chest 322.0 Time-zone syndrome 062.0 Tingling 059.0 Tired 004.0 blood 210.0 Toe (see Ankle and particular condition) blueness 212.0 Tongue bleeding 525.0 coated 525.0 mass 525.0 painful 525.0 smooth 525.0 swelling 525.0 symptoms of 525.0 Tonsils, symptoms of 527.0 Toothache 512.0 Touch, loss of sense of 059.0 Tremor 050.0 Trouble (see also Problem) breathing 306.0 eating 546.0 female 670.0 hearing 731.C job 941.0 marital 941.0 school 941.0 seeing 701.0 sleeping 062.0 walking 421.0 Twitching 050.0 eyes 708.0 Ulcer mouth 510.0 skin 113.0 tongue 525.0 vulva 663.0 Umbilical region, pain 540.0 Umbilicus discharge 126.0 draining 126.0 painful 126.0 red 126.0 Umbilicus—Con. symptoms of 126.0 unhealed 126.0 Under activity (infants) 020.0 weight 011.0 Unequal pulse 200.0 pupils 708.0 Unusual color of stools 556.0 Upper arm (see Arm and particular condition) Upper extremity, part unspecified (see Arm and particular condition) Upper quadrant, pain 540.0 Upset emotional 810.0 stomach 570.0 Urinary dysfunction 610.0 pain 604.0 symptoms NEC 610.0 Urination urinate frequent 601.0 hesitancy 610.0 inability to 603.0 painful 604.0 Urine blood 600.0 incontinence of 602.0 pus 600.0 retention of 603.0 unusual color 600.0 unusual odor 600.0 Use of orthopedic aids (instruction) 940.0 Vaccinations 910.0 Vaginal vagina atypical discharge 662.0 bleeding 662.0 brown discharge 662.0 discharge 662.0 disorders 661.0 mass 661.0 pain 661.0 swelling 661.0 Vasectomy advice regarding 930.0 request 932.0 Vertigo 069.0 Vision blurred 701.0 diminished 701.0 disturbance of 701.0 Visit, advice and instruction (see 940.0-942.0) Visit, examination eye 909.0 general medical 900.0 general psychiatric 902.0 gynecological 904.0 other 909.0 physical 901.0 pregnancy 905.0 radiological 903.0 well baby 906.0 Visit, family planning services counseling 930.0 medication 931.0 other 935.0 services 932.0 Visit followup (see Progress visit) Visit, minor surgery 950.0 Visit, preoperative 950.0 Visit, progress specified condition 980.0 unspecified condition 985.0 Visit, referral 970.0 Visit, testing laboratory 920.0 other 921.0 Visit, therapy medication 910.0 other therapy 911.0 Vitamins or hormones, injections 910.0 Voice, change in 325.0 Vomiting 572.0 blood 550.0 Vulvar disorders itching 663.0 mass 663.0 Vulvar disorders—Con. pain 663.0 swelling 663.0 ulcer 663.0 Walk, Walking difficulty in 421.0 inability to 421.0 Warts 111.0 Wasting of extremities 420.0 Watering of eye 704.0 Waxy skin 120.0 Weak blood 210.0 heart 216.0 Weakness generalized 004.0 of extremities 420.0 Weight gain 010.0 loss 011.0 under 011.0 Well-baby examination 906.0 Welts 115.0 Wheezing 307.0 Whiteheads 100.0 Worn out 004.0 Worrying 807.0 Wounds (skin) bites 116.0 blisters, nonallergic 116.0 bruises 116.0 burns 116.0 cuts 116.0 scratches 116.0 Wrinkles 110.0 Wrist (see also Arm and particular condition) drop 422.0 X-rays 903.0 35 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods research.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies. —Reports presenting analytical or interpretive studies basedon vital and health statistics, carrying the analysis further than the expository types of reports in the other series. Documents and committee veports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survev.— Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons. Data from the Institutional Population Surveys.—Statistics relating to the health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce.—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—sepecial analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HRA Rockville, Md. 20852 DHEW Publication No. (HRA) 75-1337 Series 2-No. 63 United States Life Tables by Dentulous or Edentulous Condition, 1971 and 1957-58 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration Library of Congress Cataloging in Publication Data Greville, T.N. E. United States Life tables by dentulous or edentulous condition, 1971 & 1957-58. (Data evaluation and methods research, series 2, no. 64) (DHEW publication no. (HRA) 75-1338) Includes bibliographical references. Supt. of Doc.: HE 20.6209:2/64. 1. Tooth loss—United States—Statistics. 2. United States—Statistics, Medical. I. Title. II. Series: United States. National Center for Health Statistics. Vital and health statistics. Series 2: Data evaluation and methods research, no. 64. IIL. Series: United States. Dept. of Health, Education, and Welfare. DHEW publication, no. (HRA) 75-1338. [DNLM: 1. Dental health surveys—Tables. 2. Mouth, Edentulous—Tables. W2A N148vb no. 64 1974] RA409.7145 no. 64 [RK34.U6] 312° .01°82s [312’.3°0476] ISBN 0-8406-0018-6 74-7204 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 - Price 60 cents DATA EVALUATION AND METHODS RESEARCH Series 2 Number 64 United States Life Tables by Dentulous or Edentulous Condition, 1971 and 1957-58 Numnibers of survivors by age, sex, and dentulous or edentulous condition, and average remaining lifetime by age and sex, classified by dentulous and edentulous years. Based on data collected in household interviews in 1971 and 1957-58. DHEW Publication No. (HRA) 75-1338 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration National Center for Health Statistics Rockville, Md. August 1974 NATIONAL CENTER FOR HEALTH STATISTICS EDWARD B. PERRIN, Ph.D., Director PHILIP S. LAWRENCE, Sc.D., Deputy Director JACOB J. FELDMAN, Ph.D., Acting Associate Director for Analysis GAIL F. FISHER, Associate Director for the Cooperative Health Statistics System ELIJAH L. WHITE, Associate Director for Data Systems IWAO M. MORIYAMA, Ph.D., Associate Director for International Statistical Programs EDWARD E. MINTY, Associate Director for Management ROBERT A. ISRAEL, Associate Director for Operations QUENTIN R. REMEIN, Associate Director for Program Development PHILIP S. LAWRENCE, Sc.D., Acting Associate Director for Statistical Research ALICE HAYWOOD, Information Officer OFFICE OF STATISTICAL METHODS MONROE G. SIRKEN, Ph.D., Director E. EARL BRYANT, M.A., Deputy Director Vital and Health Statistics—Series 2-No. 64 DHEW Publication No. (HRA) 75-1338 Library of Congress Catalog Card Number 74-7204 FOREWORD Life table techniques are very useful in the joint analysis of morbidity and mortality phenomena which can be viewed in cohort terms. F requently, however, the available data do not satisfy the data needs specified by the life table models and assumptions are required. In this instance, we wished to estimate dentulous longevity which depends on mortality as well as on the incidence of edentulousness, the condition of having lost all of one’s natural teeth. However, the mortality and morbidity statistics needed to estimate dentulous longevity were only partly available. We were particularly interested in comparing the sexes with respect to the lengthening of their recent dentulous longevity. Morbidity data were available for two periods, 1957-58 and 1971, from the Health Interview Survey. During each period, the edentulous prevalence rate was lower for males than females, and between the two periods there was a substantial reduction in the prevalence of edentulousness for each sex. The mortality picture was somewhat different. During 1957-58 and 1971, mortality rates were lower for females than for males, and between the two periods the reductions in the mortality were less for males than for females. This report presents selected functions of abridged edentulous life tables for 1971 and 1957-58. Life table statistics are presented for males, females, and for the combined sexes. The major findings with respect to dentulous longevity are summarized in table D on page 4. Thus, the average remaining dentulous lifetime (average length of life without having lost all of one’s natural teeth) at age 20 increased by more than 2 years between 1957-58 and 1971. During the same period, the average remaining lifetime at age 20 increased by 1 year. The average remaining dentulous lifetime was greater for females than for males during both calendar periods but the difference was greater during the latter period, indicating that females experienced greater gains than males in dentulous longevity between the calendar periods. Calculating dentulous longevity required two kinds of data which were not available, namely, age-specific incidence rates of edentulousness and age- specific mortality rates for the edentulous and for the dentulous popula- tions. The problems were resolved as follows. The incidence rates of edentu- lousness were obtained (appendix III) essentially by differencing the age-specific prevalence rates which were available from the National Health Interview Survey. (Fortunately, the prevalence rates of edentulousness in- creased montonically with advancing age.) With respect to the unavailable mortality data, it was assumed that dentulous and edentulous persons are subject to the same rates. Whether or not these assumptions are viable de- pends somewhat on the analytical objectives. If the primary objective is to compare sex differences in the lengthening of dentulous longevity, the as- sumptions are probably more tolerable than if the objective is to estimate the length of dentulous life per se. The principal justification for estimating dentulous longevity was meth- odological rather than substantive. We hope that this demonstration will stimulate applying life table techniques to other morbid conditions than edentulousness. It is noteworthy that the life table model used in this report is applicable to many kinds of morbid conditions, such as permanent impair- ments and chronic conditions, from which there is no chance of recovery. The mortality and morbidity statistics that would be required are not likely to be available for any of these morbid conditions, however, and it is doubt- ful that the assumptions that were made to overcome the problem of un- available edentulous data would be appropriate for most other morbid conditions. Therefore, other kinds of assumptions would have to be devised. Life table statistics presented in this report are subject to sampling errors and nonsampling errors. They are subject to sampling errors because the Health Interview Survey, the source of the edentulous prevalence rates is a sample data system. Sampling errors of the prevalence rates and procedures for calculating sampling errors of life table statistics are presented in appen- dix II. The life table statistics are subject to nonsampling errors including measurement errors and errors due to the unavailability of specified mortal- ity and morbidity data on edentulousness. Although estimates of nonsam- pling errors are not presented in this report, there can be little doubt that they would be larger than the sampling errors. Monroe G. Sirken, Director, Office of Statistical Methods CONTENTS Foreword Introduction Historical Survey . Interpretation of the Tables General Assumptions ‘ " Survival Rates and Transition Probabilities . Expectation of Dentulous and Edentulous Life References List of Detailed Tables’ Appendix I. Technical Notes on Methodology Source of Data 3 . Calculation of Numbers of Survivors Calculation of Average Remaining Lifetime Appendix II. Reliability of Estimates . Standard Errors of Prevalence Rates Standard Errors of Numbers of Survivors Standard Errors of Average Remaining Lifetime . Appendix III. Implied Incidence Rates Page iii 10 10 10 12 14 14 16 17 18 vi SYMBOLS Data not available----------eeeeme eee Category not applicable--------ereesmmeeoceaceaanes Quantity zero Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision--------------seceeeneeeen- * UNITED STATES LIFE TABLES BY DENTULOUS OR EDENTULOUS CONDITION, 1971 AND 1957-58 T. N. E. Greville, Ph.D. Office of Statistical Methods INTRODUCTION Data on edentulousness, the condition of having lost all of one’s natural teeth, have been obtained through the Health Interview Survey of the National Center for Health Statistics on two occasions, in 1957-58! and in 1971.2 By com- bining these data with life tables for comparable periods, one can estimate the probability that a dentulous person (that is, still having at least one natural tooth) at a specified age will be still alive but edentulous at a subsequent age. One can also estimate, for a dentulous person at a stated age, the division of the average remaining years of life into dentulous and edentulous years. It is the purpose of this report to present the results of such calculations. The comparison of the re- sults for the two time periods is of particular interest. HISTORICAL SURVEY In 1943, Klein® published life tables for teeth, based on data collected by the Committee on the Costs of Medical Care and the Public Health Ser- vice and reported by Collins.* The underlying data provided the number of natural teeth pres- ent in the mouth of each respondent, and the values in the tables were average numbers of teeth remaining at successive ages. Thus these were life tables for teeth, not for persons. From a survey of all dental practitioners, Evans and Cellier’ obtained information on all extractions performed in South Australia during a 3-week period in November 1966. From these data they estimated incidence rates of edentu- lousness by age and sex. Incidence rates like those developed in the Australian study appear to be a prerequisite for constructing the kind of tables appearing in the present report. Because only prevalence rates were available (and not incidence rates), it was necessary to make certain artificial assumptions which will be described later. Although data on prevalence of edentulous- ness have also been obtained from the Health Examination Survey® of the National Center for Health Statistics, it was preferred to utilize the data from the Health Interview Survey because they are based on a much larger sample of re- spondents and they also provide an interesting comparison, having been obtained on survey dates 13 years apart. Table A suggests that the findings of the Health Examination Survey are consistent with the data of the Health Interview Survey utilized in this report, even though data for exactly comparable age intervals are not available from the two surveys. Table A. Prevalence rates of edentulous- ness per 100 persons from different surveys, by sex: United States Health Health Interview Examination Surveys, ages Survey 3 Sex 15 and over ages 18-79 1971 1957-58 1960-62 Rate per 100 persons Both sexes-- 15,7 18.9 18,1 Female---- 16.8 20.0 19.7 Male------ 14,5 17.6 16.5 INTERPRETATION OF THE TABLES General Assumptions It has already been stated that in order to construct the kind of tables presented in this report when only prevalence rates of edentulous- ness (and not incidence rates) were available, it was necessary to make certain artificial assump- tions. In general terms, it was assumed that the prevalence rate of edentulousness in a given age interval in the hypothetical stationary popula- tion associated with the life table” is the same as in the actual sampled population, and further that age-specific mortality rates are the same for dentulous and edentulous persons. One would expect that neither assumption is in accord with actual facts. The current prevalence of edentu- lousness in a given cohort depends in a compli- cated way upon the previous history of the cohort, and it is clear from the data of the three surveys as shown in table A that the incidence of edentulousness has declined markedly in recent years. As to the second assumption, it is likely that loss of all teeth is negatively correlated with the general state of health. Therefore the values in the tables must be regarded as rough approxi- mations. A more technical discussion of these assump- tions and detailed description of the methodol- ogy employed are found in appendix I. Appen- dix II concerns the sampling variability of the prevalence rates of edentulousness and the nu- merical quantities calculated from them. Appen- dix III contains a technical discussion of the incidence rates implied by the assumptions used. Survival Rates and Transition Probabilities Tables 1 and 2 show the number of survivors at the indicated ages in a hypothetical life-table cohort starting from 100,000 births, subdivided by dentulous and edentulous condition. Table 1 is based on 1971 prevalence rates of edentulous- ness from the Health Interview Survey and on 1971 life tables.” Table 2 is based on 1957-58 prevalence data and 1958 life tables. The following illustration will help to clarify the way in which the values in these tables may be interpreted. By table 1, the proportion of females alive at age 25 who survive to age 55 is 89,291 + 97,081 = .91976. Under the assump- tion previously stated, this value applies equally to dentulous and edentulous females. Therefore the number of survivors to 55 of the 1,631 who were already edentulous at 25 is 1,631x.91976 = 1,500. The remaining 21,894 - 1,500 = 20,394 edentulous females at. age 55 are survivors of the 95,450 who were dentulous at 25. Thus if we choose to interpret such ratios as probabilities, the probability that Table B. Indicated probabilities that a dentulous person atage 20 will be alive but edentulous at ages shown by sex, based on 1971 prevalence rates and 1971 life tables: United States Both Age Sexes Female | Male Probability 45 yearSee===== - .115 .128 .103 55 years==e=we== +210 «222 .198 65 years==eeeca= «283 .308 «257 75 yearsees=ee« - «273 .344 .205 Table C. Indicated probabilities that a dentulous person at age 20 will be alive but edentulous at ages shown by sex, based on 1957-58 prevalence rates and 1958 life tables: United States Age Bosh o Female | Male Probability 45 yearseem-mecceaa .145 .152 +136 55 years-—eweecaee« .261 .281 «239 65 years~eeeeeaaa «341 «391 «293 75 yearseeeeeecea= «293 +372 +223 a dentulous female aged 25 will be alive but edentulous at 55 is 20,394 + 95,450 = .214. Tables B and C show the indicated probabili- ties, on the basis of 1971 and 1957-58 data, respectively, that a dentulous person at age 20 will be alive but edentulous at the ages shown. The probability that a person now dentulous at a given age will survive to a subsequent age but become edentulous may be regarded as the product of two probabilities: The probability of survival and the probability of being edentulous if alive at the subsequent age. In every instance the survival rate for females is greater than for males, and in all but a few cases the prevalence rate of edentulousness is also greater for females. Consequently, every probability for females in tables B and C is greater than the corresponding value for males. In general, up to age 65 the increase with age in the prevalence rates of edentulousness is more rapid than the decrease in the survival rate from age 20, while the reverse is true between ages 65 and 75. Therefore in most instances the proba- bilities in tables B and C increase with advancing age up to 65 and then decline between 65 and 75. The exception in the case of females in 1971 is due to the slow decrease in survival rates even up to age 75. On the basis of table 1, the probability that the same dentulous female at age 25 will survive to age 55 and remain dentulous is simply 67,397 + 95,450 = .706. Now, the probability that this female will survive to age 55 regardless of tooth loss is 89,291 + 97,081 =.920. Clearly this must be the sum of the two probabilities that she will survive in a dentulous and an edentulous condition. Thus the probability of edentulous survival to age 55 could have been calculated more simply as .920 —.706 = .214. This is the same value previously obtained. Expectation of Dentulous and Edentulous Life If age-sex specific rates of mortality are the same for dentulous and edentulous persons, the same is true of the average remaining lifetime, also called the expectation of life. However, it is now possible to subdivide the average remaining lifetime of a dentulous person into the years to be spent in the dentulous and the edentulous state. The results are shown in tables 3 and 4, based on the data of 1971 and 1957-58, respec- tively. It should be noted that in the computation of the average remaining years of edentulous life, those persons who never become edentulous are, in effect, credited with zero years. These figures would be much larger if only those who eventually become edentulous were included in the average. Because of the greater prevalence rates of edentulousness for females, the proportion of edentulous years shown in tables 3 and 4 is con- sistently greater for females than for males in both time intervals. For each sex, this propor- tion is substantially less for 1971 than for 1957-58, except at the oldest ages shown. Table D compares the total average remain- ing lifetime and the average remaining dentulous lifetime at selected ages, by sex, for the two time periods considered. Table D. Average remaining lifetime and average remaining dentulous lifetime of a den- tulous person by age and sex: United States, 1971 and 1957-58 Total Dentulous Survey period and age “ " Bot Bot SERGE Female | Male SHEE Female | Male 1971 Average remaining lifetime in years 20 yearsS===-===-ececmceeeme maaan 53.2 56.7 49.8 42.0 43,5 40,6 30 yearS-====m-eeccmcececccce—————— 43.9 47.1 40,8 33.8 35.3 32.4 40 yearS====-==m-ceeme een 34.7 37.7 317 26.2 27.9 24.6 50 years====---emeemmcce emcee 26.1 28.9 23.3 19.3 20.8 17.8 60 years===--=m-ccmmeme emma eo 18.5 20.6 16.1 14,1 15.2 12.8 70 years==-==-emeeeeeme meee a 12.1 13.4 10.6 9.7 10.2 9.0 1957-58 20 yearS===--=--cecemececmcmmeeean 52.2 55,2 49.4 39.7 40.4 39.1 30 yearS===----cemmecemeecn eee ———— 42.8 45,6 40.2 31.2 32.0 30.3 40 years----=--seeemmcccee———————— 33.6 36.2 31.1 23.5 24,2 22,7 50 yearS=------ccemmmmmmmmmeeeeeee 24.9 27.3 22.7 17.5 18.1 16.9 60 years===----eeemceemmcmm ena 17.3 19.1 15.6 12.6 13.3 12.0 70 years=------cmcmcmcmmmeeeeeeeon 11.1 12.1 10.1 9.1 9.3 8.7 0O0O0 REFERENCES status—life tables for teeth. J. Am. Dent. 1U.S. National Health Survey: Loss of teeth. Health Statistics. PHS Pub. No. 584-B22. Public Health Service. Washington. U.S. Government Printing Office, Sept. 1960. 2National Center for Health Statistics: Unpublished data for 1971. Division of Health Interview Statistics. Klein, H.: Tooth mortality and socio-economic 30:80-95, Jan. 1, 1943. Collins, S. D.: Frequency of dental services among 9,000 families, based on nation-wide periodic canvasses 1928 31. Pub. Health Rep. 54:629-657, Apr. 21, 1939. 5Evans, G. G. L., and Cellier, K. M.: A survey of tooth loss in South Avstralia, Aust. Dent. J. 15:299-302, Aug. 1970. 6 National Center for Health Statistics: Total loss of Assoc. teeth in adults. PHS Pub. No. 1000-Series 11-No. 27. Public Health Service. Washington. U.S. Government Printing Office, Oct. 1967. "National Center for Health Statistics: Vital Statis- tics of the United States, 1971, Vol. 11, Part A, Section 5, Life Tables. DHEW pub. No. (HRA) 74-1147. Wash- ingeon, U.S. Government Printing Office. 8 National Center for Health Statistics: Vital Statis- tics of the United States, 1958, Vol. 1, Section 5, Life Tables. Public Health Service. Washington. U.S. Govern- ment Printing Office, 1960. 9Hansen, M. H., Hurwitz, W. N., and Madow, W. G.: Sample Survey Methods and Theory, Vol. II, Theory. New, York, John Wiley & Sons, 1953. pp. 107-109. 10gpiegelman, M.: Introduction to Demography, re- vised ed. Cambridge. Harvard University Press, 1968. pp. 189-190. jordan, C.W.: Life Contingencies, 2d ed. Chicago. Society of Actuaries, 1967. Equation (14.37a), p. 280. Table LIST OF DETAILED TABLES Survivors to stated ages by sex and dentulous or States, 197 lemme mmm mm mom ee ee ee Survivors to stated ages by sex-and dentulous or States, 1957-58 mm Average remaining lifetime of a dentulous person, and edentulous years, and proportion of edentulous States, 197 lemme mmm eee Average remaining lifetime of a dentulous person, and edentulous years, and proportion of edentulous States, 1957=58== == mmm mm eee edentulous condition: United edentulous condition: United average remaining dentulous years, by sex and age: United average remaining dentulous years, by sex and age: United Page Table 1. Survivors to stated ages by sex and dentulous or edentulous condition: United States, 1971 Sex and age Total Dentulous Edentulous Both sexes Number 20 yearS==----=m-semeemeee eee ee ee em—m————————— 96,847 96,527 320 25 yearS==m--ermmmmmeme seem eee ———————————————— 96,146 94,723 1,423 30 years==-==-mecmcmemmecccee meee eee ee ————————— 95,462 91,392 3,570 35 yearS==---m--ecmeeecmecceeeeeeecccce ec ——————— 94,659 88,450 6,200 40 yearS===-==---mmmeeeeccmemee mec e em ——————— 93,499 84,832 8,667 45 yearS------------e-eeee meme mecmem————————- 91,781 80,336 11,445 50 yearS====-e-emmmemececcec meee eee ee ———————————— 89,234 73,716 15,518 55 yearS-=-=cmemmmmeeeccm eee ———————————————————— 85,439 64,874 20,565 60 years-----=---mmemmmmecmceeceeeeceeee————oo— 80,000 55,088 24,912 65 years----------m-eemmememeeceeemeeeemcc——————— 72.326 44,806 27,520 70 yearS-=-----mmeeeeee emcee eee meee —— em ——— 62,482 33,834 28,648 75 yearS-------meeememmem emcee meee eeecee—e——- 49,841 23,306 26,535 Female 20 years---------c-mmmemeeeemee meee meeee——————ao 97,440 97,128 312 25 years---------eememmmeeememe meee —- 97,081 95,450 1,631 30 years--------------meeeeeeeeeeeeeee—ee—————- 96,664 92,672 3,992 35 years=---------eeeememeeeeeeee eee cm eee ccccoa———— 96,096 89,196 6,900 40 years----------meeemeeeccceeeemcem emcee —————— 95,240 85,392 9,848 45 year§==--mmc-me-meeee emcee ee eeee——————— 93,933 81,243 12,690 50 year§----------eemcememeemeeeeeee eee —eeeeaaao 92,035 75,432 16,603 55 years---------ccmeemmmmemmmc meme mme————aa 89,291 67,397 21,89 60 years----------meeecmemmemcceeemcemeeecem—————— 85,469 58,692 26,777 65 years=-------s---emmmmemcememeeemcceecee——————— 79,951 49,777 30,174 70 years=--------emeeememmemememe me memmeeee—e——o— 72,291 39,066 33,225 75 yearSm=-=m=mccmmemeoeemeeeemeeee ee —e————e—- 61,161 27,365 33,596 Male 20 years~=---=--m---memmmemeememmeemeee—e————m———o 96,280 95,943 337 25 yearS=-----------smmmeeeemmeeemacemee—e—————= 95,229 94,020 1,209 94,280 91,131 3,149 03,223 87,769 5,454 91,776 84,333 7,443 89,642 79,459 10,183 86,435 72,009 14,426 81,582 62,345 19,237 74,555 51,495 23,060 64,812 39,911 24,901 52,974 28,791 24,183 39,004 19,202 19,802 Table 2. Survivors to stated ages by sex and dentulous or edentulous condition: United States, 1957-58 Sex and age Total Dentulous | Edentulous Both sexes Number 20 yearsS-----=----emceeeme eee eeeeeee—————— 95,991 95,127 864 25 yearS===-=- mmm eee eee eee eee eeee— eee 95,400 93,254 2,146 30 years---mrmmrmmemrmmm mmm mmm —————————————————— 94,785 91,373 3,412 35 years=---ememrmmmemecenm meee cee eee, —— er —————— 94,003 87,799 6,204 40 years---=----eeemeccccecmecececc cee c———————- 92,900 83,982 8,918 45 yearSee=memmrmmmmmm mmm ————————————— 91,191 76,600 14,591 50 yearS=-===---emmecmemem eee eeeeeeeeee em ——————— 88,578 68,737 19,841 55 yearS=-----==-mcc-cmeceee meee eeeeeeee—cee———— 84,524 58,935 25,569 60 years-------=-e--ceem eee 78,792 48,772 30,020 65 yearS=-=-=mmmmmmee eee meee meee eeeeceeemeee———- 70,791 37,696 33,095 70 years=------mmccmmeee meee e eee ceeee———————- 59,556 26,562 32,994 75 yearS=------emmmmm eee meee cee eceece—————— 46,178 17,848 28,330 Female 20 yearS=-=-m=mccmmmme eee meme mmmeeee meee 96,615 95,745 870 25 yearS=-=-=-m-mm-mm-eee meme meecemeeemmmeememenee 96,280 93,680 2,600 30 yearsS====---ememeeeeeeeecemeeeceeeeeceee——————— 95,848 91,535 4,313 35 years=----------cmmme meena 95,243 88,195 7,048 40 years-----------mcmmmmmmmme mec em eam 94,373 84,653 9,720 45 yearS=------cemecmcee eee eee ——— 93,073 77,669 15,404 50 years=-------c---mee eee m—————— 91,123 70,347 20,776 55 yearS=--mmm-memcemme meme meee 88,220 60,475 27,745 60 years-----------cmeceeeeccmmeemeeee emma 84,080 50,364 33,716 65 yearS----------m-emmeeeecemeeemceeecce——————- 78,029 39,912 38,117 70 years-------m-cmmeeeeeeeemee eee eee ————— 68,708 29,132 39,576 75 yearS-------cmccmmmme meme emcee e—————— 56,214 20,068 36,146 Male 20 years---==m-m-m--meeeeeeeeeemeememeemee——————— 95,393 94,534 859 25 year§---===-sm-meemeememmeeeeemeeeo—e-—eo-—---- 94,537 92,883 1,654 30 years----=---mmmeemmememececoeeeoo--o-—o--oe== 93,736 91,299 2,437 35 years-----=--=mmeemmmmmmceeoo-oo--o--eeeeoo== 92,772 87,484 5,288 40 years----===-mmememmmmemeemeem— esses —————————— 91,431 83,385 8,046 45 years-----=---mm-mmmecemeeeccceeecmeecee—————— 89,305 15,397 13,708 50 years=------===--sememmmmeeememeeee—eeooooo—-- 86,034 67,193 18,841 55 years----------eeemecememmemememeecccceoee———— 80,744 57,409 23,335 60 years----==--m-m-mmmemmemcceeemccceeeee—e—e=—- 73,528 47,131 26,397 65 years---====-=me-emeecemmeeceeeceoe——o————eo-- 63,767 35,486 28,281 70 years=---=---==mmmm-m-mmcccecomeee—ceceem—————— 50,985 24,065 26,920 75 years=-------=memmememceeeeeeeo omc ccccoceo——= 37,115 15,737 21,378 Table 3. Average remaining lifetime of a dentulous person, average remaining dentulous by sex and age: United and edentulous years, and proportion States, 1971 ‘ of edentulous years, Proportion of Sex and age Total Dentulous | Edentulous edentulous years Average remaining lifetime Both sexes in years 20 yearS=s=--=c--mcmceee eee 53.2 42,0 11.2 +21 25 years-=------emmm meee meee ee 48.6 37.8 10.8 vol 30 years ==-=-cemcmc meme eee 43.9 33.8 10.1 .23 35 years=------ecemce mcm 39.3 30.0 9.3 24 40 yearS-----mmmmmmm mcm 34.7 26.2 8.5 24 45 yearsS=---=--mcmee mmm 30.3 22.5 7.8 .26 50 years=-----memmmmm meme 26.1 19.3 6.8 .26 55 years --=ememcme meee 22,1 16.6 5.5 +25 60 years---=-mc-cem meme 18.5 14.1 4.4 24 65 years=----e cme 15.1 11.7 3.4 ve 70 years=---=---e- cme 12.1 9.7 2.4 .20 75 years=s=--e--memcme meme eee 9.5 8.0 1.5 .16 Female 20 yearS--=-m--emmmmm cme eee emma 56.7 43.5 13,2 23 25 yearS--=--smeeeccmm meee mee 51.9 39.2 12.7 oly 30 years-----cemcmmm meee 47.1 35.3 11.8 .25 35 years------c-cmmccemm ccc mmm eee = 42.4 31.6 10.8 .25 40 years--====-cmccmn cme 37.7 27.9 9.8 .26 45 yearS---mecme comme cmem———o 33.2 24,1 9.1 ve 50 years--==----eemcemmmm cee eee 28.9 20.8 8.1 .28 55 years ---eee comme mmm 24,7 17.9 6.8 .28 60 years---cememmcm mmm eee 20.6 15.2 5.4 +26 65 years=-sem cece 16.9 12.5 4.4 +26 70 years ------eme ccm mecca 13.4 10.2 3.2 24 75 yearS--sm=-cemmecc cnc 10.4 8.5 1.9 .18 Male 20 years ---==c-mmcmm mmm eee ao 49.8 40.6 9.2 .18 25 yearS------cmmmem emcee 45,3 36.3 2 .20 30 years=--=e--meemmmmme me meee 40,8 32.4 8.4 sal 35 years---=--memeccm mee meee 36.2 28.5 7.7 +21 40 years=---=---ceccmmmm meee 31.7 24,6 7.1 $22 45 years--=--=-cmcm mca 27.4 20.9 6.5 24 50 years ---=-emcmem meee 23.3 17.8 543 24 55 years=--==---- Smee meee meme ee 19.5 15.1 4.4 23 60 years---=cemcmm meee 16.1 12.8 3.3 .20 65 yearS--==-cecmm ccm emma 13.2 10.8 2.4 .18 70 years=-=m-memem cme emcee 10.6 9.0 1.6 .15 75 yearS-=eeemeece mmc eee 8.4 7.4 1.0 «12 Table 4, Average remaining lifetime of a dentulous person, average remaining dentulous and edentulous years, and proportion of edentulous years, by sex and age: United States, 1957-58 Proportion of Sex and age Total Dentulous | Edentulous edentulous years Average remaining lifetime Both sexes in years 20 yearS======-=--mm em eem emcee meee 52.2 39.7 12.5 .24 25 years -=--=m-m emcee emma eem 47.5 35.5 12.0 3 30 years ----=--mmcmmcmmmm eee een 42.8 31.2 11,6 «27 35 years--==-cmm cme e eee mem 38.1 27.3 10.8 «28 40 years-===-=-c-mcmcmeececceeme as 33.6 23.5 10.1 .30 45 years----=-=c-mmmce mecca 29.1 20.5 8.6 .30 50 years -====-cmmccme een 24.9 17.5 1.4 30 55 yearS===--cmecmmm emcee meen 21.0 15.0 6.0 .29 60 years -=-==-=ccmcomem emcees 17.3 12.6 4.7 vol 65 years ====-== cece meeee oa 14.0 10.6 3.4 .24 70 years --=-== mcm mce cm mmmeeeeeen 11.1 9d 2,0 . 18 75 years -=mem-m mmc eeem eee 8.6 7.3 1.3 +15 Female 20 years -======cmcecee mm mmemme ean 55,2 40,4 14.8 sind 25 years -=-=m cm cm meee mmeeceea 50.3 36.2 14.1 .28 30 years -==----mmmemme meme eee oo 45,6 32.0 13.6 .30 35 years -=-==-mccmm ecm enema 40,8 28.1 12.7 31 40 years -==== cee cem cence 36.2 24,2 12.0 33 45 years ---==ccemmcem meme eee 31.6 21,2 10.4 “33 50 years -====-cmcmccemme cme cameaan 27.3 18.1 9.2 . 34 55 yearS-=---memeem ema em eee aan 23.1 15.7 7.4 .32 60 years-=-=----ememcm emma 19.1 13.3 5.8 +30 65 years -=---m-- cme mmm ee eeeeceeeeeeeeam 15.4 11.1 4.3 +28 70 years -==-=--coemm meee meme 12.1 9.3 2.8 .23 75 years ==--=mmccm cee eee cee eee ee 9.2 7:5 1.7 .18 Male 20 years -—====-c--meee meme cme mmemee een 49.4 39.1 10.3 21 25 years -==m= mom cm ecm emcee meee mmeaaan 44.9 34,7 10.2 23 30 years ===-=comemmcee meme meee 40,2 30.3 9.9 «25 35 years ==---=m-mmccm cece mma 35.6 26.5 9.1 . 26 40 years -=---=cmmmme meee cmmmmcmcmm een 31.1 22.7 8.4 sd 45 years -=-meme comme eee 26.8 19.8 7.0 . 26 50 years --====m-ccmmm emcee memae am 22.7 16.9 5.8 .26 55 years-==--mmmm cme eee eam 19.0 14.4 4.6 .24 60 years -=-==cmcccmemm emcee eeae 15.6 12.0 3:6 «23 65 years-===-emmcccm cece mmm emem 12,6 10.1 2,5) .20 70 years -==-m--cmm em eee em mee emma 10.1 8.7 1.4 . 14 75 years ====-c-meememem mmm cm mceee an 8.0 7.1 0.9 «1A APPENDIX | TECHNICAL NOTES It is the purpose of this appendix to describe the methodology used in the construction of the detailed tables. Source of Data Data underlying this report are prevalence rates of edentulousness by sex and age for 1971 and 1957-58 obtained in the Health Interview Survey?: 1 and abridged life tables’: 8 for 1971 and 1958. The earlier prevalence rates of eden- tulousness cover the period July 1957-June 1958, while the annual abridged life tables are each based on the data of a single calendar year. A decision was made on a purely arbitrary basis to use the 1958 (rather than 1957) life tables. Differences between the life tables of the two years are negligible. The prevalence rates utilized in this report are shown (along with other data) in tables I and I. Calculation of Numbers of Survivors It was assumed in the calculations that the prevalence rates by sex and age observed in the sample apply equally to the life table popula- tion. Prevalence rates not directly available from tables I and II were estimated by straight-line interpolation. In performing this interpolation, each of the prevalence rates in tables I and II (except those for the terminal age intervals) was associated with the midpoint of the interval of exact ages to which the rate applies. For ex- ample, the age interval 45-49 (in completed 10 ON METHODOLOGY years) comprises all exact ages between 45 and 50 and the midpoint is age 47.5. The prevalence rates in 1971 for exact age 50 (needed in the calculation of edentulous survivors) was esti- mated by taking the arithmetic mean of those for age intervals 45-49 and 50-54. Such prev- alence rates were estimated for quinquennial exact ages from 20 to 75, inclusive. Prevalence rates of 1957-58 were available from the survey only for the decennial age inter- vals shown in table II. Rates for quinquennial age intervals (needed subsequently in the calcu- lation of average remaining lifetime) were esti- mated by straight-line interpolation. For example, the rate for the age interval 45-49 (midpoint 47.5) for females was estimated from those for intervals 35-44 (midpoint 40) and 45-54 (midpoint 50) by the formula 50-47.5 47.5 -40 50-40 (10.3) + 50 -40 (22.8) = .26 X 10.3 + 75 XxX 22.8 = 19.675. In the calculation of prevalence rates of 1957-58 for exact age 75, the age interval 75 and over was treated as being 75-84 (with mid- point at 80). Values of ¢ (number of survivors to age x of 100,000 live births) were taken directly from the respective life tables. They were not stand- ardized to produce a round number at age 20. The number of edentulous survivors at age x was obtained by multiplying € by the appropriate prevalence rate. The number of dentulous survi- vors is the total number minus the number of edentulous survivors. Table I. Estimated number and rate per 1,000 population of edentulous persons by sex and age: United States, 1971 Edentulous persons Total Sex and age population Rate per Number 1,000 population Both sexes 1 15010 VERB sumunuminvaasm er et SERRA NEN 19,000,939 16,210 '0.9 20-24 years =-=--= commen eee 16,255,280 93,392 5.7 25-29 13,763,688 327,978 23.8 30-34 11,418,812 582,457 51.0 35-39 10,734,663 857,640 79.9 40-44 11,510,922 1,212,850 105.4 45-49 12,094,428 1,740,707 143,9 50-54 11,151,614 2,274,141 203.9 55-59 10,133,703 2,812,180 277.5 60-64 8,383,916 2,894,655 345.3 65-69 6,881,047 2,860,527 415.7 70-74 5,163,246 2,587,790 501, 2 75-79 3,835,244 2,161,329 563.5 80-84 2,243,037 1,390,607 620.0 85-89 years === mmm em ee eee 922,063 624,327 677.1 85 years and OVer-=-==c--cemccm cece mee 1,226,544 819,008 667.7 15-19 9,555,961 117,414 10.8 20-24 8,795,072 48,685 5.5 25-29 7,110,929 198,880 28.0 30-34 5,925,961 322,774 54.5 35-39 5,648,021 502,993 89.1 40-44 5,901,517 693,769 117.6 45-49 6,285,786 958,766 152.5 50-54 5,823,709 1,213,166 208.3 55-59 5,309,179 1,497,893 282.1 60-64 4,513,193 1,554,982 344.5 65-69 3,762,027 1,543,511 410.3 70-74 2,983,182 1,517,950 508.8 75-79 2,290,609 1,350,875 589.7 80-84 1,369,923 864,271 630.9 85-89 554,103 392,203 707.8 85 years and Over -==---eeo ammo emcee ~~ 752,611 527,841 701.3 9,444,978 ,'8,796 10.9 7,460,208 44,707 6.0 6,652,759 129,098 19.4 5,492,851 259,683 47.3 5,086,642 354,647 69.7 5,609,405 519,081 92.3 5,808,642 781,941 134.6 5,327,905 1,060,975 199.1 4,824,524 1,314,287 272.4 3,870,723 1,339,673 346.1 3,119,020 1,317,016 422.3 2,180,064 1,069,840 490,7 1,544,635 810,454 524.7 873,114 526,336 602.8 367,960 232,124 630.8 473,933 291,167 614.4 1 Relative standard error is more than 30 percent. n Table II. Estimated number and percent of edentulous persons by sex and age: United States, July 1957-June 1958 Total Edentulous persons population S d in ex and age thousands Famber Rate per 100 chousands population Both sexes 15-24 21,093 194 0.9 25-34 22.738 812 3.6 35-44 22,918 2,196 9,6 45-54 19,639 4,390 22.4 55-64 14,831 5,647 38.1 65-74 years 9,627 5,329 55.4 75 years and over 4,886 3,287 67.3 15-24 years=-====mmmmmmemmme meee me meee meee ——— 11,292 102 0.9 25-34 oars mm mm mmm 11,880 535 4,5 35-44 year§=---=----mcmme eee ee meme mmm meme 11,892 1,228 10.3 45-54 year§=--=-------emmememe meme m em ——————— 10,047 2,287 22.8 55-64 yearS=-==-m==--- m-mec mmm e——————- 7,685 3,082 40.1 65-74 year§-------mmcmmmmmm meme ememm—m—ee—oommm 5,116 2,947 57.6 75 years and over--------em--mmemmceemeee em ——————— 2,755 1,957 71.0 Male -24 yearS==-==-mmmmmmm meme meee —————— 9,801 91 0.9 Ca yy meme meee m mmm mmmme—————— 10,859 277 2.6 35-44 yearS=-------m-mmeme meee mm————e——o 11,026 968 8.8 45-54 yearg-----------e-mmmemeee meee emmme—m———aen 9,592 2,103 21.9 55-64 yearS=--=-=-=----mmmmeemeemem mmm mmme————e——oo—o 7,147 2,565 35.9 65-74 years-----------ememeemme meme mmmcmm—cee——— 4,511 2,383 52.8 75 years and OVer------=--------c-emememmm—em—o—ooo- 2,131 1,330 62.4 Calculation of Average Remaining Lifetime Values of L (number between exact ages x and x + 5 in the stationary population associated with the life tables) were subdivided into den- tulous and edentulous persons by applying ap- propriate prevalence rates. These values are not shown in this report. This process is analogous to the subdivision of £ values described in the preceding section. For the 1971 data the process was carried up to age 85. The corresponding values for the terminal age intervals (Tg5 =, Lg, for the 1971 data and 7,, =_L,; for the 1957-58 data) were similarly subdivided. It will be convenient to denote the dentulous and eden- tulous components by superscripts (d) and (e). 12 For each quinquennial age x from 20 to 75, a value of Td) (the number of dentulous per- sons in the stationary population at ages x and above) was obtained by adding to SL(d) the corresponding values for all subsequent age in- tervals, including the terminal interval. In other words, the (4) column on the worksheet is ob- tained by accumulating the WASH column from the bottom up. The average remaining dentulous lifetime of a dentulous person aged x, which is denoted by eld ), was obtained by the formula od) d d 8@ = 7 o@) The justification for this formula is that the den- tulous part of the cohort is a “closed” group in the sense that a person who becomes edentulous will never again be dentulous. The second major assumption made in this report is that age-sex-specific mortality rates are the same for dentulous and edentulous persons. This implies that the average remaining lifetime at age x is the same for dentulous and eden- tulous persons. In other words, the average re- maining lifetime for dentulous persons is the same as that for all persons (of the given age and sex). Thus if é (de) denotes the average remain- ing edentulous lifetime of a dentulous person aged x this is taken to be QU) 2. 2) X Xx X Values of € were taken from the appropri- ate life tables.’> 8 Tables 3 and 4 show also values of ed and gl) which were estimated in the manner just described. 00O0 13 APPENDIX II RELIABILITY OF ESTIMATES Standard Errors of Prevalence Rates Since the numerical quantities presented in this report are based on a sample, they differ somewhat from the values that would have been obtained if prevalence rates of edentulousness had been derived from a complete census. The tables shown in appendix I and the chart included in this appendix are designed to provide measures of this sampling variability. They do not include estimates of any biases that might be present in the data. Tables I and II show, for each sex-age category, the estimated number of edentulous persons, as well as the total population (accord- ing to estimates of the U.S. Bureau of the Cen- sus) and the prevalence rate. Figure I is intended to assist the reader in determining standard er- rors of the prevalence rates. The values so ob- tained are, of course, estimates of the standard errors, but to avoid excess verbiage, they will henceforth be referred to merely as “standard errors.” In order to use this graph, one must find in the appropriate table (table I or II) the esti- mated number of edentulous persons corre- sponding to the prevalence rate in question. This estimated number must be located on the base line (horizontal axis) of the figure. The corre- sponding reading on the vertical axis is the rela- tive standard error (in percent) of this estimated number and also of the corresponding preva- lence rate. For example, consider the prevalence rate of 152.5 per 1,000 population for females aged 45-49 in 1971. By table I the correspond- ing estimate of the number of edentulous per- 14 sons is 958,766. The ordinate of the graph in figure I corresponding to an abscissa of 958,766 is approximately 5.6 percent, or .056, which is therefore the relative standard error of the prev- alence rate in question. The standard error is primarily a measure of sampling variability, that is, the variations that might occur by chance because only a sample of the population is surveyed. The chances are about 68 out of 100 that an estimate from the sample would differ from the value obtained in a complete census by less than the standard error. The chances are about 95 out of 100 that the difference would be less than twice the standard error, and about 99 out of 100 that it would be less than 2%: times as large. As a further illustration, observe that the only instances in tables I and II in which preva- lence rates for females do not exceed the corre- sponding rates for males (with the exception of some marginal cases at ages under 25) occur at ages 60-64 and 65-69 in 1971. One may ask whether these two exceptions to the general rule could reasonably be ascribed to sampling varia- bility. The following figures extracted from table I show the number of edentulous persons of each sex in each of the two age intervals in question: Female Male 60-64 ......0050mmmmsniavsrns 1,544,982 1,339,673 65-69..ccvnniiiiniiienns 1,543,511 1,317,016 Entering the chart in figure 1 with these num- bers, one can read the corresponding relative Figure |. Relative standard errors of numbers of edentulous persons and corresponding prevalence rates. RELATIVE STANDARD ERROR (%) No D- » 2 3 4 5 6 7809 2 3 4 5 6 7 839 2 3 4 5 6 7 839 100 100,000 SIZE OF ESTIMATE (IN THOUSANDS) Example of use of chart: An estimate of 1,000,000 edentulous persons (on scale at bottom of chart) has a relative standard error of 5.6 percent, or a standard error of 56,000 (5.6 percent of 1,000,000). 15 standard errors as follows: Female Male Percent B0:64.....000mmmmmissemmavins 2.4 3.0 B8-69.snsnmnsssnsmeins 2.4 3.2 These are also the relative standard errors of the corresponding prevalence rates. Consequently, the absolute standard errors of the prevalence rates per 1,000 population are: Female Male Percent BO0-648...00mmmsrrivsossnnsssn 8.3 10.4 B56 unssisvmssssrssnsins 9.8 13.5 If covariance is ignored, the standard error of the female minus male difference is the square root of the sum of the squares of the standard errors for females and males separately. The dif- ferences in prevalence rates and the correspond- ing standard errors are as follows: Difference in Standard error prevalence rates of difference Percent B80-64......00: 0000500000 1.6 13.3 05-69. siisisinemvine 12.0 16.7 It is clear that the larger prevalence rates for males in these two instances could reasonably be ascribed to sampling variation. Standard Errors of Numbers of Survivors The numbers of edentulous survivors shown in table 1 are obtained by multiplying life-table values for the United States population by prev- alence rates of edentulousness. In most in- stances, the prevalence rate in question is obtained by straight-line interpolation between two consecutive tabular prevalence rates. If the latter two rates are denoted by 7, and m, and the interpolated rate by 7 (where the circumflex accent denotes that the rates are estimates based on a sample, while its absence denotes the un- derlying true rates), and if m=tw + (l-f) mw, 16 and further if p, and p, denote the relative standard errors of 7, and ,, respectively, then the standard error of 7 is p= Llp? a2 + (1-07 p2 m2" if one neglects the covariance of 7 and m,. If t = Y%, this reduces to _lef af + p73)" m, +7, For example the figure 30,174, which is the number of edentulous survivors at age 65 for fe- males in 1971 (table 1), is obtained by multiply- ing 79,951 by .3774. The prevalence rate of 377.4 per 1,000 used in the latter calculation is the ar- ithmetic mean of the tabular rates 344.5 and 410.3. The respective standard errors of the latter rates, obtained from table 1 and figure I by the method previously described, are 8.3 and 9.8, and the formula for the case of t = % gives 3.4 percent for the relative standard error of the inter- polated rate. Thus the standard error of the number 30,174 of edentulous survivors ig 1.7 percent of 30,174, or 513. Since the number of dentulous survivors is a constant less the number of edentulous survi- vors, the absolute standard error is the same, but the relative standard error is different—in this case 513 + 49,777 = 1.0 percent. A similar calculation at age 75 gives 672 as the standard error of the number 33,596 of edentulous female survivors at that age. Now, one can observe that for males in 1971 and for each sex in 1957-58 the number of eden- tulous survivors at age 75 is less than at age 65. Thus the females in 1971 exhibit different be- havior in this respect from the other three cate- gories. Is it conceivable that this exceptional behavior could be due to sampling variability? The difference between the numbers at ages 65 and 75 is 3,422 and the standard error of this difference is approximately [(518)> + (672)%]% = 845 On the hypothesis that the numbers of eden- tulous survivors at the two ages are the same in the total population, the standardized deviate is 3,422 + 845 = 4.05. Thus it is most unlikely that sampling variability accounts for the exceptional behavior. Standard Errors of Average Remaining Lifetime The estimation of standard errors of average remaining lifetimes is more complicated, since these quantities are quotients of random varia- bles. First consider the standard error of eld), the average remaining dentulous lifetime of a dentulous person. Since the sum of PAR and é (4°) is the constant é_ (see appendix I), the standard deviations of ¢ and ¢ > (de) are the same absolute amount. The procedure for estimating the standard error of eld i is somewhat complicated and is most casily explained by reference to a specific example. Consider, therefore, the estimation of the standard error of ed for females in 1971. If | m denotes the prevalence rate of edentu- lousness in the age interval between exact ages x and x + n the formula for ol can be written sls (A-.7,. + sEgoll sd ¥ Zs (1-gs) d 2 L,+ s¥ 80 sL go + w Mgs Ts) 2 (d) €15 = Te = (sMss d 05s Henceforth the relative standard deviation of m is denoted by ny The standard error of the numerator of eld) is the standard error of the expression within parentheses in the last expression above and, ignoring covariances, can be approximated by LZ L2, ONum = [503s s¥ 5 * PL. 7 2 5 T.2]1% = 12512. we os v2, The standard error of the denominator is the same as that of gle) » which was previously shown to be 672. If one denotes the corresponding relative standard errors by py,, and Pp,» the relative standard error of a ratio of two random variables is approximated by? 2 = p? 2 P ratio - P Num + P ben 2 Num, Den’ where Y nym, pen denotes the relative covariance of the numerator and denominator. Inasmuch as a positive multiple of the denominator is essential- ly one of the components of the numerator, the covariance is almost certainly positive, and therefore < 2 \% P ratio - (Pum + P ben) In the present case, 12512/233133 = .05367, P Num P ben 672/27565 = .02438, and an upper bound to the relative standard er- ror of ¢d) is 75 [(.05367)% + (.02438)2]%* = .0589, or 5.9 percent. The true relative standard error of € ( is probably substantially less. 000 17 APPENDIX Ill IMPLIED INCIDENCE RATES In the methodology used in the construction of the detailed tables and described in appendix I, prevalence rates of edentulousness were used in a makeshift fashion because incidence rates were not available. This was made possible by some rather sweeping assumptions described in the main text and in appendix I. As incidence rates would seem to be a necessary ingredient for obtaining the kind of results presented in tables 1-4, some incidence rates must be implicit in the assumptions made. Though no actual use was made of them, it may be of some theoretical interest to discover what rates these are. The incidence rate for a condition is usually defined! ? as the number of new cases of the condition occurring during a specified period of observation divided by the average population during the period of observation. Those persons who already have the condition under study are not excluded from the population in the denom- inator. Consider the | L persons who are in the age interval x to x + n at a given moment. The number of new cases of edentulousness occur- ring in this group during the succeeding n years is, on the assumptions made in this report, x+n Le) nn x+n L©), n x 000 The first term is the number edentulous n years later at ages n years older. The subtractive term is the number of survivors n years later of those in the original group who were already edentu- lous at the initial moment. Noting that () = nlx Fx n x and dividing the number of new cases by nly one obtains for the incidence rate oe (27 xin “on One may observe that this expression is the product of two factors: an n-year survival rate for the stationary population, and a difference between successive prevalence rates. This is remi- niscent of the multiplicative theory of multiple decrement tables and the associated single decre- ment tables.!! It is interesting to note that Klein? obtained incidence rates by differencing prevalence rates without, however, offering any theoretical justi- fication for this procedure. If prevalence rates for some morbid condi- tion (from which recovery is impossible) do not increase monotonically with age, this would im- ply negative incidence rates, and the model would be untenable. Series 1. Series 2, Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. VITAL AND HEALTH STATISTICS PUBLICATION SERIES Originally Public Health Service Publication No. 1000 Programs and collection procedures. — Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods research.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies.—Reports presénting analytical or interpretive studies based on vital and health statistics, carrying the analysis further than the expository types of reports in the other series. Documents and committee reports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons. Data from the Institutional Population Surveys.— Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce,—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HRA Rockville, Md. 20852 DHEW Publication No. (HRA) 75-1338 Series 2 - Number 64 Series 2-Number 65 VITAL and HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH al Cen, a Pa fe) NCHS £3 wv Distribution and Properties of Variance Estimators for Complex Multistage Probability Samples An Empirical Distribution U.S. DEPARTMENT O HEALTH, EDUCATION, AND WELFAR Public Health Servi Health Resources Administratio Library of Congress Cataloging in Publication Data Bean, Judy A. Distribution and properties of variance estimators for complex multistage probability samples. (Series 2: Data evaluation and methods research, no. 65) (DHEW publication no. (HRA) 75-1339) “Final report.” Supt. of Docs. no.: HE 20.2210:2/65 1. Sampling (Statistics) 2. Estimation theory. 3. Monte Carlo method. I. Title. II. Series: United States. National Center for Health Statistics. Vital and health statistics. Series 2: Data evaluation and methods research no. 65. III. Series: United States. Dept. of Health, Education, and Welfare. [DNLM: 1. Population surveillance. 2. Probability. 3. Sociometric technics. W2A N148vb no. 65] RA409.U45 no. 65 [QA276.6] 312°.07°23s [519.52] ISBN 0-8406-0029-1 74-16356 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 - Price $1.10 DATA EVALUATION AND METHODS RESEARCH Series 2 Number 65 Distribution and Properties of Variance Estimators for Complex Multistage Probability Samples An Empirical Distribution This report presents results from an empirical investigation of the behavior of the replication and linearization techniques for measuring variance. The study utilizes data collected in the Health Interview Survey. Most scientific sample surveys are based on complex multistage probability designs, with design components that include unequal probabilities of selection of elements in the population, stratification, and several stages of clustering. The estimation procedures usually involve nonresponse and poststratification adjustments. These actions cause concern in the methods of estimation of standard errors and their subsequent use in constructing confidence intervals and testing hypotheses. There are several ways to study the characteristics of variance estimates. The method in this report is an empirical investigation. Data that were collected in a national sample survey become the universe and repeated samples are drawn from it. For each sample the statistics under study were calculated and sampling distributions of the estimates were generated. The material in this report was taken, in part, from Dr. Bean's doctoral dissertation. DHEW Publication No. (HRA) 75-1339 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration National Center for Health Statistics Rockville, Md. March 1975 NATIONAL CENTER FOR HEALTH STATISTICS EDWARD B. PERRIN, Ph.D., Director PHILIP S. LAWRENCE, Sc.D., Deputy Director GAIL F. FISHER, Associate Director for the Cooperative Health Statistics System ELIJAH L. WHITE, Associate Director for Data Systems IWAO M. MORIYAMA, Ph.D., Associate Director for International Statistics EDWARD E. MINTY, Associate Director for Management ROBERT A. ISRAEL, Associate Director for Operations QUENTIN R. REMEIN, Associate Director for Program Development ALICE HAYWOOD, Information Officer OFFICE OF STATISTICAL RESEARCH GEORGE A. SCHNACK, Chief, Research Coordination Staff MONROE G. SIRKEN, Ph.D., Chief, Measurement Research Laboratory Vital and Health Statistics-Series 2-No. 65 DHEW Publication No. (HRA) 75-1339 Library of Congress Card Number HRA 74-16356 CONTENTS Introduction . . . . «cc the ee ee eee eee eee Background .. +» + + 2 + +12 v sv vs ss vs sn st wns smu sw Strategy of theStudy . . . . . . .. ... vo. Highlights of the Findings . . . . . . . . ............... Review of Literature . . . . . . «cv vv vv vv vt Replication Method . . . . . ......... ooo Linearization Method . . . . . . . . ooo An Empirical Investigation of All Three General Variance EStimMators « + vv vv ve ee ee ee ee ee eee ee eee Method . . . . . «i i i i iit eee eee Description of the Population . . . . . .............. Sample Design . . . . . vc . vv ss vi vie i a eae Variables . . . . . «eee eee ee eee Estimators and Variance Estimators . . . . . . . . . . . .. . . . .. EStimators . . «vv vt ee ee ee ee ee ee eee eee eee Variance Estimators . . . . . « «vv vv vt hh he eee ee Variance Estimators for Estimator 1 . . . . . . . . . . . ..... Results . . . «cv vi i i it ttt tte sa sense Introduction . . . . . «Lo Lu eee eee eee Comparison of Replication and Linearization . . . . ....... Normality of Standardized Estimators . . . . . . . ........ Influence of Poststratification . . . . . . . . ........... BUMMABIY + « + + 5 5 + = 2 5s 2 8 2 @ 03 223 ra wes »ssnws Introduction . . . . . . «ci i ieee eee Behavior of Linearization and Replication . . . . . . .. . .... Normality of Standardized Estimators . . . . . . . . ....... Comparison of Two Ratio Estimators . . . . . . .......... Approximate Variance Estimators .. . . . . . o.oo... Influence of Poststratification. . . . . . . . . . . .. ....... Page — 12 14 14 15 16 17 17 18 18 20 20 20 24 26 27 27 29 29 29 29 30 i CONTENTS—Con. Page ConclUSIONS + vv vv vv ve ee he ee eee eee ee ee ee eee eee 30 Relevancy to HIS . . cv vv sv ss ss sn sv sms vs ss 0 2 0 8s 0 v0 31 References « + « + + v + + t+ 2 +0 tt st xv se vss most ssa ssa wats 32 Appendix I. Extent of Biasness of Two Ratio Testimators . . . . ........ 34 Introduction . . . . «i i i tt tt te eee eee 34 EStimators . . . ov vv vv vv hee ee ee ee ee ee ee eee eee 34 Comparison of Two Ratio Estimators. . . . . ................ 34 Appendix II. A Simpler Variance Estimator . . . . . . .............. 38 Introduction . . . « . «Jv tt tt th eee era eres 38 Variance Estimators for Estimator 1 . . . . . ..... ............ 38 Variance Estimators for Estimator 2 . . . . . ................. 40 Approximate Variance Estimators . . . . . ... ooo. 41 SYMBOLS Data not available---ce-eeeeeaemmeeeieeaeeeee J Category not applicable -------e-eeeecmeceeceenceas Quantity zero - Quantity more than 0 but less than 0.05---- 0.0 Figure does not meet standards of reliability or precision----------e-ccoeeeeenene- * DISTRIBUTION AND PROPERTIES OF VARIANCE ESTIMATORS FOR COMPLEX MULTISTAGE PROBABILITY SAMPLES Judy A. Bean, Ph.D., University of Iowa INTRODUCTION Background Since the early 1940’s, there has been a sub- stantial growth in the use of surveys of human populations, which has occurred in all research areas, including the biological and health sci- ences, and particularly the social sciences. Along with the increasing utilization of surveys, the purposes for which data are collected have ex- panded. Research workers first employed sur- veys to obtain specific information about groups for descriptive purposes only. Increasingly, how- ever, they have become more interested in mak- ing comparisons among subgroups of the popula- tion, in testing hypotheses about the population, or in disclosing complex relationships. Both the greater use and the change in objectives have promoted considerable theoretical and practical development in sample design which, in turn, has resulted in decreased use of the traditional method of random selection from human popu- lations. Most scientific sample surveys today are based on complex multistage probability samples, with design components often including unequal probabilities of selection for different elements in the population, stratification, and two or more stages of clustering. The estimation proce- dure may involve adjustment for nonresponse, use of concomitant variables, and poststratifica- tion. Because of these features, the assumptions of independence and equal probability of selec- tion are not valid. Therefore, analytical ques- tions arise. Theory has not been developed suffi- ciently to overcome all the difficulties evoked by the correlation induced by clusters of sample units. One analytical problem not solved in theory concerns estimation of variances and standard errors of parameter estimates. Since these are integral components of the formulas for con- structing confidence intervals and testing hypotheses, standard errors are crucial in statisti- cal inference. When variance estimates are needed for statistical inference, several choices are available. There is the option of designing and carrying out a random sample but, as mentioned pre- viously, this has become an uncommon design. The human populations to be studied today are large and widely dispersed, which makes listing and travel prohibitively expensive. However, it is frequently possible to use a reasonable approxi- mation for estimating variances when the exact formula is not known. Care must be exercised when selecting an adequate approximation. For instance, if the random sample formula PQ/n is substituted for the appropriate variance esti- mator for clustered samples, it can be shown that the variance is usually underestimated. The investigator can perhaps translate the problem for which no estimates of standard errors are obtainable into one for which esti- mates are available. For example, the 2 X 2 chi- square test can sometimes be translated into the difference between two proportions. Inferences can be made on statistics for which the true vari- ances cannot be calculated, by applying vari- ances of similar statistics. For clarification, con- sider the situation in which an analyst is inter- ested in testing whether or not k (k > 2) means are from the same population. Since the variance for the difference of two means is known, the variance for several pairs or perhaps all pairs can be computed. Then the ratio of these variances to variances that would have been obtained if the design had been a simple random sample can be calculated. The analyst then infers that the comparison of the correct variance estimate of the k& means to simple random sample variance would yield a similar ratio. Thus, he can use this ratio as an adjustment factor to the usual F ratio for testing such a hypothesis.!»2 Another approach to the problem of estimat- ing variances is to interweave within a sample design a small number of replications. Each of these replications produces an estimate of the parameter. Therefore, their comparison provides an estimate of variance for the sample. An exam- ple of this is to draw 10 independent subsamples from the same population using the same proba- bility design. Let X; denote the estimated mean of the ith subsample and __ 1 = x = ty Xi — iMs= Then from the 10 independent estimates, the simple variance of the mean X for the entire sam- ple can be computed. This estimate is VAR(X = So 10 1 0 2 x; ~ %) This method is too expensive to employ for a complex design. A similar procedure to using independent sub- samples is the random group method. To illus- trate the technique, suppose one has a simple random sample of n observations drawn with re- placement. The observations are randomly dis- tributed among ¢ groups consisting of n/t observations each. The variance estimate is where x; = the mean for the jth random group, and J 1 TL ~~ This method is suitable for multistage survey samples. If primary sampling units (PSU’s) are drawn and then second-stage units are sampled within, all the second-stage units from a primary unit constitute a single unit when forming the random groups. If there are few PSU’s, this pro- cedure is not too useful. Moreover, there is a loss of information.* The final choice is to compute the estimates of standard errors by one of the three general methods: Taylor series expansion or lineariza- tion method, pseudoreplication or replication method (originated from the methods of inde- pendent replication and random group), or jack- knife method. Unfortunately, the behavior of the variance estimates produced by these meth- ods is not thoroughly understood. Three major survey organizations—the U.S. Bureau of the Census; the National Center for Health Statistics (NCHS); and the Survey Re- search Center (SRC), University of Michigan— commonly use one of the three general methods, linearization, replication, or jackknifing, to cal- culate estimates of variances from the sample data. It is, therefore, important to become familiar with the properties of these methods. Strategy of the Study There are several possible ways to study the characteristics of variance estimates created by the general methods. One possibility is the devel- opment of applicable theory by assuming a cer- tain distributional form and obtaining exact analytical solutions. However, the mathematics involved have so far proven to be intractable. There is another drawback with this approach with a stratified multistage design: two sources of sampling error are between the first-stage units and within the first-stage units. When only one unit is chosen from a stratum, there is no consistent way of estimating the variance be- tween the first-stage units from the sample itself. In practice, the designs seldom fulfill this simple assumption of the selection of two units. An- other avenue of study is Monte Carlo sampling from synthetic populations. Such populations are difficult to construct and are of questionable representativeness. A third device is an empirical investigation. By this approach, data that have been collected in a survey become the universe, and repeated samples are drawn from it. For each sample, the statistics being studied are com- puted, and sampling distributions of the esti- mates are generated. Perhaps the most famous such study is the investigation by “Student” in 1908,5 which resulted in his derivation of the ¢ distribution. In this report, the view is that by Monte Carlo sampling of data collected in a nationwide sam- ple, variance estimators can be studied to gain insight into their properties and behavior. An empirical investigation of the behavior of the replication and linearization variance estimation methods from a viewpoint that is both broad and practical is undertaken. The study uses five variables collected on 131,575 people in the U.S. Health Interview Survey (HIS). The main objectives of the research were the following: 1. To investigate two general variance esti- mator methods? linearization and replica- tion 2. To study the distribution of the ratio of an estimated mean minus its expected value divided by its standard error 3. To measure the impact of poststratification on variance estimates Other aspects examined in the investigation were the extent of the biasness of two estimators and the feasibility of using a simpler variance esti- mator as an approximation to the correct repli- %Since the replication and linearization schemes are actually being used by several samplers, the author chose to study only these two techniques in this study. cation and linearization variance estimates (the results are given in appendixes I and II). When a study of this magnitude is described and discussed, there is the danger that the reader will lose sight of the overall objectives amidst all the details of the analysis. To facilitate report- ing, the main features of the methodology in- volved in doing the research are listed, as fol- lows: 1. Sample design. By this term is meant the components of selection of the sample units. The design used includes stratifica- tion, sampling of the first-stage units with probability proportional to size, the largest units entering the sample with probability 1, and finally, subsampling clusters of households within the first-stage units, thereby embedding two stages of clustering in the design. 2. Estimation procedure. This refers to the method for producing estimates of the health characteristics. Two different post- stratified estimators of means are used. 8. Variance calculations. This feature is the computation from the sample data of esti- mates of the variance of the estimators pro- duced in feature number 2. For each esti- mate of a characteristic, the variance is estimated in a number of different ways. It is appropriate to state the reasons why only two out of the three general variance estimators are studied here. The intent of the research is to observe the variance estimators produced by the methods in more depth than has been done pre- viously and to study the behavior of the esti- mators for types of designs and estimation pro- cedures being used today. Data handling, even for a small-scale study of a similar nature, is quite massive, and this problem is compounded by the type of design and procedures selected. Therefore, the decision was made that only two of the three variance estimator techniques would be investigated. HIGHLIGHTS OF THE FINDINGS This study was undertaken to investigate how well statistical methods important for making inferences are valid when the data have been col- lected in a multistage probability sample survey and subjected to a complex estimation proce- dure. The main points examined were the fol- lowing: 1. The behavior of the variance estimates pro- duced by the balanced half-sample replica- tion method and the linearization method 2. The distribution of the ratio of a sample estimate minus its expected value to its estimated standard error 3. The impact of the estimation technique poststratification on variance estimates Since the mathematics have not been devel- oped to answer this question of validity, the approach was to use Monte Carlo simulation from a specified universe and to determine the empirical results. The sample design was a stratified two-stage cluster design with first-stage units being selected with probability proportional to size. The estimation process involves weighting by the reciprocal of probability of selection and post- stratification. Five variables were used, and three different samples sizes were studied with 900 samples of each size drawn. Because design III is the largest sample size and, thus, the one most comparable to real sur- veys, the highlights for only this design are pre- sented. Characteristics of the universe and a typical sample are: Universe Average sample Number of PSU's 149 30 Number of segments 7,768 600 Number of persons 131,175 8,772 (approximate) Table 1 gives the major results for the variance of a ratio estimate produced by the replication method, VAR(R1S) and the linearization meth- od, VAR(L1). The proportion of times the standardized vari- able fell within certain regions was calculated for each of the five variables. These proportions were averaged across the five variables and com- pared with the expected proportions of a t dis- tribution with 19 degrees of freedom, and the normal distribution, in table 2. Table 1. A summary of the major findings for design I11 Variable Finding Restricted Proportion seeing Family income activity Physician visits Hospital days i physician days Population parameter R . . . ....... 8400.0 14.6716 4.6758 1.0557 0.6842 Average sample estimate BR, . ....... 8392.8 14.6595 4.6548 1.0597 0.6840 Sample variance of R, — sR, A 27,4457 0.9496 0.0168 7.0474 X 1072 7.8942 X 107% Average value of the variance estimate VARIBIS) iss asmismss mows 26,554.8 0.9206 0.0178 7.0842 X 10°? 8.3010 X 10°* Biasof VARIRIS) .......6m0 eas -890.9 -0.0290 0.0010 0.0368 X 10? 0.4068 X 10~* Variance of VAR(R1S) . . ......... 10,360.0 Xx 10* 1.1370 | 0.8132X 107™* 0.0129 x 10°? 0.1089 X 107% Relative bias of VAR(R1S) . ....... -0.0325 -0.0305 0.0595 0.0052 0.0515 Relative variance of VAR(R1S) . .. ... 0.1375 0.1519 0.2882 0.2596 0.1747 Average value of the variance estimate VAR(LT) ooo ieee eee 26,174.7 0.8915 0.0175 6.7380 X 10°? 8.1135 Xx 10°* Biasof VARILY) ... «viv omuecsn -1,271.0 -0.0581 0.0007 | -0.3094 X 10°? 0.2193 X 107% Variance of VARI(L1) ........... 10,480 X 10* 0.1310 | 0.7959 X 10°* 0.0113 X 10°? 0.1086 X 10~* Relative bias of VAR(L1) . ........ -0.0463 -0.0612 0.0417 -0.0439 0.0278 Relative variance of VARI(L1) . . ..... 0.1391 0.1453 0.2820 0.2275 0.1743 Relative bias of VAR(R1S) Telatve bias TVARILTS ~~" ''""" 0.7019 0.4984 1.4269 -0.1185 1.8525 Relative variance of VAR(R1S) Felative veriance of VARIL 1] ariance ST VARLD. *"'" "° 0.9885 1.0454 1.0220 1.1411 1.0023 R, -E(R,) R, -ElR,) ff = — VAR(R1S) *"° VAR(L1) with expected values under ty 9df and normal’ Table 2. Distributions of compared Standardized Average across variables deviate Normal t190f VAR(R1S) | VARI(L1) 210 Lieiiwes 0.683 - 0.675 0.668 +1640 ...... 0.900 | 0.879 0.878 0.872 +1960 ...... 0.950 | 0.932 0.928 0.925 +2570 ...... 0.990 | 0.976 0.977 0.977 The hypothesized degrees of freedom are the number of strata 19. REVIEW OF LITERATURE Replication Method Although, in the past three decades, methods for selection of sampling units and procedures for producing estimates from survey data have become sophisticated, only in relatively recent years have the methods of analyzing data for such complex surveys begun to be considered. The first articles in the literature on the topic described and discussed a variety of techniques: random groups, replicated samples, interpene- trating samples, and duplicate samples, which are the forerunners of the general replication method as applied in this study. Since this in- vestigation is concerned with the technique in its present-day form, the previous methods will not be reviewed here, but their essential features are described elsewhere.b Before giving a brief account of the develop- ment and evidence of the reliability of the repli- cation estimator of variance, its main features are outlined. The basic premise of the method is a very simple one. The estimator X' of a popula- tion parameter is calculated from the entire body of sample data (parent sample). Another estimator Y’ of the same population parameter is made, using only the data from half the sam- ple, randomly selected. The quantity (Y'- X')2 is an estimate of the variance based on one-half sample (replicate). However, this estimate, de- pending on only one difference term, has a high variance; repeated differences are required to produce a stable variance estimate. The esti- mator of variance is simply the mean of these half-sample estimates. VAR(X') = | = k Y(-X) a=1 where k is the number of replicates used. There are three other rather similar forms of variance estimators. k 1 " J Lz 0.-X ¥ a=1 where Y”, is the estimator of the parameter made from the complementary set of data for the ath replication (the half of data not used in the Y', estimator). k k 111 ’ ' 1 " ' 2. 2 pL em Xr 0 - X')? a= a=1 k 1 ’ "n 3 HT 0a P a=1 The U.S. Bureau of the Census’: was the first major survey organization to employ the method. The technique was used to estimate sampling errors of ratio estimates produced from the Current Population Survey (CPS) from 1954 through 1964.2 Gurney? worked on the theoret- ical development of the method for the U.S. Bureau of the Census. The National Center for Health Statistics draws inferences from data gathered in complex probability sample surveys. The Center uses the replication method to provide variance estimates for the estimates produced from the Health Examination Survey (HES), which consists of a direct physical examination of a probability sample of approximately 7,000 noninstitutional U.S. civilians.!? Its purposes are to make preva- lence estimates for certain medical and dental conditions and to determine the distributions of many physiological characteristics. The survey is conducted in cycles with different age groups sampled for each cycle. bees is a monthly nationwide survey of sample households conducted to provide measurements of the labor force. National estimates of employment, unemployment, and other labor force characteristics are made. The sample design of the survey includes the components stratification, clustering, and two stages of sampling. Along with this highly sophisticated sample design, an equally refined estimation process is used. Features of the pro- cedure are simple inflation of the basic data col- lected on each sample unit by the reciprocal of the probability of selecting that unit, adjustment for nonresponse, poststratification, and possibly another adjustment factor. The National Center for Health Statistics, in its efforts to provide the best estimate of vari- ance for this ratio estimate, has used several ver- sions of the replication technique and, from this experience, has developed a routine computer program to provide the analyst with prevalence estimates and their variances. The analyst, by simply requesting the estimates for the total or specified subgroups of the population, receives an estimate of the aggregate value of the health characteristic (numerator), an estimate of the population of the total or subgroup (denomi- nator), and the ratio of the two values. In addi- tion, the analyst obtains an estimate of the standard errors for these three quantities. A more detailed description of the development and use of the replication method can be found elsewhere.6,11,12 Based on research conducted by NCHS and the performance of the replication variance estimates, the technique appears not only appropriate for the type of estimate pro- duced in its surveys but also useful for variance estimates of other statistics such as regression coefficients, multiple correlation coefficients, and partial correlation coefficients; moreover, the technique has potential for being employed in other forms of statistical analysis. For in- stance, replication variance estimates are used in a modified sign test and in the calculation of a pseudo-chi-square statistic, which provides a test of independence in a two-way table. Valuable theoretical work has been done on the replication method.6 If a subset of all the possible half-samples that could be formed is selected randomly, the number necessary to pro- duce a stable variance estimate is large. By the use of a stratified design with two independent selections made from each stratum and an esti- mate of the mean, it was demonstrated that the variability among the half-sample estimates of variances comes from the between-strata contri- butions to these estimates. It was then shown how to eliminate these cross-product terms by choosing a relatively small subset of half-samples in a particular fashion that would yield an un- biased estimator of the true variance of the linear estimator. This variance estimator, in fact, is equal to the value that would be obtained if all possible half-samples were formed to calcu- late the variance estimate. The half-samples chosen in this manner are said to be orthogon- ally balanced. This set of half-samples is referred to as an orthogonal set or a balanced set. The steps involved in composing a balanced half-sample are explained. Assume a sample de- sign of two independently selected units from a stratum. First, the digit 1 or 2 is arbitrarily as- signed to each sampled unit from a stratum. Then the number of replicates needed is speci- fied. This number, k, has the restriction that it must be a multiple of 4, in order to obtain an orthogonal set. The value is equal to the number of strata, L, plus either 0, 1, 2, or 3, depending on which integer added to L will yield a number divisible by 4. The next step is to form ak X L matrix with entries a;; of either plus or minus 1 such that the columns (strata) of the matrix are orthogonal to one another. This means k 2 qj, =0 wherej# J’ a=1 Plackett and Burman!3 describe the construc- tion of such a matrix. Finally, the ath half- sample is made by including the unit numbered 1 if a;; = +1 and the unit numbered 2 if a; =-1. The names ‘balanced pseudoreplication” and “balanced repeated replication” are given to the method in which the minimum number of half- samples are needed. (Throughout this text, the term “replication method” means the full ortho- gonal balanced version unless otherwise speci- fied.) Also considered in the report® was the situa- tion in which the number of strata was so large that the number of half-samples necessary for full orthogonal balancing would be too numer- ous to be feasible because of computer costs. In this case, a method is suggested whereby the half-samples are partially balanced. To achieve partial balance, the L strata are divided into L/t groups, where t is an arbitrary integer that divides L evenly. The steps given for full orthog- onal balancing are carried out for one group of strata, which results in the elimination of the between-strata contributions for these strata. The matrix that is formed is applied to each of the remaining groups. Thus, within orthogonal sets there is no between-strata contribution, and the cross-product terms partially disappear. However, the other cross-products do not drop out. In the case of a stratified random sample and an estimated mean, this subset of half- samples has a lower estimated relative variance than a set of k randomly selected half-samples. Leel* has recently published theoretical and empirical work on a method of selecting a par- tially balanced set of half-samples. He also con- sidered the problem of the best possible value of t. There will be a larger variance when the repli- cates are only partially balanced than when they are fully balanced. However, this loss can be re- duced when the semiascending order arrange- ment (SAOA) of selecting half-samples is em- ployed. To accomplish an SAOA, the strata are first arranged in ascending order of the magni- tudes of their within-stratum weighted popula- tion variance. After this is done, the last L/2 or (L-1)/2 (if L is odd) strata are reversed. This means that these strata are put in descending order of magnitude of their within-stratum vari- ance, and the first L/t strata form the first group and so on. Another factor in reduction is to choose a small ¢ value, but the smaller the value of ¢, the more half-samples are needed. Therefore, when the investigator decides on the value of ¢, he must weigh the loss in precision against the in- crease in computer costs due to a larger number of half-samples. The behavior of the replication method when a ratio estimator is used was also examined.!5 The findings were primarily based on empirical data from the first cycle of the HES, a sample of approximately 6,600 adults. Body measure- ments were analyzed from a subsample of approximately 3,000 U.S. adult males.!6 Six- teen regression equations with age, weight, and height as the independent variables and each of the other 16 anthropometric measurements in turn as the dependent variable were tabulated. A variance estimate for a regression coefficient can be produced by the replication method. Using the data in a replicate, a regression equa- tion identical to the one computed for the par- ent sample can be calculated. Then an estimate of variance is obtained from the average of the deviations of the half-sample regression coeffi- cient estimates from the parent sample estimate. Using a balanced set of 28 replicates, estimates of variances were computed for two types of estimates: ratio estimates of population means for the physical body measurements, and multi- ple regression and correlation coefficients. According to the study, the set of balanced half- samples gave the same results as would have been obtained by employing the full set of repli- cates. Also developed was an expression for the relationship between the replication variance of the estimator of the population means and the replication variance of the average of the k& half- sample estimated population means. Because the results from a sample provide values for the quantities in the expression, an inference as to whether adequate variance estimates are being obtained can be made. It was concluded that these data showed that the replication variance estimates were sufficient. Simmons and Baird,!2 following a suggestion by Kish at the Survey Research Center, Univer- sity of Michigan, investigated whether an approximating variance formula requiring less computer time could be used in place of the exact equation. In the replication method, the data for a half-sample are subjected to the full estimation procedure in exactly the same man- ner as the data from the parent sample. This operation can use extensive computer time, es- pecially when steps such as ratio adjustment by a concomitant variable and poststratification must be carried out in each half-sample. How- ever, by weighting the individual sample observa- tions more simply than the usual way designated by the sample design, the correct replication variances can be approximated. The goal of the study was to weight the data in three ways to observe whether the approximations could be substituted for the correct variance estimate. CThese calculations were provided by the Survey Research Center, University of Michigan, in accordance with NCHS speci- fications. The data for the research were the same physical measurements as those used elsewhere.! 5 Operationally, when making the parent sam- ple ratio estimates, a ratio adjustment factor due to the use of a concomitant variable is computed for each of six residence-region classes. This was done only on the data collected in the first cycle of the HES. A poststratification factor for each of 60 age-sex-color classes is also calculated. Both adjustments are in the form of multipliers applied to the weight (reciprocal of the selection probability) for each sampled person. In the rep- lication method, the replicate estimate of the population parameter undergoes the ordinary estimating process. Hence, the multiplication factors of these adjustments should be recalcu- lated in each half-sample to bring the distribu- tion of the sample into as close agreement as possible with the distribution of the universe. By using the two multipliers from the parent sample instead of recalculating them in each half- sample, Simmons and Baird hoped that an ade- quate approximation involving less computer time could be made. The weighting schemes were to assign each observation (1) a weight of 1 as if the data were collected in a simple random sample, (2) the weight it received when the par- ent sample estimate was constructed, and (3) the appropriate weight for the specified half-sample containing the observation. A brief summary of the conclusions is offered. The estimates of vari- ance produced from the unweighted data (scheme 1) were not acceptable as approxima- tions to the true variance. The ratios of the vari- ance estimates using weighting scheme 2 to the variance estimates using weighting scheme 1 were all greater than unity, indicating that the simple random sample variance estimates are underestimates. The variability of the ratios for the different statistics was wide. Therefore, the application of a constant adjustment factor to simple random sample variances is not feasible. Simmons and Baird then compared the variance estimates resulting from weighting schemes 2 and 3. The analyses consisted of computing the ratio of replicate variances using the parent sam- ple adjustment factors to variances using unique adjustment factors. These ratios fluctuated around unity. The ratios for the ratio means of the physical body measurements and for the simple correlation coefficients were slightly greater than unity, but the ratios for the partial and multiple correlation coefficients were just below unity. The variability of these ratios was not negligible. Because of this, the investigators concluded that analysis on similar variances for more statistics was required before making a de- cision about the use of the simple weighting pro- cedure (scheme 2) in place of scheme 3. The SRC has produced variance estimates by the replication technique since 1957. Kish and Frankel! 7-18 presented the empirical evidence on the reliability of the method from SRC’s projects and from their analysis of the NCHS body measurement data for the HES. In these studies, the replication method was employed to provide standard error estimates for regression coefficients, correlation coefficients, and partial correlation coefficients. The analyses examined the square roots of the ratios of the true stand- ard errors calculated by balanced repeated repli- cations to variances assuming a simple random sample. These ratios are referred to as the design effect (DEFF). The first empirical evidence presented was the results from 20 linear regression equations calcu- lated on data for a subgroup of the population sampled in three different surveys. The subgroup was identified as husbands and wives aged 35-44 years living together, with the head of the house in the labor force and having a family income of’ $3,000 or more. There were 1,853 such defined families. Seven predictors used in various combi- nations one, two, three, or four at a time were the basis of the regression equations. An esti- mate of variance for each of 60 regression coeffi- cients was made. The mean value of the square roots of the 60 DEFF’s was slightly higher than unity, indicating that simple random sample variance estimates underestimate the true vari- ance. A total of 1,111 people were interviewed twice to gather data on the political party they supported and their political attitudes. Then, using the political party voted for as the depend- ent variable and four different attitude scales as predictors, a regression equation was con- structed. The mean value of the square roots of these four DEFF’s was higher than unity, but lower than the value in the first study. In addi- tion, the mean of the square roots of the DEFF’s for the means, the mean value for corre- lation coefficients, and the mean value for par- tial correlation coefficients all fluctuated near unity. Data collected in 3,990 interviews conducted in three national household samples were em- ployed to form six dummy variable regression equations. The predictor variables in this study were education, type of occupation, family in- come, family reserves, race, and relationship with relatives and friends. For the square roots of 64 DEFF’s, the mean value was again above unity. Kish and Frankel,!?7 under a contract with NCHS, estimated in the 16 linear regression equations based on the physiological measure- ments of 3,091 males in the HES the ratio of the replication variance to simple random sample variance. The mean value here for the regression coefficients was above 1 and higher than the pre- vious empirical values. The investigators sug- gested that this higher mean value resulted from the large clusters used in the HES. The last set of empirical results examined the variance estimates for statistics computed by the technique of multiple classification analyses. This technique is a multivariate analysis that examines the relationship of each predictor to a dependent variable that must be an interval scale or a dichotomy, with the other predictors held constant. The model is additive; and no assump- tions of equal cell frequencies, linearity of re- gression, and orthogonality are required. For each predictor, there are two estimated statis- tics, eta and beta coefficients. The squared eta coefficient is the proportion of variance of the dependent variable explained by unadjusted de- viations (cell means minus overall mean), and the squared beta is the proportion of variance of the dependent variable explained by a given pre- dictor, holding the other predictors constant. Data for a sample of 2,214 family heads were used in the multiple classification analyses equa- tion to fit a receptivity index to education of head of household, age of head, total family in- come, social participation, achievement orienta- tion, sex, and marital status. Because of the cost involved, the replication variance estimates were computed from 12 partially balanced replicates and the simple random sample variance esti- mates from 12 repetitions based on simple ran- dom splits of the data. For the six eta coeffi- cients, the mean DEFF’s were greater than 1, as were the mean DEFF’s for the corresponding betas. ; By means of a Monte Carlo approach, Levy! 9 compared the performance of the balanced replication and jackknife methods for the ratio estimator when the sample design was stratified, with two PSU’s selected from each stratum. Within each of 16 strata, a synthetic normal population with specified mean and variance was generated for two variables X, and Y,. The parameter to be estimated from the samples was R a ratio estimator. The sampling and estimation procedure was to draw from each stratum two independent estimates of X, and two independ- ent estimates of Y,. These estimates were fashioned into a combined ratio estimate, R. From a balanced set of 16 half-samples, an esti- mate of the variance of R was computed. An- other estimate of this variance was calculated by the jackknife method. In the experiment, the parameters R, were allowed to vary in succession according to the ratios 1.01, 1.05, and 1.10 for one value of the correlation between X and Y and one value of the relative variance of the population mean X. The correlation value used equaled 0.9, and the relative variance was set equal to 0.01. For analytical purposes, the estimated vari- ances, biases, and mean square errors for both the replication and the jackknife method were expressed relative to the sampling variance of the 1,200 estimates of R. The relative variance equaled the sampling variance of the variance estimates for a method divided by the square of the sampling variance of R. The relative bias squared equaled the square of the mean of the variance estimates for a method minus the sam- pling variance of R divided by the square of the sampling variance of R. The relative mean square equaled the sum of the relative variance plus the relative bias squared. The relative mean square errors for both methods increased slightly as the varying ratio increased from 1.01 to 1.10. The jackknife rela- tive mean square error estimates were lower than the estimates produced by replication for all three cases. The jackknife estimates of relative variance were also lower than replication esti- mates for all the varying ratios. When the R',’s varied by the ratios 1.05 and 1.10, the replica- tion method had the lower relative bias. These results indicate that further such studies using different parameter values are needed. Several of the design and estimation compo- nents included in complex sample surveys are unequal sampling fractions, clustering, stratifica- tion, first-stage ratio adjustment, nonresponse adjustment, and poststratification. In an empiri- cal investigation, Simmons and Bean20 meas- ured the effect these components, alone and in combination, had on replication variance. Data for the four health conditions—hypertensive heart disease, myocardial infarction, angina pec- toris, and syphilis—from the HES were used. Using six estimation procedures, six estimates of the proportion of the total population or pro- portion of a subclass of the population having a specified health condition were calculated. To implement these procedures, each examined person received a set of six weights. The weights assigned were (1) a weight of 1, as if the data were collected in a simple random sample; (2) a weight of 10,000; (3) the appropriate weight for the HES sample design; (4) the weight in 3, multiplied by a nonresponse adjustment; (5) the weight in 4, multiplied by a first-stage ratio adjustment; and (6) the weight in 5, multiplied by a poststratification factor. A simple random- sampling variance estimate was calculated for the estimate produced when the weight of 1 was used; replication variances were tabulated for the other five estimates. Ratios of variances and ratios of mean square errors formed the basis of comparison. For the ratios of variances, the numerator was the variance of one of the estimates, and the denominator was the variance of another esti- mate. For the ratios of variances, the numerator was the variance estimate produced by one method, and the denominator was the variance estimate produced by one of the other methods. To take bias into account, the ratio of the mean square error of one estimate to the mean square error of another was examined. The estimate 10 produced by the full design and estimation procedure (the last weighting scheme) was sub- stituted for the true value. Thus, an estimate of bias equaled the difference between this esti- mate and the estimate obtained for the method in comparison. The results of the comparison based on ratio of variances and the general conclusions are presented. The use of unequal sampling fractions alone showed a 27-percent increase in replica- tion variance over simple random sample vari- ance. The combined effect of unequal weighting, clustering, and stratification produced a 137-per- cent increase in replication variance over a simple random sample model. However, the variance estimate for the estimate produced by the full design and estimation procedure was only about double the simple random variance estimate. Thus, the estimation steps—non- response, first-stage ratio adjustment, and post- stratification—improved the precision of the estimates. Simmons and Bean concluded that none of the design and estimation components investigated introduced large biases in the esti- mates and that these components have substan- tial impacts on variance estimates. The strong point of the replication variance estimator method is that measurement errors, design components, and estimation procedures are all accounted for. The technique can be employed to yield variance estimates for any statistic and, thus, is flexible. It can also be adapted to a design that selects more than two PSU’s from each stratum. Linearization Method The other general method for estimating variances was developed by Keyfitz,2! who outlined a method for estimating the variance of a poststratified estimator. This estimator is derived from the fact that the variance of a sum of two independent estimates of a parameter equals the expected value of the square of the difference between them. This is VAR(X] + X;) = E(X] - X5)? where X', and X', are estimates of the param- eter X made from two independent random samples drawn with replacement. The key theorems in developing the method will be sketched without giving any proofs. This expres- sion can be extended to include estimates drawn from another stratum, assuming the selections among the strata are independent. The next term needed is the covariance between characteristics measured on the same sampled units within a stratum. Assume X'; and Y', are estimates from one random sample and X'y and Y', are estimates from another random sample. The two samples are drawn independ- ently with replacement and, thus, Cov(X';,Y"y) = Cov(X'p,Y';) = 0, E(X'}) = E(X'y), and E(Y'|) = E(Y'y). Then the covariance can be shown to equal Cov(X} +X5),(Y] + Vy) = E(X)- X5)(Y] - Ys) The key in the derivation of the linearization method is the approximation of the relative variance of a ratio estimator using the Taylor series expansion. Defining the relative variance of an estimator as the variance of the estimator divided by the square of the expected value of the estimator, then relative variance (V2) is xX’ [ x’ Y' 12 SOR Era ~ V2(X') + V2(Y') - 2V(X",Y') where > 7 = the ratio estimate of the population net aramet 2 p er VX, Y') = p(X, Y) V(X) V(Y), p(X',Y") = the correlation between X' and Y’, V(Y') = the coefficient of variation of Y', and (X') = the square root of VZ(X"). A proof of this result is in Hansen, Hurwitz, and Madow.22 Assuming again independent random samples drawn with replacement and using the above theorems, substitution, and algebraical manipu- lations, it may be shown that X) +X; Xx -X; Yi-Y; P pel 2) _ ld 2 Nn 2 Y, +Y,| CIE +Xy)T E(Y) + Yy) If selections are independent among strata, these results can be generalized to any number of strata. The last theorem gives the variance of a poststratified estimator. The sample design is two random selections drawn independently and with replacement from each stratum. The post- stratified estimator is where X" = the poststratified estimate of X charac- teristic, X,.i = the ith estimate for the ath class of the hth stratum, Y,,; = the ith estimate of population for the ath class of the Ath stratum, Y = precalculated total population in the ath group. The difficulty in determining the variance of X" is the lack of independence among the classes. The variables defining the classes are not used to stratify the population. Therefore, the covari- ance between the classes is not zero. This method evaluates the covariance between the groups separately in each stratum. These values 1 are then summed across the strata. The variance of X" is EX (X,,, + X02) VAR(X") = ED ly ————— “CEL (Yar + Vian) ’ ’ ! ’ Xha1 - Xpa2 Yar - Yao X EX, + X00) EQ (Y,, + Yia2) In this discussion, random sampling with replacement has been assumed. If sampling is without replacement, the finite population cor- rection factor can be inserted into the formulas. From these equations, formulations applicable to complex sample designs have been developed. Tepping2?3 developed a more general procedure for linear variance estimators using a Taylor series expansion and showed how the method can be employed to provide variance estimates for more complicated statistics such as regres- sion coefficients. Bean presented the theory, proofs, and description of the form of the variance estimator for the ratio estimates pro- duced by the HIS report.24 Shapiro?5 and Waltman?6 outlined the adaptation of Keyfitz’ theorems to the ratio estimation procedures employed in the CPS. In essence, the lineariza- tion method fashions a variance estimator from a linear combination of sample totals for each PSU. Statisticians of the Bureau of the Census were among the first to apply the linearization method. They investigated its use in the CPS’s sample design and estimation procedures and constructed a computer program tabulating the values needed for each PSU. Banks and Shapiro® presented and discussed the empirical basis the Bureau had acquired over the years from its variance estimates for ratio estimators. They also assembled the data on the replication variance estimates. The methodology used in the analysis was the comparison of design effects for un- biased estimates, ratio estimates, and composite estimates for a variety of labor force variables. The values (true variance as computed by linearization technique) for the unbiased esti- 12 mates and ratio estimates indicated that, with- out exception, the ratio estimate lowered the variance. In fact, for the variables with large parameter values, the decrease was substantial. When the comparison was made between the ratio estimates and the composite estimates, the latter reduced variance of some of the labor characteristics, but not all of them. Data from 1964 on 13 variables, the relative variances as estimated by linearization and repli- cation (not balanced) methods, were displayed in Banks and Shapiro. Both estimates were consistently of similar size, but there was more variability among the monthly replication vari- ance estimates. The authors concluded that the linearization estimates were more reliable. Linearization is the device by which variance estimates of the ratio-estimated characteristics of the Canadian Labor Force Survey are calcu- lated. Fellegi and Gray2?7 explained the applica- tion of this method to the survey with its particular sample design and estimation proce- dure and discussed how the analysts actually make use of the design effect in their analysis. The comparative findings of the design effects were consistent with the data from the U.S. Bureau of the Census study. The literature on linearization is small com- pared to that on replication. There are probably two reasons for this. Until Tepping’s article,3 this method was not available for the estimation of variances for statistics other than means, and the method did not generate the interest of the replication scheme. Furthermore, samplers may have used the technique without reporting its general behavior. One feature of linearization is that the variance can be estimated for each stage ‘of sampling, a fact that is important for design- ing surveys. An Empirical Investigation of All Three General Variance Estimators In the articles reviewed, there are few com- parisons of the properties and behavior of the estimates produced by the two methods when they are applied to the same data. The only extensive research along these lines has been the work of Frankel.28 Using repeated sampling from a universe modeled from real sample survey data, the validity of three fundamental assumptions, important for purposes of statisti- cal inference, was assessed. The assumptions were the following: 1. The sample estimate of the population parameter is approximately unbiased. 2. An approximately unbiased estimate of the variance of the estimate is computable from the sample. 3. The distribution of the ratio of the sample estimate minus its expected value, to its estimated standard error is reasonably approximated by the “Student” t within symmetric limits. Frankel regrouped the CPS’s March 1967 data on noninstitutional families and individuals into PSU’s consisting of three segments of approxi- mately five to six households each. These segments were the basic sampling units of CPS. The PSU’s were stratified into 6, 12, and 30 strata, and two PSU’s were selected independ- ently from each stratum for each design. For design I (6 strata) and design II (12 strata), 300 samples were drawn; for design III (30 strata), 200 samples were chosen. Thus, the sample design consisted of one stage of sampling, stratification, and single-stage clustering. The estimation procedure followed in each sample was the fitting of two linear regression equations. The total income of household head was related to the number of persons in the household under 18, the number of persons in the household, and the sex of the household head. For the second equation, the total income of the household was the independent variable, and the dependent variables were number of persons in household in labor force, age of household head, and years of school completed by household head. Estimates were made of simple means, differences of means, correlation coefficients, regression coefficients, partial cor- relation coefficients, and multiple correlation coefficients. For each of these estimates, the variance was calculated in nine different ways. The nine methods included all three of the general variance estimators: Taylor expansion dfrankel, 28 p. 2. (linearization), balanced repeated replication, and jackknife repeated replication. A descrip- tion of the specific jackknife formulation used can be found in Frankel’s?8 work. Frankel compared the behavior of the esti- mates of means, difference of means, regression coefficients, and simple, partial, and multiple correlations on the basis of relative bias. Relative bias is defined to be the difference between the estimate of the parameter and the parameter divided by the parameter. The relative biases of the mean and differences of means were all less than 1 percent, except for one difference of means that had a relative bias of 3 percent. The regression coefficients had higher relative biases than did the means. In design I, seven out of the eight regression coefficients showed relative biases of less than 4 percent; in design II, only five out of the eight were less than 4 percent; in design III, the relative biases for seven of the regression coefficients were less than 4 percent. The correlation coefficients (simple, multiple, and partial) displayed the highest relative biases. Only for design III did the mean of the relative biases for simple correlations drop below 5 percent. In analysis of assumption 2, the nine variance estimators were also considered to be mean square error estimators, and their performances in that capacity were studied. The relative bias was the main measure of evaluation. For means and differences of means, none of the nine had minimum relative bias across all the designs; for simple correlations, one of the replication meth- ods emerged with lowest relative bias. However, for the estimates of regression coefficients, a jackknife technique displayed minimum relative bias for two designs and linearization behaved best for the other design. In all three designs, the replication estimate, involving the square of the differences between half-sample estimate and complement estimate, showed the lowest rela- tive bias for partial correlation coefficients. The results provided evidence for the validity of assumption 2, but the data did not definitely €There were nine methods because each of the last two techniques provide four different possible schemes of variance calculations (the four produced by replication are outlined in the Review of Literature section). 13 resolve the issue of which variance scheme produced estimates closest to the true variance. The conjecture that the estimates produced by the three general methods are really estimates of mean square errors rather than variances was studied using another relative bias expression. For designs I and II, variance estimates for means and differences were relatively closer to the mean square error parameter than the variance parameter. More research is required before a definite answer can be given as to whether the estimates produced by the general methods are mean square error estimates rather than variance estimates. Probably the most important conclusion reached concerned assumption 3. The ratios of the sample estimate minus its expected value divided by each of its nine estimates of standard errors were computed. The evaluation of the ratios was based on the proportion of times the ratios fell within symmetric intervals around the origin. The intervals chosen were normal values. One-sided intervals were also computed, and examination of the proportions within these intervals revealed instances of asymmetry. Frankel concluded that these findings were partly real and partly the result of selecting only several hundred samples rather than several thousand samples. Because of this, he decided not to compare the proportions with values in the ¢ table. To reduce variability of the ratios, proportions were averaged for similar statistics within a design. These proportions were com- pared to the value in the ¢ table with degrees of freedom equal to the number of strata. The findings indicated that the ¢ distribution within symmetric intervals could be used to make inferences. Another vital observation was that ratios based on the replication scheme were more likely to be in agreement with the propor- tions of that distribution than the proportions based on the estimated standard errors. The Frankel investigation has contributed significantly to knowledge concerning the prop- erties of the variance estimates produced by the three general variance estimators. The validity of the assumptions investigated needs to be con- firmed for different sample designs, for different variables, and for different estimators. 14 METHOD Description of Population Morbidity data collected by the HIS in 1969 on 131,575 civilian, noninstitutional U.S. indi- viduals constitute the universe for this investiga- tion. The HIS2? is a highly stratified, multistage annual probability sample that provides esti- mates of disease incidence and prevalence and health characteristics. The same basic sampling design has been used since the survey began in 1957. The first-stage unit, PSU, is a single county or several con- tiguous counties. The entire geographical terri- tory of the United States is classified into 1,900 such units. The next step is the stratification of the 1,900 units into 357 strata, many of which consist of only one PSU. Geographically, the PSU’s are divided into subunits (each containing an expected six households) called segments. Sampling for the survey is done in two stages: one PSU is selected with probability propor- tional to population size from each of the 357 strata. The PSU’s comprising an entire stratum are included with probability 1 and are referred to as self-representing PSU’s (SR PSU’s). The other PSU’s are called non-self-representing PSU’s (NSR PSU’s). Within each selected PSU, segments are interviewed. An interview schedule on every individual residing in the household is completed. Thus, the design includes the effects of stratification and two stages of clustering, providing the model for this study. The features of the sample design employed in this empirical investigation were stratifying the PSU’s, sampling of two stages, and clustering at two levels. The steps involved in implement- ing the design were to sample the first-stage units with probability proportional to popula- tion, the largest units entering the sample with probability 1, and then to subsample within the selected first-stage units. However, before any of the details of the sample design—e.g., the deci- sion on the sampling fraction to be used—could be determined, the HIS data had to be re- grouped. The 357 selected PSU’s, each representing a stratum in HIS, were taken as the total popula- tion. An examination of the 1969 data for the original Health Interview Survey PSU’s revealed a mean of only 18 segments per PSU, excluding from the average eight PSU’s having over 100 segments. This average segment size was too small to support a two-stage design. Therefore, the 357 units were regrouped to form 149 PSU’s for the universe. Paralleling the HIS design, the 149 PSU’s were classified into 19 strata, eight consisting of only one PSU, grouped primarily by geography and population size. The original segments were retained, except the real outliers (segments with three or fewer persons) were combined with other segments to reduce varia- bility of segment size within PSU. The HIS observations were for persons, but since in this investigation everyone within a selected segment was to be in the sample (eliminating the problem of adjusting for non- response), the person records were combined to form a single segment record. To recapitulate the construction of the population, the 357 PSU’s in the HIS sample were reorganized into 149 units, the units were categorized into 19 strata, the segments were edited for outliers, and the data collected on persons in each segment were combined into a single segment record. There are several reasons for using these data as the population. The data had already been collected, edited, and made available on mag- netic tape. However, even after the completion of these operations, further extensive manipula- tions were required to put the data into the form required to perform the necessary calcula- tions. One purpose of the project was to measure the effects of stratification and cluster- ing on the variance estimators studied. The data reflect the real homogeneity or clustering that exists in human populations. The estimation component, poststratification, which was to be investigated to ascertain whether the variance estimators were sensitive to estimation proce- dures, could easily be performed on these data. Another basis for their use was the variables collected in the survey. Evidence of the behavior of the balanced half-sample replication and the linearization method is needed for a variety of variables in order to determine if observed differences between the variance estimators pro- duced by the two methods are caused by the nature of the variables themselves or by the methods. The variables available from HIS data are primarily morbidity information based on an individual’s own perception of his health. Sample Design As stated earlier, the population was divided into 19 strata, 8 self-representing and 11 non- self-representing. Characteristics of the universe are given in table 3. First-stage sampling con- sisted of the selection of the eight SR PSU’s and the selection of two PSU’s from each of the 11 NSR stratum. The number of PSU’s in each NSR stratum was small; thus, the finite population correction factor (fpc factor) would have a considerable impact on the variances. The usual procedure would have been to sample without replacement and then include the fpc factor in the formulas for the variance estimator, but unfortunately, the theory for incorporating fpc factor in these calculations has not been developed. The solu- tion was to sample with replacement. One desirable feature of a design is for the sample to Table 3. Characteristics of the population Number of Number of | Population Stratum primary segments size sampling units 1 1 323 5,219 Dos hw Ee @ 1 324 5,295 B sussmswuss 1 293 5,097 LS 1 257 4,612 B swimeswmsn 1 270 4,649 6 .......... 1 191 3,338 7 vinainiis 1 339 5,519 8 suswmsumsn 1 271 4517 9 a... 14 593 10,238 10 uiswsnmsn 15 592 10,573 1M a. 13 526 8,815 12 oisadina sa 12 482 7,638 18 swiwmiwazs 13 461 7,332 14 .......... 1 449 7,809 15 wuiwvsmine 12 481 7,785 16 .......... 1 446 7,854 VD rv oun manis 13 455 7,679 18 .usmismane 13 502 8,997 19 .......... 14 513 8,610 Total ...... 149 7,768 131,575 15 be self-weighting, which means that each ele- ment has the same probability of being selected. This feature can be included when the PSU is sampled with probability proportional to size. The first stage of sampling included the selection of two PSU’s independently from the 11 NSR strata with probability proportional to size and with replacement. To determine whether sample size plays a part in the observed differences in the variance estimates produced by the methods, three sam- ple designs were used. The number of PSU’s sampled was constant, so the subsampling rate was varied to achieve different overall sample sizes. To accomplish this, a uniform sampling fraction f, equal to the product of the PSU selection probability of selecting the PSU times the subsample fraction within the PSU, was used in each stratum. The formula is =H Xk where f = uniform sampling fraction, fi = the sampling rate of first-stage units, and Jo = the sampling rate of second-stage units. In design I, f equaled 1/75; in design II, 2/75; and in design III, 2/30. The fraction f, equaled f/f1- Thus, by setting f and knowing f;, the f, could be computed. Because two PSU’s were chosen in the NSR strata, the values 1/150, 1/75, and 1/30 were used in the computations. The expected yields from the two fractions of designs I and III were 4 segments and 20 segments, which approach the two possible extremes in sampling: simple random and ulti- mate cluster. The optimum design probably falls betwecn the two rates used. The second stage of sampling was random subsampling with replace- ment of the segments at the rate f, from the chosen PSU’s. To clarify the sampling plan, consider design III, which had an overall sampling fraction of 2 in 30. First, the eight SR strata entered the sample with probability 1. Within each, seg- ments were randomly selected at the rate f, = 16 (2/30)/1 = 2/30. Thus, for stratum 1, approxi- mately 22 segments were drawn. Next, two PSU’s were chosen with probability proportional to population from each of the remaining 11 strata. Within a selected PSU, the segments were subsampled with the rate f, = (1/30)/f;. For each design, 900 samples were independ- ently drawn. In drawing these samples, not only was the sampling distribution of the variance estimators being estimated, but also being esti- mated was the sampling distribution of the ratio estimators R;, :=1,---, 900. In repeated sam- pling, this ratio estimator will settle down to its expected value, and, thus, the simple sample variance 900 _ 2 (R; - R)2/899 will become fairly stable. The value 900 was thought to be an adequate number to achieve stability. Variables Five variables were selected from the HIS to represent a variety of basic distributional forms. These are as follows: Variable Family income Number of restricted activity days in past 2 weeks Number of physical and dental visits in past 12 months Number of days spent in short-stay hospitals in past 12 months Whether or not the person has seen a physician in last 12 months Quantity estimated Average income per person Average number of restricted activity days per person per year Average number of visits per person per year Average number of short-stay hospital days per person per year Proportion of population seeing a physician in last 12 months Two criteria for selection of the variables were that (1) they could be obtained from the person-record HIS tapes and (2) they would have a variety of distributions. Ideally, what is fEach sample person’s record contained a measure for each of the five variables; therefore, to obtain a segment record, these measures were summed across all persons in the segment. desired is for the computed variance estimates to reflect the differences of the variance estimators. Thus, the possibility that the variables them- selves cause the observed differences between the variance estimates is investigated. It is important to remember in considering the variables measured in this study that the ultimate sampling unit is the segment and that although several variables are converted to a per-person basis, no direct assessment of within- segment variability is available. Thus, distribu- tions of variables in the study directly reflect segment-to-segment variation. The income variable is a family income figure. Thus, each person within a family is assigned the same value. Income, then, was a highly clus- tered, continuous variable. However, in the HIS, income was reported only in group intervals, with the highest class being open ended. To convert from grouped data, midpoints of each interval, except for the open one, were the values given to individuals. Everyone having a family income falling into the open interval was assigned the lowest income of the class. The proportion of persons in the population visiting a physician in the past 12 months was selected as an example of a variable distributed on the unit interval. The variable, number of restricted activity days, produces an incidence statistic based on the respondent’s recall of the past 2 weeks. The respondent’s recall for the remaining two variables is 12 months. The variable, number of days spent in short-stay hospitals in the past 12 months, has a J-shaped distribution. Most individuals do not spend any days in the hospital, but there are a few people whose hospital experience lasts for months. These two extremes account for the J-shape. These variables provide an example from each type and each range class of statistics estimated in the HIS. The statistics produced by the HIS are classified by the length of the recall period and by the usual value of the measure of a health characteristic for an individual. The classes are (1) narrow range—the measure is usually 0 or 1 and occasionally 2, e.g., the number of days spent in short-stay hospitals; (2) medium range—the typical value for an individ- ual is from 0 to 5, e.g., the number of physician and dental visits; and (3) wide range—the meas- ure for an individual is usually greater than 5, e.g., the number of restricted activity days. ESTIMATORS AND VARIANCE ESTIMATORS Estimators Two estimating equations, each producing from the sample a ratio poststratified estimate of the population ratio parameter, were employed with the level at which poststratifica- tion was done as the differentiating characteris- tic. Each equation consisted of three basic operations: inflation by the reciprocal of the probability of selecting second-stage units, infla- tion by the reciprocal of the probability of selecting first-stage units, and poststratification. The purposes of stratification are to improve precision of the overall population estimates and to insure that subgroups of the population (domains of study) are in the sample in the same proportions as they are in the universe. Usually, the universe cannot be stratified into these subgroups before sampling. However, if the distribution of the subgroups in the universe is known, the sample results can be adjusted (poststratified) so that these domains of study are appropriately represented and the accuracy of the estimate increased. The subgroups in this study were the popula- tion in 24 ethnic-sex-age classes (white and nonwhite, male and female, ages 0-4, 5-14, 15-24, 25-44, 45-64, and 65+). This spread resulted in small frequencies in most of the nonwhite cells, another reason for poststratify- ing. Typically, the distribution of the subgroups is known only for the complete population, and adjustments must be made at this level. In this situation, however, the universe was totally specified and poststratification could be carried out either after combining the stratum estimates for an ethnic-sex-age class or earlier. Poststratifi- cation was applied at two different levels: universe and region. Universe level meant that the sample esti- mates for a class were combined across all the strata and then adjusted to the total population in that group. Region-level poststratification refers to the adjustment of sample estimate for 17 the PSU’s within a region to the regional distribution of classes. The PSU’s belonged to one of two regions: the combined East and North Central regions or the combined South and West regions. The first estimator of a statistic (see appendix I for the formula for Ry) is R| —poststratification at universe level Ry = ‘ y 24 2 2% a=1 Ya y where R, = the final poststratified ratio estimate of the population parameter R, x; = the poststratified estimate of health characteristic x, y = the total population size, x = the inflated estimate combined across all PSU’s of health characteristic x for ethnic-sex-age class aq, y, = the known population in the ethnic- sex-age class a, and Va = the inflated estimate across all PSU’s of the population in class a. Variance Estimators The main objective of the study is to examine the behavior of the variance estimators produced by two general variance estimation methods, replication and linearization. Ideally, the vari- ance of each estimator produced by one of the two estimating equations would be estimated using the correct application of both variance formulas and each of the approximate formulas, but due to computer costs and volume of numbers, this was not feasible. For estimator 1, variance estimates were computed by the correct 18 replication and linearization methods and two approximate schemes, while for estimator 2, only the correct replication method was used. (See appendix II for the formulas for the approximate variance estimates and for the correct replication variance estimator for Ry.) An assumption of both general methods is that two PSU’s are chosen independently from each stratum. The sample design for this study with SR PSU’s violates the assumption giving rise to the problem of how to treat these PSU’s. The solution was to divide randomly the sam- pled segments of each SR PSU into two groups and to assume that these two groups, pseudo- PSU’s, were two independent selections from the same stratum. Variance Estimators for Estimator 1 The variance for the estimate R; was esti- mated in 10 different ways. Three replication methods, each yielding three estimators, and one linearization method were used. The process for computing the correct replica- tion variance estimator is as follows: Assume that from each stratum one of the two sampled primary units is selected. Nineteen PSU’s form a half-sample. The data from these 19 PSU’s are inflated by the reciprocal of the probability of selection to their stratum level. These estimates for a particular ethnic-sex-age class are combined across all the strata. Then, within the classes, the estimates are adjusted by poststratification fac- tors. These factors are calculated, using the population estimates for the class produced by only the PSU’s in the half-sample, on the universe level, in the same manner used to calculate the factors for the parent sample. The 19 PSU’s in the complement replicate undergo an identical process to yield a similar estimate of the parameter. This is repeated for 20 such half-samples. The three variance estimates pro- duced for each sample estimate obtained from estimating R, are 20 1 oo VAR(R 1H) = 55 2 (Ri - R,)? VAR(RIC) = and VAR(R1S) = 5 [VAR(R 1H) + VAR(R1C)] NO | — where R, = the parent sample ratio estimate, R}, =the ath half-sample ratio estimate of the population parameter R by an iden- tical estimation equation to the one that yielded R; , and R{¥, =the ath complement half-sample ratio estimate. The 20 half-samples are composed from the orthogonal pattern in table 4. This design was constructed by following the method described in Plackett and Burman.!3 The pattern is determined by the rotation of the first 19 entries of the first column and by specifying the 20th row to always be minus. The linearization formula employed in this study was derived from the theorems given by Keyfitz.2! The form of the method used here can be described as follows: Remembering that R, =x|[y, the formula for the aggregate x] is 24 y x] = 2 x: = a=1 Y, a This equation can be rewritten as linear combi- nations of the simple inflated estimates of the PSU’s from each stratum. Expressed this way, the equation becomes 19 24 (*na1 + Xha2) "no_ h=1 Xp = Ya 19 4 Vhai + Yha2) a=1 h= Table 4. Orthogonal pattern for 20 replicates’ Stratum Replicate SR PSU's NSR PSU's 112 |13|4|5(6|7|8|9 (10 |11 12 13 | 14 | 15 | 16 | 17 | 18 | 19 1 sir Cwm esse HEWES s + + - - + + + + + - + = — - - + + - 2 Lm era REE EE, + =~ — a + + + - + - - - = + + — + ZB camry mmm ny - - + + + + —- + —- - - — - + + - + + 4 - + + + + - + = + - wi > - + + - + + - B i inunssssmmewss + + + + - + = - - - = + + - + + - . 8 ..inewmiisrmEwE Ly + + + - + + - - - + + - + + — = ¥ Zoro mmm mons ns mE A + + - + - + = = - - + + - + + ow. = + + 8 + - + + = - = = + + - + + - - + + + O Lissnnrrss ewan se - + - + = = = = + + - + + - - + + + + 10 iio nmrrsawpmmus + - + = = = = + + - + + - - + + + + - 1 ::issmmurss sommes - + = = = + + - + + - - + + + + - + 12 ncn rr a mm bn + = - = = + + + + - — + + + + = + - 13 him rr mmm - - = = + + - + + - - + + + + ~ - + 14 - = = + + - + + - - + + + + = + =~ - 18 iii uum sspma gs - = + + - + + - - + + + + = + = + - = 16 ..:s ues sis s mam - + + - + + = + + + + - + — + - - — 17 2:25 mbes es medi + + - + +--+ + + + - + foe = = ~ 18 + - + +--+ + + + - - + - a - + 1G och hmm ew wm oF = = + FF + + - + - - ~ -~ + 20 wis avwmunys www - = = ae ow we - a - - - = _ _ _ _ Note.—SR PSU indicates self-representing primary sampling unit; NSR PSU indicates non-self-representing primary sampling unit. 14+" denotes +1 and “'~'' denotes —1. where the inflated estimate of health charac- teristic x for class a of the 7th PSU in the Ath stratum and Xhai Vi ui = the inflated population estimate for classa of the ith PSU in the Ath stratum. From this expression, the linearization variance estimator of R, is derived. The formula is VAR(x;') VAR({L])= =———ee 2 y 19 24 sl? 2 Eloi 5a Xh2) -3 3 ni Vhat = sha) y?2 where xy; = the inflated estimate for the /th PSU of the Ath stratum, x], = the final poststratified ratio estimate of the health characteristic x for the ath ethnic-sex-age class. For more detailed formulas and discussion, see Bean.30 RESULTS Introduction The main objective of the investigation was to determine if either or both of the general variance estimator methods, replication and linearization, yields acceptable variance esti- mates for a complex ratio estimator employed in multistage probability sample surveys. The prop- erties considered critical for acceptability are the extent of the bias of the variance estimates, the degree of their variability, and the amount of their mean square error. Another very important point is whether inferences can be made when variance estimates are used. The other result presented is the effect of the estimation tech- 20 nique poststratification on variances. The bias of the two ratio estimators employed and the substitution of approximate variance estimates are discussed in appendixes I and II. Bean3? presents a more in-depth report on other proper- ties of the ratio estimators, other approximate variance estimate methods, and the results, using several different procedures for estimating the variance component from SR PSU’s. Comparison of Replication and Linearization The primary purpose of this investigation was to compare replication and linearization meth- ods of variance estimation for a sample design that includes stratification, two stages of cluster- ing, and poststratified ratio estimation. The comparisons are in terms of bias, variance, and mean square error. Only for the sample estimator R, were both the appropriate replication and linearization variance estimates computed. “Appropriate” here refers to the best application of the theory of the methods to the particular sample design and estimation process employed. For the repli- cation method, this means that the data for the PSU’s in a half-sample were (1) inflated to the PSU level, (2) inflated next to the stratum level, and (3) poststratified by adjustment factors computed from the half-sample itself. Therefore, the half-sample estimate of the population parameter contained all the elements of the sample design and estimation procedure. For the best application of the linearization theory, the basic equations by Keyfitz2! were interpreted to include inflation to PSU and stratum levels and to include an adjustment for poststratifica- tion. For each estimate of R |, one linearization and three replication variance estimates were calculated. The replication estimates VAR (R1H), VAR(R1C), and VAR(R1S) all have the same expected value. Because VAR(RI1S) is an average of the other two, its precision is ex- pected to be greater. Since the estimator VAR(R1H) is the one most often used by statis- ticians, the discussion will deal with it and with VAR(R1S). One important question is whether the meth- ods yield biased statistics. To compute true bias, the real variance of R, for this sample design is needed. Unfortunately, this value was not known, but with 900 samples, the distribution of R, should become quite stable; and the sample variance of 900 R; estimates should be a good criterion. The sample variance of R, is where Ry; By assuming s2jp , to be the true variance, measures of bias can be calculated. These are = VAR(RTH) - s% bias [VAR(R1H)] 1 where 900 2. [VAR(RIH)], VAR(R1H) = =I os bias [VAR(R1S)] = VAR(RIS) - s2, 1 where 900 VAR(RTS) = 555 2 VAR(R1S)], and bias [VAR(L1)] = VAR(LT) - s2 1 where 900 VAR(L1) = 2 [VAR(L1)]; 0 Inspection of table 5 reveals that the bias of the R1H estimator decreases across the three designs for four of the five variables. Only for the variable proportion seeing a physician does the bias, which goes from -6.083 X 10-6 in design I to 1.614 X 1073 in design II and then down to 4.206 X 10-6 in design III, not follow the pattern. For physician visits and hosp‘ral days, the bias is always positive, while for family income and restricted activity days, it switches from positive to negative. Table 5. The bias of the R1H, R1S, and L 1 variance estimators of the sample estimator R, of the population parameter R Variable Variance estimator Restricted Proportion seein Family income st Physician visits Hospital days p Saf 9 activity days a physician Design | VARIBIH) .cumnsisvansmms 3.857 X 10° 2.736 X 10°! 2.751 X 10°? 3.750 x 10° -6.083 X 10° VARIRIS) ..:uvirisevnmas 4.103 xX 10° 2.783 XxX 10°! 2.558 X 10°? 3.765 xX 1072 -7.095 X 10°¢ VARILYY i. inv svivnmvmas -3.191 xX 10° | -1.085X 10! -4.050X 10°* | -2.242%X 10°? -4.944 X 107% Design II VAR(RIH) ............... 1.489 xX 10° | -4.337.x 10? 2.693 x 10°? 1.929 x 1073 1.614 xX 107° VAR(R1S) ............... 1.640% 10° | -4.092 X 10°? 2.688 x 10°? 1.785% 10°? 1.581 xX 107° VARILYY ..vovwucsissvnmwmws -3.211X 10? | -1.598 X 10°! 1.022%X 107% | -6.848X 107° 6.945 X 10°¢ Design 11 VAR(RIH) ............... -8.772X 10* | -2.934 X 10°? 1.073 x 10°? 4490 X 10°° 4.206 X 10°¢ VAR(R1S) ............... -8.909 X 10? | -2.900 X 10°? 1.042% 1073 3.684 xX 107° 4.069 X 10°¢ VARI ..owwwmiossmmmps -1.271 x 10° 7.121 X10™* | -3.094 X 10°* 2.193 X 10°¢ -6.818X 102 21 For family income, restricted activity days, and hospital days, the bias of VAR(R1S) also decreases with increasing sample size. For the first two variables mentioned, the bias is positive in design I, but in design III, it is negative. For physician visits, the bias in design II is slightly greater than the bias in design I, while for design III, it is at its smallest value. VAR(R1S)’s bias shows a different trend for proportion seeing a physician. The bias is = 7.095 X 106 in design I, which increases to 1.581 X 10-7 in design II and then drops to 4.069 X 1076 in design III. This is the only example for VAR(R1S) and VAR (R1H) for which the bias first is negative and then becomes positive. VAR(L1) exhibits a different pattern. The biases for physician visits and proportion seeing a physician are negative for design I. These biases decrease across the designs. However, the method goes from underestimating to over- estimating the variance. The bias is the smallest in design II for family income and hospital days. Comparing the biases of the three methods does not yield a consistent pattern of one of the methods having less bias than the other. The important fact is that for all three estimators, the bias is small and tolerable. However, turning to the variances of the methods in table 6, more of a trend is found. The variances of the variance estimates are: VAR[VAR(R1H)] = an 900 2 X 3 {[VAR(R1H)],;- VAR(R1H)} J=1 VAR[VAR(R1S)] = 565 900 2 XY. {[VAR(R1S)] ;- VAR(RIS)} j=1 VAR[VAR(L1)] = 555 900 2 X 3 {[VAR(L1)];- VAR(LI)} i=1 The first point to consider when examining the variance of estimators is to check to make sure the variance does decrease with increasing sample size. This is true for all three methods. Table 6. The variance of the R1H, R1S, and L 1 variance estimators of the sample estimator R, of the population parameter R Variable Variance LU Family income Restricted Physician visits | Hospital days Proportion Seeing activity days a physician Design | VARIRIH) . +: wows sss 3 amm oss wu 1.144 xX 10° 2.052 2859X 107 | 2.428% 10°? 1.733Xx 107% VARIRBISY .:issnmssinsmmmens ss 1.111 X 10° 1.976 2.763 Xx 107? 1.902% 10°? 1.579 x 107% VARILY) ...svomsisannmderssn 1.062 X 10° 1.487 2.380%X 107° | 2.833x 107* 1.265X 10°* Design II VARIRIHY .::ivsmwisssmmmsasss 3.217 x 10% 0.521 8.306 X 10° 1.199 X 10° 3.471 X 10° VAR(R1S) . .... 3.201 xX 10° 0.505 | 8.274 x 10°* 1.042 Xx 1074 3.307 Xx 107° VAR(LT) 3.258 x 10° 0.460 | 8.250 X 10°* 5.651 X 10° 3.338% 10° Design 111 VARIRIH) . ..: sucess mmmeisis 1.047 X 10% 0.140 | 8.271 Xx 10% 1.319x 107% 1.104 X 10°° VARIBRISY . .:. sss visimemussiss 1.036 X 10° 0.137 8.132X 10°* 1.288 X 10°% 1.089 X 107° VAR(LT) 1.048 x 10° 0.131 7.959 X 107° 1.135% 10% 1.086 X 10°° 22 The variance of the VAR(R 1H) is always larger than the variance of VAR(R1S). This provides empirical evidence that R1S is more precise than the other forms of replication variance esti- mators. With three exceptions—family income in designs II and III and proportion seeing a physician in design II—the variance of the VAR(L1) method is less than the variance of VAR(R1S). It is preferable to judge the merits of the three methods on the basis of the mean square error, which takes into consideration both prop- erties of bias and variance. The ideal estimator would be one that produces minimum mean square error for all values of the parameter o*R, . Perhaps such an estimator does not exist, but the empirical results here provide evidence on the behavior of the mean square error for the R1H, R1S, and L1 estimator methods. The formulas for the mean square errors (MSE) are MSE[VAR(R1H)] = VAR[VAR(R 1H)] + bias? [VAR (R 1H)] = VAR[VAR(R1H)] + [VAR(R 1H) - s2 12 1 MSE[VAR(R1S)] = VAR[VAR(R1S)] + bias? [VAR(R1S)] = VAR[VAR(R1S)] + [VAR(R1S) - ix 2 and MSE[VAR(L1)] = VAR[VAR(L1)] + bias? [VAR(L1)] = VAR[VAR(L1)] + [VAR(L) - 52 12 1 As can be observed in table 7, the mean square errors for the methods for each variable decrease with increasing sample size. When looking at the values for the procedures pair- wise, the results indicate that VAR(R1S) always has a smaller mean square error than VAR (R1H), which is easily explained by the less Table 7. The mean square error of the R1H, R1S, and L 1 variance estimators of the sample estimator R, of the population parameter R Variable Variance bsumator Family income Restricted Physician visits | Hospital days Proporgion iia activity days a physician Design | VABRIBIHY : cwv sss nmma vs ss mun 1.159 X 10° 2.127 2.866 X 10°? 2.442% 107? 1.736 X 1078 VABIRIS) . vv vcs a mod 68 4 8 no 1.128 X 10° 2.053 2.770X 107? 1.916 Xx 1073 1.584 X 107% VARIILT) ooo 1.072 X 10° 1.499 2.396 X 107? 2.883 X 107* 1.510X 1078 Design II VARIBIH) i vv isis muwsissnmens 3.239 X 10° 0.522 8.379 X 10°* 1.236 X 107* 3.731 xX 107° VARIRIS) ova iis omemmes is mp 3.228 X 10° 0.507 8.323 X 10°* 1.074 Xx 10°* 3.557 X 10°? VABRILY) ius vrssmammesss wimp 3.259 x 10% 0.485 7.653 x 10°* 5.651 X 107° 3.386 X 107° Design Il VAR(RIH) . .. 1.055 x 10° 0.140 8.387 x 10°¢ 1.320 107° 1.122% 107° VARIRISY vw ss so mums as wmmms 1.044 x 10° 0.138 | 8.240x 107° 1.288 X 10° 1.106 X 107° VARILY) svc iss nme sssss wmmmss 1.064 X 10° 0.135 8.010 xX 10° 1.145 x 10° 1.091 x 107° 23 variability of VAR(R1S). The other finding is that VAR(L1) has a smaller mean square error than VAR(R1S) aside from the variable family income in designs II and III. The mean square error is less even though VAR(L1) did not always have the smaller bias. For the only two cases where the mean square error of VAR(L1) is not lower, the variance of VAR(L1) was greater. In the other situation, proportion seeing a physician in design II, of greater variance for the L1 method, the difference in the bias was large enough to cause VAR(L1) to have the smaller mean square error. The behavior of the replication and lineariza- tion methods for the properties of bias, variance, and mean square error is very good, which implies the methods are yielding acceptable variance estimates. However, there are other properties to consider when deciding on a variance method; one of these is examined next. Normality of Standardized Estimators As mentioned previously, scientists are in- creasingly concerned with making inferences from data collected in scientific sample surveys rather than just calculating descriptive statistics. Ideally, an investigator would like to construct confidence intervals for population parameters. The statement that the average annual income per person lies between $7,500 and $9,400 with 95 percent confidence is more useful than a point estimate of the average income. Such a statement using the normal distribution could be made if one is willing to assume that the variable is sufficiently close to being normally distrib- uted and that the departure from random sampling is not a severe limitation. The objective here is to study empirically the ratio estimate minus its expected value divided by its estimated standard error. The analysis consists of computing the proportion of times the statistic falls into certain regions. By this procedure, the applicability of confidence state- ments based on the normal distribution can be approached. In this section, only the statistics for which VAR(R1H), VAR(R1S), and VAR(L1) were 24 used as the variance estimates will be discussed. These ratios are R, - E(R,) , VAR(R 1H) R, - ER) 3 \/VAR(R1S) and where " 1 900 _ E(Ry) = 900 2 Ry; 1 Such ratios for each estimate in every sample and the proportion of times the ratio fell within the regions—(-1.000, 1.000), (-1.000, 0), (0, 1.000), (-1.645, 1.645), (-1.645, 0), (0, 1.645), (-1.960, 1.960), (-1.960, 0), (1.960, 0), and (-2.576, 2.576)—were computed. These regions were chosen for purposes of comparison with the normal distribution. The proportion of area under the normal curve for the four symmetric intervals are 0.6827, 0.90, 0.95, and 0.99, respectively. The proportion of times the ratio fell within the limits is given in tables 8-10. Considering just the symmetric intervals, one immediately notices that intervals based on this type of statistic usually have a lower confidence than the normal level, but the differences are gener- ally very small. There are instances in which the proportion of time the statistic based on these different variance estimates fell within the limits exceeded the corresponding normal values. These results are good news, indeed, because coupled with the findings of Frankel?8 they justify the type of inferences consumers of data collected in multistage complex sample surveys often wish to make. For example, a health planner concerned with the use or expansion of health care facilities would most likely need to draw an inference from the estimated propor- tion of the population who saw a physician at least once in the past 12 months. The data indicate that such an inference based on the normal distribution is really not a bad approxi- mation and could be used with reasonable confidence. A word of caution about these findings is appropriate. Inspection of the one-sided inter- Table 8. Proportion of times the sample estimate minus its expected value divided by R1H estimate of standard error is within the stated limits Limits Variable +1.000 | -1.000, O 0, 1.000 +1.645 | -1.645,0 0, 1.645 +1.960 | -1.960,0 0, 1.960 +2576 Design | Family income 0.6789 0.3411 0.3378 | 0.8833 0.4500 0.4333 | 0.9356 0.4744 0.4611 0.9800 Restricted activity days 0.6778 0.3322 0.3456 | 0.8811 0.4411 0.4400 | 0.9367 0.4711 0.4656 0.9789 Physician visits . . . .| 0.6444 0.2989 0.3456 | 0.8744 0.4300 0.4444 | 0.9233 0.4633 0.4600 0.9822 Hospital days 0.6556 0.2933 0.3622 | 0.8622 0.4156 0.4467 | 0.9122 0.4489 0.4633 0.9578 Proportion seeing a physician 0.6567 0.3433 0.3133 | 0.8689 0.4400 0.4289 | 0.9300 0.4633 0.4667 0.9811 Design 11 Family income 0.6944 0.3567 0.3378 | 0.8833 0.4467 0.4367 | 0.9389 0.4733 0.4656 0.9744 Restricted activity days ....| 0.6356 | 0.2956 0.3400 | 0.8844 0.4356 0.4489 | 0.9300 0.4622 0.4678 0.9778 Physician visits . . . .| 0.6733 | 0.3244 0.3489 | 0.8967 0.4478 0.4489 | 0.9367 0.4744 0.4622 0.9811 Hospital | days 0.6722 | 0.3233 0.3489 | 0.8733 0.4367 0.4367 | 0.9289 0.4722 0.4567 0.9767 Proportion | seeing a physician 0.6833 | 0.3489 0.3344 | 0.9089 0.4589 0.4500 | 0.9522 0.4789 0.4733 0.9822 Design III Family income 0.6656 | 0.3189 0.3467 | 0.8656 0.4167 0.4489 | 0.9278 0.4522 0.4756 0.9767 Restricted activity | days ....| 0.6511 0.3322 0.3189 | 0.8700 0.4456 0.4244 | 0.9244 0.4733 0.4511 0.9756 Physician visits . . . .| 0.6844 0.3333 0.3511 0.8989 0.4456 0.4533 | 0.9456 0.4678 0.4778 0.9822 Hospital | days 0.6789 0.3422 0.3367 | 0.8789 0.4467 0.4322 | 0.9244 0.4722 0.4522 0.9656 Proportion seeing a physician 0.6900 0.3511 0.3389 | 0.8756 0.4378 0.4378 | 0.9222 0.4556 0.4667 0.9833 25 Table 9. Proportion of times the sample estimate minus its expected value divided by R1S estimate of standard error is within the stated limits Limits Variable +1.000 | -1.000,0 | 0,1.000 | +1.645 | -1.645,0 | 0,1.645 | +1.960 | -1.960,0 | 0, 1.960 | +2.576 Design | Family income . | 0.6800 0.3456 0.3344 | 0.8822 0.4478 0.4344 | 0.9344 0.4733 0.4611 0.9789 Restricted activ- ity days 0.6722 0.3278 0.3444 | 0.8811 0.4356 0.4456 | 0.9378 0.4700 0.4678 0.9789 Physician visits . | 0.6556 0.3067 0.3489 | 0.8733 0.4300 0.4433 | 0.9244 0.4622 0.4622 0.9778 Hospital days . . | 0.6611 0.2956 0.3656 | 0.8644 0.4178 0.4467 | 0.9156 0.4533 0.4622 0.9578 Proportion seeing a physician 0.6578 0.3433 0.3144 | 0.8656 0.4356 0.4300 | 0.9322 0.4667 0.4656 0.9811 Design I1 Family income . | 0.6911 0.3567 0.3344 | 0.8822 0.4456 0.4367 | 0.9378 0.4733 0.4644 0.9756 Restricted activ- ity days 0.6344 0.2956 0.3389 | 0.8833 0.4333 0.4500 | 0.9311 0.4644 0.4667 0.9756 Physician visits . | 0.6722 0.3233 0.3489 | 0.8978 0.4500 0.4478 | 0.9367 0.4733 0.4633 0.9778 Hospital days . . | 0.6800 0.3244 0.3556 | 0.8744 0.4389 0.4356 | 0.9344 0.4744 0.4600 0.9767 Proportion seeing a physician 0.6767 0.3456 0.3311 0.9133 0.4611 0.4522 | 0.9522 0.4800 0.4722 0.9822 Design 111 Family income . | 0.6678 0.3200 0.3478 | 0.8667 0.4189 0.4478 | 0.9256 0.4500 0.4756 0.9778 Restricted activ- ity days 0.6511 0.3311 0.3200 | 0.8689 0.4456 0.4233 | 0.9222 0.4722 0.4500 0.9778 Physician visits . | 0.6878 0.3356 0.3522 | 0.8978 0.4422 0.4556 | 0.9422 0.4656 0.4767 0.9822 Hospital days . . | 0.6789 0.3422 0.3367 | 0.8789 0.4478 0.4311 0.9267 0.4722 0.4544 0.9644 Proportion seeing a physician 0.6900 0.3511 0.3389 | 0.8778 0.4389 0.4389 | 0.9256 0.4589 0.4667 0.9833 vals indicates that they are not symmetric. Thus, any statements about one-sided intervals cannot be made, and confidence intervals com- puted are not of minimum length. An area of further study could be the direct corroboration of the skewness and leptokurtosis noted in the tables by computing and examining the third and fourth sample moments. Influence of Poststratification Another purpose of the study is to determine the effect the estimation component poststratifi- cation has on variances. The reason for adjusting is to lower the variance. Normally, to reduce variance a larger sample is drawn, but if the same thing can be accomplished by the estimation 26 procedure, there can be a savings in cost by using a smaller sample size, unless, of course, the savings in field costs are offset by costs of implementing the estimation procedure. Vari- ance reduction has been demonstrated primarily using simplified methods and theoretical analy- sis, while the studies of Simmons and Bean20 and Banks and Shapiro? gave empirical evidence. This investigation examines the effect for the particular design used here. The VAR(R 1H) and VAR(R1C) schemes call for the poststratification factors to be computed within each half-sample and each complement half-sample, while for the methods VAR (R2H) and VAR(R2C), the parent sample adjustment factors are used. (See appendix II for formulas for VAR(R2H) and VAR(R2C).) The orthog- Table 10. Proportion of times the sample estimate minus its expected value divided by L1 estimate of standard error is within the stated limits Limits Variable +1.000 | -1.000,0 | 0,1.000 | 1.645 | -1.645,0 | 0,1.645 | +1.960 | -1.960,0 | 0,1.960 | +2.576 Design | Family income . | 0.6733 0.3400 0.3333 | 0.8622 0.4378 0.4244 | 0.9211 0.4667 0.4544 0.9722 Restricted activ- ity days 0.6433 0.3189 0.3244 | 0.8611 0.4244 0.4367 | 0.9244 0.4667 0.4578 0.9711 Physician visits . | 0.6211 0.2867 0.3344 | 0.8489 0.4222 0.4267 | 0.9044 0.4522 0.4522 0.9711 Hospital days . . | 0.6200 0.2844 0.3356 | 0.8344 0.4011 0.4333 | 0.8933 0.4411 0.4522 0.9600 Proportion seeing a physician 0.6189 0.3211 0.2978 | 0.8378 0.4244 0.4133 | 0.9022 0.4511 0.4511 0.9711 Design 11 Family income . | 0.6811 0.3522 0.3289 | 0.8833 0.4444 0.4389 | 0.9289 0.4667 0.4622 0.9744 Restricted activ- ity days 0.6178 0.2867 0.3311 | 0.8644 0.4211 0.4433 | 0.9211 0.4567 0.4644 0.9711 Physician visits . | 0.6533 0.3167 0.3367 | 0.8900 0.4467 0.4433 | 0.9356 0.4744 0.4611 0.9767 Hospital days . . | 0.6511 0.3156 0.3356 | 0.8533 0.4256 0.4278 | 0.9200 0.4678 0.4522 0.9700 Proportion seeing a physician 0.6544 0.3378 0.3167 | 0.9022 0.4522 0.4500 | 0.9467 0.4744 0.4722 0.9800 Design 111 Family income . | 0.6644 0.3167 0.3478 | 0.8600 0.4133 0.4467 | 0.9256 0.4489 0.4767 0.9767 Restricted activ- ity days 0.6400 0.3256 0.3144 | 0.8611 0.4411 0.4200 | 0.9156 0.4700 0.4456 0.9756 Physician visits . | 0.6811 0.3311 0.3500 | 0.8956 0.4444 0.4511 | 0.9378 0.4622 0.4756 0.9822 Hospital days . . | 0.6689 0.3322 0.3367 | 0.8733 0.4456 0.4278 | 0.9233 0.4744 0.4489 0.9644 Proportion seeing a physician 0.6867 0.3489 0.3378 | 0.8722 0.4367 0.4356 | 0.9244 0.4567 0.4689 0.9844 onal pattern used for the methods was the same, producing identical composition. The ratio of the variance estimator secured from VAR(R1H) to the estimator produced by VAR(R2H) does measure the relative impact of poststratification. The average of these ratios was calculated. The ratios are defined to be oS 1 900 [VAR(R1)]; 900 = [VAR (R2))) average ratio of variances for method ¢ $M where ¢=H,C,S. Only the third replication scheme is discussed. Computing the poststratification weights in each half-sample reduces variance estimates for family income and physician visits (see the ratio of VAR(R1S) to VAR(R2S) in table 11. This is true for all three designs. A reduction is achieved for all the remaining variables in design III. These data provide evidence that poststratifica- tion does improve precision. SUMMARY Introduction Increasing use of scientific survey sampling has led to the development of a wide variety of techniques. These have been produced in re- 27 Table 11. The relative impact of poststratification on the variance estimates Average value of the ratio of Warez VAR(R1H) | VAR(RIC) | VAR(R1S) VAR(R2H) | VAR(R2C) VAR(R2S) Design | FOMIY TACOIME . oo o vown svn st mmm bs sb mma od ss v8 3 ad sada amudassss mn 0.833 0.833 0.833 Restricted activity days . . . . . oi vi ee ee ee ee ee ee 1.030 1.030 1.031 PRhySICIAN VISITS . . . ot i i i tt te ee ee ee ee ee ee ee eee 1.001 1.000 1.000 Hospitalidays , + 2 5 www 0 0 58 Bae £2 5 8 FUE 8 FEE FEY £ FI 0 HHE® EE EE EE 1.035 1.036 1.035 Proportion seeing a Physician . . . . . oc ci ie ee ee ee ee 1.034 1.026 1.030 Design II Family income . . . . . LL ee ee eee 0.840 0.844 0.842 - Restricted activity days . . . . . Le ee ee ee ee eee 1.002 1.005 1.003 Physician VISITS . . . . «cv i i i i tt ee ee ee 0.973 0.973 0.973 Hospital days . . . . oo ee ee eee 1.015 1.004 1.009 Proportion seeing a PRYSICIAN . . . uw vs « 5 5s 5 4 6 2 5 5 Foe & 55 8 Www tt aw bn 0.985 0.984 0.984 Design 111 Family INCOME . . . Lo ot tt et ee ee ee ee ee ee eee 0.851 0.850 0.850 Restricted activity days . . . . . . « c i i ee ee ee ee ee ee ee ee ee eee 0.980 0.982 0.981 Physician VISITS , o 4 wv wm s+ » % swims » 8 3 Www # £3 5 8 Emm Fk uw mE EE Lk 8a WE 0.961 0.961 0.961 Hospital days . . u sc wiv o 2 5 5 Bum Es 8% HEE EEE EP AEE EER ME Es EE We 0.967 0.966 0.967 Proportion seeing a physician . . . . oo ot iii ie ee ee ee 0.975 0.973 0.974 sponse to the need to sample populations that are geographically scattered, requiring large PSU’s. The use of poststratification to adjust the distribution of the sample for certain demo- graphic characteristics has produced further methodologic refinements. Such features justify the generic term ‘“‘complex multistage probabil- ity samples” to describe the methods. Because of the complexity of these samples, variance formulas for estimators are themselves compli- cated, and often only approximate expressions can be obtained. Also, since the criteria of simple random sampling, independence, and normality are not met, classical statistical proce- dures must be examined to determine their applicability in such an environment. This in- vestigation has employed an empirical approach to consider various problems of estimating vari- ances and making inferences for complex sam- ples. By the procedure of Monte Carlo sampling from a completely specified universe, the follow- ing points were investigated: 28 1. Behavior of the two general variance esti- mator methods: linearization and replica- tion 2. Distribution of the ratio of an estimated mean minus its expected value divided by its standard error with the normal distribu- tion 3. Study of the bias of two estimators 4. Investigation of a simpler variance estima- tor as an approximation to the correct replication and linearization procedures 5. Measurement of the impact of poststratifi- cation on variance estimates produced by the two general methods The study considered a particular sample design and estimation process. The design was complex, involving stratification and two stages of cluster- ing with poststratification. The results are for 900 samples drawn for three different sample sizes and, thus, are based on a considerable body of evidence. Behavior of Linearization and Replication The variance of a poststratified ratio esti- mator was estimated by the replication and linearization methods. Three different variance estimators were computed from the replication method. For each of 900 samples drawn for three sample sizes, two ratio estimates and their variances were computed. The variance estimates produced by these techniques VAR(RI1H), VAR(R1S), and VAR(L1) were compared on the bases of bias, variance, and mean square error. For the sample design and estimation proce- dure employed here, both the replication and linearization methods provide generally ade- quate variance estimates. Almost no severe abnormalities were observed. The biases of the methods are satisfactory, and the measures of total mean square error for the methods are adequately small with one exception. The vari- ance of the VAR(L1) method is generally less than the variance of VAR(R1S), especially for the variable hospital days. Normality of Standardized Estimators The distribution of the ratio sample estimate —E(sample estimate) \/VAR (sample estimate) was studied. When the sample design and estima- tion procedure result in a complex survey, the exact distribution of this ratio is unknown. For this study, such ratios were computed for each estimator, VAR(R1H), VAR(R1S), and VAR(L1). To examine confidence interval state- ments, the proportion of times the ratio fell within stated limits was computed. The sym- metric limits chosen were the normal deviates for the 99th, 95th, 90th, and 68.27th quantiles of a normal distribution. Also calculated was the proportion of times the ratio fell in the one- sided intervals (-1.000, 0) (0, +1.000), (- 1.645, 0) (0, 1.645), (-1.960, 0), and (0, 1.960). Proportions for the symmetric intervals with VAR(R1S), VAR(R1H), and VAR(L1) as the variance estimator were exceedingly close to corresponding normal values. For the most part, the proportions for the three methods were smaller than the proportions for the normal distribution. Examination of the one-sided inter- vals showed that the proportions on either side of zero are not the same. This fact and the indication of leptokurtosis provided evidence that the distributions of the ratios are not symmetric. This body of empirical data indicates that approximate interval estimates based on normal distribution theory can be made. Other impor- tant points are that (1) the true confidence level is somewhat lower than the nominal level of the normal, (2) one-sided intervals cannot be con- structed, and (3) because of asymmetry, these intervals are not minimum length. Comparison of Two Ratio Estimators Sampling theory states that ratio estimators are biased.2># In the present study, two esti- mators, distinguished by the level of poststratifi- cation, were investigated: R; was adjusted to the universe totals of a class, and R, was a combined ratio estimator with the poststratifica- tion adjustment at the region level. These estimators were calculated for each of the five variables for every selected sample. The bias and relative bias were reported. The differential bias between 7, (average of replicate and complement estimators of R) and R; was also determined using the first replication equa- tion. None of the relative biases was greater than 1 percent. Thus, these ratio estimators are essentially unbiased for the sample sizes studied. Approximate Variance Estimators Two computationally simpler methods were included in the study to determine whether one of them could be used to produce approximate variance estimates of R,. The first method was a replication method in which the parent sample adjustment factors were used in place of sepa- rate factors for each half-sample. The second method produced an overall variance estimate for R; by first estimating the variances of Rgnc,; and Rgy ; and then combining the two correctly. Included in this discussion were 29 the results of estimating variances correctly for R,. Only the first approximation VAR(R2S) can be seriously considered for possible use. This method was more biased than the correct procedure, but the mean square errors were within tolerable bounds. The proportions pro- duced when this variance estimator was used do not have large deviations from the normal values. However, the direction of the deviations varied more than for the VAR(R 1S) method. The behavior of VAR(R4S), which is a linear combination of subuniverse estimators, was ade- quate. The difference between this method and the approximate one, VAR(R3S), is that VAR (R4S) is the appropriate variance estimator for the ratio estimator Ro. The ratio estimator R, is poststratified to the regional distribution. There- fore, the variances are weighted averages of the variance estimators of the region estimators. The bias, variance, and mean square error were found to be acceptable and not to be extremely large. The findings on the ratio of the sample esti- mator minus its expected value divided by its standard error were that for VAR(R4S) the empirical proportions for the symmetric inter- vals compared favorably to the proportions for the normal distributions. Influence of Poststratification Theory states that the benefits derived from adjusting sample results to population totals for selected characteristics lie in improvement of the precision of the ratio estimator by insuring that domains of study have the same distribution in the sample as in the population. Assessment of the effect of poststratification on variances was a natural byproduct of the calculations done for the other methods. The estimates of variances produced by the first replication method were compared with those secured from the second equation. In the first scheme, poststratification factors were computed within each half-sample and each complement using half the sample data, while the parent sample factors were applied when making these estimates in the second scheme. VAR(R1S) showed a reduction in variance for family income in design I; for family income, 30 physician visits, and proportion seeing a physi- cian in design II; and for all the variables in design III. Conclusions The outstanding finding was that both the replication method and the linearization method are highly satisfactory methods for estimating variances of poststratified ratio estimators ob- tained from complex probability surveys of the type studied here. Features of the sample design employed were unequal sampling fractions, stratification, and two stages of clustering. Since the methods yield adequate variance estimates, they provide the means for drawing valid infer- ences with known precision from such surveys. For these data, the linearization variance estima- tion method had slightly lower variance and mean square error than the replication method. Neither method showed substantial bias. The proportions based on the ratios of the sample estimate minus its expected value divided by its standard error as estimated by the replication method were closer to the corre- sponding normal values than the proportions with the linearization method as the variance estimator. However, the ratios with the L1 variance estimator had a more consistent pattern than the ratios with the standard error estimated by replication method. The consistency was evidenced by the difference between empirical proportions and the corresponding ones for the normal distribution getting smaller as sample size increased. The applicability of confidence intervals by using normal distribution theory has been shown to be adequate. The true level of confidence is usually lower than the confidence stated, which should be brought out when making inferences. Another item that should be mentioned is that the distributions of the ratios are not symmetric. Thus, one-sided intervals are not valid and the intervals constructed are not of minimum length. The biases of the ratio poststratified estima- tors were negligible for the sample sizes studied. The VAR(R2S) estimator proved to be an excellent approximation, and its use is recom- mended. However, there is a tradeoff involved when using an approximation. The gain in simpler formulas may be offset by larger biases. Also, the proportions produced when the VAR(R2S) was used to estimate standard errors varied widely in the direction of its deviation from the proportions of the normal distribution. The investigator should consider that consumers of the statistics may make such inferences whether or not he does. Then each must weigh the losses and the consequences when deciding about a variance estimation method. The effect of the estimation component poststratification was to lower variance and mean square error. This reduction is important in making inferences from complex sample surveys and can be converted into cost savings. Relevancy to the HIS The striking feature of this investigation has been to provide empirical evidence that data collected in complex multistage probability sur- veys such as the HIS can be used for analytical purposes. Variance estimates, which are essential in analyses, can be calculated by using either the replication or the linearization variance esti- mator method. Having adequate variances opens avenues of further research in developing testing techniques and other refined analyses, perhaps along the lines of multivariate procedures, com- parable to classical ones. Equally rewarding as finding that suitable variance estimates can be computed was the result that tests and confidence intervals can be approximated by normal distribution theory. This allows one to draw inferences about the morbidity estimates published by the HIS. The HIS has now a choice of either of the two methods to use in computations. The variability of the linearization method was slightly lower than that of the replication method, but neither is very biased. The linearization equations can easily be interpreted to obtain estimates of components of variance, which are crucial in planning such surveys. Another factor to con- sider when choosing a variance estimator method is that the replication technique is easily applicable to the estimation of variances for many types of estimators without deriving new theory or new computer programs for each estimator. This is not true for the linearization method. 000 31 REFERENCES IKish, Leslie: Confidence intervals for clustered samples. Am Soc. Rev. 22(2):154-165, Apr. 1957. 2Kish, Leslie: Survey Sampling. New York. John Wiley & Sons, Inc., 1965. 3Deming, William E.: Some Theory of Sampling. New York. John Wiley & Sons, Inc., 1950. 4Hansen, M. H., Hurwitz, W. N., and Madow, W. G.: Sample Survey Methods and Theory. Vol. 1. New York. John Wiley & Sons, Inc., 1953. 5¢gtudent.” (W. S. Gossett): The probable error of a mean. Biometrika 6(Pt. 1):1-25, Mar. 1908. 6National Center for Health Statistics: Replication: An approach to the analysis of data from complex surveys. Vital and Health Statistics. Series 2, No. 14. DHEW Pub. No. (HSM) 738-1269. Health Services and Mental Health Administration. Washington. U.S. Government Printing Office, Apr. 1966. 7U.S. Bureau of the Census: The Current Population Survey: A report on methodology. Tech. Paper No. 7. Washington. U.S. Government Printing Office, 1963. 8 Banks, Martha J., and Shapiro, Gary M.: Variances of the Current Population Survey, including within- and between-PSU components and the effect of the different stages of estimation. Proc. Soc. Statist. Sec. Am. Statist. A., 1971, 40-49. 9Gurney, Margaret: The variance of the replication method for estimating variances for the CPS sample design. Washington. U.S. Bureau of the Census, 1962. (Dittoed memorandum.) 10National Center for Health Statistics: Cycle I of the Health Examination Survey: Sample and Response, United States, 1960-1962. Vital and Health Statistics. PHS Pub. No. 1000-Series 11-No. 1. Public Health Service. Washington. U.S. Government Printing Office, 1964. McCarthy, Philip J., Simmons, Walt R., and Losee, Garrie J.: Replication techniques for estimating the variances from complex surveys. Paper presented at the Joint Session of the Joint Epidemiology and Statistics Section of the American Public Health Association, Chicago, Ill., Oct. 18, 1965. 126immons, Walt R., and Baird, James T., Jr.: Pseudo- replication in the NCHS Health Examination Survey. Proc. Soc. Statist. Sec. Am. Statist. A., 1968, 19-30. 13packett, R. L., and Burman, J. P.: The design of optimum multifactorial experiments. Biometrika 33(Pt. IV):305-325, June 1946. 141 ce, Kok-huat: Partially balanced designs for half-sample replication method of variance estimation. J. Am. Statist. A. 67(338):324-334, June 1972. 15National Center for Health Statistics: Pseudo-replication: Further evaluation and application of the balanced half-sample technique. Vital and Health Statistics. Series 2, No. 31. DHEW Pub. No. (HSM) 73-1270. Health Services and Mental Health Administration. Washington. U.S. Government Printing Office, Jan. 1969. 16McCarthy, Philip J.: Pseudo-replication: Half samples. Rev. International Statist. Inst., 37:239-264, 1969. 17Kjsh, Leslie, and Frankel, Martin: Balanced repeated replications for analytical statistics. Proc. Soc. Statist. Sec. Am. Statist. A., 1968, 2-10. 18g ish, Leslie, and Frankel, Martin R.: Balanced repeated replication for standard errors. J. Am. Statist. A., 65:1071-1094, 1970. 19 evy, Paul S.: A comparison of balanced half-sample replication and jackknife estimators of the variances of ratio estimates in complex sampling with two primary sampling units per stratum. Rockville, Md. National Center for Health Statistics, 1971. (Dittoed report.) 20gimmons, Walt R., and Bean, Judy A.: Impact of design and estimation components on inference. New Developments in Survey Sampling. (N. L. Johnson and H. Smith, Jr., eds.) New York. Wiley-Interscience, 1969, pp. 601-628. 21K eyfitz, Nathan: Estimates of sampling variance where two units are selected from each stratum. J. Am. Statist. A., 52(280):503-510, Dec. 1957. 22Hansen, M. H., Hurwitz, W. N., and Madow, W. G.: Sample Survey Methods and Theory. Vol. 2. New York. John Wiley & Sons, Inc., 1953. 23Tepping, Benjamin J.: Variance estimation in complex surveys. Proc. Soc. Statist. Sec. Am. Statist. A., 1968, 11-18. 24National Center for Health Statistics: Estimation and sampling variance in the Health Interview Survey. Vital and Health Statistics. Series 2, No. 38. DHEW Pub. No. (HRA) 74-1288. Health Resources Administration. Washington. U.S. Government Printing Office, June 1970. 258 hapiro, Gary M.: Keyfitz method of estimating variance g and its application to the Current Population Survey. Washing- ton. U.S. Bureau of the Census, 1966. (Dittoed report.) 26Waltman, Henry: Variance estimation formulas to be used in computing total NSR PSU variances for the 449-area design. Washington. U.S. Bureau of the Census, 1968. (Dittoed report.) 27 Fellegi, I. P., and Gray, G. B.: Sampling errors in periodic surveys. Proc. Soc. Statist. Sec. Am. Statist. A., 1971, 28-35. 28Frankel, Martin R.: Inference from Survey Samples: An Empirical Investigation. Ann Arbor. Institute for Social Re- search, University of Michigan, 1971. 29y.S. National Health Survey: The statistical design of the Health Household-Interview Survey. Health Statistics. PHS Pub. No. 584-A2. Public Health Service. Washington. U.S. Govern- ment Printing Office, July 1958. 30Bean, Judy A.: Behavior of replication and linearization variance estimators for complex multistage probability samples. Unpublished doctoral dissertation. University of Texas, 1973. 33 APPENDIX | EXTENT OF BIASNESS OF TWO RATIO ESTIMATORS Introduction Another objective of the study was to determine the extent of biasness of two ratio estimators. The formulas used and the results follow. Estimators ’ The two estimators of a scatistic follow: 1. R, —poststratification at universe level. The formula is given in the section Estimators and Variance Estimators. 2. Ry —poststratification at region level x" +x" > ENC,2 SW, 2 R — 2 y 24 24 3 i YENC,a 4 3 , Yswa C4 XENC,a 4 XSW,a _ = YENC,a * Ysw,a y where ENC = East and North Central regions, SW = South and West regions, Ry = poststratified ratio estimate of the pop- ulation parameter R, YENC,a ’ ) YENC,a 24 n —- ' XENGC,2™ 2 XENC,a XENC,a = the sample estimate for class a in the ENC Region, YENC,a = the population size in class a in the ENC Region, 24 " _ > ’ YSW,a Xsw,2 = Ss XSW,a a= YSW,a Xgw,, = the sample estimate for class a in the SW Region, Ysw,e = the population size in class a in the SW Region, Ysw,a = the sample estimate of population size for class a in the SW Region. Comparison of Two Ratio Estimators Ratio estimators are commonly employed in complex multistage sample surveys and usually contain the features of unequal sam- pling fractions and poststratification. The distributions of these ratio estimators are un- known. In this investigation, two such ratio estimators were employed in order to study the extent of their biasness. Each estimation procedure included inflation by the reciprocal of the selection probability and poststratification by 24 ethnic-sex-age classes. This probability of selection is the product of the probabilities of selection from each step of selection: PSU and segment. The level at which poststratification was performed differentiates the estimators. The R; estimator was adjusted to a universe level; Ro was adjusted to the distribution of subgroups within the ENC Region and within the SW Region. Ratio estimators are biased estimators. The extent of this bias has been studied both theoretically and empirically. Since these ratio estimates were calculated for each of five varia- bles for 900 samples for three sample sizes, sample empirical evidence was available to in- vestigate the magnitude of the bias of the poststratified ratio estimators employed here. The universe was completely specified, so the parameter R was known. A measure of the bias 1s bias (R;) = R; - R 1 where R= Pn value, 900 2, Ry, Table 1 shows that neither of the two esti- mators consistently had the smaller bias. In RB; = wa #=1,2 Table I. The bias of two ratio estimators design I, with the exception of the variable proportion seeing a physician, R, has the smaller value. For the other two designs, R, does not compare as well, and in fact, only exhibits the smaller bias for the variable propor- tion seeing a physician in design II and for hospital days in III. The relative bias of the estimators used for another evaluation of the estimators is defined to be relative bias (R;) = where R = the population parameter, R; = the sample mean of the estimates pro- 1 duced by ratio estimator z, 900 Sim, i=1,2 =k “ Table II shows the values for the two esti- mates for each variable and design. The relative bias is always less than 1 percent for any of the Table II. The relative bias of two ratio estimators Estimator Variable [S—eEr——— R, | R, Design | Family income .................. -0.0047 | -0.0124 Restricted activity days ............ 0.0049 0.0021 Physician visits. ................. -0.0135 -0.0197 HOSPIE) 0BYS sims misaisamssmines 0.0097 0.0082 Proportion seeing a physician ....... 0.0059 -0.0035 Design II Family income ................... 0.0055 0.0047 Restricted activity days ............ -0.0009 -0.0004 Physician Visits ................... -0.0119 -0.0112 HOSOI BYE oo vv v5 oi 5 60 5 iti # ki 3 0001 3 0.0016 0.0009 Proportion seeing a physician . ....... 0.0035 0.0041 Design 111 Family income ...............0... -0.0082 -0.0064 Restricted activity days ............ -0.0012 -0.0010 Physician visits ................... -0.0210 -0.0198 Hospital days ..isiavsmmswasaisimss 0.0040 0.0045 Proportion seeing a physician ......., 0.0021 0.0007 Estimator Variable = = R, R, Design | Formily INCOME «uu svismies ms sims ion sy -0.0006 -0.0015 Restricted activity days ............ 0.0031 0.0014 PHYSICIAN VISES + ous mis wiv wis os mma 4 8 -0.0029 -0.0040 Hospital days «. een iwesmsoimsnmss 0.0093 0.0078 Proportion seeing a physician ........ 0.0009 -0.0004 Design 11 Family INCOME . wiv scission om» wins 5» 0.0007 0.0006 Restricted activity days ............ -0.0006 -0.0003 Physician VISES ...:sissismisnsmsss -0.0025 -0.0024 HOSPHAl dBYS: « « wiv» wis 5 wv 4% # wim » 51m 5 0.0015 0.0009 Proportion seeing a physician ........ 0.0006 0.0006 Design 111 Family INCOME . vu» sins wma sams nin ss -0.0010 -0.0008 Restricted activity days ............ -0.0008 -0.0007 Physician visits ................0... -0.0045 -0.0042 Hospital daVs «swiss viv s vies ms ows wo sw 0.0039 0.0043 Proportion seeing a physician ........ -0.0003 0.0001 35 Table I11. Distribution of differential bias of r, and R, as a fraction of the estimated standard error of r, for replication method 1 Variable r,-R, . : VY VAR(r,) Family Bois Physician Hospital Proportion income visits days days physician Design | Lessthan =30 . . . . . .. ee 4 4 5 7 18 “VIE =TB ss rm EE EI HEE REI MEER EEE NEE 4 6 5 3 12 “BW =20 rir TE EE ARAN EE REE EER men 5 8 2 9 22 S200 TE wis i hs hE ss mmm rr ne mm kw wk 18 29 13 10 40 1510-10 LL 45 49 64 42 97 “PE 0B wv vs 3 sv ums ss FREEBIE OHEE BEES 142 139 129 115 178 “OBI =O0 ww os rvs SRE ESSE EEE LIER RBA hE me a 345 265 304 291 254 ODDIE ow vv 75 5 BE EA EES mE me en mn a x mr. 249 257 267 264 173 OB YOGTO) 5. 0m i 45% 5 moi wn aw mm nen mm ne x wn 66 98 86 113 74 ace Ey 17 30 20 27 22 510.20 LL LL 2 12 3 10 6 2010.25 LLL 3 2 1 6 1 Tio es TEE EEE ITT TTY 0 1 1 3 2 Greater than .30 . . . . 5 v8 55 55 3m E558 mame ova umm. 0 0 0 0 1 Design |1 Lessthan —.30 . . . . . 0 0 0 1 0 =30t0-.25 LLL 0 0 0 0 0 2B ~20 wn ps is REE EE A MRM ER CE AMEE SLE HEE 0 0 0 1 0 ~2010=18 uu sss 0 pwn 0s EB HEE ER EER EE SEES Eee 0 2 3 3 2 (Rin 24 ; Ya 2 XSW,a 7 a= * = Z where the sum is over Ysw PSU’s in the SW Region, xgw,1 = the poststratified ratio estimate of health characteristic x for the SW Region. Then the variance estimator is yéncVAR(Renc,1) +ydw VAR(Rsw 1) y2 The two orthogonal patterns used are in tables IV and V. Another feature of this approxima- tion is the level at which poststratification was done for the half-sample estimators. Since the estimator R, is poststratified at a universe level, these factors should be computed within each half-sample at a universe level, in order for the half-sample estimates to be equivalent. However, this cannot be done when two different-sized patterns are used. The solution was to apply the parent factors. Formulas for these variances are VAR(R,) = r; 2 - Rgne,i ) 12 1 A 2 ) = 752 (REnc,o ~ Renc,1) VAR(Rgnc,1:S) = SIVAR (Rec, 1 :H) + VAR(Rgx 1:0) 12 : I VAR(Rsw 1:H) = 15 2 12 VAR(Rgy ;:C) = HZ VAR(Rgy 1:5) = VAR(R3H) VAR(R3C) 1 = 5 [VAR(Rsw H = yEnc [VAR(Rgnc, 1: H = anc VAR(Rpnc ' r 2 (Rsw o ~ Rsw,1) I 2 (RSw a - Rw 1) ) + VAR(Rgy ;:C)] + vdy [VAR (Rg | tH] CY) +20 [VARR,,, ,:C) 39 Table IV. Orthogonal pattern for ENC Region variance estimates’ Strata in the ENC Region Replicate 112) 3|4|5| 91011 ]12]|13 1...... + + - + + - - - + 2 waswia + - + + + = = & + - I - + + + - = - + - + 4 ...... + + + = = = + - + + 5 nesmas + + - = = + - + + - Q woonin + = = = + = + + - + 7 ...... Li i + - + + - + + 8 ...... - - + - + + - + + + OQ sosman - + - + + - + + + - 10 vuvnns + - + + - + + + - = mM... - + + - + + + - - - 12 viamss - = = = = = = - - te 144 denotes +1 and -*' denotes —1. § Table V. Orthogonal pattern for SW Region variance estimates’ Strata in the SW Region Replicate 6|7|8|14|15|16 | 17 | 18 | 19 Yin - + = - - + + - 2 airmenmy + - + - - - + + + 3........ + + - + - - - + + 4 ........ - + + - + < - - + 8 uisnsnns + - + + - + a = - 6 ........ + + = + + - + - - Taivneins + + + - + + - + - Bursmenss - + + + - + + = + 9... - + + + - + + - 10 ..00i000 - = = + + + - + + V1 comms mss + - = - + + + - + 12 ......... - = = - - - - ~ - ' 4’ denotes +1 and ’-'’ denotes —1 and VAR(R3S) = F[VAR(R 3H) + VAR(R3C)] where Rinc.o = the ath replicate ratio estimate of health characteristic x for the ENC Region with parent sample post- stratification factors used, Rgy o = the ath replicate ratio estimate of health characteristic x for the SW Region with parent sample post- stratification factors applied. Variance Estimators for Estimator 2 The variances for the estimates R, were estimated by the correct replication method only. For R,, the variance estimators are VAR(R4H), VAR(R4C), and VAR(R4S). This scheme consists of estimating the variances for the regional estimates Rg, o and Roy. 9 and then combining them for a variance estimate of Ry. This process is appropriate since, for the parent sample estimate, poststratification was at the regional level. For the half-sample and the complement half-sample estimates, the regional poststratification factors were calculated each time. The orthogonal patterns in tables IV and V were used in producing the variances for Rpne, 9 and Rey 9. The formulas are - Yenclenc,2 tT Yswhsw,2 R, = y where Ryne o = the ratio estimate for the ENC Re- "gion with poststratification at a re- gional level, Bors = the SW Region ratio estimate with "" poststratification at the regional level, VAR(R,) = YENc [VARRgyc,2 | + Yew [VAR(Rgy 2) y? 1 12 VAR(Rg yc, 9H) = 12 Zinc o«" Rene,2) 2 1 o VAR(Rgyc,2:€) = 124 Ring, - Rgye,o) 1 " VAR (Ryn 21S) = 5 [VAR(Rgyc oH) + VAR(Rp nc 5:0) 1 12 » v st ! » 2 VAR(Rgy o:H) 12% (Rw Rgy o) 1 12 > . EE nn » 2 VAR(Rgy 5:C) - 12 (RSw sw,2) 1 VAR(Rgy 5:5) = 5[VAR(Rgy o:H) + VAR(Rgy »:C)] = I I 1 VAR(R4H) {inc [VAR Rypg 5H) +33 [VAR (Ry 5: Hf =, : - 1 VAR(R4C) = dine [VAR(Rg yc 2:C)] + ¥3w [VAR(Rgy, ,: an, y VAR(R4S) = 5 [VAR (R4H) + VAR(R4C)] where these measures are the same as given in the Results section. Inspection of table VI shows the biases for VAR(R3S) for every variable decreas- Reno = the ath half-sample ratio estimate ing with increasing sample size. The VAR(R2S) for the ENC Region adjusted to re- gion level, Ry , = the ath half-sample ratio estimate "for the SW Region adjusted to re- gion level. Approximate Variance Estimators Investigators would like to obtain estimates of variance by simple methods. The results of two approximate methods thought to be solutions to this problem are discussed in this report£ One of the methods, VAR(R2H), has been used both by the NCHS and the University of Michigan SRC. The discussion is based on the results of the third replication form. For this evaluation, bias, variance, and mean square error were computed. The definitions of &Two more approximations of the correct variance estimates were also investigated. These results are given in Bean. method displays a different pattern. For family income and restricted activity days, the biases of VAR(R2S) reduce as the sample size increases; for the other three variables, the values increase, going from design I to design II, but then drop to their lowest in design III. VAR(R2S) does sometimes underestimate the variance of R,. This occurs in design I for the variable propor- tion seeing a physician and in designs II and III for the variable restricted activity days. Regard- less that the biases for three variables increase, going from design I to design II, the R2S method behaves better than the R3S approxima- tion. It always has the lowest bias. Both the quantities variance and mean square error in tables VII and VIII indicate that the R2S method outperforms VAR(R3S) without any question. This procedure always has the smaller value of variance or mean square error just as in the base of bias. A possible explanation for why VAR(R3S) is not a good approximate variance estimator is offered. VAR(R3S) is a variance estimator for 41 Table VI. The bias of the approximate variance estimators of the sample estimator R, Variable Variance estimator Family Restricted Physician Hospital PAIRORIOR income activity days visits days seeing 2 physician Design | VARIR2ZS)Y . : sons is ssmuw ins vam 2.469 X 10* 2.164 X 107! 3.677 Xx 1073 1.621 X 1073 | -1.078 X 107° VAR(R3S) .... 1.222 X 10° 5.885 X 107! 3.498 X 1072 | 3.647 X 107? 6.536 X 107 Design |1 VARIBIS) ;: c sovms iss mmm ini mms 1.219 X 10* | -2.705 X 10°? | 3.984 X 10°? | 1.656 X 1073 | 2.027 X 107° VAR(R3S) . .. 5.893 xX 10* 1.334 xX 10°! 1.884 Xx 1072 | 2.489 X 1073 3.344 X 1074 Design I11 VARIR2SY ; » s vusn sss smme 638 sun 4.173 xX 10° | -2912x 1073 1.097 X 107% | 3.593 X 10°* 6.765 X 10°¢ VARIR3SY . ..:vs5sisasmacss ams 2.461 x 10* 6.099 Xx 107% | 8.263 X 1072 | 7.021 X 10°* 1.437 Xx 107* Table VII. The variance of the approximate variance estimators of the sample estimator R, Variable Variance estimator Family Restricted Physician Hospital Prononen income activity days visits days hysicion Design | VAR(R2S) . .... 1.5694 x 10° 0.190 X 10 2989 Xx 107* | 5.298 x 10°* 1.541 xX 107% VAR(R3S) oii 2.678 X 10'° | 0.244 X 10 5.020 X 10°* | 5.904 X 10°* 9.937 X 1077 Design I1 VARIAZST uw vrssomeme sis mmm sissy 4.800 xX 10° | 5.320 Xx 10°! 9.071 X 107* | 8.409 X 107° 3.729 X 107° VARIR3SY .v::ssmumssismEadanss 4.633 X 10° 5.901 xX 10°! 1.418X 107° | 8.606 X 10°° 1.810 x 1077 Design 111 VARIB2S) ..: iis mwmss sommes 1.364 X 10% 1.495 X 10°! 8.956 X 10° 1.373 xX 107° 1.109 X 107° VARIB3S) ...::iswwmesssnmmeres 1.106 X 10° 1.594 X 10°! 1.807 Xx 10°* 1.447 X 107° 4.402 X 1078 R,, the parent sample estimator, obtained from a weighted average of variance estimators of Rgnc,1 and Rgy ;. The estimating equation poststratifies the estimators to universe levels and, thus, these subuniverse estimators are ad- justed to an overall distribution of ethnic-sex- age, rather than to subuniverse distribution. Questions that should be asked are What is the magnitude of the bias in Rgn¢ ; and Rgy ;? and If the bias is not negligible, what is the 42 effect on the variance estimator? These ques- tions cannot be directly answered here because the biases of the subuniverse estimators were not calculated. However, VAR(R 4S) is the appropri- ate weighted variance estimator for a ratio estimator poststratified to regional levels, and its performance can be studied. The analysis was based on comparison of bias, variance, and mean square error. The formulas for these measures follow: bias [VAR(R4S)] = VAR(R4S) - 5% Rg VAR[VAR(R4S)] = 1 900 2 505 Le {[VAR(R4S)]; - VAR(R4S)} i=1 MSE[VAR(R4S)] = VAR[VAR(R4S)] + bias? [VAR(R4S)] where 900 VAR(R®) = gs 2 [VAR(R4S)] The biases, variances, and mean square errors for each variable and design are given in table IX. For design I, the R4S method has a range of 4.813 X 10-5 for proportion seeing a physician to 9.182 X 10-3 for family income. In designs II and III, the smallest value is again for proportion seeing a physician and the largest bias for family income. Unlike the appropriate variance esti- mator VAR(R1S) for the ratio estimator R, the biases here decrease with increasing sample size, which is a desirable property. The range for the variance of the variance estimator R4S is 1.191 X 10? for family income to 2.077 X 10-8 for proportion seeing a physician in design I; in design II, the range is 3.297 X 108 to 3.748 X 109; and in design III, it is 1.008 X 108 for the variable family income to 1.072 X 10°? for proportion seeing a physician. The mean square errors exhibit a similar pattern. Another consideration is the distribution of the sample estimate minus its expected value divided by its estimated standard error. The ratio computed was R, - E(R,) VVAR(R4S) where - | 900. E(R,) = R, = 500 2 "ai j= The proportion of times these ratios fell within stated limits of the normal distribution are displayed in table X. Only 12 times out of a possible 60 did the empirical proportions for VAR(R4S) surpass the normal values. The mag- nitudes of VAR(R4S)’s deviations are compara- ble to the amounts for VAR(R1S) with no discernible tendency for the differences to be- come smaller across the designs. The total error of the variance method is satisfactory, but there Table VIII. The mean square error of the approximate variance estimators of the sample estimator R | Variable Variance estimator Family Restricted Physician Hospital haps ian income activity days visits days ahydicion Design | VARIR2S) . is smumisssmwmmensa 2.204 X 10° 0.195 X 10 3.003% 107? 5.324 X 107* 1.652 X 10°® VARIB3S) ..:si sommes sssmvmyussss 3.763 x 10'° | 0.278 X 10 6.245% 10°* | 6.037 xX 10°* 1.421 xX 107° Design | VAR(R2S) . «iii ieee 6.289 x 10° 5.328 X 10°! 9.229X 107* | 8.684 x 107° 4.140Xx 10°? VARIB3S) sure vor anmmme noes sma 8.110 X 10° 6.079 X 10°! 1.774 Xx 10°* | 9.226 X 10 °* 2.930% 10°77 Design Il] VARIR2S) + wus sss mmmesss moms» 1.539 x 10° 1.495 X 10°! 9.320 xX 107° 1.386 X 10° 1.155 x 107° VARIABS) vowv ss 83 hmm ees» Baws s 1.713% 10° 1.631 X 10°! 2.490 X 107* 1.496 X 10° 6.470 X 10°" 43 Table IX. The bias, variance, and mean square error of R4S variance estimator of R, Variable Bias Variance Magen square error Design | Pomily Income s « s o sm@ o 5 2 BE 1 E52 FRB EL LE HMMA EB 8 Fn g 9.182 xX 10° 1.191 X 10° 1.257 X 10° Restricted GOLIVILV QBYS .... o 5 5 2.05 44 + 3 3 SMM 6 & + a HOKE E48 Poa En ¥ 3.944 x 10°! 0.209 X 10 0.224 X 10 PRYSICIAN VISITS . . oo tv to te ee 4517%X 107% | 3.173X 107? 3.193 x 10°? Hospital days . . «ov vv oe ee ee ee ee 3.398 x 10°? 1.961 x 107? 1.973% 107? Proportion seeing a physician . . . . . . 4813 x 107° | 2.077x 10° 2.308 X 10° Design II Family INCOME . . oo ott te ee eee ee 2.425% 10° 3.297 x 10° 3.356 X 10° Restricted activity days . . . . . «vo ieee ee eee 4.057 Xx 10°? | 5.433x 10°! 5.450 X 10°! Physician visits , o so 5 2 5 # 90% 4 8 8 8 #5 # #0 9% www » 6 8» ma ww x vo» 3.626 x 10°? 9.455 X 1074 9.587 X 107 HOSPITBI DOYS « : « win w & 3 + 3 986 £5 5 9 HEH § 8 8 SHWE 2 2 4 www « « 2.759 X 107? 2.095 xX 107 2.171 X 10° Proportion seeing a physician . . . . . . uit 2476 Xx 107% | 3.748 X 107° 4.362% 107° Design I11 Family INCOME ©. «oo oe ee ee eee -8.598 X 10? 1.008 xX 10° 1.015 x 10° Restricted activity days . . . . . «oie ee eee -1.357 X 10? 1.405 X 10°! 1.407 X 10°! PRySICIan VISTIY » «iow o 2 8 3 Bem 5 3 55 SEE + + 85 Hoa EF SL Pam ¥ oe 1.214X 107° | 8.227 X 10°* 8.374 X 107% HOSPItAL JBYS « o +o & i 5 5 Mm® ¥ 35 » BAe #8 28 $0 v § 88 Bui0ls #88 2 3.161 xX 10°* 1.594 X 10°* 1.604 X 107% Proportion seeing a physician . . . . . . . . i 5.098 X 10° | 1.072X 10° 1.098 X 107° are indications that the size of the bias is a problem. Although it would be difficult to state any general rules about drawing inferences when the stated confidence level could be in error in either direction, constructing such intervals is conceivable. The method VAR(R2S) can offer a savings in computer time, since the process of calculating new poststratification weights within each half- sample is eliminated. Because the comparable measures used show this method to be a good approximation, it is also compared with VAR (R1S). The bias of VAR(R2S) is less in the three designs for restricted activity days and in designs I and II for the variable hospital days. In the case of variance, VAR(R2S) has the lower value for variables restricted activity days, hospital days, and proportion seeing a physician in design I and also for hospital days in design II. The mean square errors of VAR(R2S) are lower for the same designs and variables as the variances. For a better understanding of the comparison between the two methods, the relative biases within a design are averaged over the variables. The relative bias is defined as 44 VAR(R1S) - s> R relative bias [VAR(R1S)] = 2 ! Ry VAR(R2S) - 52 2 R, relative bias [VAR(R2S)] = k The average relative biases for VAR(R 1S) for the three designs are 0.0782, 0.0656, and 0.0363 as compared to 0.0862, 0.1231, and 0.3150, the averages for VAR(R2S). Thus, these averages show that as the sample size increases, the magnitude of the bias in VAR(R2S) in- creases, whereas for VAR(R 1S) it decreases. To complete this discussion, the proportion of times the ratio of the sample estimate minus its expected value divided by the R2S estimated variance is within certain limits is in table XI. The observation made is that 16 out of 60 empirical proportions are larger than the same proportions for the normal distribution. Con- Table X. Proportion of times the sample estimate minus its expected value divided by R4S estimate of standard error is within the stated limits Limits Variable +1.000 | -1.000,0 | 0,1.000 | +1645 | -1.645,0 | 0,1.645 | +1.960 | -1.960,0 | 0, 1.960 | +2.576 Design | Family income . | 0.6878 0.3478 0.3400 | 0.8967 0.4533 0.4433 | 0.9422 0.4756 0.4667 0.9844 Restricted activ- ity days 0.6733 0.3300 0.3433 | 0.8967 0.4456 0.4511 | 0.9444 0.4767 0.4678 0.9789 Physician visits . | 0.6578 0.3144 0.3433 | 0.8722 0.4289 0.4433 | 0.9356 0.4700 0.4656 0.9800 Hospital days . . | 0.6722 0.3211 0.3511 | 0.8478 0.4222 0.4256 | 0.9033 0.4644 0.4389 0.9556 Proportion seeing a physician 0.6911 0.3522 0.3389 | 0.9011 0.4456 0.4556 | 0.9478 0.4678 0.4800 0.9811 Design 11 Family income . | 0.6922 0.3489 0.3433 | 0.8833 0.4333 0.4500 | 0.9344 0.4633 0.4711 0.9778 Restricted activ- ity days 0.6556 0.3044 0.3511 | 0.8856 0.4256 0.4600 | 0.9344 0.4567 0.4778 0.9778 Physician visits . | 0.6878 0.3400 0.3478 | 0.8967 0.4578 0.4389 | 0.9422 0.4833 0.4589 0.9800 Hospital days . . | 0.6789 0.3189 0.3600 | 0.8833 0.4356 0.4478 | 0.9367 0.4711 0.4656 0.9756 Proportion seeing a physician 0.7011 0.3644 0.3367 | 0.9167 0.4678 0.4489 | 0.9544 0.4867 0.4678 0.9856 Design [11 Family income . | 0.6689 0.3211 0.3478 | 0.8667 0.4167 0.4500 | 0.9233 0.4456 0.4778 0.9800 Restricted activ- ity days 0.6578 0.3356 0.3222 | 0.8700 0.4456 0.4244 | 0.9256 0.4733 0.4522 0.9744 Physician visits . | 0.6867 0.3344 0.3522 | 0.9033 0.4467 0.4567 | 0.9444 0.4689 0.4756 0.9844 Hospital days . . | 0.6856 0.3367 0.3489 | 0.8811 0.4456 0.4356 | 0.9278 0.4744 0.4533 0.9667 Proportion seeing a physician 0.6822 0.3378 0.3444 | 0.8856 0.4378 0.4478 | 0.9322 0.4589 0.4733 0.9867 trastingly, the VAR(R1S) estimates exceeded the normal proportions only six times. This comparison is done because, ideally, when an analyst uses such an approximation, he would like to know in which direction he could be making an error. A somewhat unexpected observation is that even though the differences are not in the same direction, the average deviations are smaller for VAR(R2S). These average differences for each statistic across the five limits and three designs are +.0145,-.0197, -.0081, -.0015, and -.0097. The average dif- ferences of VAR(R1S) are -.0168, -.0200, and -.134. This body of empirical evidence supports the proposal of using VAR(R2S) as an approxi- mation. 45 Table XI. Proportion of times the sample estimate minus its expected value divided by R2S estimate of standard error is within the stated limits Limits Variable +1.000 | -1.000,0 | 0,1.000| +1.645{ -1.645,0| 0,1.645| *+1.960| -1960,0| 0,61.960 | +2.576 Design | Family income .| 0.7433 0.3811 0.3622 | 0.9122 0.4644 0.4478 | 0.9544 0.4822 0.4722 0.9933 Restricted activ- ity days 0.6689 0.3289 0.3400 | 0.8844 0.4367 0.4478 | 0.9333 0.4656 0.4678 0.9767 Physician visits .| 0.6567 0.3100 0.3467 | 0.8744 0.4333 0.4411 0.9244 0.4644 0.4600 0.9756 Hospital days . .| 0.6689 0.3022 0.3667 | 0.8644 0.4200 0.4444 | 0.9200 0.4556 0.4644 0.9589 Proportion seeing a physician 0.6489 0.3400 0.3089 | 0.8544 0.4311 0.4233 | 0.9256 0.4667 0.4589 0.9833 Design I1 Family income .| 0.7256 0.3711 0.3544 | 0.9133 0.4611 0.4522 | 0.9500 0.4800 0.4700 0.9844 Restricted activ- ity days 0.6433 0.3000 0.3433 | 0.8878 0.4411 0.4467 | 0.9333 0.4611 0.4722 0.9733 Physician visits .| 0.6811 0.3311 0.3500| 0.9011 0.4522 0.4489 | 0.9378 0.4744 0.4633 0.9853 Hospital days . .| 0.6722 0.3267 0.3456 | 0.8800 0.4389 0.4411 0.9344 0.4767 | 0.4578 0.9756 Proportion seeing a physician 0.6900 0.3533 0.33671 09144 0.4622 0.4522 | 0.9522 0.4800 0.4722 0.9833 Design 111 Family income .| 0.7033 0.3378 0.3656 | 0.9011 0.4344 0.4667 | 0.9489 0.4611 0.4878 0.9900 Restricted activ- ity days 0.6533 0.3344 0.3189 | 0.8744 0.4478 0.4267 | 0.9289 0.4733 0.4556 0.9767 Physician visits .| 0.7011 0.3411 0.3600 | 0.9056 0.4433 0.4622 | 0.9489 0.4700 0.4789 0.9833 Hospital days . .| 0.6933 0.3500 0.3433 0.8889 0.4533 0.4356 | 0.9322 0.4789 0.4533 0.9689 Proportion seeing a physician 0.7011 0.3556 0.3456 | 0.8822 0.4400 0.4422 | 0.9300 0.4589 0.4711 0.9844 OOO 46 #* U. S. GOVERNMENT PRINTING OFFICE : 1975 584-528/37 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22, VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1000 Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods research.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies. —Reports presenting analytical or interpretive studies based on vital and health statistics, carrying the analysis further than the expository types of reports in the other series, Documents and committee veports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survev.—Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons, Data from the Institutional Population Surveys. .— Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality,—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses, Data on natality, marriage, and divovce.—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HRA Rockville, Md. 20852 (Re 7% NCHS RAP RN SORES / < CRIN Comparability of Mortality Statistics for the Seventh and Eighth Revisions of the International Classification of Diseases, UN ETE EG U. S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration Library of Congress Cataloging in Put lication Data Klebba, A. Joan. Comparability of mortality statistics for the seventh and eighth revisions of the International classification of diseases, United States. (Vital and health statistics: Series 2, Data from the national vital statistics system; no. 66) (DHEW publication; no. (HRA) 76-1340) Includes bibliographical references. Supt. of Docs. no.: HE 20.6209:2/66 1. Nosology. 2. Death—Causes. 3. United States— Statistics, Medical. I. Dolman, Alice B., joint author. IL United States. National Center for Health Statistics. International classification of diseases, adapted for use in the United States. IIL Title: Comparability of mortality statistics . . . IV. Series: United States. National Center for Health Statistics. Vital and health statistics: Series 2, Data evaluation and methods research; no. 66. V. Series: United States. Dept. of Health, Education, and Welfare. DHEW publication; no. (HRA) 76-1340. [DNLM: 1. Classification. 2. Mortality —U.S. W2 A N148vh no. 66] RA409.U45 no. 66 [RB115] 312°.07°23 [312’.2°0723] ISBN 0-8046-0039-9 75-619043 DATA EVALUATION AND METHODS RESEARCH Series 2 Number 66 Comparability of Mortality Statistics for the Seventh and Eighth Revisions of the International Classification of Diseases, United States DHEW Publication No. (HRA) 76-1340 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Resources Administration National Center for Health Statistics Rockville, Md. October 1975 NATIONAL CENTER FOR HEALTH STATISTICS HAROLD MARGULIES, M.D., Acting Director ROBERT A. ISRAEL, Acting Deputy Director GAIL F. FISHER, Associate Director for the Cooperative Health Statistics System ELIJAH L. WHITE, Associate Director for Data Systems EDWARD E. MINTY, Associate Director for Management PETER L. HURLEY, Acting Associate Director for Operations JAMES M. ROBEY, Ph.D., Associate Director for Program Development ALICE HAYWOOD, Information Officer DIVISION OF VITAL STATISTICS JOHN E. PATTERSON, Director ALICE M. HETZEL, Deputy Director ROBERT J. ARMSTRONG, M.S., Chief, Mortality Statistics Branch Vital and Health Statistics-Series 2-No. 66 DHEW Publication No. (HRA) 76-1340 Library of Congress Catalog Card Number 75-619043 PREFACE It has been the practice to revise the International Classification of Diseases every 10 years since 1900 to keep abreast of medical knowledge. Fach decennial revision has produced breaks in the comparability of cause-of-death statistics. The degree of discontinuity resulting from the introduction of the Eighth Revision, used beginning with data year 1968, has been considerable for a number of causes of death. An adequate presentation of mortality trends without reference to these discontinuities is impossible. The authors acknowledge with pleasure that special thanks are due to Bernice Wilkins, formerly Statistician, National Center for Health Statistics, for her assistance in identifying some of the important changes between the Seventh and Eighth Revisions that resulted in these discontinuities. Grateful appreciation is also expressed to Jeffrey D. Maurer and Evelyn J. Glass for help in the preparation of the tables in this report. CONTENTS Page INMOAUCLION. wos urr anim ntns reo tare CREB ER RIB EI SRIF NEU NIB UI HORT 2a 1 Scope of thE Report sus snvvnursnrensv women mosdassatabissms anaes 1 Changes in Methodology . ......... iii. 2 Meaning of Comparability Ratios ................................. 3 Use of Ratios as Revision FPACLOrS « xr vvsn ve nnsnnin autos sarin nmvinmin 4 1. Diseases of Heart . . ...... iit tein 5 All Diseases of Heart (ICDA Nos. 390-398, 402, 404, 410-429) ............. 5 Major Components of Diseases Of Heart «cc avin vanis onion ioiomeinina 7 2. Malignant Neoplasms . . ...... iii 17 All Malignant Neoplasms (ICDA Nos. 140-209) .............c.uuiuuen... 17 Major Components of Malignant Neoplasms .......................... 19 8. Cerebrovascular DISCAses « «sv ven usa namos saab amarsnnnnrmmennes wns 32 ES 36 5, Influenza and PRotitionia. «curs ninss son sus sans sss sw sons smusnns a 39 6. Certain Causes of Mortality in Early Infancy .......................... 42 7. Didhetes Mellitus viv ssn vin ainid sas tons ssnneosarsmorsnsmmenwes 44 8. Arteriosclerosis .. o.oo... eee eee 44 9. Bronchitis, Emphysema, and ASIAMIA ; css ss un snssnninsisnnsmsnisioss ssa 45 All Bronchitis, Emphysema, and Asthma (ICDA Nos. 490-493) ............. 45 Major Components . . . out it tee 47 10. Cirhosis Of LAVERY vu v sv sv smears ss musnpinstonumsnsssasmusnmsns 49 JL. 50ICHAE sonata nia mise ia bad ss nmaantmmemnr mer mms mm any 50 12. Congenital Anomalies . . . . oo. ea 50 13. HOMICIAE nv un ss nw sms sas sn Fs sss Ros nado hiv ammsmnransamesss ws 52 14. Nephritis and Nephrosis . ... oo ei 52 15. Peptic VICer uv run rusrnntanenminenmsameasssuinssansavsanss 58 vi List of Detaled TabIEE « « voc ssn vam ns sa so ssnssaomeurnsbmemmen naman 61 Appendix I. Components of Comparability Ratios for 15 Leading Causes of Death and Their Major Subcategories . ....... outta 68 Appendix IL, Techical NOES + coc nv inminssmnsanan ads nd amass ses oem 78 Death SIatistits » s vu vunssnss sv rive ins tasdsvisramrmmemmssmenwesn 78 RACE wns summsusstsattomtanssnrssntsrmmoraswrmupssrasesssesios 78 Population Bases . ......... ee 78 J 79 Appendix III. The United States Standard Certificate of Death, 1968 Revision . . . . . . 80 Appendix IV. Brief Summary of Statistical Design... ..................... 82 General Plan for Deaths at ANLAGES « . ci sora vsnssasmsnva saint nnsmesn 82 Appendix V. Comparability Ratio for Major Cardiovascular Diseases . ........... 85 Appendix VI. Notes for Use in Primary Mortality Coding for the Eighth Revision .... 86 Appendix VII. Assignment by the Eighth Revision of Deaths That by the Seventh Revision Were Assigned to Malignant Neoplasm of Other and Unspecified Sites (ICD NO. 199) «4 svt cet s nnn orrarsosassnmssnsenennsenssossns 92 SYMBOLS Data not available------seeeeemmeeeeeeeeeeeee Category not applicable-----------ceeeeommmreeeaeooe Quantity zero Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision-------------ssseeeeeeeenens * COMPARABILITY OF MORTALITY STATISTICS FOR THE SEVENTH AND EIGHTH REVISIONS OF THE INTERNATIONAL CLASSIFICATION OF DISEASES A. Joan Klebba, Statistician, Division of Vital Statistics, and Alice B. Dolman, Medical Codification Specialist, Office of International Statistics INTRODUCTION Scope of the Report During 1900-1969 causes of death were clas- sified according to eight revisions of the Inter- national Classification of Diseases (ICD). These revisions are made about every 10 years to reflect progress in medical knowledge. The revi- sion in use in the United States beginning with data year 1968 is the Eighth Revision Inter- national Classification of Diseases, Adapted for Use in the United States, 1965 (hereinafter denoted by ICDA). Table A. Deaths and death rates for the 15 leading causes 1969 This report presents the principal changes in classification and coding procedures for the 15 leading causes of death in 1969 and their major components. These 15 leading causes and their components are included among the 48 causes shown in tables 1 and 2. Measures of the comparability of mortality statistics between 1967 and 1968 are shown for these 48 causes in appendix I. The 15 leading causes accounted for 89 percent of the 1,921,990 deaths that occurred in the United States in 1969 (table A). A description of the sources of data and of rates and other terms is given in appendix II, of death: United States, [Rates per 100,000 estimated population] Cause of death (Eighth Revision, International Classifica- Raak tion of Diseases, Adapted, 1965) Number Rage All causeS-------======--m-----oe-----o--o-o-ss-oo-ss-os 1,921,990 951.9 1| Diseases of heart-------==--comecoo-uo- 390-398,402,404,410-429 739,265 366.1 2| Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tissues------===--mccoocoooonnoonommon 140-209 323,092 160.0 3| Cerebrovascular diseases----==--==c-cmmmcomcoooocoonmm= 430-438 207,179 102.6 4| Accidents -=-==m-mmmmcmcccme meme memmme mom m———mm mee E800-E949 116,385 | 57.6 Motor vehicle accident§==-=------ccmmmmmmnomnnaamm= E810-E823 55,791 27.6 All other accidents----===-==-eeececco=-- E800-E807,E825-E949 60,594 30.0 5| Influenza and pneumonia--=---===--===---------- 470-474 ,480-486 68,365 33.9 6| Certain causes of mortality in early infancy-====-===--cccmcmmcmmmaaan 760-769.2,769.4-772,774-778 43,171 21.4 7 Diabetes mellituS=======c-c-cccccecmemem mee mceoom m= 250 38,541 19.1 8| Arteriosclerosis--==-==------cmemcemmemme me ecem mm eom——= 440 33,063 16.4 9| Bronchitis, emphysema, and asthma=-----==-=----c------= 490-493 31,144 15.4 10| Cirrhosis of liver----------ccccemmmmmommoom ccm m mmm mmm 571 29,866 14.8 11| Suicide==--mmmmmmemcemce eee mmmmemmmm mm meme E950 -E959 22,364 11.1 12| Congenital anomalies=-=--==-=--mc=c-oooooooocomonnnno—- 740-759 17,008 8.4 13| Homicide=--=mmmmcmmcme mcm meee cmmmm mm mmmm meen mmo E960-E978 15,477 7.7 14| Nephritis and nephrosis-----=--=----c--ccoocommommnn- 580-584 9,417 4.7 15| Peptic ulcer----=-==---memememmmmeee mma em ——————————————— 531-533 9,312 4.6 ..| All other causeS======-=c-cccececcccmcmemmmmmmm—mm—— Residual 218,341 108.1 “rm 1 fechnical Notes.” A copy of the standard certificate of death adopted with some modifica- tions by most States is shown in appendix III. A brief summary of the statistical design to esti- mate the numerators of the comparability ratios used in this report is given in appendix IV; and the comparability ratio for the group title Major cardiovascular diseases (ICDA Nos. 390-448) is given in appendix V. Notes for use in primary mortality coding for the Eighth Revision are shown in appendix VI. Changes in Methodology A review of earlier approaches made to provide guidelines to the comparability of the titles from revision to revision of the Interna- tional Classification of Diseases (ICD) makes it clear that progress has been made over the years. But undoubtedly even the systematic method for evaluating breaks in comparability of mortality statistics in the transition from the Seventh to the Eighth Revision will leave the reader with unanswered questions that point to the need for further development of the method for evaluating more details of the disruption of mortality trends caused by revisions of the ICD. As shown below, there have now been eight revisions of the ICD: Years in use in United States Year of Conference by which adopted Revision of the International Classification of Diseases 1900 1900-1909 1909 1910-1920 1920 1921-1929 1929 1930-1938 1938 1939-1948 1948 1949-1957 1955 1958-1967 1965 1968 to date The search for a systematic method for evaluation of classification changes and other changes in the transition from one to the next revision of the ICD began with the introduction of the Second Revision, adopted in 1909 and used during 1910-20. Dunn and Shackley, in their report on the comparability of mortality statistics between the Fourth Revision (adopted in 1929 and in use 1930-38) and the Fifth Revi- sion (adopted in 1938 and in use 1939-48), included a definitive review of earlier attempts to assess the effects on the comparability of mortality statistics resulting from the intro- duction of a new revision.! Essentially, the approach in these earlier attempts was to compare the frequencies of deaths assigned to a given title of the new revision for the first year it was in use with the frequencies assigned to the most nearly com- parable title of the prior revision for the last year it was in use. Dunn and Shackley were the first to apply the dual-coding method. In their attempt to measure the loss of comparability with the introduction of the Fifth Revision, they first classified all certificates for deaths occurring in 1940 by the Fifth Revision (in use 1939-48) and then classified the same certificates by the Fourth Revision (in use 1930-38). They pre- sented the results in two sets of figures: one showing the number of certificates assigned to each cause by the Fifth Revision and the other showing the number of certificates assigned by the Fourth Revision. Instead of computing com- parability ratios, however, these authors com- puted percentages to show the extent to which comparability had been lost for each cause by the introduction of the Fifth Revision. The International Conference for the Sixth Decennial Revision of the International Lists recommended that deaths for a country as a whole in the year 1949 or 1950 should be coded according to both the Fifth and Sixth Revisions. In the United States 1950 was selected as the transition year. The first set of comparability ratios, based on coding the same deaths occurring in 1950 by both the Fifth and Sixth Revisions, was published by the Vital Statistics Division of the National Center for Health Sta- tistics (NCHS) in 1964.2 The second set of com- parability ratios measured the effects of changes between the Sixth and Seventh Revisions. The ratios in the present report, computed to measure breaks in comparability of cause-of- death statistics resulting from the introduction of the Eighth Revision, make only the third set of ratios for the United States. The Sixth Revision represented a more sweeping change than any of the previous revi- sions. As stated by Faust and Dolman,? up until the Sixth Revision (in use 1949-57) .. selection of a cause of death for tabula- tion was made by reference to a set of prior- ity tables published in the Manual of Joint Causes of Death. This manual, which was developed for use in the United States, was followed until 1949, when an international procedure for joint-cause selection was adopted. The Manual of Joint Causes of Death consists of priority tables based on a series of decisions made by clinicians re- garding the relative importance of various diseases as causes of death. By using these priority tables when more than one cause was jointly reported, the primary cause for tabulation was selected. The new interna- tional rules adopted for use in the United States in 1949 made the medical practitioner responsible for indicating the underlying cause of death. If the medical certification is properly made, the physician’s statement of the underlying cause of death is accepted for tabulation. In general the changes incorporated in the Seventh Revision (in use 1958-67) were of limited scope and consisted of only essential changes and amendments of errors and incon- sistencies. The ICD has been made increasingly specific over these eight revisions. In the First Revision the categories and subcategories of discases and injuries numbered about 200; by 1939, when the Fifth Revision was put in use, the categories and subcategories numbered 425; and by 1968, when the Eighth Revision was first used, the Detailed List consisted of 671 categories of diseases and morbid conditions plus 182 cate- gories for classification of external causes of injuries and 187 categories for characterization of injuries according to the nature of lesion. Meaning of Comparability Ratios The comparability ratios presented in this report are based on coding the same deaths occurring in 1966 by both the Seventh and Eighth Revisions. More specifically, the denom- inator of a ratio for any particular cause is the number of deaths in 1966 assigned to that cause in accordance with the Seventh Revision, and the numerator is the estimate of the number of deaths in 1966 assigned to the comparable cause in accordance with the Eighth Revision. The estimates of the deaths assigned by the Eighth Revision are based on a random sample of all deaths in 1966 stratified by cause (appendix IV). The year 1966 was selected because it was the most recent year for which final mortality statistics according to the Seventh Revision were available at the time of this study. A comparability ratio of 1.00 indicates that the same number of deaths was assigned to a particular cause or combination of causes whether the Seventh or Eighth Revision was used. A ratio showing perfect correspondence (1.00) between the two revisions does not necessarily indicate that the cause was unaf- fected by changes in classification and coding procedures because the changes may compensate for each other. A ratio of less than 1.00 results from one of two situations: (1) a decrease in assignments of deaths to a cause in the Eighth Revision as com- pared to the Seventh, resulting, for example, from the transfer of inclusion terms, or (2) the cause as described by the Eighth Revision is only a part of the Seventh Revision title with which it is compared. Usually a ratio of more than 1.00 results from an increase in assignments of deaths to a cause in the Eighth Revision as compared with the comparable Seventh Revision cause. At times, however, the increase may result from the fact that the Eighth Revision cause is not the equivalent of that described by the Seventh Revision title with which it is compared. For example, consider the following division of the Eighth Revision group title Diseases of arteries, arterioles, and capillaries (440-448): Arterio- sclerosis (440) and Other diseases of arteries, arterioles, and capillaries (441-448). A ratio of 1.549 is obtained by comparing the group title corresponding to ICDA Nos. 441-448 with the combination of the following three titles in the Seventh Revision: Aortic aneurysm, non- syphilitic, and dissecting aneurysm (451); Seventh Revision Eighth Revision Eighth Revision title Number Estimated Category Category b of number number deaths number of deaths JL 15,213 --- 23,567 Other diseases of arteries, arterioles, and CAOPINBEIBE. ..ouvvsrusrrimmmssrnemassipsnmmissinsumssssRaes EETTRAS ERR 451-456 15,213 441-448 123,567 Aortic aneurysm (nonsyphilitic)........ccccouvveeeeneeennnnes 451 11,270 441 12,196 GaANGIENE.....ccivieiireeiireeeieeee cree saeessrreesaeseeesssneeenns 455 348 445 2,720 All other diseases of arteries, arterioles, and CAPMIBPIBS ...ocireivrrmnmvssmsnsmmsssssssmsvaniemssmirssossisssassiuinss 452-454 ,456 3,595 442-444,446-448 8,555 I'The subtotals do not add up exactly to the total because of the application of different sampling fractions. Gangrene of unspecified cause (455); and Other arterial diseases (452-454, 456). The estimated number of deaths assigned to each of the three- digit subcategories in the Eighth Revision under 441-448 together with the number of deaths assigned to each of the comparable three-digit subcategories in the Seventh Revision are shown above. As this table shows, the comparability ratio of 1.549 (appendix I) results from an increase in assignments of deaths to each of the three subcategories in the Eighth Revision as com- pared with the corresponding assignments in the Seventh Revision. About 86 percent of the 1,434 deaths assigned in the Seventh Revision to Aneurysm of aorta (022) under Syphilis and its sequelae (020-029) are assigned in the Eighth Revision to Aortic aneurysm (nonsyphilitic) (441). Also, about 83 percent of the 2,342 deaths assigned in the Seventh Revision to General arteriosclerosis with mention of gan- grene as a consequence (450.1) are assigned in the Eighth Revision to Arteriosclerotic gangrene (445.0). Finally, about 21 percent of the 10,078 deaths assigned in the Seventh Revision to Hernia and intestinal obstruction (560, 561, 570) are transferred in the Eighth Revision to Arterial embolism and thrombosis of mesenteric artery (444.2). Use of Ratios as Revision Factors The following table illustrates the applica- tion of comparability ratios to determine whether changes for two death rates were real or resulted from the adoption of the Eighth Revi- sion. The figures used are from the 10-percent sample for January-June 1967 and January-June 1968. In each instance the reported death rate in column (3) was multiplied by the ratio in Death rate per 100,000 population: January-June Eighth Revision title and category numbers Comparability 1968: 1967: Seventh Revision Eighth Revised by ratio Revision Reported in column (1) (2) (3) (4) Hypertensive heart disease with or without renal OISBASE 1. svusismarmrni sss T ss TS IR RRR NRE hrs 402,404 0.3941 9.7 26.6 10.5 Active rheumatic fever and chronic rheumatic heart CIISBBBR 1: cusvvnssrmmmmrsnssisssinisavavsva RTARTA SER 390-398 1.1519 8.7 7.6 8.8 column (1) to obtain the death rate for January-June 1967 that is most nearly com- parable to the death rate for January-June 1968. A comparison of the revised death rates for the earlier period with the corresponding death rates 1. DISEASES All Diseases of Heart (ICDA Nos. 390-398, 402, 404, 410-429) As shown in the table below, the net effect of changes introduced with the Eighth Revision on the number of deaths classified to Diseases of heart, based on coding the above-described sample of records for deaths occurring in 1966 by both the Seventh and Eighth Revisions, was an increase from 727,002 to 730,261. This gives a comparability ratio of 1.0045 (appendix I). An estimated 10,711 deaths classified under this title by the Eighth Revision were classified elsewhere by the Seventh Revision, and 7,452 deaths classified under this title by the Seventh Revision were classified elsewhere by the Eighth. In other words, by the Eighth Revision there was a gain to Diseases of heart of 10,711 deaths and a loss of 7,452 deaths, resulting in a net gain of 3,259 deaths. The changes responsible for the largest losses among the estimated 7,452 deaths no longer assigned to Diseases of heart are presented below under the discussions relating to the major com- ponents of the diseases of heart for which the losses occurred. The four major changes respon- sible for the largest gains among the estimated 10,711 deaths now assigned to Diseases of heart came from a number of Seventh Revision cate- gories and were distributed over the components of heart disease. These four major changes are presented below. (1) Among the 10,711 deaths assigned to Diseases of heart by the Eighth but not by the Seventh Revision were an estimated 1,773 that had been assigned by the Seventh to accidental causes. Of these 1,773 deaths, an estimated 1,422 were assigned by the Eighth Revision to for January-June 1968 shows a small decrease in the death rate for hypertensive heart diseases and no significant change in the death rate for active rheumatic fever and chronic rheumatic heart disease. OF HEART Ischemic heart disease (ICDA Nos. 410-413). These assignments occurred because the Seventh Revision rules that under certain circumstances resulted in coding an accidental cause in prefer- ence to other reported conditions were changed in the Eighth Revision. These Seventh Revision rules applied when two or more conditions, one of which was an accident, poisoning, or violence (but not a late effect of these), were entered on the death certificate in such a way that none could be regarded as the underlying cause of death.5 The Eighth Revision rules, however, lead to the selection of the first-mentioned condition when causes of death are entered on the death certificate in this manner. The following example of entries in the medical certification section of a death certificate illustrates the effect of this change in the coding rules. The death described was classified to accidental fall by the Seventh Revision rules and to coronary arteriosclerosis by the Eighth Revision rules. I (a) Coronary arteriosclerosis (b) Fracture of hip due to (c) Accidental fall (For further discussion on transfers by the Eighth Revision from accidental causes, espe- cially falls, see section 4, “Accidents.”) (2) Also among the 10,711 deaths assigned to Diseases of heart by the Eighth but not by the Seventh Revision were an estimated 2,326 that were assigned by the Seventh Revision to Other hypertensive disease (ICD Nos. 444-447). These transfers resulted from a change in the procedure for coding causes of death involving arteriosclerotic heart disease (without mention of coronary artery disease) jointly reported with Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of : Cause of death and category number deaths pnat went 10 Beatis imeniuma 2) in 19686 (ICDA Nos. 390-398, other causes 402,404,410-429) (1) (2) (3) (4) AN CAUSE sos ssmmssvmmrosssmmsios Eves sis hess 001-E999 1,863,149 730,261 1,132,888 Diseases of heart.... 400-402,410-443 727,002 719,550 7,452 Rheumatic fever.................... eeeer...400-402 456 449 7 Chronic rheumatic heart disease.................... 410416 14,556 14,436 120 Diseases of mitral valve...........ccccceeeeeeeuueuuereeennn. 410 5,718 5,666 52 Diseases of aortic valve specified as rheumatic....411 1,713 1,693 20 Diseases of pulmonary valve and other endocarditis, specified as rheumatic ........... 413,414 773 773 - Other rheumatic heart diseases............. 412,415,416 6,352 6,304 48 Arteriosclerotic heart disease, including coronary CIISOBSE «ov uunsrms rus mrss sss rE TA TEE nara names d 420 573,191 570,220 297 Arteriosclerotic heart disease so described....... 420.0 158,802 158,012 790 Heart disease specified as involving coronary BILBIIBS wovvinisniistnsisenvmmiismisinsmsissammnnnssnnnnnnnenntad 420.1 414,101 411,920 2,181 Angina pectoris without mention of coronary AISEASE eevee eres, 420.2 288 1288 L Chronic endocarditis not specified as rheumatic...421 3,785 3,560 225 Of mitral valve, specified as nonrheumatic ...... 421.0 166 157 9 Of aortic valve, not specified as rheumatic ...... 421.1 2,311 2,261 50 Of other valves, not specified as rheUMALIC coisa 421.2421.4 1,308 1,142 166 Other myocardial degeneration ...................c........ 422 49,796 49,175 621 With arteriosclerosis ............... cree 422.1 41,935 41,464 471 Without mention of arteriosclerosis.......422.0,422.2 7,861 7,711 150 Other diseases of heart ............covveeeveeeeeeeee.... 430-434 31,042 28,277 2,765 Acute and subacute endocarditis........................ 430 757 706 51 Acute myocarditis and acute pericarditis, not specified as rheumatic .......c.ccceoeeeveveeennenn.. 1,065 1,051 14 Functional disease of heart.................... 10,008 7,693 2,315 Other and unspecified diseases of heart 19,212 18,827 385 Hypertensive heart disease.................oc.uuvnn.... 54,176 53,433 743 Hypertensive heart disease with arteriolar NBPNIOSCIBIOSIS isis ivirininvinasnssnsnrsnssnsomssnnemasmad 11,606 11,345 261 Other hypertensive heart disease 42,570 42,088 482 Other CAUSES ......cceeeeeeeeeeiiieeeeee eee ee eeeeee esses 1,136,147 10,711 1,125,436 I There were no deaths in the sample assigned to the Seventh Revision title Angina pectoris without mention of coronary disease (ICD No. 420.2). NCHS nosologists state, however, that with a possible few exceptions these deaths would be assigned by the Eighth Revision to Angina pectoris (ICDA No. 413). The exceptions, if any, resulted from the dropping of the preference in the Seventh Revision of angina pectoris over cardiovascular diseases. hypertensive disease. The Seventh Revision had no special provision for classifying this com- bination of diseases, and under certain circum- stances the hypertensive disease was coded as the cause of death. For example, reports of arteriosclerotic heart disease (ICD No. 420.0) due to a condition classifiable to ICD Nos. 444-447 (Other hypertensive disease) were classified to the appropriate hypertensive disease category. In the Eighth Revision arteriosclerotic heart disease is classified as an ischemic heart disease, and four-digit subdivisions have been provided under the ischemic heart disease cate- gories for jointly reported hypertensive disease. (3) A third group among these 10,711 deaths assigned by the Eighth Revision to Diseases of heart included an estimated 1,743 that were classified by the Seventh Revision to Vascular lesions affecting central nervous system (ICD Nos. 330-334). The largest number of these 1,743 causes, an estimated 1,249, were assigned to Chronic ischemic heart disease (ICDA No. 412). In addition, an estimated 166 were assigned to Hypertensive heart disease (ICDA No. 402), 164 to Acute myocardial infarction (ICDA No. 410), and 82 to Sympto- matic heart disease (ICDA No. 427). The assignments to ICDA No. 412, Chronic ischemic heart disease, resulted primarily from a change in the procedure for coding arterio- sclerotic heart disease, hypertension, and an intracranial vascular lesion when jointly reported as causes of death. As mentioned above, no special provision was made in the Seventh Revi- sion for coding Arteriosclerotic heart disease (ICD No. 420.0) jointly reported with hyper- tension, but provision was made for coding hypertension jointly reported with an intra- cranial vascular lesion to the intracranial vascular lesion. In the Eighth Revision provision is made for coding hypertension jointly reported with arteriosclerotic heart disease as well as with intracranial vascular lesions. The result of this change in coding procedure is that some deaths that were attributed to intracranial vascular lesions by the Seventh Revision are classified to chronic ischemic heart disease with hypertension by the Eighth Revision. Two examples of entries in the medical cer- tification part of the death certificates that illustrate this change in assignment are given below: I (a) Arteriosclerotic heart disease and cerebral hemorrhage (b) Hypertension (c) I (a) Cerebral embolism (b) Arteriosclerotic heart disease (c) Hypertension By the Seventh Revision both of the deaths reported above would have been assigned to intracranial vascular lesions, but by the Eighth Revision, they are assigned to Chronic ischemic heart disease. (4) Another substantial group among the 10,711 deaths assigned by the Eighth Revision to Diseases of heart is an estimated 1,242 causes that were coded by the Seventh Revision to General arteriosclerosis (ICD No. 450). Almost all of these (1,209) were transferred to Chronic ischemic heart disease (ICDA No. 412). In the Eighth Revision provision is made for coding arteriosclerotic heart disease (which is classified to ICDA No. 412) in preference to arterioscle- rosis when the two conditions are jointly re- ported on the death certificate. According to the provisions of the Seventh Revision preference was given to arteriosclerotic heart disease only when this condition was specified as due to arteriosclerosis. Therefore causes of death certi- fied in the following manner are classified to Chronic ischemic heart disease (412) by the Eighth Revision and to General arteriosclerosis (450) by the Seventh Revision. I (a) Hypostatic pneumonia (b) General arteriosclerosis with arterio- sclerotic (c) heart disease Although the comparability ratio for the entire group of Diseases of heart is close to 1.000, some major components of this group have comparability ratios that differ substan- tially from 1.000. Factors pertinent to the ratios shown in the table in appendix I for the five major components of diseases of heart are pre- sented in the following paragraphs. Major Components of Diseases of Heart Active rheumatic fever and chronic rheu- matic heart disease (ICDA Nos. 390-398).—In the Seventh Revision diseases of the aortic valve not qualified as rheumatic were included in ICD No. 421.1, Chronic endocarditis of aortic valve, not specified as rheumatic. This category was not included under the heading Chronic rheu- matic heart disease (ICD Nos. 410-416). In the Eighth Revision, however, aortic valve diseases, unless specified as nonrheumatic, are included under Chronic rheumatic heart disease (ICDA Nos. 393-398). As a result of this change in clas- sification almost all of the deaths that were Seventh Revision Eighth Revision Estimated number of deaths in column (2) } Number of that went to Active Estimated number of Cause of death and category number deaths rheumatic fever and deaths in column (2) in 1966 chronic rheumatic that went 10 heart disease (ICDA other causes Nos. 390-398) (1) (2) (3) (4) All CAUSES oovvovemnimmmis sivas iim savas sas 001-E999 1,863,149 17,293 1,845,856 Comparable causes .................. 400-402,410416,421 1 17.323 16,652 671 Rheumatic fever and chronic rheumatic heart disease! o.oo, 15,012 14,724 288 Rheumatic fever 456 443 13 Chronic rheumatic heart disease 14,556 14,256 275 Diseases of mitral VBIVE....uumsasivmmmmmsimimsind 5,718 5,600 118 Diseases of aortic valve specified as rheumatic..411 1,713 1,686 27 Diseases of pulmonary valve and other endocarditis, specified as rheumatic ......... 413,414 773 759 14 Other rheumatic heart diseases........... 412,415,416 6,352 6,236 116 Chronic endocarditis of aortic valve, not specified BS PHBMITIBTIC .oiniivirvmmses cumsvisssnssssisnssermsasvivsnmed 421.1 2.311 1,928 383 Other CAUSES ........uuueiirineiiiiiieeiinieeeraeeeesannnns Residual 1,845,826 641 1,845,185 I This is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. assigned to category 421.1 by the Seventh Revi- sion were classified under chronic rheumatic heart disease by the Eighth. This change accounts in great part for the fact that a larger number of deaths (an estimated 17,293) were classified by the Eighth Revision to Active rheu- matic fever and chronic rheumatic heart disease (ICDA Nos. 390-398) than were classified by the Seventh Revision (only 15,012 deaths) to the similar title Rheumatic fever and chronic rheu- matic heart disease (ICD Nos. 400-402, 410-416). This gives an adjustment factor of 1.1519 for these two titles. (See appendix I.) As shown above, 1,928 of the 2,311 deaths assigned by the Seventh Revision to ICD No. 421.1 were classified by the Eighth Revision to ICDA Nos. 390-398. During the period 1950-67 the death rate for Rheumatic fever and chronic rheumatic heart disease (ICD Nos. 400-402, 410-416) declined from 14.8 to 7.2 deaths per 100,000 population. The apparent rise in the death rate for these causes in 1968 to 8.2 per 100,000 population is attributable in great part to the above- mentioned classification change. Applying the comparability ratio 1.1519 to the 1967 death rate of 7.2 raises it to 8.3 deaths per 100,000 population, which is very close to the 1968 rate of 8.2. The downward trend continued through 1971, when the rate was 7.1 per 100,000 popu- lation. Hypertensive heart disease with or without renal disease (ICDA Nos. 402, 404).— The com- parability ratio for Hypertensive heart disease with or without renal disease (ICDA Nos. 402, 404) is only 0.3941. This ratio is obtained by dividing 21,350, the estimated number of 1966 deaths classified by the Eighth Revision to this group of causes, by 54,176, the number clas- sified to the most nearly comparable Seventh Revision title, Hypertensive heart disease (ICD Nos. 440-443). (See appendix 1.) The following table shows the results of coding a random sample of 2,389 of these 54,176 deaths. As shown in the table below, an estimated 33,741 of the 1966 deaths that were classified by the Seventh Revision to Hypertensive heart disease (ICD Nos. 440-443) were not classified Eighth Revision Seventh Revision Estimated Cause of death and category number number Category Number number of deaths of deaths TOMA ciseisn inion 54,176 440-443 54,176 Hypertensive heart disease with or without renal disease: 10,717 9,718 32,636 1,105 Hypertensive disease: 1:7 — 42,570 440,441,443 42,570 10,717 46 31,050 757 Hypertensive heart and renal disease: Total ..ocinesisamnnianss 11,606 442 11,606 B02 ciinisisassmmns - NOK. onssporvergrne 9,672 Chronic ischemic heart disease with hypertensive QISBASE .........essssssnenssssvssassvanenssrnssnss A12.0..coiiimimminis 1,587 Other Eighth Revision titles........coirssimsisys seviesssins Other .smmnsmses 347 to the most nearly comparable Eighth Revision title, Hypertensive heart disease with or without renal disease (ICDA Nos. 402, 404). Of these, about 97 percent (32,636 of the 33,741) were classified to Chronic ischemic heart disease with hypertensive disease (ICDA No. 412.0) by the Eighth Revision. As shown in the next table, there were a total of 42,570 deaths assigned to Other hypertensive heart disease (ICD Nos. 440, 441, 443) in 1966 when the Seventh Revision was in use. In the dual-coding study, it was found that about 75 percent of these deaths (31,807) were not assigned to Hypertensive heart disease with or without renal disease (ICDA Nos. 402, 404) by the Eighth Revision. Of the 31,807 deaths, an estimated 31,050 were classified to Chronic ischemic heart disease with hypertensive disease (ICDA No. 412.0). The table below also shows that a total of 11,606 deaths in 1966 were assigned to Hyper- tensive heart disease with arteriolar nephro- sclerosis (ICD No. 442) by the Seventh Revision. In the dual-coding study it was found that about 17 percent of these deaths (1,934) were not assigned to the Eighth Revision title Hyper- ‘tensive heart disease with or without renal disease (ICDA Nos. 402, 404). Again, an esti- mated 1,587 of these 1,934 deaths were classi- fied to Chronic ischemic heart disease with hypertensive disease (ICDA No. 412.0). These differences in code assignments between the Seventh and Eighth Revisions resulted primarily from the two changes in clas- sification that are described below. One of the two changes is for Arterioscle- rotic heart disease jointly reported with hyper- tensive disease. The Seventh Revision had no special provision for classifying Arteriosclerotic heart disease so described (ICD No. 420.0) jointly reported with hypertensive disease. Con- sequently, some deaths attributable to this com- bination of diseases were classified by the Seventh Revision to Hypertensive heart disease (ICD Nos. 440-443). In the Eighth Revision arteriosclerotic heart disease is classified as an ischemic heart disease, and four-digit sub- divisions are provided under the ischemic heart disease categories for jointly reported hyperten- sive disease. As a result of this change in classifi- Seventh Revision Eighth Revision Estimated number of deaths in column (2) that went to Estimated number of Number of Hypertensive heart deaths in column (2) Cause of death and category number deaths di ith that 1 to in 1966 isease with or at wen without renal other causes disease (ICDA Nos. 402,404) (1) (2) (3) (4) All CAUSES evecare 001-E999 1,863,149 21,350 1,841,799 Hypertensive heart oisease....iuusv suisse 440-443 54,176 20,435 33,741 Hypertensive heart disease with arteriolar NOPNTOSCIBIOBIS visiisssiivisivrisssrsmsssprivssnsiisssasssvrasanss 442 11,606 9,672 1,934 Other hypertensive heart disease ............ 440,441,443 42,570 10,763 31,807 Other CAUSES .......cceevvvvuiiieeeeeeeeieiceeee eee eeaeaes Residual 1,808,973 915 1,808,058 cation, the following examples of cause-of-death certifications were classified to a hypertensive heart disease (specifically to ICD No. 443) by the Seventh Revision and to Chronic ischemic heart disease with hypertensive disease (ICDA No. 412.0) by the Eighth Revision: I (a) Arteriosclerotic heart disease (b) Hypertensive heart disease I (a) Hypertensive and arterio- (b) sclerotic heart disease The second change is for Other myocardial degeneration with arteriosclerosis, jointly re- ported with hypertensive disease. In the Seventh Revision category 422.1, Other myocardial degeneration with arteriosclerosis, included the conditions listed below, and provision was made for coding any condition in this category jointly reported with a hypertensive disease (ICD Nos. 440-447) to Hypertensive heart disease (440-443). Cardiosclerosis Cardiovascular: arteriosclerosis degeneration disease 10 Cardiovascular sclerosis Myocardial degeneration with arteriosclerosis or synonym in ICD category 450 In the Eighth Revision the above-listed condi- tions are included in ICDA category 412, Chronic ischemic heart disease, and provision is made for classifying a jointly reported hyperten- sive disease in a four-digit subdivision of this category, 412.0. Specific examples illustrating the effect of this change in classification are given below. I (a) Cardiovascular disease [7th, ICD No. 422.1; 8th, ICDA No. 412.9] (b) Hypertension [7th, ICD No. 444; 8th, ICDA No. 401] As stated above, the Seventh Revision provided for the classification of cases such as this to hypertensive heart disease (ICD No. 443), which includes cardiovascular disease (ICD No. 422.1) jointly reported with hypertension (ICD No. 444). In the Eighth Revision this example is classified to category 412.0, Chronic ischemic heart disease with hypertensive disease, which includes cardiovascular disease unqualified (ICDA No. 412.9) jointly reported with hyper- tension (ICDA No. 401). I (a) Myocardial degeneration [7th, ICD No. 422.2; 8th, ICDA No. 428] (b) Arteriosclerosis [7th, ICD No. 450; 8th, ICDA No. 440.9] [7th, ICD No. 446; 8th, ICDA No. 403] II Nephrosclerosis In the Seventh Revision arteriosclerosis (ICD No. 450) with myocardial degeneration (ICD No. 422.2) was classified to 422.1, Other myocardial degeneration with arteriosclerosis. The above example was assigned by the Seventh Revision, however, to category 442, Hyperten- sive heart disease with arteriolar nephrosclerosis, with includes conditions classifiable to category 422.1 jointly reported with nephrosclerosis. In the Eighth Revision, arteriosclerosis (ICDA No. 440) with myocardial degeneration (ICDA No. 428) is classified to 412.9, Chronic ischemic heart disease without mention of hypertensive disease. This example, however, is assigned to category 412.0, which includes conditions classifiable to 412.9 jointly reported with nephrosclerosis. Table B. om specified ratios Estimated number of deaths in 1966 assigned to specified {ICDA) Sosording to the Eighth Revision: based on a stratified random r e During 1958-67 the death rate for Hyperten- sive heart disease (ICD Nos. 440-443) declined from 42.7 to 25.3 deaths per 100,000 popula- tion. Applying the comparability ratio of 0.3941 to the 1967 death rate of 25.3 lowers it to 10.0. This adjusted rate is close to the 1968 rate of 8.8 deaths per 100,000 population. The down- ward trend for this cause continued through 1971, for which year the rate was 6.8 deaths per 100,000 population. Ischemic heart disease (ICDA Nos. 410-413).—The most nearly comparable title for this group of diseases in the Seventh Revision is Arteriosclerotic heart disease, including coro- nary disease (ICD No. 420) (table B). Based on the 1966 comparability study, an estimated 14.57 percent more deaths (totaling 83,500) were assigned by the Eighth Revision to this title than were assigned to Arteriosclerotic heart disease, including coronary disease (ICD No. 420) by the Seventh Revision (appendix I). As shown in the table, page 12, the largest single group of these 83,500 deaths (an estimated 41,228) were assigned by the Seventh Revision to Other myocardial degeneration with arterio- category numbers sample drawn venth Revision category numbers for provisional set of comparability [Some figures in this table differ slightly from those in other tables because they are based on a preliminary sample that was smaller in size] Eighth Revision ICDA numbers Seventh Revision ICD numbers Total 400,401 | 402-404 410 411 412 413 Total-=--=cccecee- 687,083 3,238 27,443 | 366,951 6,888 | 282,309 254 420,0=--c-mcemcm cman 157,748 - - 3,160 - | 154,588 - 420. meme cma 411,678 - - | 360,794 6,785 43,857 242 422.1 meme emma 41,229 - = 236 -| 40,993 - 430-434 cmc meme 4,095 - 27 1,090 53 2,925 - 440-4430 cmcmmm meee 53,657 184 20,543 69 - 32,861 - Ly ata 11,454 - 9,780 - - 1,674 - 440, 441, 443 --cneceemn 42,203 184 10,763 69 - 31,187 - GO4-447] mmm meme meee 10,959 2,857 6,189 46 - 1,867 - 446mm m mmm mmm meme eee 6,620 5 5,705 11 - 899 - 444 L445, 447 -—--amemn 4,338 2,851 484 35 - 968 - All other----------cc--- 7,717 197 684 1,556 50 5,218 12 11 Seventh Revision Eighth Revision Estimated number of Number of deaths in column (2) Estimated number of umber-o that went to deaths in column (2) Cause of death and category number deaths . in 1966 Ischemic heart that went to n disease (ICDA Nos. other causes 410-413) (1) (2) (3) (4) Al) COUSES suivsssvssnmssssmrsssrerrnsesiirerssssssvasss 001-E999 1,863,149 656,691 1,206,458 Comparable causes .................. 420,422.1,440,441,443 657,696 642,199 15,497 Arteriosclerotic heart disease, including coronary ISAS on. eeeeeeeeree eee en ene 420 573,191 569,715 3,476 Arteriosclerotic heart disease so described....... 420.0 158,802 157,749 1,053 Heart disease specified as involving coronary BIEBUIBS corsivsranin consensus amas eatin aus iors ianiyerd 420.1 414,101 411,678 2,423 Angina pectoris without mention of coronary GISRESE oisnmimimria RAL 420.2 288 2288 2 Other myocardial degeneration with ArLErIOSCIBIOBIS con visisisssmmsnsnsssssssnarsisemsniesssunsesy 422.1 41935 41,228 707 Other hypertensive heart disease ............ 440,441,443 42,570 31,256 11,314 Other CAUSES .......eerreiriniieerniieereeieerenieeeensieeenns Residual 1,205,453 14,492 1,190,961 I This is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. There were no deaths in the sample assigned to the Seventh Revision title Angina pectoris without mention of coronary disease (ICD No. 420.2). NCHS nosologists state, however, that with a possible few exceptions these deaths would be assigned by the Eighth Revision to Angina pectoris (ICDA No. 413). The exceptions, if any, resulted from the dropping of the preference in the Seventh Revision of angina pectoris over cardiovascular diseases. sclerosis (ICD No. 422.1), and of these 41,228 deaths an estimated 40,993 were classified by the Eighth Revision to Chronic ischemic heart disease (ICDA No. 412). Among the terms clas- sified by the Eighth Revision to Chronic ischemic heart disease (ICDA No. 412) that were included under Other myocardial degeneration with arteriosclerosis (ICD No. 422.1) in the Seventh Revision are the following: Cardiovascular: arteriosclerosis degeneration disease sclerosis Questions have been raised about the appropriateness of including cardiovascular diseases in category 412, Chronic ischemic heart disease, since this classification implies that cardiovascular disease means disease of the coro- nary arteries. A study conducted several years 12 ago of death certificates reporting “cardio- vascular disease” indicated that this frequently is not the case. In anticipation of a possible change in the classification of these terms and for the benefit of those interested in doing further study, NCHS has created special four-digit sub- categories under category 412 that are used instead of those listed under category 412 in the Eighth Revision, ICDA (table 3). These special four-digit subcategories permit the separation of deaths due to cardiovascular disease without evidence of a disease of the coronary arteries from those due to other conditions classifiable to chronic ischemic heart disease with or with- out mention of cardiovascular disease. They are shown in the table below. As shown below, of the 303,362 deaths in 1969 assigned to Chronic ischemic heart disease, a total of 235,807 were described in terms that led to their assicnment to numbers 412.1 and 412.3. The remaining 67,555 deaths were de- scribed in terms that led to their assignment to cardiovascular disease without mention of 1969 1971 Cause of death and Eighth Revision category number Number of Number of danths Percent deaths Percent Chronic ischemic heart disease .................eevuves 412 303,362 100.0 312,351 100.0 Chronic ischemic heart disease with or without cardiovascular disease with hypertensive DSBS ovvnes sins tii aT AS TRA AER RSS 412.1 19,884 6.6 19,832 6.3 Cardiovascular disease without mention of chronic ischemic heart disease with hypertensive UISOABH ovis mmmnsmims sis TRE ARS TT SATA FRAN ERAT 412.2 23,486 7.7 22,610 7.2 Chronic ischemic heart disease with or without cardiovascular disease without mention of hypertensive QiSESE ......uuinrisirnsvimassimsssssessons 412.3 215,923 71.2 221,404 70.9 Cardiovascular disease without mention of chronic ischemic heart disease without mention of WY RErtenSiVe THSEASE ......uisisrsriivatsisssassstuviisne 412.4 44,069 14.5 48,505 15.5 chronic ischemic heart disease, that is, to numbers 412.2 and 412.4. (It should be noted that subcategories 412.1 and 412.2 together are equivalent to ICDA category 412.0, and subcate- gories 412.3 and 412.4 together are equivalent to ICDA category 412.9.) For 1969 the death certificates for only 44,069 of these 67,555 deaths, however, did not mention hypertension. Because the Seventh Revision title Other myocardial degeneration (ICD No. 422) excluded conditions with men- tion of hypertensive disease, no more than these 44,069 deaths may be assumed to have been transferred from Other myocardial degeneration with arteriosclerosis (ICD No. 422.1). Also, as shown in the above table, the percentage of deaths from Chronic ischemic heart disease assigned to category number 412.4 increased from 14.5 for 1969 to 15.5 for 1971. Of deaths assigned to causes included in 412, the percentage assigned to category number 412.4 is greater for persons of races other than white than for white persons (table 3). The 1966 comparability study also showed an estimated 31,256 deaths classified to Ischemic heart disease (ICDA Nos. 410-413) by the Eighth Revision that had been assigned to Other hypertensive heart disease (ICD Nos. 440, 441, 443) by the Seventh Revision. The vast majority of these (an estimated 31,050) were classified to Chronic ischemic heart discase with hypertensive disease (ICDA No. 412.0). As previously stated, these differences in code assignments between the two revisions resulted primarily from changes in the classification of deaths involving arteriosclerotic heart disease jointly reported with hypertensive heart disease and of those involving other myocardial degener- ation with arteriosclerosis jointly reported with hypertensive disease. It was also found that out of a total of 10,008 deaths classified to Functional disease of heart (ICD No. 433) by the Seventh Revision, an estimated 2,805 were assigned to Chronic ischemic heart disease without mention of hypertensive disease (ICDA No. 412.9) by the Eighth Revision. These differences in code assignments resulted primarily from a change in the Seventh Revision procedure which gave priority to Functional disease of heart (ICD No. 433) over Other myocardial degeneration with arteriosclerosis (ICD No. 422.1). In the Eighth Revision functional heart diseases are included in category 427, Symptomatic heart disease; and category 412, Chronic ischemic heart disease, includes conditions classifiable to category 422.1 in the Seventh Revision. Also, a condition classifiable to category 412 is given priority over one classifiable to 427. The distribution by the Eighth Revision of the 414,101 deaths assigned by the Seventh Revision to Heart disease specified as involving 13 Cause of death and Eighth Revision Eytiniatea fests in category number ICD No. 420.1 Ischemic heart disease .................. 410413 411,678 With hypertensive disease (.0) .............. 37,073 Without mention of hypertensive disease (.9) .iuuvrirreerrnriiieeerereerraereinaes 374,605 Acute myocardial infarction.............. 410 360,794 With hypertensive disease (.0) ............... 31,985 Without mention of hypertensive 130852 LD) ...ervcerresmrrrensissssaiunainssssusnsan 328,809 Other acute or subacute forms of ischemic NEart diSease....ivirssiinsriesnserrnrenerenss 411 6,785 With hypertensive disease (.0) ............... 485 Without mention of hypertensive AISEASE 1D) «crrrrrcrrrerenrsssnsassamsassnssisiusting 6,300 Chronic ischemic heart disease .......... 412 43,857 With hypertensive disease (.0) ............... 4,361 Without mention of hypertensive disease (.9) .uoeiiiiirerriiiiieeire creer 39,496 ANGiNa PECLONS .evvvreiereeniaruninneiennenanad 413 1242 With hypertensive disease (.0) ............... 1242 Without mention of hypertensive Ai30888 LO) .....cxonivnminissisversnsmvrmssssssesive - Oher CAUSES cummins srcsssnsavuversarunnss Residual 2,423 This figure may be unreliable because there was only one sample death assigned to this cause. coronary arteries (ICD No. 420.1), as found by the 1966 comparability study, is shown above. About 88 percent of the 360,794 deaths assigned to Heart disease specified as involving coronary arteries (ICD No. 420.1) by the Seventh Revision were assigned to Acute myocardial infarction (ICDA No. 410) by the Eighth. Since category 410 includes, however, deaths involving acute myocardial infarction jointly reported with chronic ischemic heart disease, the number of these deaths associated with chronic ischemic heart disease cannot be ascertained. Adjusting the 1967 death rate of 289.7 deaths per 100,000 population for Arterio- sclerotic heart disease, including coronary disease (ICD No. 420), to the level it would have been if the 1967 deaths had been coded by the Eighth Revision raises it to 331.9 per 100,000 (289.7 multiplied by 1.1457). This adjusted rate is closer to the 1968 rate of 337.6 deaths per 100,000 population for Ischemic heart disease, which is based on coding deaths by the Eighth Revision. The 1971 death rate for this cause is 327.0 per 100,000 population. 14 Chronic disease of endocardium and other myocardial insufficiency (ICDA Nos. 424, 428).— Figures for 1966 and 1967 are shown in tables 1 and 2 for Seventh Revision categories 421, Chronic endocarditis not specified as rheu- matic, and 422, Other myocardial degeneration, which appear to be comparable to the above ICDA categories. As shown in appendix I and in the table below, however, only about 16 percent of the deaths assigned to ICD categories 421 and 422 by the Seventh Revision were assigned to ICDA categories 424 and 428 by the Eighth. This occurred primarily because terms included under Other myocardial degeneration with arteriosclerosis (ICD No. 422.1) were transferred by the Eighth Revision to Chronic ischemic heart disease (ICDA No. 412). The result was that an estimated 41,228 of the 41,935 deaths that were assigned to ICD No. 422.1 by the Seventh Revision were assigned to Ischemic heart disease (ICDA Nos. 410-413) by the Eighth Revision. Another important factor was the change between the Seventh and Eighth Revisions in the classification of diseases of the aortic valve. As stated under the discussion of Active rheumatic fever and chronic rheumatic heart disease (ICDA Nos. 390-398), diseases of the aortic valve not specified as rheumatic were included in category 421.1 in the Seventh Revision. In the Eighth Revision aortic valve diseases, unless they are qualified as nonrheumatic, are included under Chronic rheumatic heart disease. As a result of this change an estimated 1,922 deaths that were assigned to ICD No. 421.1 by the Seventh Revi- sion were assigned to ICDA No. 395.9 by the Eighth Revision. The 1967 death rate for Chronic endo- carditis not specified as rheumatic together with Other myocardial degeneration (ICD Nos. 421 and 422) was 26.6 per 100,000 population. Adjusting this rate to the level it would have been if 1967 deaths had been coded by the Eighth Revision reduces it to 4.8 per 100,000 (26.6 multiplied by 0.1823). This adjusted rate is close to the 1968 rate of 3.9 deaths per 100,000 population, based on coding deaths by the Eighth Revision. For 1971 the death rate for this cause dropped to 3.0 per 100,000 popula- tion. Seventh Revision Eighth Revision Estimated number of deaths in column (2) z Number of that went to Chronic A Rumer oF : 7 eaths in column (2) Cause of death and category number deaths disease of endocardium that went to in 1966 and other myocardial Other causss insufficiency (ICDA Nos. 424, 428) (1) (2) (3) (4) Al ICOUSES iovvusnvisisiinsesnmmmrsnmssssnsanmesennnn 001-E999 1,863,149 9,768 1,853,381 Nonrheumatic chronic endocarditis and other myocardial degeneration! IN 421,422 53,581 8,688 44,893 Comparable causes .........c..ccoueeeunen... 421,422.0,422.2 11,646 8,452 3,194 Chronic endocarditis not specified as PRBUINBILIC 1. vv aimminnsnivsis ini vs sii Te ante ssa nsmnnse 421 3,785 1,469 2,316 Of mitral valve, specified as nonrheumatic ....421.0 166 129 37 Of aortic valve, not specified as rheumatic ....421.1 2.311 283 2,028 Of other valves, not specified as rheumatic .....ceeuueeeeiiieeeeeieeeeeeeeevena 421.2421.4 1,308 1,057 251 Other myocardial degeneration without mention OF BrterioSCIBrosiS i ivisiviriscssssssssssnnes 422.0,422.2 7,861 6,983 878 Other causes SRA EIT Residual 1,851,503 1,316 1,850,187 Other myocardial degeneration with arterio- SCIOTOSIS civ. crimes nmi RT TR ATR Tas R brand 422.1 41,935 236 41,699 IThis is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. All other forms of heart disease (ICDA Nos. 420-423, 425-427, 429). Introduction of the Eighth Revision resulted in assigning about 19 percent fewer deaths to this group title than were assigned to the most nearly comparable title, Other diseases of heart (ICD Nos. 430-434), by the Seventh Revision. This gives a comparability ratio of 0.8104 (appendix I). As shown in the table below, an estimated 5,664 of the 10,008 deaths that had been assigned to Functional disease of heart (ICD No. 433) by the Seventh Revision were not assigned to All other forms of heart disease (ICDA Nos. 420-423, 425-427, 429) by the Eighth Revision. An estimated 2,805 of these 5,664 deaths were classified to Chronic ischemic heart disease with- out mention of hypertensive disease (ICDA No. 412.9), an estimated 1,498 to Arteriosclerosis (ICDA No. 440), and an estimated 272 to Cere- brovascular diseases (ICDA Nos. 430-438) by the Eighth Revision. These differences in assign- ments resulted from the following changes in coding procedures: 1) 2) Functional heart disease jointly reported with Other myocardial degeneration with arteriosclerosis.—In the Seventh Revision provision is made for coding Functional disease of heart (ICD No. 433) in preference to Other myocardial degeneration with arteriosclerosis (ICD No. 422.1). In the Eighth Revision this combination of diseases is classified to Chronic ischemic heart disease without mention of hypertensive disease. Functional heart disease jointly reported with arteriosclerosis.—The Seventh Revi- sion provides for the classification of this combination of diseases to functional heart disease. In the Eighth Revision no special preference is given to functional heart disease. Therefore, under certain circumstances assignment is to Arterio- sclerosis, e.g., when functional heart disease is reported as due to arteriosclerosis. 15 Seventh Revision Eighth Revision Estimated number of NUE Of deaths in column (2) Estimated number of c f death and b ow my ° that went to All deaths in column (2) 2use OF deathiand category number . pr other forms of heart that went to n disease (ICDA Nos. other causes 420-423,425-427,429) (1) (2) (3) (4) All CAUSES ..ociccnminsmimimsa stom smi smessvey 001-E999 1,863,149 25,156 1,837,993 Other diseases Of NEArt ..............c..coovvvereesnnes 430-434 31,042 23375 | 7,667 Acute and subacute endocarditis..............eeuvunnnnn. 430 757 654 103 Acute myocarditis and acute pericarditis, not specified as rheumatic .......cccevvvveriieiieeenenennnn. 431,432 1,065 892 173 Functional disease of heart 10,008 4,344 5,664 Other and unspecified diseases of heart ................ 434 19,212 17,485 1,727 OLNEY CAUSES. ..ooivisirissistssssisissamsamersests eosvsnns Residual 1,832,107 1,781 1,830,326 3) Functional heart disease jointly reported Cerebrovascular disease with arteriosclerosis and cerebrovascular disease.—In both the Seventh and Eighth Revisions provision is made for (a) Arrhythmia and cerebrovascular coding cerebrovascular diseases in pref- erence to arteriosclerosis. As noted above, however, the provision for coding functional heart disease in preference to arteriosclerosis was dropped in the Eighth Revision. As a result of this change the following examples are clas- sified to cardiac arrhythmia by the Seventh Revision and cerebrovascular disease by the Eighth Revision: I (a) Cardiac arrhythmia (b) Arteriosclerosis (c) 16 disease (b) General arteriosclerosis (c) The adjusted 1967 death rate for Other diseases of heart (ICD Nos. 430-434) is 12.8 deaths per 100,000 population. This is obtained by multiplying the recorded death rate of 15.8 by the comparability ratio of 0.8104. The death rates for 1968-71 are 14.0, 15.0, 15.9, and 16.6, respectively, per 100,000 population. 2. MALIGNANT NEOPLASMS All Malignant Neoplasms (ICDA Nos. 140-209) The introduction of the Eighth Revision resulted in an estimated net increase of less than 0.2 of 1 percent in the number of deaths assigned to this group of causes. Based on the sample of 1966 deaths coded by both the Seventh and Eighth Revisions, there were 2,342 deaths assigned to categories under this title by the Eighth Revision that were classified else- where by the Seventh. On the other hand, there were 1,816 deaths assigned to this title (ICD Seventh Revision Eighth Revision Estimated number of deaths in column (2) Number of that went to Malignant Estimated number of Cause of death and category number deaths neoplasms, including deaths in column {2) . neoplasms of lymphatic that went to in 1966 and hematopoietic other causes tissues (ICDA Nos. 140-209) (1) (2) (3) (4) AN CAUSES wusussmisrmsisimmmsssisiaisiismisenassanne 001-E999 1,863,149 304,262 1,558,887 Selected causes.................... 140-205,294,295,297-299 305,013 302,559 2,454 Malignant neoplasms, including neoplasms of lym- phatic and hematopoietic tissues! .ovoverennnnnn. 140-205 303,736 301,920 1,816 Malignant neoplasm of buccal cavity and PRAY NX. euiiicieieeeccceeeeeeeee ee crreeeeeeeee 140-148 6,800 6,744 56 150 143 7 1,629 1,614 15 Of other and unspecified parts of buccal CBVIRY curs issnnmms neanses pm Ra STE SSRI 142-144 2,224 2,205 19 Of PharyNX...ueeeeeeecvieeeeeecceeeeecieeee eevee 145-148 2,797 2,782 15 Malignant neoplasm of digestive organs and peritone- um, not specified as secondary...150-156A,157-159 95,079 94,559 520 Of esophagus 5,505 5,461 44 Of StOMACh....ccccieeireccreeeceee cece, 17,623 17,623 - Of small intestine, including duodenum 693 652 41 Of large intestine, except rectum..........ceeuuu.... 32,811 32,616 195 Cecum, appendix, and ascending colon 4,605 4,568 37 Transverse colon, including hepatic and splenic Lr ————. 153.1 1,472 1,397 75 Descending colon ............ 153.2 1,151 1,151 - Sigmoid colon .......ceeeevverreennnnn. 153.3 5,814 5,814 - Multiple parts of large intestine................... 163.7 78 78 - Large intestine (including colon), part UNSPBCIFIBU .uvinsvimsnissssiionisinsrrmmmmmnsnsnsns esse 18,155 18,117 38 Intestinal tract, part unspecified.. 1,536 1,491 45 Of rectum... ..ueeeeeeccciereeecneee cere, " 10,663 10,538 125 Of biliary passages and of liver (stated to be Primary Site) .....eueeeiveureeeriierereeesrneessessseneeesnns 155 6,584 6,543 41 LVEF. ccteieesrrrrrenrrreennnnaeea sears 155.0 2,112 2,112 - Other and multiple sites of biliary PASSAYCS vruusrnasvssnirsinssainmss piv cH RISS 155.1,155.8 4,472 4,431 41 Of liver, not stated whether primary or SBCONTBLY scrsivismsimmimrmrissrissthssmannnnmmunrrenssrsss 3,149 3,149 - Of pancreas 16,360 16,286 74 Of peritoneum and of unspecified digestive IP IBIYS coarse SEARS IAERS 158,159 1,691 1,691 - I This is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. 17 Continued from previous page. Seventh Revision Eighth Revision Estimated number of deaths in column (2) NUIBEF 6k that went to Malignant Estimated number of Cause of death and category number deaths neoplasms, including deaths in column (2) in 1966 neoplasms of lymphatic that went to and hematopoietic other causes tissues (ICDA Nos. 140-209) (1) (2) (3) (4) Malignant neoplasm of respiratory system, not specified as secondary ..........ccccecuvveereeennnnn.. 160-164 54,934 54,841 93 OF 1ary NX eee cuisine 161 2,623 2,623 - Of bronchus and trachea, and of lung specified Cd TT n., 162 20,913 20,891 22 Of lung, unspecified as to whether primary or SECONTBIY .cvovsiissssivivnrsisnsnnnnnnsrnnsrssassssssrsssercunns 30,565 30,519 46 Of other parts of respiratory system.......... 833 808 25 Malignant neoplasm of breast.........ccccceceeeueeennn.. 27,533 271.377 156 Malignant neoplasm of genital organs.. . 40,378 40,175 203 OF CErviX Uteri.uuenurereeeueeeieieeieeeeeeeeeeee ena 7,665 7614 51 Of other and unspecified parts of uterus ...172-174 5,731 5,697 34 Of ovary, fallopian tube; and broad ligament ...175 9,163 9,145 18 Of other and unspecified female genital OrGaNS..uiesiss 865 850 15 Of prostate 15,941 15,856 85 Of all other and unspecified male genital ONDINE: cousin srsirsvrssansnsmnissnnssnssenunssnessspasaresssns 1.013 1,013 - Malignant neoplasm of urinary organs. - 14,166 14,073 93 OF KidNeY weve eee eee 5,841 5,827 14 Of bladder and other urinary organs 8,325 8,246 79 Malignant neoplasm of other and unspecified SITS tuuereeeiiureeeeeeie eee careree ee eeaaees 156B,165,190-199 35,032 34,667 365 Of skin ..190,191 4,560 4,545 15 OF BY6 cvisnniummmminm insets tes pas fess bia isan 192 358 358 - Of brain and other parts of nervous system...... 193 7,355 7,152 203 Of thyroid gland............. 1,008 1,008 - Of bone................ 1,792 1,792 - Of connective tissue 1,318 1,318 - Of other specified sites, not stated to be secondary 1,422 1,408 14 Of unspecified sites .... 1568B,165,198,199B 17,219 17,086 133 Leukemia and aleukemia............c.cooeevuveerenueennn.. 204 14,012 13,926 86 Lymphosarcoma and other neoplasms of lymphatic and hematopoietic tissues.................... 200-203,205 15,802 15,558 244 Lymphosarcoma and reticulosarcoma ..200 7,563 7,548 15 Hodgkin's disease 3,412 3,398 14 Other neoplasms of lymphatic and hematopoietic TISSUE... ....orvin rrp suisssm at 202,203,205 4,827 4,612 215 Other diseases of blood and blood-forming OFGONS ci vussniniormmimmss inane iassi stensos 294,295,297-299 1,277 639 638 Other CauSes...........eccueeeeeeeeeeeeeeeeeeeeeeeeenen Residual 1,558,136 1,703 1,556,433 Nos. 140-205) by the Seventh Revision that were classified elsewhere by the Eighth. Among the 2,342 deaths that were assigned to Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tissues, by the Eighth Revision but to other categories by the 18 Seventh were about 639 deaths that were classi- fied under Other diseases of blood and blood- forming organs (ICD Nos. 294, 295, 297-299) by the Seventh Revision. A change in the classi- fication of polycythemia (vera) from ICD No. 294 in the Seventh Revision to ICDA No. 208 in the Eighth Revision accounted for most of these differences in assignments. Also included in the 2,342 figure were an additional 382 deaths that were classified under Diseases of the blood and blood-forming organs (ICD Nos. 290-299) by the Seventh Revision. These differences in assignments were due primarily to the transfer of terms from Leuko- erythroblastic anemia (ICD No. 292.3) in the Seventh Revision to Myelofibrosis (ICDA No. 209) in the Eighth Revision. Major Components of Malignant Neoplasms Malignant neoplasms of buccal cavity and pharynx (ICDA Nos. 140-149).— There was a net increase of 405 deaths assigned to this group of causes by the Eighth Revision. Of the 567 deaths classified under the above title by the Eighth Revision and under other titles by the Seventh, the largest component was assigned by the Seventh Revision to Malignant neoplasm of unspecified sites (ICD Nos. 1568, 165, 198, 199B). More specifically, it is esti- mated that 415 of these deaths were assigned by the Seventh Revision to Malignant neoplasm with primary site not indicated (category number 199B). These differences in assignments resulted primarily from a change in the procedure for classifying malignant neoplasm of multiple sites with no indication as to which was the primary site. The rules in effect with the two revisions are given below. It should be noted, however, that these rules do not apply to malignant neo- plasm of liver or lymph nodes without specifica- tion as primary jointly reported with another site. Under such circumstances the neoplasm of liver or lymph nodes is assumed to be secondary. Seventh Revision.—“If there is no indication as to which was the primary site (for example, if sites are entered on the same line or in a sequence which does not point to one as the primary), assignment should be to malignant neoplasm of multiple sites (199), except where the classification provides specifically for multiple sites within three- digit categories (140.8, 141.8, etc.).”d Eighth Revision.—“If there is no indication as to which was the primary site (for example, if sites are entered on the same line or in a sequence which does not point to one as the primary), prefer a defined site to an ill-defined site in category 195 and of two or more defined sites prefer the first men- tioned.”6 The effect of this change in coding procedure can be illustrated by the following example, which was classified to Malignant neoplasm with primary site not indicated (category number Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of that went to Malignant deaths in column (2) Cause of death and category number deaths neoplasns of buccal that went to in 1968 cavity and pharynx other causes (ICDA Nos. 140-149) (1) (3) (4) All CAUSES .evuneeieeeeeiiiieeeeiieeeee eee eines 001-E999 1,863,149 7,205 1,855,944 6,800 6,638 162 150 126 24 1,629 1,603 26 Of other and unspecified parts of buccal CAVITY teviriiiiiieieeeee eee en esas sees ree eee an aa eens 142-144 2,224 2,194 30 Of PhAryNX..oocoeuuiiiineeeee eee cee eee 145-148 2,797 2,715 82 Other Causes ums ios iiss ior Residual 1,856,349 567 1,855,782 19 Table C. Comparable category numbers for Malignant neoplasms of digestive peritoneum according to the Eighth and Seventh Revisions and comparability these causes of death: United States, 1966 organs and ratios for Eighth Seventh Compara - . Revision Revision bility Bighth Revision ICDA numbers ICD numbers ratiol category title 1) (2) 3) Malignant neoplasms of digestive organs and peritoneum=----------e-cmmemccmmeonna- 150-159 | 150-156A,157-159 0.9660 Of esophagus=---==--ccccccmmcccm meme 150 150 0.9909 Of stomach=------ccsmcacmcc mcm c ccc cc ccee 151 151 1.0165 Of small intestine, including duodenum=--- 152 152 0.9806 Of large intestine, except rectum==------ 153 153 0.9994 Of rectum and rectosigmoid junction=------ 154 154 1.0012 Of liver and intrahepatic bile ducts, specified as primary----------=cccec-n-o- 155 155.0 0.9814 Of gallbladder and bile ducts=----------- 156 155,1,155.8 0.9860 Of pancreas------====-c----c-cm-moccooooo- 157 157 1.0019 Of peritoneum, retroperitoneal tissue, and unspecified digestive organs-------- 158,159 158,159 0.8639 Ratio of deaths assigned according to the Eighth Revision cording to the Seventh Revision. 199B), by the Seventh Revision and to Malig- nant neoplasm of nasopharynx (ICDA No. 147) by the Eighth Revision. I (a) Carcinoma of nasopharynx and larynx (b) (c) Notes for use in primary mortality coding for the Eighth Revision are shown in appendix V1. Estimates based on 1968 data that are con- sistent with the estimated number of 1966 deaths that were assigned by the Seventh Revi- sion to category number 199B and by the Eighth Revision to ICDA Nos. 140-149 are presented in appendix VII. The rise in the death rate for Malignant neo- plasms of buccal cavity and pharynx between 1967 and 1968 from 3.4 to 3.6 deaths per 100,000 population is attributable to the above- mentioned change in coding procedure between the Seventh and Eighth Revisions. Applying the comparability ratio of 1.0596 to the 1967 death rate for this group of causes raises it to 3.6, the same as that for 1968. For 1969 the rate was 3.7 deaths per 100,000 population. 20 to deaths assigned ac- Malignant neoplasms of digestive organs and peritoneum (ICDA Nos. 150-159).—The intro- duction of the Eighth Revision resulted in the assignment of about 3.4 percent fewer deaths to categories under the above title than were assigned to categories under the comparable Seventh Revision title. The comparability ratios for this group of causes and for the nine subcate- gories into which it has been divided are shown in table C. As shown in the table below, an estimated 4,415 of the 95,079 deaths that were assigned to the comparable Seventh Revision title Malignant neoplasm of digestive organs and peritoneum, not specified as secondary (ICD Nos. 150-156A, 157-159), were not assigned to Malignant neoplasms of digestive organs and peritoneum (ICDA Nos. 150-159) by the Eighth Revision. Conversely, an estimated 1,184 deaths assigned to this group by the Eighth Revision were classi- fied elsewhere by the Seventh. Among the 4,415 deaths classified by the Seventh Revision under Malignant neoplasm of digestive organs and peritoneum, not specified as secondary, but classified elsewhere by the Eighth Revision were an estimated 3,149 deaths that were assigned to category number 156A by the Seventh Revision and to ICDA Nos. 194-199 by the Eighth Revision (table D). Most of these Seventh Revision Eighth Revision Estimated number of NumGer of deaths in column (2) Estimated number of Cause oF degthiand category number deaths that went to Malignant deaths in column (2) in 1966 neoplasms of digestive that went to organs and peritoneum other causes (ICDA Nos. 150-159) (1) (3) (4) A COUBES ..ciasies siiniinninmmmmsmresssssrsrsssrnnnnnnn 001-E999 1,863,149 91,848 1,771,301 Malignant neoplasm of digestive organs and peritoneum, not specified as secondary ............. 150-156A,157-159 95,079 90,664 4,415 Of esophagus ....150 5,505 5,461 44 OF STOMACH ..coicisnnnissvssiminivsssrinsanansunnstnsarsesassasssssss 151 17,623 17,547 76 Of small intestine, including duodenum ............... 152 693 652 41 Of large intestine, except rectum ...163 32,811 32,418 393 Cecum, appendix, and ascending colon ...153.0 4,605 4,494 111 Transverse colon, including hepatic and splenic FIEXUI®S coerce ieee ee trees 1,472 1,397 75 Descending colon 1.151 1,151 - SigMOTId COIN icinriivimmmmsrsiven rose vsssnsessssess 5,814 5,774 40 Multiple parts of large intestine ” 78 78 L Large intestine (including colon), part unspeci- FROG au vivrinssnmrnmmmssnsnrosmons dims RAAB ERE FRA 18,155 18,078 77 Intestinal tract, part unspecified. 1,536 1,446 90 OF PRCHUIN cs sesmsinismrimsvinssmnmsimmmessnssennnpresmmmssannnsann 10,663 10,455 208 Of biliary passages and of liver (stated to be Primary Site) .......cccueeeeeeveureiiiiiereeeeesiieeeeeeessnns 155 6,584 6,398 186 I — 155.0 2,112 2,008 104 Other and multiple sites of biliary PBISATRE «ouusirsvnissumsssssimnssmmmenssnenmpsesmusssns 155.1,155.8 4,472 4,390 82 Of liver not stated whether primary or secondary?.... ...156A 3,149 z 3,149 OF PONGIOAS cu.uvismsmmimmivsivsvivivesenesrsimanmrssssantosss 157 16,360 16,248 112 Of peritoneum and of unspecified digestive OF OBIS rims aE A TR AT NT rare a RR. 158,159 1,691 1,485 206 Other CAUSES .......cceeervrurnrrinrnrrrneereeeeesessesessnanns Residual 1,768,070 1,184 1,766,886 IThere were no deaths in the sample assigned to the Seventh Revision title Malignant neoplasm of multiple parts of large intestine (ICD No. 153.7). NCHS nosologists state, however, that these deaths would be distributed by the Eighth Revision over the four-digit categories under Malignant neoplasm of large intestine, except rectum (ICDA No. 153). 2In computing a provisional comparability ratio (0.9991) for Malignant neoplasm of digestive organs and peritoneum (ICDA Nos. 150-159), the Seventh Revision title Malignant neoplasm of liver not stated whether primary or secondary (ICD No. 156A) was not included in the set of titles most nearly comparable to Malignant neoplasm of digestive organs and peritoneum. Category number 156A was thus excluded because most deaths assigned to this category by the Seventh Revision were removed by the Eighth Revision from under the title classifying the system of the body attacked by the neoplasm (in this case, from under digestive organs and peritoneum) and placed under a subsection of Section II, “Neoplasms”: Malignant neoplasm of other and unspecified sites (ICDA Nos. 190-199). More specifically, these deaths were transferred to the new Eighth Revision title: Malignant neoplasms of liver, unspecified (ICDA No. 197.8). differences in assignments were due to the transfer of malignant neoplasm of liver, not stated whether primary or secondary, from cate- gory number 156A in the Seventh Revision to ICDA No. 197.8 in the Eighth. A satisfactory indication of how many of the 3,149 deaths were assigned to ICDA No. 197.8 and how many were assigned to ICDA Nos. 194-199 other than 197.8 may be obtained from the distribution of deaths occurring in 1968 in the three- and four- digit categories of ICDA Nos. 194-199. This distribution shows that the 2,981 deaths assigned to ICDA No. 197.8 constituted 14.35 percent of the total number (20,768 deaths) assigned to ICDA Nos. 194-199. It may be postulated, therefore, that approximately 14.35 percent (or 2,808 deaths) of the estimated 19,570 deaths in 1966 assigned by the Eighth Revision to ICDA Nos. 194-199 were assigned to 197.8. The difference between 3,149 and 2,808 21 Table D. Number of deaths liver, intrahepatic in 1966 coded by the bile ducts, gallbladder, or extrahepatic bile ducts is Seventh Revision to each title in which malignant included in the title, whether the neoplasm of title was for neoplasm specified as primary or secondary or was unqualified, distributed by the titles to which assigned by the Eighth Revision, and number of deaths in 1966 and 1968 assigned to these Eighth Revision titles: United States Seventh Revision Eighth Revision Estimated number of deaths in col. (2) Number that went to ICDA Nos. 155, 156, and Estimated of 194-199 (primarily to 197.7 and 197.8) number of deaths Number of deaths in in 1968 ICD No. deaths in col. (2) that Cause of death 1966 Number | that went | went to of to other ICDA ICDA No. deaths causes Nos. in in 1966 col. (3) (1) (2) 3) (4) (5) (6) All causeS========== 001-E999 1,863,149 155,156,197.7,197,8 10,982 1,852,168 11,251 Specified causes-------- 155,156 11,083 155,156,197.7,197.8 10,898 185 11,251 Malignant neoplasm of biliary passages and of liver (stated to be primary site)----- 155 6,584 155,156 6,398 185 6,805 Lives am 155.0 2,112 SE | 2,008 104 1 Gallbladder and extrahepatic gall ducts, including 151.1 41 ampulla of Vater--- 155.1 4,471 156 4,349 81 4,609 156.0 vere winin 2.777 156.1 vee es 1,204 156.2 vee cee 397 156.9 es “es 231 Multiple sites------ 155.8 23 2155 or 156 71 2. ven Malignant neoplasm of liver (secondary and unspecified)---=----- 156 4,499 | 194-199, primarily to 197.7,197.8 4,499 - 4,446 Malignant neoplasm of liver not stated whether primary or secondary-========= 156A 3,149 194-199, primarily to 197.8 3,149 - 2,981 Malignant neoplasm of liver, secondary- 156B 1,350 194-199, primarily to 197.7 1,350 - 1,465 Other causes--==--=------ Residual | 1,852,066 155,156 84 | 1,851,982 6,805 IThe titles corresponding to these ICDA numbers are as follows: Malignant neoplasms of liver specified as pri- mary and intrahepatic bile ducts (ICDA No. 155); Malignant neoplasms of gallbladder and bile ducts (ICDA No. 156); Mali No. 197.7); and Malignant neoplasms of liver, unspecified (ICDA No. 197.8). 2According to the coding rules or 156, 22 in effect with the Eighth Revision, this death was ant neoplasms of other sites (ICDA Nos. 194-199); Malignant neoplasms of liver, specified as secondary (ICDA assigned to either No. 155 (341) may be taken as the minimum estimate of the number of deaths assigned to 156A hy the Seventh Revision that were assigned by the ‘Eighth Revision to ICDA Nos. 194-199 other than 197.8. This estimate of 341 deaths is qualified as ‘minimum because the number of 1968 deaths assigned to 197.8 (which con- stituted 14.35 percent of the total number in ICDA Nos. 194-199) may have included some deaths assigned by the Seventh Revision to cate- gories other than 156A. Among the estimated 341 deaths assigned to 156A by the Seventh Revision and to ICDA Nos. 194-199 other than 197.8 by the Eighth were those involving malignant neoplasm of liver, not stated whether primary or secondary, jointly reported with malignant neoplasm of abdomen. In the Seventh Revision malignant neoplasm of the abdomen and intra-abdominal cancer were included under Malignant neoplasm with primary site not indicated (category number 199B), and provision was made for coding Malignant neoplasm of liver, not stated whether primary or secondary (156A), in pref- erence to a condition classifiable to 199B. In the Eighth Revision a separate subcategory is provided for Abdomen, intra-abdominal cancer (ICDA No. 195.0) under Malignant neoplasms of ill-defined sites (ICDA No. 195), and provision is made for coding malignant neoplasm of an ill- defined site in preference to malignant neoplasm of liver, not stated whether primary or second- ary. As a result of this change in coding proce- dure the following example would be assigned to category number 156A by the Seventh Revision and to ICDA No. 195.0 by the Eighth Revision. I (a) Carcinoma of liver [7th, category no. 156A; 8th, ICDA No. 197.8] (b) Abdominal carcinoma [7th, category no. 199B; 8th, ICDA No. 195.0] In the relatively small sample of 1966 deaths coded by both revisions none of the deaths assigned by the Seventh Revision to category number 156A were assigned to ICDA numbers other than 194-199 by the Eighth Revision. It is known from other sources, however, that some deaths assigned to 156A by the Seventh Revi- sion may have been assigned to numbers other than 194-199 by the Eighth Revision. For example, in 1968 a review was made of 103 death certificates on which cancer of the lung was mentioned. Among these certificates was the following: I (a) Cancer of liver ~~ [7th, category no. 156A; 8th, ICDA No. 197.8] (b) Metastatic cancer of lung [7th, ICD No. 165; 8th, ICDA No. 162.1] Because of a charge in the interpretation of “metastatic” neoplasm of lung, the above record would be coded to Malignant neoplasm of liver, not stated whether primary or secondary (cate- gory number 156A), by the Seventh Revision and to Malignant neoplasms of bronchus and lung (ICDA No. 162.1) by the Eighth. The coding rules in use with the Seventh Revision provided for the classification of malignant neo- plasm of lung qualified as “metastatic” to secondary neoplasm of lung. Malignant neo- plasm of a site not specified as secondary (including liver) was coded in preference to a secondary neoplasm. According to the Eighth Revision rules, malignant neoplasm of the lung qualified as “metastatic” is not considered to be specified as secondary. A malignant neoplasm of the liver or lymph nodes without specification as primary is assumed to be secondary when it is jointly reported with malignant neoplasm of another site that is not specified as secondary. In addition to the deaths coded to category number 156A by the Seventh Revision and to 197.8 by the Eighth Revision, the comparability study also showed that some deaths coded to Seventh Revision ICD No. 155.0 were assigned to 197.8 by the Eighth Revision. It is likely that most of the 104 deaths assigned by the Seventh Revision to Malignant neoplasm of liver (stated to be primary site) (ICD No. 155.0) and not reassigned by the Eighth Revision to primary malignant neoplasm of liver were coded to Malignant neoplasms of liver, unspecified (ICDA No. 197.8) (table D). The 1955 multiple-cause study shows that a significant number of deaths attributable to Malignant neoplasm of biliary passages and of liver (stated to be primary site) (ICD No. 155) were associated with cirrhosis of liver. According to the coding rule in use with 23 the Seventh Revision, cancer of liver, unqual- ified, indicated by the certifier to be due to cirrhosis of liver was coded to primary cancer of liver (ICD No. 155.0). This rule, was not incorporated in the coding instructions for use with the Eighth Revision in time to be appli- cable to any data year prior to 1975. Prior to this time deaths certified in the above-described manner were coded by the Eighth Revision to Malignant neoplasms of liver, unspecified (ICDA No. 197.8). The largest component of the estimated 1,184 deaths assigned to Malignant neoplasms of digestive organs and peritoneum by the Eighth Revision but coded elsewhere by the Seventh was an estimated 727 deaths assigned by the Seventh Revision to Malignant neoplasm, primary site not indicated (category number 199B). These differences in assignments resulted primarily from a change in the procedure for classifying malignant neoplasm of multiple sites with no indication as to which was the primary site. For example, a report of “cancer of stomach and pancreas” was coded to Malignant neoplasm with primary site not indicated (cate- gory number 199B) by the Seventh Revision and to the first-mentioned site, cancer of stomach (ICDA No. 151.9), by the Eighth Revision. A more detailed discussion pertaining to the classi- fication of malignant neoplasms of multiple sites appears above under Malignant neoplasms of buccal cavity and pharynx (ICDA Nos. 140-149). Estimates based on 1968 data that are con- sistent with the estimated number of 1966 deaths assigned by the Seventh Revision to 199B and by the Eighth Revision to ICDA Nos. 150-159 (727 deaths) are presented in appendix VIL In 1968 the death rate for Malignant neo- plasms of digestive organs and peritoneum was 46.8 per 100,000 population (table 1). In 1967 the death rate for this group of causes was 48.2 per 100,000 population. Based on the dual coding of 1966 data, however, there were 91,848 deaths assigned to ICDA Nos. 150-159 by the Eighth Revision and 95,079 deaths assigned to ICD Nos. 150-156A, 157-159 by the Seventh Revision. The comparability ratio for these causes, 0.9660, is obtained by dividing 91,848 by 95,079. Adjusting the 1967 death rate to the level it would have been if the 1967 deaths had been coded by the Eighth Revision (48.2 multiplied by 0.9660) lowers it to 46.6 deaths per 100,000 population. Malignant neoplasms of respiratory system (ICDA Nos. 160-163).—The net effect of changes introduced with the Eighth Revision on the number of deaths classified to Malignant neoplasms of respiratory system was an increase from 54,934 to 56,668. The comparability ratio is 1.0316 (appendix I). Seventh Revision Eighth Revision Estimated number of Number of deaths in column (2) Estimated number of that went to Malignant deaths in column (2) Cause of death and category number deaths neoplasms of that went 10 in 1966 respiratory system other causes (ICDA Nos. 160-163) (1) (3) (4) AN) COUSES osm sivs 710 TERR TRAE TR ERAS 001-E999 1,863,149 56,668 1,806,481 Malignant neoplasm of respiratory system, not specified as secondary ........cccccevvvviiieninieeeenens 160-164 54,934 54,584 350 OF IBY NK cvisinsirin inser savas TEV AE Ey Yams ET et esa heen ies 161 2,623 2,597 26 Of bronchus and trachea, and of lung specified BE PPHNIALY currre ruvsaiuiiinnnits bedi sEisin Emmis ass bana pas 162 20,913 20,801 112 Of lung, unspecified as to whether primary or SBCEORABTY. ccrverrrrrrrsmrmnrenenmmssnsmsssssnrsnnnssd sts ea EE EII EY 163 30,565 30,403 162 Of other parts of respiratory system.............. 160,164 833 783 50 Other CAUSES ......evveueeeeiieineererneeeesnieeerasaeeaeenes Residual 1,808,215 2,084 1,806,131 24 As shown in the table above, an estimated 2,084 deaths classified under the above title by the Eighth Revision were classified to causes other than Malignant neoplasm of respiratory system, not specified as secondary (ICD Nos. 160-164), the comparable Seventh Revision title. Included in the 2,084 figure were an esti- mated 1,572 deaths. The majority of these were assigned to Malignant neoplasm of thoracic organs (secondary) (ICD No. 165) by the Seventh Revision. A limited number of these 1,572 deaths probably were assigned to Malig- nant neoplasm with primary site not indicated (category number 199B) by the Seventh Revi- sion (appendix VII). A change in the classifica- tion of “metastatic” neoplasm of the lung from Malignant neoplasm of thoracic organs (second- ary) (ICD No. 165) in the Seventh Revision to Malignant neoplasms of bronchus and lung (ICDA No. 162.1) in the Eighth Revision accounted for most of these differences. The remaining differences resulted from a change in the procedure for coding malignant neoplasm of multiple sites with no indication as to primary site. For example, a report of “carcinoma of lung and bladder” was coded to Malignant neoplasm with primary site not indicated (cate- gory number 199B) by the Seventh Revision and to the first-mentioned site, carcinoma of lung (coded to ICDA No. 162.1), by the Eighth Revi- sion. (A more detailed discussion pertaining to the classification of malignant neoplasm of multiple sites appears under Malignant neo- plasms of buccal cavity and pharynx (ICDA Nos. 140-149), page 19.) Also among the 2,084 deaths transferred to Malignant neoplasms of respiratory system by the Eighth Revision were an estimated 56 deaths assigned by the Seventh Revision to Benign neo- plasm of respiratory system. These differences resulted from a change in the classification of “mesothelioma of the pleura” from Benign neoplasm of respiratory system (ICD No. 212) in the Seventh Revision to Malignant neoplasms of pleura (ICDA No. 163.0) in the Eighth Revi- sion. The third group among the 2,084 transferred deaths comprised an estimated 53 deaths that had been assigned to Tuberculosis of respiratory system (ICD Nos. 001-008) by the Seventh Revi- sion. These differences in assignments occurred because the Seventh Revision rules which under certain circumstances resulted in coding an infectious or parasitic disease in preference to other reported conditions were changed in the Eighth Revision. These Seventh Revision rules applied when two or more conditions, one of which was an infectious or parasitic disease, were entered on the death certificate in such a way that none could be regarded as the under- lying cause. The Eighth Revision rules lead to the selection of the first-mentioned condition when causes of death are entered on the death certificate in this manner. A death attributable to “lung cancer and tuberculosis of lung,” for example, was assigned to tuberculosis of lung by the Seventh Revision rules and to cancer of lung by the Eighth Revision rules. The assignment to other categories of about 350 deaths that had been assigned to Malignant neoplasm of respiratory system, not specified as secondary, by the Seventh Revision partly offset the gain of about 2,084 deaths to the Eighth Revision title Malignant neoplasms of respira- tory system. An estimated 184 of these deaths were assigned to Malignant neoplasms of abdomen, intra-abdominal cancer (ICDA No. 195.0), and to Malignant neoplasms of pelvis, pelvic viscera, rectovaginal septum (ICDA No. 195.1), by the Eighth Revision. These differ- ences in assignments resulted from a change in the procedure for coding malignant neoplasms of the abdomen and pelvis. In the Seventh Revi- sion these neoplasms were included under Malig- nant neoplasm with primary site not indicated (category number 199B), and provision was made for coding malignant neoplasm of a specified site in preference to a condition classi- fiable to 199B. In the Eighth Revision malignant neoplasms of the abdomen and pelvis are included under Malignant neoplasms of ill- defined sites. When malignancy of another site is indicated to be due to malignant neoplasm of an ill-defined site, assignment is to the ill-defined site. As a result of this change in coding pro- cedure the following example would be assigned to carcinoma of lung (coded to ICD No. 163) by 25 the Seventh Revision and to abdominal carcinoma (coded to ICDA No. 195.0) by the Eighth Revision. I (a) Carcinoma of lung (b) Abdominal carcinoma (c) Also among the 350 deaths not reassigned to malignant neoplasm of respiratory system were an estimated 74 deaths coded by the Eighth Revision to malignant neoplasm of other sites. These transfers probably resulted from a change in the interpretation of the rules used to deter- mine whether there was more than one primary neoplasm. When the Seventh Revision was in use, a malignant neoplasm of a site not specified as either primary or secondary was assumed to be secondary when jointly reported with a primary malignant neoplasm, regardless of the position of the causes on the medical certifica- tion form. When the Eighth Revision went into effect, this interpretation was changed. The fact that a malignant neoplasm of one site is specified as primary is not considered evidence that neoplasms of other reported sites are secondary. As a result of this change in coding procedure, the following example was assigned to Malignant neoplasm of larynx (ICD No. 161) by the Seventh Revision and to Malignant neo- plasms of breast (ICDA No. 174) by the Eighth. I (a) Cancer of breast (b) (c) II Primary cancer of larynx Applying the comparability ratio of 1.0316 to the 1967 death rate of 29.4 deaths per 100,000 population for Malignant neoplasm of respiratory system, not specified as secondary (ICD Nos. 160-164), raises this rate to 30.3 deaths per 100,000 population. This is the level the 1967 death rate would have been if deaths for that year had been classified by the Eighth Revision. This adjustment of the 1967 rate reduces the percentage rise between 1967 and 1968 from 8.16 percent to 4.95 percent. For 1969 the death rate for Malignant neoplasms of respiratory system (ICDA Nos. 160-163) was 32.7 deaths per 100,000 population. Malignant neoplasms of breast (ICDA No. 174).—The introduction of the Eighth Revision caused no serious break in the comparability of mortality statistics for this cause. The compara- bility ratio is 0.9913 (appendix I). As shown in the table below there were an estimated 312 deaths in 1966 that were assigned by the Seventh Revision to Malignant neoplasm of breast (ICD No. 170) and to other causes by the Eighth Revision. Approximately one-half of these deaths were assigned by the Eighth Revi- sion to malignant neoplasm of other sites, mainly to sites of the respiratory system and to abdomen and pelvis. Some of these differences in assignments occurred because of a change in the classification of metastatic neoplasm of the lung from Malignant neoplasm of thoracic organs (secondary) (ICD No. 165) in the Seventh Revision to Malignant neoplasms of bronchus and lung (ICDA No. 162.1) in the Eighth Revision. As a result of this change the Seventh Revision Eighth Revision Estimated number of Estimated number of Number of deaths in column (2) ; Cause of death and category number deaths that went to Malignant deaths in column (2) in 1966 neoplasms of breast Other Causes (ICDA No. 174) (1) (3) (4) All CAUSES ..evvveeeeeeeeeeiiiiieiiieesiiieeeereeeeneaes 001-E999 1,863,149 27,293 1,835,856 Malignant neoplasm of breast...............ccecevuueerrnnnnns 170 27,533 22.21 312 OHIVEr COUSES vii vvunisesvisens viminvsivnssssnamaaiae ssa Residual 1,835,616 72 1,835,544 26 following example would be classified to car- cinoma of the breast by the Seventh Revision and to carcinoma of the lung by the Eighth. I (a) Carcinoma of breast (b) Metastatic carcinoma of lung (c) The previously mentioned change in the pro- cedure for coding malignant neoplasms of the abdomen and pelvis resulted in some deaths being assigned to malignant neoplasm of the breast by the Seventh Revision and to malignant neoplasm of the abdomen or pelvis by the Eighth Revision. Malignant neoplasms of the abdomen and pelvis were included under Malig- nant neoplasm with primary site not indicated (category number 199B) in the Seventh Revi- sion, and provision was made for coding malig- nant neoplasm of a specified site in preference to a condition classifiable to 199B. In the Eighth Revision malignant neoplasms of abdomen and pelvis are included under Malignant neoplasms of ill-defined sites, and the rules provide for the assignment of malignant neoplasm of a specific site. which is reported as due to malignant neo- plasm of an ill-defined site to the ill-defined site. Thus the following example would have been assigned to carcinoma of the breast (under ICD No. 170) by the Seventh Revision and to carcino- matosis of abdomen (under ICDA No. 195.0) by the Eighth Revision. I (a) Carcinoma of breast (b) Carcinomatosis of abdomen (c) Malignant neoplasms of genital organs (ICDA Nos. 180-187).—As shown in the table below, there was no serious break in the con- tinuity of mortality statistics for Malignant neo- plasms of genital organs as a result of the intro- duction of the Eighth Revision. The comparability ratio for this cause is 1.0034 (appendix I). Inasmuch as the rates in table 1 are based on the total population (male and female) they are of limited value. Data from the comparability study of deaths occurring in 1966 are given below. The comparability ratio for malignant neoplasms of female genital organs is 0.9963. The corresponding comparability ratio for malig- nant neoplasms of the male genital organs is 1.0133. Malignant neoplasms of urinary organs (ICDA Nos. 188, 189).—The comparability ratio for causes classified under this title is 1.0171 (appendix I). As shown in the table, page 29, Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of that went to Malignant deaths in column (2) Cause of death and category number deaths neoplasms of genital that went to in 1966 organs (ICDA Nos. other causes 180-187) (1) (3) (4) All CAUSES coerce, 001-E999 1,863,149 40,516 1,822,633 Malignant neoplasm of genital organs.............. 171-179 40,378 39,969 409 OF COTVIX WB vcuveumas imi msnamiiis visi 171 7,665 7,562 103 Of other and unspecified parts of uterus....... 172-174 5,731 5,680 51 Of ovary, fallopian tube, and broad ligament ....... 175 9,163 9,092 al Of other and unspecified female genital organs........... 865 835 30 Of prostate 15,941 15,806 135 Of all other and unspecified male genital OFQGBINIS coos iiss aR SEARS Enns sass an sass wR 178,179 1,013 994 19 Other CAUSES .......cceeeeeeeeieeeeeeeeeeeeeeereaaaeenn Residual 1,822,771 547 1,822,224 27 Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of i : that went to Malignant deaths in column (2) Cause of death and category number Sane neoplasms of female Hot Want1o n genital organs (ICDA other causes Nos. 180-184) (1) (2) (3) (4) All CAUSES ..vvvveerereeereeeineeieienaieninnneeneeeaes 001-E999 1,863,149 23.337 1,839,812 Malignant neoplasm of female genital organs...171-176 23,424 23,169 255 OF COIVIX UREBTE ciovuiimmiiviisisminimsamis srs mira ss seni 171 7,665 7,562 103 Of other and unspecified parts of uterus....... 172-174 5,731 5,680 51 Of ovary, fallopian tube, and broad ligament ....... 175 9,163 9,092 7 Of other and unspecified female genital OFGANS. cu eeeiiiireereeeeeeaeenssasnassasasanaeeaeaeeaeeseeaans 176 865 835 30 OTHEE COUBES ....onniresrorsnusmmenimnisbnbssirmmiivsss sEaT 40 Residual 1,839,725 168 1,839,657 Seventh Revision Eighth Revision Estimated number of Number of deaths in column (2) Estimated number of c t death and cab b um > 0 that went to Malignant deaths in column (2) ause:OF Leal anc Category: humber . Ra neoplasms of male that went to in genital organs (ICDA other causes Nos. 185-187) (1) (2) (3) (4) AN CAUSES ..uviersssssrssirinsssssersansrserervenrissivin 001-E999 1,863,149 17,179 1,845,970 Malignant neoplasm of male genital organs...... 177-179 16,954 16,800 154 Of PLOBTBIE vovvvinvirremras staan at REE REA IN TATA RR SR 177 15,941 15,806 135 Of all other and unspecified male genital OFGANS. .eeeeeeeeieriunnrnrerereereeeseeeesesssesssessssnsssssnnns 178,179 1,013 994 19 Other causes .... Residual 1,846,195 379 1,845,816 there were 14,166 deaths assigned to this title by the Seventh Revision and 14,408 deaths as- signed to it by the Eighth Revision. The largest component of the estimated 421 deaths classified under Malignant neoplasms of urinary organs by only the Eighth Revision included about 193 deaths that were classified to Malignant neoplasm with primary site not indicated (category number 199B) by the Seventh Revision. These transfers resulted from a change in the procedure for classifying malig- nant neoplasm of multiple sites with no indica- 28 tion as to which was the primary site. For example, a report of “cancer of kidney and lung” was coded to Malignant neoplasm with primary site not indicated (category number 199B) by the Seventh Revision and to the first- mentioned site, cancer of the kidney (coded to ICDA No. 189.0), by the Eighth Revision. (A more detailed discussion pertaining to the classi- fication of malignant neoplasm of multiple sites appears under Malignant neoplasms of buccal cavity and pharynx (ICDA Nos. 140-149), page 19.) Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of i Cause of death and category number deaths that went to Malignant deaths in column 2) in 1966 neoplasms of urinary that went to n organs (ICDA Nos. other causes 188,189) (1) (2) (3) (4) All CAUSES eee eee eee 001-E999 1,863,149 Co 14,408 1,848,741 Malignant neoplasm of urinary organs............. 180,181 14,166 13,987 179 OF KTANBY un isisnsnsnsvuninnsis 254s oa mov ns EA SA SACO Re 180 5,841 5,787 54 Of bladder and other urinary organs..................... 181 8,325 8,200 125 OLhEr CAUBES......omversmmmmsnmpmmmssssnspassmnansihsbn ssa Residual 1,848,983 421 1,848,562 Malignant neoplasms of all other and unspecified sites (ICDA Nos. 170-173, 190-199). —~About 2 percent more deaths were classified under the above title by the Eighth Revision than had been classified under the com- parable Seventh Revision title (ICD Nos. 156B, 165, 190-199) (appendix I). Data from the comparability study show that an estimated 4,478 deaths classified under this title by the Eighth Revision were classified elsewhere by the Seventh Revision. On the other hand, an esti- mated 3,736 deaths classified to these causes by the Seventh Revision were classified elsewhere by the Eighth Revision. The result was a net increase of 742 deaths. Among the 4,478 deaths classified under Malignant neoplasms of all other and unspecified sites by the Eighth Revision were an estimated 3,149 deaths which were classified by the Seventh Revision to Malignant neoplasm of liver, not stated whether primary or secondary (cate- gory number 156A). Most of these differences in assignments were due to the transfer of malig- nant ncoplasm of liver, unspecified as to whether primary or secondary, from category number 156A in the Seventh Revision to cate- gory 197.8 in the Eighth Revision. As shown in the table below, the 3,736 deaths assigned to Malignant ncoplasm of other and unspecified sites by the Seventh Revision but to other causes by the Eighth Revision included 3,307 deaths assigned by the Seventh Revision to Malignant neoplasm of unspecified sites (ICD Nos. 156B, 165, 198, 199B). An estimated 1,572 of these 3,307 deaths were coded to Malignant neoplasm of thoracic organs (secondary) (ICD No. 165) by the Seventh Revision and to Malignant neoplasms of respiratory system (ICDA Nos. 160-163) by the Eighth. As previously mentioned, these differ- ences resulted primarily from a change in the classification of metastatic neoplasm of the lung from the Seventh Revision title Malignant neo- plasm of thoracic organs (secondary) (ICD No. 165) to the Eighth Revision title Malignant neoplasms of bronchus and lung (ICDA No. 162.1). Also included in the 3,307 figure were an estimated 1,587 deaths coded to category 199B by the Seventh Revision and to malignant neo- plasm of specified sites other than those classi- fiable to ICDA Nos. 160-163, 170-173, 190-194 by the Eighth Revision (appendix VII). These differences resulted from a change in the pro- cedure for classifying malignant neoplasm of multiple sites with no indication as to which was the primary site. The remaining 148 of the 3,307 deaths were assigned by the Eighth Revision to causes other than malignant neoplasms. 29 Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of that went to Malignant d . eaths in column (2) Cause of death and category number deaths neoplasms of all other thot Wert to in 1966 and unspecified sites other couses (ICDA Nos. 170-173, 190-199) (1) (2) (3) (4) All CAUSES ..ccoerereerennrennrnneneennnnneeneeneaeaeens 001-E999 1,863,149 35,774 1,827,375 Comparable causes ................ 156A,156B,165,190-199 38,181 34,445 3,736 Malignant neoplasm of other and unspecified SHES] oer reer errr res cere snens 156B,165,190-199 35,032 31,296 3,736 OF SKIN ..ccrvmmmmmssmsmimisinmms ans 190,191 4,560 4,498 62 Of eye............... 358 358 - Of brain and other parts of nervous system.. 7.355 7.137 218 1,008 1,008 - 1,792 1,698 94 1,318 1,318 - Of other specified sites, not stated to be SRCONVIATY sursuvsusisssvessnsmmimssnseissnesissoni aa ann ss 195,199A 1,422 1,367 55 Of unspecified sites .................. 156B,165,198,1998 17,219 13,912 3,307 Malignant neoplasm of liver, not stated whether PrIMArY OF SECONKONY wussnsimssmirsisisiosssnimssssss 156A 3,149 3,149 —- Other CAUSES.......cuuveeeeeeeeeeeeremreeeeeieeeeeeeeenennns Residual 1,824,968 1,329 1,823,639 LThis is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. Leukemia (ICDA Nos. 204-207). There was no appreciable break in the comparability of mortality statistics for this cause as a result of implementation of the Eighth Revision. The comparability ratio is 0.9974 (appendix I). As shown in the table below, there were 14,012 deaths assigned to Leukemia and aleukemia (ICD No. 204) by the Seventh Revision. An estimated 13,976 deaths were assigned to the comparable title Leukemia by the Eighth Revi- sion. Other neoplasms of lymphatic and hema- topoietic tissues (ICDA Nos. 200-203, 208, 209). -About 5 percent more deaths were assigned by the Eighth Revision to this group of causes than were assigned by the Seventh Revi- sion to the comparable title Lymphosarcoma and other neoplasms of lymphatic and hema- topoictic tissues (ICD Nos. 200-203, 205) (appendix I). As shown in the table, page 31, this increase was due in large part to the assign- ment of an estimated 585 deaths to ICDA Nos. Seventh Revision Eighth Revision Numb § Estimated number of Estimated number of c f death and cat b un 2 o deaths in column (2) deaths in column (2) aise 0} -ceath and category number . pn that went to Leukemia that went to in (ICDA Nos. 204-207) other causes (1) (2) (3) (4) All CAUSES «eevee eeaaeeas 001-E999 1,863,149 13,976 1,849,173 Leukemia and aleukemia.........cccoeeeeuiiiieinniienennnnnnnn, 204 14,012 13,834 178 OLNEI COUSES oii usiinriismnsivssmssrorsneississgsonsens sie Residual 1,849,137 142 1,848,995 30 200-203, 208, 209 by the Eighth Revision that were assigned ICD Nos. 294, 295, 297-299 by polycythemia (vera) from ICD No. 294 in the Seventh Revision to ICDA No. 208 in the Eighth the Seventh Revision. Most of these differences Revision. in assignments resulted from the transfer of Seventh Revision Eighth Revision Estimated number of deaths in column (2) : Number of that went to Other Estimated numbier of ; deaths in column (2) Cause of death and category number deaths neoplasms of lymphatic hat Went 1o in 1966 and hematopoietic Other causes tissues (ICDA Nos. 200-203,208,209) (1) (2) (3) (4) AN CAUSES ssvivsivmimmimsssninimmmnmnsemsnrssrmsens 001-E999 1,863,149 16,602 1,846,547 Comparable causes....... 200-203,205,294,295,297-299 17,079 16,063 1,016 Lymphosarcoma and other neoplasms of lymphatic and hematopoietic tissues! ooovoeeeernnnn, 200-203,205 15,802 15,478 324 Lymphosarcoma and reticulosarcoma ................ 200 7,563 7,497 66 Hodgkin's diSEase ..........euuuneieeunieeiireieereeenaeeeennnns 201 3,412 3,398 14 Other neoplasms of lymphatic and hematopoietic BISSUBS. cuveveereeiieeeeeiiireeeeireeeeanae es eeannnns 202,203,205 4,827 4,583 244 Other diseases of blood and blood-forming OEGANS, ovr rvrsrvmrvessisisivorsssinepmssisnine 294,295,297-299 1,277 585 692 OtNEI CAUSES ....cevveveieeeniieeeieeeieeeeeeeeeiieeeeeaees Residual 1,846,070 539 1,845,531 I This is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. 31 3. CEREBROVASCULAR DISEASES The comparable Seventh Revision title for Cerebrovascular diseases (ICDA Nos. 430-438) is Vascular lesions affecting central nervous system (ICD Nos. 330-334). In the Seventh Revision this title was included in Section VI, “Diseases of the nervous system and sense organs,” but the Eighth Revision title Cerebrovascular diseases (ICDA Nos. 430-438) is in Section VII, “Diseases of the circulatory system.” The comparability ratio between these two titles for cerebrovascular diseases is 0.9905 (appendix I). Although this ratio is close to 1.00, there were some important changes in coding procedures, which in part compensated for each other. The table at bottom of page shows that 4,516 deaths in 1966 assigned to ICD Nos. 330-334 in the Seventh Revision (2.2 percent) were transferred by the coding pro- cedures in effect with the Eighth Revision to categories other than Cerebrovascular diseases (ICDA Nos. 430-438). On the other hand, about 2,569 deaths not assigned to Seventh Revision categories 330-334 were assigned by the coding procedures in effect with the Eighth Revision to Cerebrovascular diseases. The largest groups transferred from cerebro- vascular diseases were an estimated 1,249 deaths to Chronic ischemic heart disease (ICDA No. 412) and about 582 deaths to All other diseases of arteries, arterioles, and capillaries (ICDA Nos. 442-444, 446-448). The changes in coding pro- cedure resulting in the transfers to ICDA No. 412 are described above in the section “Diseases of Heart.” Most of the 582 transfers of deaths to the Eighth Revision title All other diseases of arteries, arterioles, and capillaries (ICDA Nos. 442-444, 446-448) resulted from the linking of the conditions under this title by a provision in the classification to Arteriosclerosis (ICDA No. 440) when it is reported as the underlying cause of the conditions under the following categories of this group title: 443-444, 446. For example, a frequently encountered certificate contains the following combination of diseases. Entries in Part I and Category number Part II of certificate 7th Rev. | 8th Rev. I (a) Mesenteric infarction | 570.2 444.2 (b) Arteriosclerosis 450.0 440.9 II Cerebral thrombosis 332 433 Final code 332 444.2 Seventh Revision Eighth Revision Estimated number of g E ber of Number of deaths in column (2) Boimata numa of Cause of death and category number deaths that went to Cerebro- that Went 1O in 1966 vascular diseases (ICDA Other Couses Nos. 430-438) (1) (2) (3) (4) All CAUSES ..eeveierereeeeeiieiieincasannennninesaeeees 001-E999 1,863,149 202,894 1,660,255 Vascular lesions affecting central nervous BYBUBIVY cou cevivssiniminssissssssnsavansnass snavwnsnvnsrravmmss ,...330-334 204,841 200,325 4,516 Subarachnoid hemorrhage ...........eeeevieiieieniennennnne. 8,631 8,287 244 Cerebral Nermnorrhage assess sesssssvenssisy 111,446 109,400 2,046 Cerebral embolism and thrombosis 63,943 62937 1,006 Other vascular lesions affecting central nervous BY STBITY uinvonsss ra PAE EAE REARS Emam abe 20,921 19,701 1,220 Other causes 1,658,308 2,569 1,655,739 32 By the Seventh Revision, after arterioscle- rosis was selected as the presumptive underlying cause of death (taken to be the underlying cause of the mesenteric infarction), the assignment was modified by the provision that if arterioscle- rosis appeared on the certificate with mention of any condition in ICD Nos. 330-332, 334, the combination was coded. That is, the death was assigned to the specified cerebrovascular disease. By the linking provided by the Eighth Revision between arteriosclerosis and mesenteric infarc- tion, the presumptive underlying cause is taken to be the mesenteric infarction. Thus, despite the fact that both the Seventh and Eighth Revisions provide for the linkage of cerebrovascular disease with mention of arterio- sclerosis, the final assignment by the Eighth Revision of the above certificate is to the first condition linked with arteriosclerosis—that is, the assignment is made to Arterial embolism and thrombosis of mesenteric artery (ICDA No. 444.2). The largest transfers to Cerebrovascular discases (ICDA Nos. 430-438) from Seventh Revision categories other than 330-334 were from: Accidental falls (ICD Nos. E900-E904), about 377 deaths (with other accidents only about 47 deaths); Other arterial diseases (ICD Nos. 452-454, 456), about 311 deaths; Func- tional disease of heart (ICD No. 433), about 272 deaths; General arteriosclerosis (ICD No. 450), 331 deaths; and Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tissues (ICD Nos. 140-205), about 178 deaths. The changes in coding procedures resulting in the transfers from Accidental falls and from Functional disease of heart are described above in the section “Diseases of Heart” (page 5). The assignment of about 311 deaths from Other arterial diseases (ICD Nos. 452-454, 456) in the Seventh Revision to Cerebrovascular diseases (ICDA Nos. 430-438) in the Eighth Revision resulted primarily from the reclassifica- tion of occlusion of carotid artery from Arterial embolism and thrombosis (ICD No. 454) to Occlusion of precerebral arteries (ICDA No. 432). In addition, about 64 deaths assigned by the Seventh Revision to Other aneurysm, except of heart and aorta (ICD No. 452), were assigned by the Eighth Revision to Subarachnoid hemor- rhage (ICDA No. 430). This change resulted from the reclassification of berry and miliary aneurysm (of the brain) from ICD No. 452 to ICDA No. 430. Most of the transfers from General arterio- sclerosis (ICD No. 450), an estimated 248 out of the 331 deaths, were coded by the Eighth Revision to Other cerebrovascular diseases (ICDA Nos. 437, 438). This transfer resulted from shifting the terms “cerebral ischemia” and “cerebral insufficiency” to the Eighth Revision title Generalized ischemic cerebrovascular disease (ICDA No. 437). In accordance with the Seventh Revision these terms were included under Other diseases of brain (ICD No. 355). By the Seventh Revision, when a cause classified under Other diseases of brain (with the exception of cerebral atrophy) was reported on the certificate as due to arteriosclerosis, the death was coded to arteriosclerosis and not to cerebrovascular disease. Both the Seventh and Eighth Revisions provide for the modification of the provisionally selected underlying cause when such modifica- tion results in a more useful and informative condition for tabulations of mortality data. For example, as discussed above, if arteriosclerosis, a generalized disease, has been provisionally selected as the underlying cause for a certificate on which a cerebrovascular disease also is reported, both the Seventh and Eighth Revisions provide for the assignment of the death to the cerebrovascular disease. The provisionally selected underlying cause (arteriosclerosis) is “linked” by the rules for classification to cerebrovascular diseases (and to a number of other diseases), and the combination of arterio- sclerosis and the cerebrovascular disease is coded to the cerebrovascular disease. Thus when the terms “cerebral ischemia” and “cerebral insuf- ficiency” were transferred to Other cerebro- vascular diseases (ICDA Nos. 437, 438), the certificates on which these terms appeared with arteriosclerosis were assigned by the Eighth Revision to the indicated cerebrovascular dis- ease. The transfers from Malignant neoplasms, including neoplasms of lymphatic and hema- topoietic tissues (ICD Nos. 140-205), to Cerebrovascular diseases (ICDA Nos. 430-438) 33 resulted for the most part from a modification of guides for determining the probability of sequences. In accordance with coding proce- dures in effect with the Seventh Revision (beginning with 1959 for this particular proce- dure), any intracranial disease in ICD Nos. 330-334 reported as due to a malignant neo- plasm was classified to the neoplasm as the underlying cause of death. But with the Eighth Revision the coding procedure was changed so that not every one of the intracranial vascular diseases reported as due to a malignant neoplasm would be assigned to the malignant neoplasm. Although the introduction of the Eighth Revision for data year 1968 did not produce any appreciable break in the comparability of mortality statistics for the entire group of cerebrovascular diseases, it did produce a con- siderable degree of discontinuity for the com- ponents of this group of diseases. One of the reasons for the lack of comparability for the components is that while the Sixth and Seventh Revisions were in use (1949-67) cerebrovascular diseases were distributed among only five components, but according to the Eighth Revi- sion they are distributed among nine compo- nents (table E). Table E. Deaths from Eighth Revision category Cerebrovascular diseases, distributed by the Seventh Revision categories: United States, 1966 [Deaths are those occurring within the United States. Seventh Revision categories for which the number of deaths was less than 5 percent of the number assigned to the Eighth Revision category or was less than 100 are not shown | Eighth Revision | Seventh Revision Final — = ee . i eee pe — ea count of deaths ; Col. for Beble (6) as Seventh ot Esti- b a per- | Revision hi Runber cent cate- a per: ICDA | mated of of ori cent of No Category title | number ICD No. Category title deaths col | as col. : | of going to ~ 8 | deaths Eighth 3) he be 5 Revision | a | Revision | category | category | | (1) (2) | 3) (4) (5) (6) 7) (8) 9) 430-438 Cerebrovascular dis- €AaSeS=mmmmmmm meme ——- 202,942 330-334 Vascular lesions affect- ing central nervous SyStem=--==coecoooaaaao 200, 364 98.7 204,841 97.8 | 450 General arteriosclero- BlEnmme annem m mm mm——— 331 0.2 | [ 452-454 ,456 Other arterial diseases-- 312 0.2 433 Functional disease of heart-----ccoomomoao_ 268 0.1 | E903 Fall on same level------- 176 0.1 @ 341- 344,352, 354-369, 380- 384,386, 388- | 390, 394-398 Other diseases of nerv- ous system and sense OrgansS===-===-=ccemmmmoan 106 0.1] @ | E900-E902 Fall from one level to another-------ceeooao__ 101 0.0 E904 Unspecified falls-------- 101 0.0 | -—— Other categories--------- 1,183 0.6 | Total 202,942 100.0 @ = Category not comparable to Eighth Revision category. 34 Table E. Deaths from Eighth Revision category Cerebrovascular diseases, distributed by the Seventh Revision categories: United States, 1966—Con. | Deaths are those occurring within the United States. Seventh Revision categories for which the number of deaths was less than 5 percent of the number assigned to the Eighth Revision category or was less than 100 are not shown| Eighth Revision Seventh Revision Final i. = e count of deaths Col. for ti- Fett. (6) as | Seventh te Esti- number | 2 Der- Revision a per- ICDA mated of cen ca cent of Category title number ICD No. Category title deaths 0 gories col. No. of going to ool ral (8) deaths Eighth & TE fr Revision R gh category eyxsion category 1) 2) 3) (4) (5) (6) 7) (8) 9) 430 Subarachnoid hemor- rhage----------------=- 8,723 330 Subarachnoid hemorrhage-- 8,459 97.0 8,531 99.2 --- Other categories--------- 264 3.0 @ Total 8,723 | 100.0 431 Cerebral hemorrhage---- 52,616 331 Cerebral hemorrhage------ 52,247 99.3 | 111,446 46.9 --- Other categories--------- 369 0.7 @ Total 52,616 | 100.0 433 Cerebral thrombosis---- 57,600 332 Cerebral embolism and thrombosis------=----=-= 56,311 97.8 63,943 88.1 --- Other categories--------- 1,289 2.2 @ wae Total 57,600 | 100.0 434 Cerebral embolism------ 1,272 332 Cerebral embolism and thrombosis-------------= 1,244 97.8 63,943 1.9 -—- Other categories--------- 28 2.2 @ Total 1,272 100.0 432,435 | Other specified acute - oo cerebrovascular dis- EABEE mmm mmm mmm ———— 3,776 332 Cerebral embolism and thrombosis-------------- 2,903 76.9 63,943 4.5 331 Cerebral hemorrhage------ 498 13.2 @ 452-454 ,456 Other arterial diseases-- 220 5.8 @ -——- Other categories--------- 155 4.1 @ Total 3,776 100.0 436 Acute but ill-defined oo 1 oo cerebrovascular dis- ease---==-=========---- 51,168 331 Cerebral hemorrhage------ 46,773 91.4 | 111,446 45.9 333,334 Other vascular lesions affecting central nerv- ous system-------------- 3,815 1.5 @ -—- Other categories--------- 580 1.1 @ Total 51,168 | 100.0 437,438 | Other cerebrovascular oT] ) - diseases-------------- 27,787 333,334 Other vascular lesions affecting central nerv- ous system-----------=-= 14,762 53.1 20,921 70.6 331 Cerebral hemorrhage------ 10,449 37.6 111,446 9.4 332 Cerebral embolism and thrombosis -—— 1,659 6.0 @ 450 General arteriosclerosis- 248 0.9 @ -— Other categories--------- 669 2.4 @ wie Total 27,787 | 100.0 @ = Category not comparable to Eighth Revision category. 35 4. ACCIDENTS The comparability ratio for Accidents (with motor vehicle and other accidents taken to- gether) between the Seventh Revision (ICD Nos. E800-E962) and Eighth Revision (ICDA Nos. E800-E949) was 0.9570 (appendix I). As shown in the three tables below, this reduction in the number of deaths assigned to Accidents by the Eighth Revision over the number assigned by the Seventh Revision resulted primarily from transfers from accidents Seventh Revision Eighth Revision Estimated number of Estimated number of Number of i 2 Cause of death and category number deaths SSeatisin alumi 10} A realm @ in 1966 (ICDA Nos. E800-E949) other causes (1) (2) (3) (4) All CBUSES ....covvivmsimimmismmmmmansrannarssnrrsssnses 001-E999 1,863,149 108,683 1,754,466 Accidents .......cceeeeecuueeeeeeiinennnn. ..E800-E962 113,563 106,455 7,108 Railway accidents........ ..E800-E802 1,027 1,000 27 Motor vehicle accidents................. ..E810-E835 53,041 52,456 585 Motor vehicle traffic accidents................ E810-E825 51,933 51,373 560 Motor vehicle traffic accident involving collision WAH rANWEY TBI. ums immmsssimmsssansnnres E810 1,800 1,757 43 Motor vehicle traffic accident to pedestrian .. E812 8,675 8,612 63 Other motor vehicle traffic accidents involving CONISTON. ....eeiiiiiieeceeeee cen E811,E813-E819 22,641 22,297 344 Motor vehicle noncollision traffic BOCHHBIVIS ..ovenss cunviviieninisssssiistensmmsnnnnnmes E820-E824 14,291 14,202 89 Motor vehicle traffic accident of unspecified PBRUPR.cviivaininiiimmmmssssssssesensererennanserresessntons E825 4,526 4,505 21 Motor vehicle nontraffic accidents.......... E830-E835 1,108 1,083 25 Other road vehicle accidents..................... E840-E845 292 292 - Water transport accidents..............ooeo....... E850-E858 1,630 1,630 - Aircraft accidents.............ccceeevveeeeennnnnn... E860-E866 1,510 1,486 24 Accidental poisoning by solid and liquid SUDSTaNCeS ......cuvveeereee cece cece, E870-E888 2,283 1,806 477 Accidental poisoning by gases and vapors... E890-E895 1,648 1,523 125 Accidental falls...........eeeeeeoueeeeeeeeeeaneeeeanns E900-E904 20,066 16,906 3,160 E900-E902 5,772 5,312 460 RT ERRATA Tam Fob mr eames nnnss E903 5,593 4,839 754 .... E904 8,701 6,755 1,946 Blow from falling or projected object or missile... E910 1,459 1,432 27 Accident caused by machinery ....... E912 2,070 1,989 81 Accident caused by electric current E914 1,025 1,025 — Accident caused by fire and explosion of combustible INBLBYIBN ooeivustsvinsnsimmmvimimminss vss inmsninssss sammsnmunsannnn E916 8,084 7,813 271 Accident caused by hot substance, corrosive liquid, steam, and radiation ..............cceeeeeevnnnnnnnns E917,E918 409 337 72 Accident caused by firearm.............cccvvvvenn.n. E919 2,558 2,189 369 Inhalation and ingestion of food or other object causing obstruction or suffocation .......... E921,E922 1,831 1,672 159 Accidental drowning .........cccceeeuuu.... E929 5,687 5,431 256 Excessive heat and insolation ...........c..cceeeuu...... E931 531 430 101 Complications due to nontherapeutic medical and surgical procedures, therapeutic misadventure, and late complications of therapeutic ProCEUUIES. ...ovusisciiniriresssvmissssisssnsisasnunnnn E940-E959 1,411 1,129 282 All other accidents....... E911,E913,E915,E920,E923- E928,E930,E932-E936,E960-E962 7,001 5,909 1,092 Other CAUSES ........c.veeeeeeeeeeeeeeeeeeeee sesso, Residual 1,749,586 2,228 1,747,358 36 Seventh Revision Eighth Revision Estimated number of 3 Number of deaths in column (2) Estimated number 27 Cause of death and category number deaths that went to Motor gathis incolumn 3 . : that went to in 1966 vehicle accidents (ICDA other causes Nos. E810-E823) (1) (2) (3) (4) AN COUSES cussssssssrninsirsrsssssssssnsnnsarnrennanne 001-E999 1,863,149 52,622 1,810,527 Motor vehicle accidents..........cccceeeeecunennnes E810-E835 53,041 52,250 791 Motor vehicle traffic accidents ................. E810-E825 51,933 51,243 690 Motor vehicle traffic accident involving collision with railway train.......cccccevveeeieieeiiieeeieeenenennn. E810 1,800 1,757 43 Motor vehicle traffic accident to pedestrian .... E812 8,675 8,570 105 Other motor vehicle traffic accidents involving COIISION. cee eeeeeree eases E811,E813-E819 22,641 22,274 367 Motor vehicle noncollision traffic ACCTHRIVIS cocsrsssmsrnmmmrnssemmmnsvrssnsservsssisisens E820-E824 14,291 14,158 133 Motor vehicle traffic accident of unspecified DADTAIT Russ sunemavsvasnsoso nahn smmenesssivnsun is trssubanisuneis buts E825 4,526 4,484 42 Motor vehicle nontraffic accidents. ...E830-E835 1,108 1,007 101 OLNEY COUSHS. con enioinsssinsas ivr hartbarmmin sins sous Sorsaes Residual 1,810,108 372 1,809,736 other than motor vehicle accidents. The com- parability ratio for Motor vehicle accidents (ICDA Nos. E810-E823) between the Seventh and Eighth Revisions was close to 1.00—actually 0.9921, but the comparability ratio for All other accidents (ICDA Nos. E800-E807, E825-E949) was only 0.9250 (appendix I). Three major changes between the Seventh and Eighth Revisions that resulted in the trans- fer of a substantial number of deaths formerly attributed to All other accidents (ICD Nos. E800-E802, E840-E962) are summarized below. (1) An estimated 1,923 of the 6,751 deaths transferred from these accident categories were assigned by the Eighth Revision to the following new title introduced for classifying deaths for which it was not possible for the certifier to determine whether the injuries were accidentally or purposely inflicted: Injury undetermined whether accidentally or purposely inflicted (ICDA Nos. E980-E989). These 1,923 deaths that were formerly assigned to accidents (excluding motor vehicle accidents) constitute about 63 percent of this new category. (Deaths assigned by the Seventh Revision to Suicide make up about 31 percent of the new category; and those assigned to Homicide, about 2 per- cent.) (2) The second largest group transferred from these accidents was 1,773 deaths that were assigned to the Eighth Revision title Diseases of heart. As stated above in section 1 of this report, “Diseases of Heart,” an estimated 1,422 of the deaths transferred from All other accidents were assigned by the Eighth Revision to Ischemic heart disease (ICDA Nos. 410-413), and the important changes in the coding procedures for accidents that resulted in this transfer of deaths are summarized in that section. It is believed these same changes in coding procedures between the Seventh and Eighth Revisions account for the most part for an additional estimated 424 deaths that were assigned by the Seventh Revision to these accidents being trans- ferred by the Eighth Revision to Cerebrovascular diseases (ICDA Nos. 430-438). In summary, an estimated 3,188 of these 6,751 transferred deaths were fatalities resulting from falls. About half of these 3,188 deaths were transferred by the Eighth Revision to Ischemic heart disease (1,171 deaths with falls involved) and to Cerebrovascular diseases (377 deaths with falls involved) (appendix I). (3) An estimated 319 of the 6,751 trans- ferred deaths were assigned by the Eighth Revi- sion to Pneumonia (ICDA Nos. 480-486). About 37 Seventh Revision Eighth Revision Estimated number of : Number of deaths in column (2) Estimated number of eaths in column (2) Cause of death and category number deaths that went to All other . i that went to in 1966 accidents (ICDA Nos. Oihel Causas E800-E807,E825-E949) (1) (2) (3) (4) AI CAUBRE...; 01000155 rnnesmsssssaminii es sissies ass 001-E999 1,863,149 55,983 1,807,166 All other accidents................... E800-E802,E840-E962 60,522 53,771 6,751 Railway accidents... .E800-E802 1,027 1,000 27 Other road vehicle accidents. E840-E845 292 292 - Water transport accidents......... E850-E858 1,630 1,630 - Aircraft acCidemtS ...umimrisesssiss .... EB60-E866 1,510 1,486 24 Accidental poisoning by solid and liquid SUBSTANCES .ovovsisvcrsmmmimmsmssssissnsstesessnesnannane E870-E888 2,283 1,806 477 Accidental poisoning by gases and vapors. E890-E895 1,648 1,523 125 Accidental falls...........ccooerureecererenennnene.. E900-E904 20,066 16,878 3,188 Fall from one level to another E900-E902 5,772 5312 460 Fall on same level E903 5,693 4,839 754 Unspecified falls ... E904 8,701 6,727 1,974 Blow from falling or projected object or missile... E910 1,459 1,405 54 Accident caused by machinery ee 2,070 1,936 134 Accident caused by electric current................... E914 1,025 1,025 - Accident caused by fire and explosion of combustible material .............cceeeeeeeeececnnnennn. E916 8,084 7,788 296 Accident caused by hot substance, corrosive liquid, steam, and radiation..................... E917,E918 409 337 72 Accident caused by firearm..............ccceeuueennnn..e. E919 2,558 2,189 369 Inhalation and ingestion of food or other object causing obstruction or suffocation .......... E921,E922 1,831 1,672 159 Accidental drowning..................... 5,687 5,431 256 Excessive heat and insolation ... 531 430 101 Complications due to nontherapeutic medical and surgical procedures, therapeutic misadventure, and late complications of therapeutic ProCedUres..........ceeeeeeeeeeeeeeeeieennnnenaaenenns E940-E959 1,411 1,129 282 All other accidents ...... E911,E913,E915,E920,E923- E928,E930,E932-E936,E960-E962 5,814 1,187 Other Causes ........ccuueeeeeeeeeeeeeeeeeeeeeeeeeeeeaeeaann, Residual 1,802,627 2,212 1,800,415 53 of these 319 deaths were assigned by the Seventh Revision to Inhalation and ingestion of food or other object causing obstruction or suffocation (ICD Nos. E921, E922); and it is possible that an additional 119 of them were assigned by the Seventh Revision to Foreign body entering other orifice (ICD No. E923). The resulting total of 172 deaths were transferred from these accidents to the Eighth Revision category Pneumonia, unspecified (ICDA No. 486). These transfers were probably made as a result of the provision that when such a disease as Pneumonia, unspecified (ICDA No. 486), is the resulting disease condition of such accidents, the assignment is to be made to the resulting 38 disease condition and not to the accident. (In both revisions if this aspiration of food resulted from a disease which presumably affects the ability to control the process of swallowing, for example, cancer of the throat or a disease resulting in paralysis, the assignment is to the stated underlying disease. Also in both revisions asphyxia from aspiration of mucus or vomitus which resulted from a disease is coded to the disease.) Partly compensating for the 6,751 deaths transferred from these titles for accidental fatal- ities were 2,212 deaths transferred by the Eighth Revision from other causes to these titles. 5. INFLUENZA AND PNEUMONIA The introduction of the Eighth Revision resulted in assigning about 4 percent more deaths to Influenza and pneumonia (ICDA Nos. 470-474, 480-486) than had been assigned by the Seventh Revision to the most nearly com- parable title, Influenza and pneumonia, except pneumonia of newborn (ICD Nos. 480-483, 490-493) (appendix I). About 5,355 deaths not assigned to ICD Nos. 480-483, 490-493 by the Seventh Revision were transferred by the coding procedures in effect with the Eighth Revision to ICDA Nos. 470-474, 480-486, while about 2,557 deaths that were assigned by the Seventh Revision to ICD Nos. 480-483, 490-493 were transferred by the coding procedures in effect with the Eighth Revision to categories other than ICDA Nos. 470-474, 480-486. As shown in the table below, the largest group transferred to Influenza and pneumonia by the Eighth Revision was an estimated 2,232 deaths that had been assigned by the Seventh Revision to Pneumonia of newborn (ICD No. 763). This transfer reflects a change in classifica- tion providing that only diseases specific to the newborn (e.g., hemolytic disease of the new- born) be included under the title Certain causes of mortality in early infancy (ICDA Nos. 760-769.2, 769.4-772, 774-778), and that condi- tions not different from those classified outside the perinatal classification, such as pneumonia and diarrhea, be excluded. Tabulations of deaths assigned by the Eighth Revision to Pneumonia (ICDA Nos. 480-486) for all infants will give the number for this age group dying from pneu- monia. The Seventh Revision title Pneumonia of newborn (ICD No. 763) was limited to deaths from Pneumonia (ICD Nos. 490-493) and inter- stitial pneumonia (unspecified), an inclusion term under ICD No. 525, at ages under 28 days. Therefore tabulations of 1968 deaths at ages under 28 days assigned by the Eighth Revision to Pneumonia (ICDA Nos. 480-486) will give the approximate number that would have been assigned to Pneumonia of newborn (ICD No. 763) if the Seventh Revision had been used for 1968. Another large group transferred to Influenza and pneumonia by the Eighth Revision was about 1,472 deaths that had been assigned by the Seventh Revision to Other chronic inter- Seventh Revision Eighth Revision Estimated number of i Number of deaths in column (2) Eye be Cause of death and category number deaths that went to Influenza WAC WEA in 1966 and pneumonia (ICDA other causes Nos. 470-474,480-486) (1) (2) (3) (4) All CAUSES ..vuvrerenninreiereeeieeeeene es ee ee einennn 001-E999 1,863,149 66,413 1,796,736 Selected causes...........ceeen.n.. 480-483,490-493,525,763 70,363 64,762 5,601 Influenza and pneumonia, except pneumonia of Newborn! o.oo 480-493 63,615 61,058 2,557 INI BNZ inns vinnssn mis aT HR ATR 480-483 2,830 2,683 147 Pneumonia, except pneumonia of newborn.490-493 60,785 58,375 2,410 Lobar pneumonia... ....ccoeeveveveeuerennneeneeeeeeeennenes 490 8,864 8,565 299 BronChopNeuITYONIa wuss ives wsrsisssasssnees 491 33,276 31,949 1,327 Primary atypical pneumonia..........c.cc...e —— 492 5,729 5,409 320 Pneumonia, other and unspecified.........ccc........ 493 12,916 12,452 464 Other chronic interstitial pneumonia.................... 525 4,271 1,472 2,799 Pneumonia of newborn .........ccoeiiiiiiiiiiiiieiieeieeens 763 2,477 2.232 245 Other CAUSES ......ueeveerineiiieerieeeaieeeiieeenereaeennaans Residual 1,792,786 1,651 1,791,135 IThis is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. 39 stitial pneumonia (ICD No. 525). The 4,271 deaths in 1966 assigned to this Seventh Revision title were reassigned by the Eighth Revision primarily to the following titles: (1) Other chronic interstitial pneumonia (ICDA No. 517), about 2,656 deaths, and (2) Acute interstitial pneumonia (ICDA No. 484), about 1,400 deaths. The transfer of these 1,400 deaths to the latter title resulted from shifting the terms “Interstitial pneumonitis’ and “interstitial pneu- monia,” not otherwise specified, to the Eighth Revision title Acute interstitial pneumonia (ICDA No. 484). The transfers from Influenza and pneumonia by the Eighth Revision are quite widely distrib- uted over the classification. Each of the three transfers, however, involved over 200 deaths. One of these three resulted from the change in assignment of the combination of pneumonia and alcoholism (with alcoholism in the “due to” position on the certificate) from Pneumonia, other and unspecified (ICD No. 493), hy the Seventh Revision to Alcoholic addiction (ICDA No. 303.2) by the Eighth Revision. Another of these changes was the transfer of about 213 deaths from Pneumonia (ICD Nos. 490-493) to Nutritional marasmus (ICDA No. 268) and to Other and unspecified nutritional deficiency (ICDA No. 269.9). In the Seventh Revision the terms athrepsia, cachexia, extreme wasting, marasmus, and inanition were con- sidered to be ill-defined conditions and deaths from these conditions at ages 1 year and over were assigned to causes under Senility and ill- defined diseases (ICD Nos. 790-795). Con- sequently, in accordance with the Seventh Revision, in the event that the provisionally selected underlying cause was one of these conditions and some condition other than an ill- defined condition was also reported, e.g., pneu- monia, the other condition was usually reselected as the underlying cause of death. But by the Eighth Revision, inasmuch as the terms athrepsia, cachexia, extreme wasting, marasmus, and inanition are no longer classified under Symptoms and ill-defined conditions (ICDA Nos. 780-796), no reselection of another condi- tion on the certificate as the underlying cause of death is made. Instead, by the Eighth Revision, if the condition selected as the underlying cause 40 is athrepsia, cachexia, extreme wasting, or marasmus, the death is assigned to Nutritional marasmus (ICDA No. 268), and if the condition selected as the underlying cause is inanition, the death is assigned to Other and unspecified nutri- tional deficiency (ICDA No. 269.9). About 208 deaths transferred from pneu- monia in the Seventh Revision to Other diseases of respiratory system (ICDA Nos. 501-508, 512, 514-516, 519) in the Eighth constituted the third group of transferred deaths. It is believed that most of these transfers are attributable to the shift of a number of terms included by the Seventh Revision under the pneumonias to other categories by the Eighth Revision. Among such terms were “lipoid pneumonia,” which by the Eighth Revision is included under Other diseases of lung (ICDA No. 519.2), and “adynamic pneu- monia” and ‘‘asthenic pneumonia,” which by the Eighth Revision are included under Pul- monary congestion and hypostasis (ICDA No. 514). The 1968 death rate (36.8 deaths per 100,000) for Influenza and pneumonia (ICDA Nos. 470-474, 480-486) is 27.8 percent higher than the 1967 death rate (28.8 deaths per 100,000) for Influenza and pneumonia, except pneumonia of newborn (ICD Nos. 480-483, 490-493). Applying the factor 1.044 to the 1967 death rate, however, raises it to 30.1 deaths per 100,000, the level it would have reached if deaths in 1967 had been coded by the Eighth Revision rather than by the Seventh. The remaining increase in this death rate for 1968 (836.8 minus 30.1, or 6.7 deaths per 100,000 population) may be attributable to the wide- spread influenza epidemic in 1968. The comparability ratio for Influenza between the Seventh Revision (ICD Nos. 480-483) and Eighth Revision (ICDA Nos. 470-474) is 0.9572 (appendix I). As shown in the table below, this reduction in the number of deaths assigned to influenza by the Eighth Revi- sion (2,709 deaths) from the number assigned by the Seventh Revision (2,830 deaths) amounted to 121 deaths. This reduction is attributable in great part to the dropping by the Eighth Revision of the priority given by the Seventh Revision to influ- enza when two or more conditions are entered Seventh Revision Eighth Revision Nuh f Estimated number of Estimated number of c F death 20d b uw or ° deaths in column (2) deaths in column (2) use Of geath and category number . ans that went to Influenza that went to 0 (ICDA Nos. 470-474) other causes (1) (2) (3) (4) All COUSES «oiivesmmmsissinss viassnssssssasussunsemsine 001-E999 1,863,149 2,709 1,860,440 INFIUBNZA evenniieiiii ieee eerie eerie eee eree eee eae ee eeans 480-483 2,830 2,676 154 OLNEY CAUSES «cor venersmrmsnsmemsiisisninsaisssataiassnnnsnns Residual 1,860,319 33 1,860,286 on the certificate in such a way that none of them can be regarded as the underlying cause. Most of these transferred deaths assigned by the Seventh Revision to influenza were assigned by the Eighth Revision to Acute myocardial infarc- tion (ICDA No. 410) (32 deaths), Chronic ischemic heart disease (ICDA No. 412) (32 deaths), and to the remainder of the diseases of the circulatory system (38 deaths). Most of the transfers resulting in breaks in continuity of mortality statistics for Pneumonia (ICDA Nos. 480-486) have been described above in the discussion of Influenza and pneumonia. As shown in the table below, an estimated 4.8 percent more deaths were assigned by the Eighth Revision to Pneumonia (ICDA Nos. 480-486) than were assigned by the Seventh Revision to Pneumonia, except pneumonia of newborn (ICD Nos. 490-493) (appendix I). Seventh Revision Eighth Revision Nuinber of Estimated number of Estimated number of deaths in column (2) deaths in column (2) Cause of death and category number deaths . . that went to Pneumonia that went to in 1966 (ICDA Nos. 480-486) other causes (1) (2) (3) (4) All CAUSES ..evvvveiieeeeeeeeeeeieiiiiiieeeseeeaenes 001-E999 1,863,149 63,704 1,799,445 Selected CAUSES .......uueueeeeeereeerrirnennnnns 490-493,525,763 67,533 62,079 5,454 Pneumonia, except pneumonia of newborn !.490-493 60,785 58,375 2,410 Lobar pneumonia ......ceeeuueeiirieeeereniicereieerienn anes 490 8,864 8,565 299 Bronchopneumonia.......... ...491 33,276 31,949 1,327 Primary atypical pneumonia.... ..492 5,729 5,409 320 Pneumonia, other and unspecified.... ...493 12,916 12,452 464 Other chronic interstitial pneumonia...........c.c.cee... 525 4.271 1,472 2,799 Pneumonia of NeWDOIT .u..aiiarin sess sissssesssnessnss 763 2,477 2,232 245 Other CAUSES ....ouimsvrsrssrrssssmmrasnsssssmrespnessnsnn Residual 1,795,616 1,625 1,793,991 I This is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. 41 6. CERTAIN CAUSES OF MORTALITY IN EARLY INFANCY The Eighth Revision reduced deaths assigned to Certain causes of mortality in early infancy (ICDA Nos. 760-769.2, 769.4-772, 774-778) by 8.17 percent from the number assigned by the Seventh Revision to the comparable cause Certain diseases of early infancy (ICD Nos. 760-776). This percentage, based on coding the sample of deaths in 1966 by both the Seventh and Eighth Revisions, represents a reduction from 51,644 deaths by the Seventh Revision to 47,425 deaths by the Eighth Revision (appendix I and table below). It may be assumed that the 8.17 percent reduction found for the sample of 1966 deaths also is close to the percent reduction between 1967 and 1968 that is attributable to the change to the Eighth Revision. The number of deaths in 1967 coded to this cause by the Seventh Revi- sion was 48,314, and the number for 1968 coded to the comparable group of causes by the Eighth Revision was 43,840 deaths—a reduction of 4,474 deaths. Inasmuch as 8.17 percent of the 48,314 deaths in 1967 is 3,947 deaths, the difference between 4,474 and 3,947 deaths (527 deaths) is the estimated maximum number that may be attributed to an actual lowering of the death rate for this group of causes. A more precise measure of the true decline in mortality among infants from these causes may be obtained by applying the comparability ratio of 0.9183 to the 1967 infant mortality rate from these causes (1,371.0 deaths per 100,000 live births). Multiplying the rate of 1,371.0 by 0.9183 gives a death rate of only 1,259.0 per 100,000 live births. The difference between the 1968 rate (1,248.2 deaths per 100,000 live births) and the adjusted 1967 rate (1,259.0) is 10.8, constituting a real decrease in mortality of only 0.86 percent. As stated above under section 5, “Influenza and Pneumonia,” the reduction in the assign- ment of deaths to the Eighth Revision title Certain causes of mortality in early infancy reflects primarily the change in classification to provide that only diseases specific to the new- born (e.g., hemolytic disease of the newborn) be included under this title, and that conditions not different from those classified outside the peri- natal classification, such as pneumonia and diar- rhea, be excluded. About 2,232 of the 2,477 deaths assigned by the Seventh Revision to Pneumonia of newborn (ICD No. 763) were assigned by the Eighth Revi- sion to Pneumonia (ICDA Nos. 480-486). Tabulations for the age group under 1 year of deaths assigned to this latter cause by the Eighth Revision give the number of infants dying from pneumonia. Tabulations are also available Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estinaied Purber of Number of that went to Certain 7 wi deaths in column (2) Cause of death and category number deaths causes of mortality in that went to in 1966 early infancy (ICDA Stor Coutas Nos. 760-769.2, 769.4-772,774-778) (1) (2) (3) (4) All CAUSES eevee cece 001-E999 1,863,149 47,425 1,815,724 Certain diseases of early infancy........c............ 760-776 51,644 46,706 4,938 Pneumonia of newborn 2,477 191 2,286 Diarrhea of newborn ...........cccccvvcuvvveeeiveeineieeinnaan, 250 5 245 Remainder of Certain diseases of early IEBNCY conning 760-762,765-776 48,917 46,510 2,407 Other CAUSES .....ccueeeeeeiiiieeeeciieeeeeeeeeeee ee eres Residual 1,811,505 719 1,810,786 42 showing the number of these deaths that occurred at ages under 28 days. About 233 of the 250 deaths that were assigned by the Seventh Revision to Diarrhea of newborn (ICD No. 764) were assigned by the Eighth Revision to Enteritis and other diarrheal diseases (ICDA Nos. 008, 009). As for the pneumonias, tabulations of deaths for the age group under 1 year assigned to Enteritis and other diarrheal diseases (ICDA Nos. 008, 009) give the number of infants dying from this cause. Tabulations are also available showing the number of these deaths that occurred at ages under 28 days. The estimated 2,407 deaths that were assigned by the Seventh Revision to Certain diseases of early infancy (other than Pneumonia of newborn and Diarrhea of newborn) and were not assigned by the Eighth Revision to the comparable title Certain causes of mortality in early infancy were distributed among Eighth Revision titles as follows: 0 Number of Title and category number deaths SOPHICRINIA coco covisscormminsiriimssmissssrsinsen si saws» Sass 4 038 555 Congenital anomalies........cceeieercsescasnasssssinn 740-759 288 Symptoms and ill-defined conditions.......... 780-796 226 Other avitaminoses and nutritional deficiencies ........cccueeeeeeeeeneeeannnnnnns 260,261,263-269 329 All other endocrine and metabolic diseases......... 251,252,256-258,270-272,273.1-279 82 Other diseases of respiratory SYSIBINY ...ovavssssassasinssnnnnsn 501-508,512,514-516,519 206 Diseases of the skin and subcutaneous TISSUE «eeveeeeeeneeeeemnaeeeaeaeaaaeaasaanannnns 680-686,690-709 82 OLOGY CAUSES ...ovisesvmimmirisitsiss ius s stamens tmisssmsss gees; § 2 v 639 An estimated 719 deaths not coded by the Seventh Revision to Certain diseases of early infancy were shifted from a number of cate- gories to the Eighth Revision title Certain causes of mortality in early infancy (ICDA Nos. 760-769.2, 769.4-772, 774-778). Among these were an estimated 203 deaths that in the Seventh Revision were assigned to Other diseases of heart (ICD Nos. 430-434). This change resulted from the transfer of such terms as ‘“‘cardiac arrest of newborn” and “cardiac respiratory arrest of newborn” from ICD No. 433.0 of the Seventh Revision to the Eighth Revision category number 778.9, under Other conditions of fetus or newborn (ICDA No. 778). Also among these 719 deaths were an esti- mated 158 deaths that in the Seventh Revision were assigned to Symptoms, ill-defined, and unknown causes (ICD Nos. 780-793, 795). This shift resulted primarily from the transfer of terms under the category of the Seventh Revi- sion for ill-defined conditions to the Eighth Revision title Other conditions of fetus or new- born (ICDA No. 778). Deaths to newborns resulting from hemorrhage of a number of sites (e.g., liver and kidney) that by the Seventh Revision were coded to categories other than Hemorrhagic disease of newborn (ICD No. 771) were transferred to this title by the Eighth Revision (ICDA No. 778.2). Data from the comparability study based on deaths in 1966 show that transfers from the Seventh Revision title Certain diseases of early infancy (ICD Nos. 760-776) to the Eighth Revi- sion title Congenital anomalies (ICDA Nos. 740-759) resulted in a decrease of 315 deaths assigned to the former cause; and transfers from the Seventh Revision title Congenital malforma- tions (ICD Nos. 750-759) to Certain causes of mortality in early infancy (ICDA Nos. 760-769.2, 769.4-772, 774-778) resulted in an increase of 148 deaths to the latter cause. 43 7. DIABETES MELLITUS There were no serious breaks in continuity for mortality statistics for Diabetes mellitus (ICDA No. 250) between 1967 and 1968. The comparability ratio between the Seventh and Eighth Revisions is 0.9971 (appendix I). Seventh Revision Eighth Revision NUE £ Estimated number of Estimated number of Ca f death and b Y io © deaths in column (2) deaths in column (2) mse OF Ceath and caiegory number . Bathe that went to Diabetes that went to in mellitus (ICDA No. 250) other causes (1) (2) (3) (4) All causes ...........cee....... ’ ....001-E999 1,863,149 34,496 1,828,653 Dialyetes MBIILUS .cumsrarimmeisismiss itnmas sist 260 34,597 33,870 727 Other CAUSES .......ccueeeeeeereeeeeeeeeeeeeeeeneeessenns Residual 1,828,552 626 1,827,926 8. ARTERIOSCLEROSIS As shown in the table below, the break in trend between the Seventh and the Eighth Re- vision, introduced for data year 1968, for Arteriosclerosis (ICDA No. 440), was appre- ciable, resulting in a comparability ratio for the Seventh and Eighth Revisions of only 0.8963 (appendix I). The three largest components of the esti- mated 6,824 deaths in the 1966 comparability study that were assigned by the Seventh Revi- sion to General arteriosclerosis (ICD No. 450) but were not reassigned by the Eighth Revision to the comparable title Arteriosclerosis (ICDA No. 440) were as follows: (1)an estimated 2,418 deaths that were transferred by the Eighth Revision to the list title All other diseases of arteries, arterioles, and capillaries (ICDA Nos. 442-444, 446-448); (2) an estimated 1,938 deaths that were transferred from the four-digit Seventh Revision title General arteriosclerosis, Seventh Revision Eighth Revision Estimated number of Estimated number of Number of deaths in column (2) . Cause of death and category number deaths that went to degths incolumn (9) + . . that went to in 1966 Arteriosclerosis (ICDA other causes No. 440) (1) (3) (4) BICAUSES 1..nrnmnmmssvmriissiiamsvsinstunnere 001-E999 1,863,149 34,873 1,828,276 General arteriosclerosis. ............cocuveeueeeerneeessueesnenn. 450 38,907 32,083 6,824 Other Causes........ccceevueeeeueeeireeeecreeecaeeeeaneenns Residual 1,824,242 2,790 1,821,452 with mention of gangrene as a consequence (ICD No. 450.1), to the Eighth Revision title Arterio- sclerotic gangrene (ICDA No. 445.0); and (3) an estimated 1,209 deaths that were transferred from General arteriosclerosis (ICD No. 450) to Chronic ischemic heart disease (ICDA No. 412). The reason for the transfer of the 2,418 deaths from General arteriosclerosis to All other diseases of arteries, arterioles, and capillaries (ICDA Nos. 442-444, 446-448) is that the Eighth Revision provides that if Arteriosclerosis (ICDA No. 440) is reported as the underlying cause of any condition in ICDA Nos. 443, 444, and 446 , the combination of arteriosclerosis and the condition is to be coded, and the death assigned to the condition. As mentioned above in section 1, “Diseases of Heart,” the estimated 1,209 deaths were transferred from General arteriosclerosis (ICD No. 450) to Chronic ischemic heart disease (ICDA No. 412) because the Eighth Revision provides that if Arteriosclerosis (ICDA No. 440) is jointly reported with Chronic ischemic heart disease (ICDA No. 412), the combination is coded, and the death is assigned to Chronic ischemic heart disease (ICDA No. 412). The major shift in the opposite direction from the changes described above was the transfer of an estimated 1,498 deaths from the Seventh Revision title Functional disease of heart (ICD No. 433) to the Eighth Revision title Arteriosclerosis (ICDA No. 440). As also mentioned under section 1, “Diseases of Heart,” this resulted primarily from the dropping in the Eighth Revision of the priority given in the Seventh Revision to Functional disease of heart (ICD No. 433) over General arteriosclerosis (ICD No. 450). 9. BRONCHITIS, EMPHYSEMA, AND ASTHMA All Bronchitis, Emphysema, and Asthma (ICDA Nos. 490-493) The comparability ratio for this group of causes between the Seventh and Eighth Revi- sions is 1.0034 (appendix I). Although this ratio is close to 1.000, the corresponding factors for the three components of this group of diseases vary considerably from 1.000. Before describing the classification changes and trends of the death rates for each of these three components, the reasons for the introduction by the United States for data year 1969 of a fourth component will be given. This new four-digit component is Chronic obstructive lung disease (ICDA No. 519.3). The assignment of deaths by physicians and other medical certifiers to the general title Seventh Revision Eighth Revision Estimated number of ; E Number of deaths in column (2) stimiaiad numberof “ deaths in column (2) Cause of death and category number deaths that went to Bronchitis, that Went to in 1966 emphysema, and asthma other causes (ICDA Nos. 490-493) (1) (2) (3) (4) A COURS co iivivsiininivismoimissiiniibannnisnninnpine 0Q1-E999 1,863,149 29,841 1,833,308 Comparable causes ............c.ccevunnnne. 241,501,502,527.1 29,740 28,602 1,138 ASTM. iii ieee eee eee e eee e sae esa ee eae 241 4,324 4,089 235 Bronchitis, chronic and unqualified............... 501,502 5,164 4,894 270 Emphysema without mention of bronct.itis ...... 527.1 20,252 19,619 633 Other COUSES ..uvsimeusmsmrmisssmisisissenssrormsissiness Residual 1,833,409 1,239 1,832,170 45 “chronic obstructive lung disease” (COLD) or to ‘““chronic obstructive pulmonary disease” (COPD) did not start to become fashionable until the late 1960’s or early 1970’s. For the years the Seventh Revision was in use (1958-67) “chronic obstructive lung disease” was indexed as an inclusion term under Other (ICD No. 527.2), which in turn was a four-digit code under Other diseases of lung and pleural cavity (ICD No. 527). The number of deaths attributed to ICD No. 527.2 remained almost stable between 1958 and 1967, increasing only from 2,482 for 1958 to 2,596 for 1967. It should be stressed that there were a con- siderable number of other terms in addition to “chronic obstructive lung disease” that were indexed as inclusion terms under ICD No. 527.2. Some, but not all of these terms, are shown in the section below, extracted from the Inter- national Classification of Diseases (Seventh Revi- sion): Acute oedema of lung Acute pulmonary oedema Hernia of lung Mediastinitis (acute) (chronic) Stenosis of: bronchus trachea Ulcer of bronchus heart disease NOS or | without mention of heart failure This title excludes acute oedema of lung with mention of any condition in 434.4 or 782.4 (434.2) and chronic or unspecified pulmonary oedema (522). Nevertheless, as stated above, there were only 2,482 deaths for 1958 and 2,596 deaths for 1967 assigned to ICD No. 527.2. During the late 1960s persons involved in coding the medical section observed that “chronic obstructive lung disease” was appearing with increasing frequency on the death certificate. By this time the Eighth Revision had already been adopted by the 1965 International Conference for the Revision of the International Classification of Diseases. The Index to the 46 Eighth Revision (Volume 2) provided that chronic obstructive lung disease be assigned, again along with a number of other inclusion terms, to Other diseases of lung (ICDA No. 519.2). The National Center for Health Statis- tics, to separate out of ICDA No. 519.2 deaths attributed to chronic obstructive lung disease, introduced with data year 1969 the following special four-digit category: *Chronic obstructive lung disease (ICDA No. 519.3) (which is marked with an asterisk that indicates it was introduced independently by the United States). As data for later years became available it was found that the number of deaths attributed to this cause increased rapidly, rising from 2,704 deaths for 1969 to 6,321 deaths for 1971. To provide that deaths would not be assigned to *Chronic obstructive lung disease (ICDA No. 519.3) if a more specific diagnosis such as chronic bronchitis, emphysema, or asthma also appeared on the death certificate, the coding procedures were updated for 1971 and 1972 data years in accordance with the link- ages below:7 *519.3 Chronic obstructive lung disease with- out mention of asthma, bronchitis, or emphysema Excludes conditions in 519.3 with condi- tions in: 490 Bronchitis (491)(Chronicbronchitis) 491 (Chronic bronchitis) (491) 492 (Emphysema) (492) 493 (Asthma) (493) But the limitation imposed by these linkage provisions did not alter the upward trend in the number of deaths assigned to *Chronic obstruc- tive lung disease without mention of asthma, bronchitis, or emphysema (ICDA No. 519.3). The number of deaths assigned to ICDA No. 519.3 increased from 6,321 for 1971 to 11,334 for 1973. This special four-digit subdivision for *Chronic obstructive lung disease without mention of asthma, bronchitis, or emphysema (519.3) includes the following inclusion terms: *519.3 Chronic obstruction (disease): airway (s) lung (s) Without pulmonary mention of respiratory asthma (493) bronchitis (chronic) (491) emphysema (492) only Obstructive (chronic): airway disease broncho-pulmonary disease lung disease pulmonary disease respiratory disease Medical textbooks for decades have included a section on ‘“‘chronic obstructive lung disease,” a title that embraces a number of clinical syn- dromes of varying etiology and pathology, with the common feature of increased hindrance to the flow of air out of the lungs resulting from an intrapulmonary condition—including asthma, bronchitis, and emphysema. But until recent years physicians almost always entered on the death certificate a specific diagnosis such as emphysema, bronchitis, or asthma instead of this generalized term. Rarely in the history of classification of diseases has such a reversal as this one been observed—with an increasing number of medical certifiers entering on the certificates a group title for a complex of diseases instead of a specific diagnosis. It is believed that the general term “chronic obstructive lung disease” is most often used by medical certifiers when they are in doubt as to whether the specific underlying cause of death was emphysema or chronic bronchitis. Major Components Chronic and unqualified bronchitis (ICDA Nos. 490, 491).—Based on the sample of deaths in 1966 coded by the Seventh and Eighth Revi- sions, there were 5,484 deaths assigned to this cause by the Eighth Revision and 5,164 deaths assigned to the comparable title (ICD Nos. 501, 502) by the Seventh Revision, giving a ratio of 1.0620 (appendix I and table below). The deaths not assigned by the Seventh Revision to ICD Nos. 501, 502 that were assigned by the Eighth Revision to ICDA Nos. 490, 491 were trans- ferred from a number of Seventh Revision titles including Asthma (ICD No. 241) and Bronchi- ectasis (ICD No. 526). An estimated 41 deaths assigned by the Seventh Revision to Asthma (ICD No. 241) were transferred by the Eighth Revision to Chronic bronchitis (ICDA No. 491). By the Seventh Revision only that asthma not indicated as allergic with mention of bronchitis (acute or chronic) was considered to be linked with bronchitis: that is, a death attributed to such a combination of causes was assigned to bronchi- tis. But by the Eighth Revision allergic asthma, as well as asthma not indicated as allergic, when selected as the presumptive underlying cause of death, is linked with bronchitis: that is, the death is assigned to bronchitis. About 132 of the deaths assigned by the Eighth Revision to Chronic bronchitis (ICDA No. 491) were assigned by the Seventh Revision to Bronchiectasis (ICD No. 526). In the Seventh Revision bronchiectasis was linked with mention Seventh Revision Eighth Revision Estimated number of deaths in column (2) Estimated number of Number of 5 . that went to Chronic deaths in column (2) Cause of death and category number Lun and unqualified that Went to n bronchitis (ICDA Nos. other causes 490,491) (1) (3) (4) BIALGAUSES wvvvvraninramsssnmmins ssi mis a ies 001-E999 1,863,149 5,484 1,857,665 Bronchitis, chronic and unqualified................. 501,502 5,164 4,851 313 OLRBr CAUSES ....ovvisvimmssrsmsssansrssnsasnussasnsensansssnens Residual 1,857,985 633 1,857,352 47 Seventh Revision Eighth Revision Estimated number of Estimated number of Number of s v deaths in column (2) deaths in column (2) Cause of death and category number eal, that went to Emphysema that went to tn (ICDA No. 492) other causes (1) (2) (3) (4) All CAUSES ...evveereenraeeeeeisiereeeesssneesasesssnaes 001-E999 1,863,149 21,350 1,841,799 Comparable CauSs .....ccursssssisssnsississssrsasssns 241,527.1 24,576 20,980 3,596 ASHI cui iinmmmsnssiimsie sia ws HE RER EER SYST SATA RS 241 4,324 1,425 2,899 Emphysema without mention of bronchitis! ....527.1 20,252 19,655 697 Other CAUSES ......uuuvieerrrneeernrieeernneersnnnreessnnnaees Residual 1,838,573 370 1,838,203 IThis is the title to which deaths in tables 1 and 2 were assigned during 1950-67, when the Sixth and Seventh Revisions were in use. of any condition under Bronchitis (ICD Nos. 500-502); but this linkage was dropped by the Eighth Revision. For example, the deaths de- scribed by the following entries were classified by the Seventh Revision to Bronchiectasis (ICD No. 526) and by the Eighth Revision to Chronic bronchitis (ICDA No. 491): I (a) Chronic bronchitis and bronchiectasis (b ( ~~ 0 I (a) Bronchiectasis (b) Chronic bronchitis (c) Actually, as indicated above, all 132 deaths were assigned to Chronic bronchitis by the Eighth Revision. Emphysema (ICDA No. 492).—There were 21,350 deaths assigned by the Eighth Revision to Emphysema (ICDA No. 492) and 20,252 deaths assigned by the Seventh Revision to the nearly comparable title Emphysema without mention of bronchitis (ICD No. 527.1), giving a ratio of 1.0542 (appendix I). As shown in the table above, this increase of about 5.42 percent in the deaths assigned to emphysema resulted primarily from the transfer of about 1,425 deaths that were assigned by the Seventh Revi- sion to Asthma (ICD No. 241). By the Seventh Revision asthma was not linked with emphysema; while by the Eighth Revision it is linked with emphysema: that is, if asthma is the presumptive underlying cause and there is mention of emphysema on the certifi- cate, the death is assigned to emphysema. Applying the factor 1.0542 to the 1967 death rate for emphysema (10.6) raises it to 11.2—the level it would have reached if the deaths in 1967 had been coded by the Eighth Revision. Asthma (ICDA No. 493).—Based on the sample of deaths in 1966, there were only 3,007 deaths assigned to this cause by the Eighth Revi- sion but 4,324 deaths assigned to it by the Seventh Revision my Eighth Revision Estimated number of Estimated number of Number of : A deaths in column (2) deaths in column (2) Cause of death and category number Regs that went to Asthma that went to (ICDA No. 493) other causes (1) (3) (4) All CAUSES ..evveeeeireieeeecireeesssieeee seinen eas 001-E999 1,863,149 3,007 1,860,142 ASTING co uivs vivir sn i RR I SRE RASS HRT 241 4,324 2,623 1,701 Other CouSBS...ouvmmsmmmmsrvssaussssrsmosssstesssninsss Residual 1,858,825 384 1,858,441 48 Seventh Revision, giving a ratio of 0.6954. (Appendix 1.) As stated above, an estimated 1,425 of the 1,701 deaths that were assigned to asthma by only the Seventh Revision were transferred to An estimated 60 deaths assigned by the Seventh Revision to Asthma were assigned by the Eighth Revision to Acute myocardial infarc- tion (ICDA No. 410), with the greatest number of them (49 deaths) going to the four-digit title Acute myocardial infarction, without mention Emphysema (ICDA No. 492) as a result of the of hypertensive disease (ICDA No. 410.9). provision in the Eighth Revision for linkage of asthma with emphysema. 10. CIRRHOSIS OF LIVER between 1967 and 1968. The comparability ratio is 1.0055 between the Seventh and Eighth Revisions (appendix I). As shown in the table below, there was no serious break in comparability of mortality sta- tistics for Cirrhosis of liver (ICDA No. 571) Seventh Revision Eighth Revision Numb £ Estimated number of Estimated number of c # death and cat b - i 9 deaths in column (2) deaths in column (2) ause of death and category number . pri that went to Cirrhosis that went to n of liver (ICDA No. 571) other causes (1) (2) (3) (4) All CAUSES ..ceeeverreieieeeieeeeeeeeeeeeeeecee eee 1,863,149 26,839 1,836,310 Cirrhosis of lIVer.....cccceiiiieiiiiiieeeeee cece eee 26,692 26,203 489 Without mention of alcoholism 17,320 16,962 358 With alcoholism.........ceeeeeeeiiineenennnn, RY : 9,372 9,241 131 Other causes............ 1,836,457 636 1,835,821 49 11. SUICIDE The comparability ratio for Suicide (ICDA Nos. E950-E959) between the Seventh and Eighth Revisions was 0.9472 (appendix I). About 31 percent of the 3,059 deaths in 1966 assigned by the Eighth Revision to the new cate- gory Injury undetermined whether accidentally or purposely inflicted (ICDA Nos. E980-E989) had been assigned by the Seventh Revision to Suicide. Assignments to this new title are also discussed under Accidents (ICDA Nos. E800- E949) on page 36 and under Homicide (ICDA Nos. E960-E978) on page 52. Seventh Revision Eighth Revision Nurnber of Estimated number of Estimated number of Cc f death and b 7 vay © deaths in column (2) deaths in column (2) Buse OF death.and category number : : that went to Suicide that went to n (ICDA Nos. E950-E959) other causes (1) (2) (3) (4) Al CAUSES covrirsimmmmmissrsrisssms mira est ermrssss 001-E999 1,863,149 20,158 1,842,991 SUICIAR vevvreeieriinneieicrnne es reneee er scieneens E963,E970-E979 21,281 20,046 1,235 Suicide by pPoiSONING .......ccccevueeeernrnneeeeennns E970-E973 5,588 5,192 396 Suicide by hanging and strangulation ................ E974 2,863 2,825 38 Suicide by firearm and explosive .......cccceeevereeenes E976 10,407 10,072 335 Suicide by all other means..... E963,E975,E977-E979 2,423 1,957 466 OLNEY CAUSES vivsvsisnssissnssansiiimrvsssrss TaN Ts es srsERaS Residual 1,841,868 112 1,841,756 12. CONGENITAL ANOMALIES Based on the sample of 1966 deaths coded by both revisions, there were about 18,529 deaths assigned by the Eighth Revision to Con- genital anomalies (ICDA Nos. 740-759) and 18,158 deaths assigned by the Seventh Revision to the most nearly comparable title Congenital malformations (ICD Nos. 750-759), giving a comparability ratio of 1.0204 (appendix I). If the 17,328 deaths in 1967 were increased by this factor, the adjusted figure would be 17,681. The difference between the figure for 1968 (16,793) and the adjusted figure for 1967 (17,681) is 888 deaths. But this reduction of 888 deaths also reflects fewer infants at risk resulting from the decrease in live births in 1968 as compared with the number in 1967. 50 More precise annual measures of the true decline in mortality may be obtained by limiting consideration to deaths occurring at ages under 1 year from Congenital anomalies per 100,000 live births. (About 66 percent of the total number of deaths assigned to this cause occur at ages under 1 year.) Based on the dual-coding study of deaths occurring in 1966, an estimated 12,644 infant deaths were assigned by the Eighth Revision to Congenital anomalies (ICDA Nos. 740-759) and 12,200 infant deaths were assigned by the Seventh Revision to Congenital malformations (ICD Nos. 750-759). Use of the comparability ratio 1.036 (obtained by dividing the 12,644 deaths by the 12,200 to adjust the 1967 Seventh Revision Eighth Revision Estimated number of . Number of deaths in column (2) Estimated number of deaths in column (2) Cause of death and category number deaths that went to Congenital th ) . at went to in 1966 anomalies (ICDA Nos. —— 740-759) (1) (3) (4) Al CBUSBS cvccorrsssinassisismseamisimsinsssininsisnanse 001-E999 1,863,149 18,529 1,844,620 Congenital malformations ............cceeeeeeeeeeeennen 750-759 18,158 17,453 705 Spina bifida and meningocele..........ccceeeueemeenenn.. 751 1,151 1,125 26 Congenital hydrocephalus and other congenital malformations of nervous system and sense organs, 752,753 1,756 1,655 101 Congenital malformations of circulatory system ..754 9,020 8,819 201 Other congenital malformations............. 750,755-759 6,231 5,853 378 Mental deficiency 325 511 256 255 Other CauSes..........ceceeeeeeerrreeeeereneeeeessnnasasaans Residual 1,844,480 820 1,843,660 infant mortality rate for congenital malfor- mations (330.4 per 100,000 live births) to the level it would have had if the Eighth Revi- sion had been used for 1967 results in an adjusted rate of 342.3. The difference between the 1968 rate (315.6 per 100,000 live births) and the adjusted rate for 1967 is 26.7 deaths per 100,000 live births, constituting a real decrease of 7.8 percent between 1967 and 1968 in the risk of dying from this group of causes at ages under 1 year. One of the changes involving congenital anomalies was the transfer of about 256 of the 511 deaths assigned by the Seventh Revision to Mental deficiency (ICD No. 325) to Down’s dis- ease (ICDA No. 759.3). These transferred deaths were assigned by the Seventh Revision to the four-digit subtitle Mongolism (ICD No. 325.4). 51 13. HOMICIDE There was no sizable disruption in the com- parability of statistics for Homicide (ICDA Nos. E960-E978) with the introduction of the Eighth Revision (with a comparability ratio of 0.9969) (appendix I). In the dual-coding study of deaths occurring in 1966 it was found that about 52 of the 211 deaths that were not reassigned to homicide were transferred to the new Eighth Revision title Injury undetermined whether accidentally or purposely inflicted (ICDA Nos. E980-E989). Assignments to this new title are also discussed under Accidents (ICDA Nos. E800-E949) on page 36 and under Suicide (ICDA Nos. E950- E959) on page 50. Deaths assigned by the Seventh Revision to homicide constituted only about 2 percent of all deaths assigned to this new Eighth Revision title. Seventh Revision Eighth Revision Number of Estimated number of Estimated number of c f death and cat b Hn oe ° deaths in column (2) deaths in column (2) Buse of death. anv category Number ; ee , that went to Homicide that went to n (ICDA Nos. E960-E978) other causes (1) (2) (3) (4) AY CAUSES... osiirsisrins sits s its Cees ons vena insen 001-E999 1,863,149 11,570 1,851,579 Homicide, mimosa E964,E980-E985 11,606 11,395 211 Assault by firearm and explosive.........ccccceeennnee E981 6,855 6,773 82 Assault by cutting and piercing instruments...... E982 2,330 2,300 30 Assault by other means .........ccccuuuee E964,E980,E983 2122 2,023 99 Injury by intervention of police .........ccccceeerinns E984 298 298 - EE XBOUEON ccuivssssnininsinsnrors Ritossa ssn mat narunevsssessss E985 1 1 OMNOT CAUSES cssrsssriisnsssivsssmsrrinsineesrrmmmunsunnsnnanss Residual 1,851,543 175 1,851,368 14. NEPHRITIS AND NEPHROSIS The dual-coding study of deaths occurring in 1966 showed that an estimated 10,227 deaths were assigned to Nephritis and nephrosis (ICDA Nos. 580-584), whereas the number assigned by the Seventh Revision to this title (ICD Nos. 590-594) was 11,540 deaths, giving a compara- bility ratio of only 0.8862 (appendix I). If the 10,941 deaths in 1967 were decreased by applying this factor, the adjusted figure would be 9,696, giving a reduction between 1967 and 1968 of only 4.0 percent. Because of the increase in the population, a more precise measure of the amount of reduc- 52 tion in mortality may be obtained by applying the comparability ratio to the death rate rather than to the number of deaths. Using the adjusted death rate for 1967 (4.9 deaths per 100,000) gives a reduction between 1967 and 1968 of 4.1 percent. As shown in the table below, an estimated 1,476 of the 10,376 deaths that were assigned by the Seventh Revision to Chronic and unspecified nephritis and other renal sclerosis (ICD Nos. 592-594) were not reassigned by the Eighth Revision under Nephritis and nephrosis (ICDA Nos. 580-584) (appendix I). An examina- Seventh Revision Eighth Revision Estimated number of 3 Number of deaths in column (2) Becimaras numbers) Cause of death and category number deaths that went to Nephritis 0 . that went to in 1966 and nephrosis (ICDA Other CouTas Nos. 580-584) (1) (2) (3) (4) All CAUSES .....covirieeeeiereeeeecee eee 001-E999 1,863,149 10,227 1,852,922 Nephritis and NephrOSiS...............rrroorrvovvvnn.. 590-594 11,540 9,955 1,585 Acute NePhritiS.....cccccuueeeeeiiureeeieieeeeeeeeaeeeeeesennnnn 590 485 462 23 Nephritis with edema, including nephrosis ........... 591 679 593 86 Chronic and unspecified nephritis and other renal sclerosis...... RETR a TERRA SRR TEE 592-594 10,376 8,900 1,476 Other CaUSES.......ccceeeeeeeeecereeeeeeeeeeeeeee essen, Residual 1,851,609 272 1,851,337 tion of the major components of these 1,476 deaths shows that they were assigned by the Eighth Revision as follows: (1) an estimated 986 of the deaths were transferred to the list title Other diseases of urinary system (ICDA Nos. 591, 593, 595-599), and (2) an estimated 152 deaths were transferred by the Eighth Revision to Hypertensive renal disease (ICDA No. 403). It should be noted that the subgroup of categories under the title “Other diseases of urinary system” does not include Infections of kidney (ICDA No. 590). Association of Hypertension with Nephritis and nephrosis.— About 84 percent of all deaths attributed to Nephritis and nephrosis (ICDA Nos. 580-584) were assigned to chronic and unqualified nephritis (ICDA Nos. 582, 583). These conditions are frequently associated with hypertension. The percentage of deaths attributed to category numbers 582 and 583 was somewhat higher for the population of races other than white (85.9 percent) than for the white population (83.1 percent). Physicians report that since the kidney is frequently involved in forms of hypertension, it is often difficult to be sure of the difference between primary change causing the hyper- tension and secondary damage because of the hypertension. Although hypertension may complicate any form of renal disease, clinical histories show that it is particularly common and severe in glomer- ular diseasc.8 Multiple-cause data based on a sample of deaths in 19559 show that for the white popula- tion hypertension was reported as a contribu- tory condition in 12.7 percent of all deaths assigned to Chronic and unspecified nephritis and other renal sclerosis (ICD Nos. 592-594). The corresponding percentage for the popula- tion of races other than white was even higher— 18.1 percent. It is believed that figures for hypertension as a contributory condition of nephritis and nephrosis are underreported, particularly for the population of races other than white. It should be noted that the coding proce- dures in effect with the Eighth Revision provide, as did the procedures in effect with the Seventh Revision, that conditions of renal damage classi- fiable to nephritis and nephrotic syndrome (ICDA Nos. 580-583) be selected as the under- lying cause of death when the renal condition was a consequence of Essential benign hyper- tension (ICDA No. 401). Within this fact may be found an explanation of why the mortality from Nephritis and nephrosis continues to be higher for the population of races other than white, for it is well known that this population is afflicted relatively more frequently with hypertension than is the white population. Consistent with this prevalence differential by color, hyper- tension is also reported as the underlying cause of death relatively more frequently for the population of races other than white than for the white population (figure 1). 53 ae — — 20.0 r 19.0 — a Foe "3 Ch 5 \ 17.0 \ oo Other male \ : FY / r \ 2 5 16.0 — N 5s 5 r \ raat 15.0 |— \ - \ A \ / X \ / 14.0 |— \ / \ : \. 7 \ \/ he 13.0 ~ ~, Zz Ny y e 120 — Other female » ™ < r N\ = A gS 11.0 LY oo \ | 8 10.0 — L~ 5 \ 8 NN. - ~ | x 90 \ | w \ a \ w = 80 I— < «© | | | | | | | | | { | 1.0 1 Ll 1 | 1 1 1 | 1 1 dl | 1 L 1 1 "| | 1950 1955 1960 1965 1970 | YEAR | | [For 1968 and 1969 rates are based on deaths assigned to category numbers 400, 401, 103 of the Eighth Revision International Classification of Diseases, Adapted for Use in the United States; and for 1950-67 they are based on deaths assigned to | category numbers 444-447 of the Sixth and Seventh Revisions, adopted in 1948 | and 1955, respectively. Rates were adjusted by the direct method, using the total | population of the United States in 1940 as the standard population. The popu- | lation is classified in the following age groups (in years): under 1, 1-4, 5-14, 15-24, 25-34, 35-44, 45-54, 55-64, 65-74, 75-84, 85 and over.] Figure 1. Age-adjusted death rates for Hypertension, by color and sex: United States, 1950-69. Except for the break in continuity of mor- tality statistics between 1957 and 1958 (with the introduction of the Seventh Revision) and the lesser break between 1967 and 1968 (with the introduction of the Eighth Revision), the mortality trend for hypertension as the under- lying cause of death was clearly downward 54 during 1950-69. The comparability ratio for hypertension between the Sixth and Seventh Revisions was 1.33, resulting primarily from the dropping of the preference given by the Sixth Revision to Arteriosclerotic heart disease so described (ICD No. 420.0) over Other hyper- tensive disease (ICD Nos. 444-447).5 Between the Seventh and Eighth Revisions the compara- bility ratio was 0.8199. An estimated 1,238 deaths assigned by the Seventh Revision to Hypertension with arterio- lar nephrosclerosis (ICD No. 446) were trans- ferred by the Eighth Revision to titles other than Hypertension (ICDA Nos. 400, 401, 403). About 899 of these 1,238 deaths were trans- ferred to Chronic ischemic heart disease with or without cardiovascular disease with hypertensive disease (ICDA No. 412.%1). The transfer of these 899 deaths constituted a reinstatement of the Sixth Revision priority that Arteriosclerotic heart disease so described (ICD No. 420.0) had over Hypertension with arteriolar nephrosclerosis without mention of heart (ICD No. 446). This priority given to ICD No. 420.0 was dropped by the coding rules introduced with the Seventh Revision for 1958. It should be noted that the Seventh Revision also eliminated the words “without mention of heart” from the Sixth Revision title corre- sponding to ICD No. 446: thus the Seventh Revision title is reduced to Hypertension with arteriolar nephrosclerosis (ICD No. 446). As used with the Sixth Revision, the words “with- out mention of heart” were not to be inter- preted literally but were defined to mean without mention of only the following types of heart disease: Other myocardial degeneration (ICD No. 422), Functional disease of heart (ICD No. 433), and Other and unspecified diseases of heart (ICD No. 434). In other words the phrase was not to be taken to mean “without mention of coronary heart disease.” Comparison of mortality from Infections of kidney and from Nephritis and nephrosis. -A number of researchers have called attention to the fact that while the total death rate for Nephritis and nephrosis decreased, the death rate for infections of the kidney turned upward, at least until about the middle of the 1960’. Waters!0 states: Seventh Revision Eighth Revision — ye i een Estimated number of . Number of deaths in column (2) Estimated number of deaths in column (2) Cause of death and category number deaths that went to q 3 that went to in 1966 Hypertension (ICDA Nos. other causes 400,401,403) (1) (2) (3) (4) BI CBUSOS vrvsssiursrmmmsnsbsssniiss cn smaiss ss sonmmns 001-E999 1,863,149 9,331 1,853,818 Other hypertensive disease ............cccueeeeuuen.... 444-447 11,380 8,638 2,742 Hypertension with arteriolar nephrosclerosis........ 446 6,873 5,635 1,238 Essential hypertension and other hypertensive disease without mention of heart .......... 444,445,447 4,507 3,003 1,504 Other CauSes.......c..ceeeveeecueeecieeeeeee eee eeereannn Residual 1,851,769 693 1,851,076 In England and Wales age-specific death rates for nephritis and nephroses have decreased while those for infections of the kidney have increased during the same period (1949-65). This seems to be due at least in part to a change in diagnostic termi- nology. The death rates for nephritis and nephrosis and infections of the kidney com- bined show a decrease in men at all ages and in women below the age of 65 years. Reasons for the increase in deaths recorded as from infections of the kidney include the possibility that infection is more commonly sought for now that it is amenable to treat- ment, the increased number of cases coming to necropsy, and the association between analgesics and renal disease. In the United States, as in England and Wales, the death rate for Infections of kidney (ICDA No. 590) did rise throughout the 1950’s and the early 1960’. This rise continued for 1958 and the early years of the 1960s, despite the fact that with the introduction of the Seventh Revision for 1958 about 2 percent fewer deaths were assigned to Infections of kidney than had been assigned to this title by the Sixth Revision. But during the carly sixties the upward trend in the United States for Infec- tions of kidney leveled off and then started downward (figures 2, 3A, and 3B). The coding procedures in effect with the Eighth Revision, introduced for data year 1968, resugt in the assignment of about 3 percent more deaths to Infections of kidney (ICDA No. 590) than had been assigned to this cause by the Seventh Revision. The transfer of 120 deaths from Other diseases of urinary system (ICD Nos. 601, 603, 605-609) to the Eighth Revision title Infections of kidney (ICDA No. 590) is believed to reflect primarily the transfer of a number of inclusion . oo] | v Er | 9.0 | — ae 7 r Sa 7 z 80 (— Other male, ,«xs+ “ Sr ° . LW < 70 |— 2 wi Other female = | i a = | | 8 | 6.0 [=] | 8 o S 50 | « bs a TOTAL ( © 4.0 + White male | e met =~ ~ «c Lf ~ | 3.0 |— XxX” _ ~~ ~I -3 [ > -— V/hite female 20 — - = = | what tide sg Loa ww wd ep [ 1950 1955 1960 1965 1970 YEAR [For 1968 and 1969 rates are based on deaths assigned to category number 590 of the | Eighth Revision International Classification of Diseases, Adapted for Use in the United States; and for 1951-67, they are based on deaths assigned to category number 600 of the Sixth and Seventh Revisions, adopted in 1948 and 1955, respectively. Rates were adjusted by the direct method, using the total population of the United States in 1940 as the standard population. The population is classi- fed in the following age groups (in years): under 1, 14, 5-14, 15-24, 25-34, 35-44, | 45-54, 55-64, 65-74, 75-84, 85 and over.) sr —— = — — — _ Figure 2. Age-adjusted death rates for Infections of kidney, by color and sex: United States, 1951-69. 55 56 MALE FEMALE 200.0 — 200.0 — 85 and over \ 85 and over 100.0 100.0 |— 90.0 — 90.0 — \ 80.0 (— 80.0 | 70.0 — 70.0 60.0 |— ~ 60.0 | / —— willl Se- 50.0 — >’ 50.0 |— 2 — Lu 40.0 (— / 40.0 | Ville i / J 30.0 — pr 30.0 | _7 J Re ’ ’ 20.0 | 200 | af / ’ 100" 100 | ra 9.0 f— 9.0 | ~ 8.0 | 80 | 5.0 ~~ > 5.0 J i 0 \ 0 — 1 40 | 40 a 45-54 30H 3.0 — 20 ,\ RATE PER 100,000 POPULATION \ > \ ® 201 A \ . \ oY i~_’ ’~ / Under 1 3 36-44 / ~ VV AN FO A mm \ oN AAA — \ 10 TNs \ fd IN), iy 10 | ot A — NA (a \ i NTN 2 \ 9 fee road” v= ON I 9 [= wa \ ! IN a \ Bile: = \/ VN 8 | . . / > L \ 7 / \ \ : A 7 / \ Y 4 . Fs A 25.34 ~ oN p 6+ 7 / ’ 7 — Lay | T 6 [= —-—y\ / \ : . or : . b+ 1 sl | y \./ Pc vo 5 pS ~\ / . / Al \ Under 1 \ +f : ns pererrieeeees J NJ hy TEA oN Js \ iy N15 : / 3 al § { Fi ye wld \ the errensiiiat 1528 § ¥ Xx a A No \ A : \ ! \ NERY ‘\ JN \ A 4 ; A FN MY , \ | Po FY OINTV Vo } ! R PA FN von \/ 2)! ered WL YY La 2} ¥ \ vA abo bbe a 3d Mews igo a lov ca bn YX ded 1950 1955 1960 1965 1970 1950 1955 1960 1965 1970 YEAR YEAR [For 1968 and 1969 rates are based on deaths assigned to category number 590 of the Eighth Revision International Classification of Discases, Adapted for Use in the United States; and for 1951-67 they are based on deaths assigned to category number 600 of the Sixth and Seventh Revisions, adopted in 1948 and 1955, respectively. Because of low frequencies, rates are not shown for children aged 1-14 years.) Figure 3A. Death rates for Infections of kidney among the white population, by age and sex: United States, 1951-69. MALE FEMALE 200.0 — 200.0 — 85 and over 100.0 |— 100.0 — 85 and over 90.0 — 90.0 80.0 — 80.0 70.0 — 70.0 60.0 — 60.0 50.0 — 50.0 40.0 — 40.0 30.0 — 30.0 20.0 — 20.0 ‘ - —_— < < N— 10.0 : 10.0 9.0 |— - \ ; 9.0 8.0 (— . / \ 8.0 720 ; 7.0 z 6.0 / I\, i 6.0 E | I As ’ ? i “VW < 58 i ISON | 50 5 Ao ON 5 a A AI vA NL oe < J \/ \; | Vo I, 8 NJ Wy L ! Under 1\ / \ ’ i AN Sg so 1 I dT " I x 3.0 2 i 4 \ z iN \ > 20} i /\ A ! 20 3 I NT VS ; I \ [Y 1.0 y 1.0 == | 9 B= 1 8 ol ~ ~ 3 ] ”\ N 611 dN dN 6 Vo FY I 3 5 oy \ \ Jd N 5 [— 1824} yf 5 ! X } \ J | \ 4 \ ¥ 4 ; } ! ] \ 3 cesnncrnen 3 } 3 \ \ yo \ \ bod \ \ 2 i 2 vo \ FE Vd VY \ | } | \ }} \) \! }\} \ Tt St tintin) J aly oe by yy 1950 1955 1960 1970 1950 1956 1960 1965 1970 YEAR YEAR [For 1968 and 1969 rates are based on deaths assigned to category number 590 of the Eighth Revision International Classification of Diseases, Adapted for Use in the United States; and for 1951-67 they are based on deaths assigned to category number 600 of the Sixth and Seventh Revisions, adopted in 1948 and 1955, respectively. Because of low frequencies, rates are not shown for children aged 1-14 years. ] Figure 3B. Death rates for Infections of kidney among the population other than white, by age and sex: United States, 1951-69. 57 Seventh Revision Eighth Revision Estimated number of : Number of deaths in column (2) Estimated aimiber of : deaths in column (2) Cause of death and category number deaths that went to Infections i ® that went to in 1966 of kidney (ICDA other causes No. 590) (1) (2) (3) (4) AA CAUSES vin ivsiommissvsvmisistavassspampesnsnsiassass 001-E999 1,863,149 9,772 | 1,853,377 Infections OF KIAN@Y .u..nismssesssmssissnssvesrseienvenenee 600 9,498 9,157 341 Other diseases of urinary system ....... 601,603,605-609 3,207 120 3,087 OUNBE CBUSES ...ocorrrrrmmrarssrmmsssarsmsssssmsbbsasodsbbnsssss Residual 1,850,444 495 1,849,949 terms under the Seventh Revision title Other discases of kidney and ureter (ICD No. 603) to the Eighth Revision title Other pyelonephritis, pyelitis, and pyelocystitis (ICDA No. 590.1). Among these transferred terms were necrosis, kidney, and renal papillitis. Applying the comparability ratio between the Seventh and Eighth Revisions of 1.0288 to the 1967 death rate for Infections of kidney (4.6 deaths per 100,000 population) gives an adjusted rate for 1967 of 4.7—about the same rate as that for 1968 (table 1). The puzzling relationship between the trends for mortality from Nephritis and nephrosis and Infections of kidney may be made clearer with the availability of multiple-cause trends for these two groups of causes. Multiple-cause data from the study of deaths in 1955 (the most recent year for which such data have been published) show that only about 32 percent of the deaths for which the certificates have an entry of Infec- tions of kidney are assigned to this causc.? The same study shows that about 52 percent of the deaths for which the certificates have an entry of Nephritis and nephrosis are assigned to this cause. 15. PEPTIC ULCER A reduction in the assignments of deaths to the Eighth Revision title Peptic ulcer (ICDA Nos. 531-533) from the number that had been assigned to the comparable Seventh Revision title Ulcer of stomach and duodenum (ICD Nos. 540, 541) resulted in a comparability ratio of 0.9856 (appendix I). The major shift in this reduction (based on the study of deaths occurring in 1966) was the transfer of an estimated 498 deaths from those assigned by the Seventh Revision to Ulcer of stomach (ICD No. 540) to the Eighth Revision title Gastrojejunal ulcer (ICDA No. 534). The transfer of these 498 deaths resulted from a 58 change in assignment of gastrointestinal hemor- rhage when the site of the ulcer was not specified. By the Seventh Revision certificates with such entries were assigned to Ulcer of stomach (ICD No. 540). Another estimated 51 deaths that were assigned by the Seventh Revi- sion to Ulcer of duodenum (ICD No. 541) were assigned to Gastrojejunal ulcer (ICDA No. 534). These changes are reflected in the great increase in the number of deaths for 1968 assigned to Gastrojejunal ulcer (ICDA No. 534) (721 deaths) over the number of deaths (196) in 1967 assicned to the same Seventh Revision title Gastrojejunal ulcer (ICD No. 542). Seventh Revision Eighth Revision Estimated number of . Number of deaths in column (2) Estimated number of : deaths in column (2) Cause of death and category number deaths that went to Peptic : that went to in 1966 ulcer (ICDA Nos. oihier couses 531-533) (1) (2) (3) (4) All CAUSES «cee 001-E999 1,863,149 10,172 1,852,977 Ulcer of stomach and duodenum..................... 540,541 10,321 9,429 892 Ulcer of stomach... 5,599 4,896 703 Ulcer of duodenum 4,722 4,533 189 Other CAUSES.......cccueeeeeeeecnreee cee ceeeeeeeaeeeaeeenes Residual 1,852,828 743 1,852,085 The Eighth Revision group title Peptic ulcer (ICDA Nos. 531-533) has the following three subtitles (shown with the number of deaths assigned to them for 1969): Number of deaths in 1969 Subtitle and category number 3,719 4,381 1,212 It should be noted that the number of deaths coded by the Eighth Revision to Ulcer of stomach (ICDA No. 531) is greatly reduced from the number assigned to this same title (ICD No. 540) by the Seventh Revision. The study based on deaths occurring in 1966 shows that the comparability ratio for Ulcer of stomach (ICDA No. 531) is 0.6515. The major component of the 2,084 deaths assigned to Ulcer of stomach (ICD No. 540) by the Seventh Revision but not reassigned to this title by the Eighth Revision was 1,293 deaths that were assigned by the Eighth Revision to the new title Peptic ulcer, site unspecified (ICDA No. 533). By the Seventh Revision, when ulcer (ruptured) with site unspecified was reported with indication that the ulcer was internal, such as resulting peritonitis, the ulcer was considered “peptic,” with assignment to Ulcer of stomach (ICD No. 540). As mentioned above, about 498 of the 2,084 transferred deaths were assigned by the Eighth Revision to Gastrojejunal ulcer (ICDA No. 534). Seventh Revision Eighth Revision Estimated number of Estimated number of Number of . . deaths in column (2) deaths in column (2) Cause of death and category number es that went to Ulcer of that went to n stomach (ICDA No. 531) other causes (1) (2) (3) (4) 1,863,149 3,648 1,859,501 Ulcer of Stomach ..........ooeeeeeeiiiieeeeeeeeeeeeeeeanns 540 5,599 3,515 2,084 Other (CAUSES... cms mmmmarssinssisivosmanansssassranse Residual 1,857,550 133 1,857,417 REFERENCES 1 U.S. Bureau of the Census: Comparison of cause-of- death assignments by the 1929 and 1938 Revisions of the International Lists, deaths in the United States, 1940, by H. L. Dunn and W. Shackley. Vital Statistics - Special Reports, Vol. 19, No. 14. Washington, D.C, June 1944. 2National Center for Health Statistics: Comparability ratios based on mortality statistics for the Fifth and Sixth Revisions: United States, 1950, by M. M. Faust and A. B. Dolman. Vital Statistics Special Reports, Vol. 51, No. 3. Public Health Service. Washington, D.C., Feb. 1964. 3National Center for Health Statistics: Comparability of mortality statistics for the Fifth and Sixth Revisions: United States, 1950, by M. M. Faust and A. B. Dolman. Vital Statistics Special Reports, Vol. 51, No. 2. Public Health Service. Washington, D.C., Dec. 1963. 4 National Center for Health Statistics: Comparability of mortality statistics for the Sixth and Seventh Revi- sions: United States, 1958, by M. M. Faust and A. B. Dolman. Vital Statistics—Special Reports, Vol. 51, No. 4. Public Health Service. Washington, D.C., Mar. 1965. 5World Health Organization: Manual of the Interna- tional Statistical Classification of Diseases, Injuries, and Causes of Death, Based on the Recommendations of the Seventh Revision Conference, 1955. Geneva. World Health Organization, 1957. “Medical certification and rules for classification” used with the Seventh Revision are shown on pp. 357-371. 6World Health Organization: Manual of the Interna- tional Statistical Classification of Diseases, Injuries, and 60 Causes of Death, Based on the Recommendations of the Eighth Revision Conference, 1965. Geneva. World Health Organization, 1967. “Medical certification and rules for classification” used with the Eighth Revision are shown on pp. 415-436. "National Center for Health Statistics: 1973 instruc- tion manual for underlying cause of death, Nosology Guidelines. Rockville, Md. Public Health Service, Nov.- Dec. 1972. 8Wrong, O. M.: Glomerular disease, in P. B. Beeson and W. McDermott, eds., Textbook of Medicine (Cecil- Loeb), Vol. 11. Philadelphia. W. B. Saunders Company, 1971. pp. 1177-1198. 9 National Center for Health Statistics: Vital Statistics of the United States, 1955, Supplement: Mortality Data, Multiple Causes of Death. Public Health Service. Wash- ington. U.S. Government Printing Office, 1965. 10Waters, W. E.: Trends in mortality from nephritis and infections of the kidney in England and Wales. The Lancet. Feb. 3, 1968. pp. 241-243. '{ansen, M. H., Hurwitz, W. N., and Madow, W. G.: Sample Survey Methods and Theory, Vol. 1, 2d printing. New York. John Wiley and Sons, Inc., 1957. ch. 4, p. 122. L2Cramer, H.: Mathematical Methods of Statistics. Princeton. Princeton University Press, 1971. table 2, p. 558. 13 Percy, C., Garfinkel, L., Krueger, D. E., and Dolman, A. B.: Apparent changes in cancer mortality, 1968. Public Health Rep. 89(5), Sept.-Oct. 1974. LIST OF DETAILED TABLES Page Table 1. Death rates for 48 selected causes: United States, 1966-69===-cccccccccccaaaa- 62 2. Deaths for 48 selected causes: United States, 1966-=69=====cccccccccaacaaanaaa- 63 3. Death rates for Chronic ischemic heart disease and for four special four-digit categories of this disease, by age, color, and sex: United States, 1968-71---- 64 61 Table 1. Death rates for 48 selected causes: United States, 1966-69 [Rates per 100,000 population. Numbers after causes of death are category numbers of the Eighth Revision International Classification of Diseases, Adapted for Use in the United States, adopted in 1965. Rates for 1968 and 1969 are based on deaths assigned to these Eighth Revision category numbers; rates for 1966 and 1967 are based on deaths assigned to the category numbers of the Seventh Revision, adopted in 1955. These category numbers are shown in the last column of this table. | Category numbers Cause of death and category numbers of Eighth Revision 1969 1968 1967 1966 Seoopaing Io the Revisions All CAUSES === === mmm mmm meee 951.9 965.7 935.7 950,3 | 4. Enteritis and other diarrheal diseases-------=--=c-- 008,009 1.3 1.5 3.8 3.9 543,571,572 Tuberculosis, all formg------=----commoaaoano --010-019 2.8 3.1 3.5 3.9 | 001-019 Syphilis and its sequelae 0.3 0.3 1.2 1.1 | 020-029 Other infective and parasitic diseases-Remainder of 000-136 4.0 4.0 3.5 3.8 | 001-138,Remainder of Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tissues-------ccomocmoaomaaanoo 140-209 160.0 159.4 157.2 155.1 | 140-205 Malignant neoplasms of buccal cavity and pharynx--140-149 3.7 3.6 3.4 3.5 140-148 Malignant neoplasms of digestive organs and Per itoneum= == momo eee een 150-159 46.5 46.8 48.2 48.5 150-156A,157-159 Malignant neoplasms of respiratory system- --160-163 32.7 31.8 29.4 28.0 160-164 Malignant neoplasms of breast-=------coeomomooooaooooo 174 14.4 14.6 14.3 14.1 170 Malignant neoplasms of genital organs-- --180-187 20.3 20.5 20.4 20.6 171-179 Malignant neoplasms of urinary organs------------- 188,189 7.4 7.4 7.4 7.2 180,181 Malignant neoplasms of all other and unspecified EE tt 170-173,190-199 18.7 18.6 18.6 17.9 156B,165,190-199 Leukemia === moon meee eee 204-207 7:2 7.2 7.2 7.2 | 204 Other neoplasms of lymphatic and hematopoietic tissues —= =m mmm eee 200-203,208,209 9.0 8.9 8.3 8.1 | 200-203,205 Benign neoplasms and neoplasms of unspecified MAUL == == == =m mm mee ee emo 210-239 2.3 2.5 2:5 2.5 | 210-239 Diabetes mellitus --== coco mcm ome 250 19.1 19.2 17.7 172.7 260 Anemias === o-oo meee oo --280-285 1.6 1.7 1.7 1.8 | 290-293 Meningitis momo m meme el 320 0.9 0.9 1.0 1.2 | 340 Major cardiovascular diseases----=---=cecoeoomooo_oo 390-448 501.7 512.1 506.5 516.1 | 330-334,400-468 Diseases of heart----------cceeooo 390-398,402,404,410-429 366.1 372.6 364.5 371.2 | 400-402,410-443 Active rheumatic fever and chronic rheumatic heart disease---=-=--=-ocommmo mee 390-398 7.6 8.2 7.2 7.7 | 400-402,410-416 Hypertensive heart disease with or without renal disease - === o-oo m meme 8.1 8.9 25.3 27.7 | 440,441,442 ,443 Ischemic heart disease 331.7 337.6 289.7 292.7 | 420 Chronic disease of endocardium and other myocardial insufficiency ---=--eoomommm ieee. 424,428 3.7 3.9 26.6 27.4 | 421,422 All other forms of heart disease -420-423,425-427,429 15.0 14.0 15.8 15.8 | 430-434 Hypertension--=----ceccemamaaooooo 400,401,403 4,2 4.5 5.6 5.8 | 444-447 Cerebrovascular diseaseS---=--=-ccmoomooomooomano 430-438 102.6 105.8 102.2 104.6 330-334 Arteriosclerosis ---=-oo om mmmm eo 440 16.4 16.8 19.0 19.9 | 450 Other diseases of arteries, arterioles, and capillaries -=-=--commom eee 441-448 12.4 12.4 15.1 14.6 | 451-468 Acute bronchitis and bronchiolitig---==-ceccemmaaaaaao 466 0.6 0.7 0.5 0.5 500 Influenza and pneumonia-=--==--==--- 470-474 ,480-486 33.9 36.8 28.8 32.5 | 480-493 Influenza -=== == comme eee ee eee 470-474 3.0 33 0.7 1.4 | 480-483 Pneumonia --=--m ooo ee 480-486 30.9 33,2 28.0 31.0 | 490-493 Bronchitis, emphysema, and asthma--------=-ccoeomu__ 490-493 15.4 16.6 15.3 15.2 | 501,502,527.1,241 Chronic and unqualified bronchitig-=--===ceeaauaao- 490,491 2.9 3.1 2.7 2,6 | 501,502 Emphysema - == = coo cc mmm eee 492 11.4 12.1 10.6 10.3 527:1 ASEM = == = =m oe ee ee eee 493 1.2 1.3 2.1 2,2 | 241 Peptic ulcer ----mmom moomoo 531-533 4.6 4,7 5.0 5.3 | 540,541 Hernia and intestinal obstruction 550-553,560 3.7 3.0 5.0 S51 560,561,570 Cirrhosis of liver-------ccomomm mee 571 14.8 14.6 14.1 13.6 581 Cholelithiasis, cholecystitis and cholangitis------- 574,575 2.1 2.2 2.2 2.3 | 584,585 Nephritis and nephrosis-------coocmmmmm eo 580-584 4.7 4.7 5.5 5,9 590-594 Infections of kidney--=-=ccoomomom meee 590 4.3 4.7 4.6 4.8 600 Hyperplasia of prostate----=---cocommmmmmm ee 600 1,2 1.3 1.6 1.6 | 610 Congenital anomalies---=-=ccecmomm momo 740-759 8.4 8.4 8.8 9.3 | 750-759 Certain causes of mortality in early infancy -------cccmomme ooo 760-769.2,769.4-772,774-778 21.4 21.9 24.4 26.4 | 760-776 Symptoms and ill-defined conditions----=-=cccmauoaao 780-796 13.0 11.8 12,2 12,2 | 780-795 All other diseases=---=-=-=c--momommmmm eC Residual 50.9 51.1 34.4 34,7 | Residual Accidents =--mmm om om eee E800 -E949 52.6 57.5 57.2 58.0 | EB800-E962 Motor vehicle accidents------ceoeomommamae E810-E823 27.6 27.5 26.7 27.1 | E810-E835 All other accidents------=cocommeaaa_o E800-E807,E825-E949 30.0 30.0 30.4 30.9 | EB800-E802,E840-E962 Suicide =m-m mmo ee el E950-E959 11.1 10.7 10.8 10.9 | E963,E970-E979 Homicide----v--ccmcoaoao -E960-E978 7.7 13 6.8 5.9 | E964,E980-E985 All other external causeS=-------=c-coo-o- -E980-E999 2.5 2.2 --- --- | Residual 62 Table 2. Deaths for 48 selected causes: United States, 1966-69 [Numbers after causes of death are category numbers of the Eighth Revision International Classification of Diseases, Adapted for Use in the United States, adopted in 1965. Deaths for 1968 and 1969 are those assigned to these Eighth Revision category numbers; deaths for 1966 and 1967 are those assigned to the category numbers of the Seventh Revision, adopted in 1955. These category numbers are shown in the last column of this table. | Category numbers Cause of death and category numbers of Eighth Revision 1969 1968 1967 1966 SE Revisions All caAuSeS=---ecremmcceemeeceemeeceeee mmm 1,921,990 | 1,930,082 | 1,851,323 | 1,863,149 ver Enteritis and other diarrheal diseases=-----=-ceecee-= 008,009 2,612 2,940 7,504 74332 543,571,572 Tuberculosis, all forms==re-e-cacecaaan 5,567 6,292 6,901 7,625 | 001-019 Syphilis and its sequelae 543 586 2,381 2,193 020-029 Other infective and parasitic diseases-Remainder of 000-136 8,069 7,958 6,964 7,496 001-138,Remainder of Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tiSSu€S=mmmmcramcamcaacccaccaaann 140-209 323,092 318,547 310,983 303,736 140-205 Malignant neoplasms of buccal cavity and pharynx--140-149 7,553 7,29 6,718 ,800 140-148 Malignant neoplasms of digestive organs and peritoneums meee ecmeeccmmeccceceec ec cmmcmee———— 93,986 93,563 95,320 95,079 150-156A,157-159 Malignant neoplasms of respiratory Sete. 66,038 63,485 58,086 54,934 160-164 Malignant neoplasms of breaste=-eeescecea-- 29,083 29,081 28,217 27,533 170 Malignant neoplasms of genital oTgAnE—. 41,008 40,936 40,408 40,378 171-179 Malignant neoplasms of urinary organs 14,897 14,792 14,656 14,166 130,181 Malignant neoplasms of all other and unspecified 170-173,190-199 37,825 37,247 36,761 35,032 156B,165,190-199 Leukemi@memm mm ocmoo mecca cacao 204-207 14,450 14,375 14,336 14,012 204 Other neoplasms of lymphatic and hematopoietic tiSSUESmm mmm meme emo 200-203,208,209 18,252 17,774 16,481 15,802 | 200-203,205 Benign neoplasms and neoplasms of unspecified NAEUTE mmm mmm momen meme c oom m mcm moe mmmoeeoeeo 4,677 4,948 5,013 4,923 210-239 Diabetes mellitus- 38,541 38,352 35,049 34,597 | 260 Anemiase------ -- 3.318 3,494 3,460 3,452 | 290-293 MeningitiSmememmmc mercer reer meee 1,719 1,707 2,046 2,324 340 Major cardiovascular diseases=--=--=c---o-c-conmoooano 390-448 [1,013,015 | 1,023,399 | 1,002,111 | 1,010,812 | 330-334,400-468 Diseases of hearte-eeececocacacanan 390-398,402,404,410-429 739,265 744,658 721,268 727,002 | 400-402,410-443 Active rheumatic fever and chronic rheumatic heart diseaseeecceccccacccmccc acim eeeean 390-398 15,432 16,358 14,176 15,012 | 400-402,410-416 Hypertensive heart disease with or without renal diseases-==ecmrenmreeennn 16,286 17,698 49,975 54,176 | 440,441,442 ,443 Ischemic heart disease 669,829 674,747 573,153 573,191 420 Chronic disease of endocardium and other myocardial insufficiency-=-e--ceccmcmmm ence 424,428 7,475 7,836 52,697 53,581 | 421,422 All other forms of heart disease==---420-423,425-427,429 30,243 28,019 31,267 31,042 | 430-434 Hypertensione=-re-e-cmcecemmcceannann 400,401,403 8,426 9,063 11,151 11,380 444-447 Cerebrovascular diseaseS==meececcmmocccccocoaaoannn 430-438 207,179 211,390 202,184 204,841 330-334 Arteriosclerosis=em-mmecccmcmmcmce ieee en 440 33,063 33,568 37,564 38,907 | 450 Other diseases of arteries, arterioles, and | capillariesee=cemmccocomcm mamma cmcmemem aaa 441-448 25,082 24,720 29,944 28,682 | 451-468 Acute bronchitis and bronchiolitiSmemmeccacccccacanancnn 466 1,286 1,432 958 987 500 Influenza and pneumonia 470= 474, 480-486 68,365 73,492 56,892 63,615 480-493 Influenza=-=-- 470-474 5,971 7,062 1,475 2,830 | 480-483 Pneumoniaseeeeeeecccccaca cece cacemcccceeaeeeaaa 480- 486 62,394 | 66,430 55,417 60,785 | 490-493 Bronchitis, emphysema, and asthmaeeseeeecceceacaccans 490-493 31,144 | 33,078 30,318 29,740 501,502,527.1,241 Chronic and unqualified bronchitis-=- -=490, i 5,843 6,205 5,306 5,164 501,502 Emphysema=eeeeee= Ses eseeeeeecscscmcscescscsssseeaman- 22,939 24,185 20,875 20,252 527.1 ASthmaseecemcccacaan ncaa cca ra cca mm reece cc cece 493 2,362 2,688 4,137 4,324 241 Peptic ulcer-eemeeececcccacacccannnn mmmmmmemmmeee———— 531-533 9,312 9,460 9,825 10,321 540,541 Hernia and intestinal obstructione=---« -550-553,560 7,500 7,758 9,814 10,078 560,561,570 Cirrhosis of livereseseeececacecccccacccccccoccccccaanan 571 29,866 29,183 27,816 26,692 581 Cholelithiasis, cholecystitis and cholangitis -574,575 4,262 4,385 4,383 4,592 584,585 Nephritis and nephrosise==-eeecececacaanacann- --580-584 9,417 9,311 10,941 11,540 5902594 Infections of kidney==-e-mmececccamacm mca cane ee ean 590 8,750 9,395 9,006 9,498 | 600 Hyperplasia of prostatess-ee-cecemecccmcocomneononaann- 600 2,499 2,647 3,136 3,217 610 Congenital anomalieS==seseceeecoremerace mem necan anne 740-759 17,008 16,793 17,328 18,158 750-759 Certain causes of mortality in early infancy-eseseececccacnccnanaaa. 760~ 769.2,769.4-772,774-778 43,171 43,840 48,314 51,644 760-776 Symptoms and ill-defined conditiong=-==--==--ec-eua- 780-796 26,160 23,656 24,098 23,960 780-795 All other diseases=-emmeecccmccmcccmroaoccnnnnaaan Residual 102,724 102,192 68,163 67,947 Residual AccidentSmeemmmmc armen m ee E800-E949 116,385 114,864 113,169 113,563 | E800-E962 Motor vehicle accidents E810-E823 55,791 54,862 52,924 53,041 | E810-E835 All other accidentS=e==e-e-cescarannnn E800-E807,E825-E949 60,594 60,002 60,245 60,522 | E800-E802,E840-E962 Suicidemm-meannn --- -E950-E959 22,364 21,372 21,325 21,281 | E963,E970-E979 Homicideseescomeaereccnnann E960-E978 15,477 14,686 13,425 11,606 E964 ,E980-E985 All other external causeS==----eececcacacconcannnx E980-E999 5,147 4,315 -——— -—— Residual 63 Table 3. Death rates for Chronic ischemic heart disease and for four special four-digit categories of and sex: United States, 1968-71 [Rates per 100,000 population in specified groups] this disease, by age, color, PLN ®oNOWn 64 Eighth Revision Cause of death, color, sex, and year Total! gn 1-4 number year | years 412 Chronic ischemic heart disease 0.1 0.0 0.1 0.0 0.3 0.0 0.4 0.1 0.1 0.1 0.1 0.1 0.3 0.0 0.3 0.0 0.1 . 0.4 0.0 0.3 0.0 0.4 0.3 - 0.1 0.7 0.1 0.7 - 0.4 - 1.0 0.2 412.12 Chronic ischemic heart disease with or without cardiovascular disease with hypertensive disease 9.6 - - 9.7 - 0.0 9.8 - - 10.2 0.0 - White male: 1971 8.4 - - 8.5 - - 8.5 - - 9.0 0.1 - 10.9 - - 10.9 - - 11.1 - - 1.2 - - 8.7 - - 8.8 - 0.1 9.7 - - 10.0 - - 9.8 - - 10.2 - - 10.3 - - 11.3 - - 412.2? Cardiovascular disease without mention of chronic ischemic heart disease with hypertensive disease 11.0 0.1 - 11.4 - - 11.6 - - 12.4 - - 7.7 0.1 - 8.0 - - 8.2 - - 8.5 - - 10.3 - - 10.6 - - 10.7 - - 11.5 - - See footnotes at end of table. 5-9 years © oo © oo oooo rs) NY olor oI co ooo oOo © oo © He oOoOoOo ooo0o o 11 ot oir 10 o [| o Le oo aa 11 oo oo "100 oo 1 coor R= Vion oii o Le ol 10 [ET ena 15-19 years OrFFN ocooo OHHN ocooo ooo I Or oooo NNN oOoOoOo Rubs oooo o o o te vi toro tio 1o1o oooo ror o o 101 0 o re © ocooo HO oocoo Oro oooo oocoo 25-29 | 30-34 years | years 0.8 2,8 1.0 2.8 0.9 2.8 0.7 2.7 0.9 i 1.1 2.8 0.9 2.9 0.7 2.7 0.4 1.1 0.4 1.0 0.3 0.8 0.2 0.7 3.2 11.6 4.1 11.0 3.9 14.1 4.2 10.9 1.6 4.9 1.5 6.8 2.8 6.5 1.7 9.9 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.0 0.1 0.1 - 0.1 - 0.1 0.0 0.0 0.0 0.0 - 0.0 0.1 0.7 0.1 0.1 - 0.5 0.4 1.2 0.2 - 0.1 0.5 - 0.1 - 0.1 0.2 0.5 0.2 0.7 0.3 0.8 0.2 0.8 0.1 0.1 0.1 0.3 0.2 0.3 0.0 0.3 0.1 0.2 0.0 0.3 0.1 0.2 0.1 0.2 Table 3. Death rates for Chronic ischemic heart disease and for four special four-digit categories of this disease, by age, color, and sex: United States, 1968-71—Con. [Rates per 100,000 population in specified groups] 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85 years years years years years years years years years years years and over 8.4 20.4 40.3 76.0 136.7 238.6 406.3 746.1 1,362.3 2,440.7 »203. 1 8.6 19.5 41.2 76.3 133.1 242.4 424.6 753.7 1,383.2 2,480.4 4,800.4 | 2 8.4 19.3 38.7 75.8 137.9 249.5 437.2 800.2 1,348.1 2,457.2 5,547.1] 3 7.6 19.3 39.0 76.2 140.6 255.0 447.5 819.7 1,377.4 2,470.6 +332 4 9.9 26.4 55.2 103.9 189.6 325.0 539.6 937.1 1,643.0 2,846.6 5,675.31 5 10.2 24.8 55.7 103.5 189.9 328.0 557.2 943.2 1,659.7 2,867.8 5,090.2 6 10.0 24.5 52.2 102.8 187.5 336.2 552.2 988.7 1,618.8 2,760.1 5,640.1 7 9.3 23.9 50.8 103.0 186.4 339.4 558.2 1,012.4 1,630.9 2,752.7 5,620.51 8 2.6 6.3 13.7 29.5 60.9 125.7 252.1 550.0 1,155.3 2,268.5 5,228. 9 2.4 5.8 13.4 30.5 60.7 126.1 265.1 560.5 1,175.5 2,302.1 4,925.3|10 2.4 5.9 12.1 29.5 61.2 126.6 271.6 590.5 1,155.4 2,339.4 6,065.2 (11 2.1 5.8 13.3 28.6 62.6 130.3 280.9 606.2 1,193.7 2,367.2 ,968.6 (12 33.3 65.6 124.2 216.4 353.9 515.7 755.9 1,231.9 1,685.3 2,163.0 3,324.5|13 32.5 71.9 133.6 210.7 347.3 541.7 808.6 1,253.3 1,745.9 2,112.3 2,872.5|14 34.6 70.0 133.3 229.6 365.3 588.1 942.6 1,581.5 1,649.2 1,975.0 2,465.6 |15 28.9 72.8 133.3 224.8 388.4 606.1 982.9 1,630.6 1,599.2 1,925.7 2,785.5|16 16.6 41.5 69.3 131.4 205.8 348.1 562.0 987.5 1,356.3 1,721.4 3,228.9(17 19.9 38.8 76.0 138.2 226.1 374.7 594.0 959.1 1,364.7 1,929.7 2,931.7118 16.6 41.3 72.3 130.8 231.1 432.0 775.1 1,152.2 1,217.2 1,787.8 2,741.3(19 16.6 43.9 82.2 147.6 260.5 460.2 834.0 1,112.6 1;279.% 1,806.3 2,980.6 (20 0.4 0.9 1.8 4,2 8.6 15.9 28.5 53.6 97.5 158.2 271.7121 0.4 0.8 2.2 4.3 8.6 16.4 30.1 56.2 96.3 161.7 261.022 0.4 1.0 1.9 4,2 8.7 16.5 31.9 61.3 97.3 166.1 295.123 0.4 1.0 2.1 4.7 92.3 18.6 36.1 61.8 103.2 167.8 293.1 (24 0.4 1.0 2.4 4.6 10.5 18.2 32.1 55.8 97.1 147.3 234.025 0.4 0.7 2.5 4.5 9.6 19.2 33.5 56.8 96.1 154.6 217.526 0.2 1.0 2.1 4.4 9.8 19.1 33.0 63.2 89.9 151.0 239.227 0.2 1.1 2.1 5.5 10.4 21.5 36.4 63.7 100.7 151.7 249.028 0.) 0.3 0.5 2.2 4.5 10.3 20.9 46.5 93.9 168.7 300.729 0.1 0.4 0.8 2.1 4.9 10.7 22.8 50.1 92.8 168.1 295.3(30 0.2 0.3 0.9 2.1 4.8 10.2 23.2 53.6 98.2 179.4 354.8|31 0.1 0.3 0.8 2.2 5.0 10.3 28.5 53.2 102.9 182.1 339.5(32 2.6 4.5 i 13.9 24.9 35.3 54.0 84.7 121.7 102.7 159.2133 2.2 3.0 7.4 12.0 24.6 32.1 55.7 88.7 119.5 138.5 167.834 1.8 4.5 6.5 18.3 25.1 40.4 76.5 115.3 128.9 131.6 132.835 2.3 5.3 9.1 12.8 26.5 49.9 84.0 111.9 111.0 140.5 145.5(36 1.2 2.3 4.1 10.2 16.6 31.9 51.9 90.2 129.3 143.7 228.9137 0.9 2.9 6.3 14.5 19.7 34.5 50.7 92.4 126.9 151.4 208.5|38 1:1 2.8 4.5 9.9 21.2 36.7 80.1 105.5 127.0 146.9 182.5|39 1:7 2.6 6.6 13.8 25.5 50.1 78.1 117.0 122.9 142.1 215.3 (40 2.7 3.2 5.8 10.9 16.4 25.8 39.6 62.1 92.1 132.6 204.141 1.7 3.4 6.7 10.8 17.9 27.3 42.4 62.3 95.9 141.4 198.242 1.8 3.8 6.8 10.7 18.4 29.8 44.0 67.6 94.5 138.2 232.443 1:7 3.9 7:2 11.8 19.7 30.5 49.5 74.3 102.4 144.9 246.1 |44 0.8 1.6 3.4 7.0 12.9 22.0 33.8 51.4 77.2 109.9 143,745 0.6 1.6 3.8 8.0 13.2 23.4 35.9 52.2 80.3 105.1 139.6(46 0.7 1.7 3.7 7.8 14.5 24.8 35.8 56.8 77.8 105.1 151.747 0.7 1.7 4.5 8.3 14.2 24.9 38.5 61.5 81.9 107.6 160.448 0.4 0.8 2.6 4.6 8.4 14.4 26.5 48.8 85.4 132.6 218.9|49 0.6 1.1 2.4 4.5 9.1 15.5 27.5 49.6 88.0 144.7 214.5|50 0.5 1.3 2.8 4.6 8.2 16.2 28.2 51.8 88.1 145.5 269.5|51 0.5 1.3 2.9 5.0 10.2 16.7 32.5 59.0 97.6 154.4 280.6152 65 Table 3. Death rates for Chronic ischemic heart disease and for four special four-digit categories of this disease, by age, color, and sex: United States, 1968-71—Con. [Rates per 100,000 population in specified groups] sw Non Eighth Revision number Cause of death, color, sex, and year Total’ 412,22 412.3? 412.4 Cardiovascular disease without mention of chronic ischemic heart disease with hypertensive disease—Con. All other male: 1971 Chronic ischemic heart disease with or without cardiovascular disease without mention of hypertensive disease Cardiovascular disease without mention of chronic ischemic heart disease without mention of hypertensive disease Vo no PONS “nos Ooo Hoo PownN Non NowWww» woNN wv =n OVW OY Soo Under 1-4 5-9 10-14 | 15-19 | 20-24 | 25-29 | 30-34 1 year | years | years | years | years | years | years | years - - - - - 0.4 1.5 4,2 - " - - 0.1 0.1 2.3 4.2 - - - - 0.1 0.9 1.3 6.5 - - - - 0.1 0.3 L.7 4.1 0.7 - - 0.1 0.1 0.1 0.7 1.7 - - - - - - 0.6 3.8 - - - - 0.2 0.2 1.6 4.3 ” ws - - - 0.1 1.4 5.8 0.0 0.0 0.0 0.1 0.1 0.2 0.3 2,0 0.1 0.0 0.0 0.0 0.1 0.2 0.6 1.8 0.2 0.0 0.0 0.0 0.1 0.1 0.5 3.6 0.3 0.0 0.0 0.0 0.0 0.1 0.4 1.6 - 0.0 0.0 0,1 0.1 0.2 0.7 2.6 0.1 0.0 - 0.0 0.0 0.2 0.9 2.3 0.3 0.0 - 0.0 0.1 0.1 0.7 2.2 0.3 0.0 0.0 0.0 0.0 0.2 0.6 2.1 - - 0.0 0.0 0.0 0.1 0.2 0.8 0.1 - 0.0 0.0 0.0 0.1 0.3 0.6 0.3 - - 0.0 0.1 0.1 0.1 0.5 0.1 0.0 0.0 0.0 - 0.0 0.1 0.5 0.4 0.3 0.2 - 0.1 0.8 1.4 5.7 - - - - 0.1 0.5 1.3 Set - - - 0.1 0.1 0.7 2:3 5.5 0.3 0.1 0.1 - 0.1 0.6 2.1 4.5 - - 0.1 - 0.3 0.1 0.7 2.2 0.4 - - 0.1 0.2 0.4 0.7 2.4 - - 0.1 - 0.2 - 1.1 1.7 1.0 0,2 - - 0.1 0.1 0.3 3.5 - 0.0 , 0.0 0.0 0.0 0.0 0.1 0.3 - 0.0 0.0 - 0.0 0.0 0.1 0.2 0.1 0.0 - - - 0.0 0.1 0.3 0.1 0.0 0.0 - 0.0 0.0 0.1 0.2 - 0.0 - 0.0 - 0.0 0.1 0.3 - 0.0 0.0 - 0.0 0.0 0.1 0.2 0.1 0.0 - - - 0.0 0.0 0.3 - - - - - 0.0 0.1 0.2 - - 0.0 0.0 0.0 0.0 0.0 0.1 - - - - - 3 0.0 0.1 0.1 0.0 - - - 0.0 0.0 0.1 0.1 0.0 - - - - 0.0 0.0 - - - - 0.1 0.1 0.1 1.0 - - - - - 0:1 0.3 1.0 - - - - - - 0.3 1.6 0.3 - - - 0.1 0.2 - 1.2 - - - - - 0.1 - 1.0 - - - - 0.1 0.2 0.1 0.1 - - - - - 0.1 0.1 0.4 - - 0.1 - - - - 0.4 ! Figures for age 2 In anticipation further study, NCHS the Eighth Revision without evidence of disease with or without mention of cardiovascular disease. of a possible change in the ICDA. These special four-digit a disease of the not stated are included in total but not distributed among age groups. classification of these terms has created these special four-digit subcategories under category 412 that are used instead of those listed in subcategories permit the coronary arteries from those due to other conditions classifiable to chronic ischemic heart Subcategories 412.1 and 412,2 together are equivalent to ICDA category and for the benefit of those interested in doing separation of deaths due to cardiovascular disease 412.0, and subcategories 412.3 and 412.4 together are equivalent to ICDA category 412.9. Table 3. Death rates for Chronic ischemic heart disease and for four special four-digit categories at this disease, by age, color, and sex: United States, 1968-71—Con. 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85 years years years years years years years years years years years and over I — mete yr 10.7 17.6 32.1 56.6 79.7 103.3 138.9 180.6 186.0 224.7 263.3 8.6 22.7 42.4 54.1 82.4 110.1 147.1 181.3 219.5 255.6 245.4 11.7 24.2 40.5 56.8 25.7 127.0 164.3 254.8 227.3 244.7 237.7 11.0 24.4 39.8 63.6 99.5 138.8 198.1 283.8 231.5 221.6 280.0 8.8 18.7 26.9 49.8 67.9 96.0 127.7 197.7 235.9 249.5 359.2 11.0 16.5 33.2 50.6 81.9 101.3 140.6 193.5 247.4 319.7 341.9 9.5 21.3 35.3 51.2 79.4 132.0 190.4 245.8 217.8 242.9 327.3 9.3 23.1 37.1 56.5 86.8 128.1 218.4 242.2 241.8 283.2 381.9 5.3 14.0 A 50.3 92.0 162.2 279.8 518.7 965.6 15759.7 4 5.8 13.2 27.5 51.4 92.8 165.9 294.6 529.0 984.0 1,783.4 3,554.7 | 10 5.5 12.5 24.9 50.5 93.2 170.7 303.1 561.3 958.3 1,782.4 4,123.4 | 11 4.8 12.2 25.2 50.0 93.7 174.6 305.7 574.9 977.7 L, 4 12 7.8 21.3 42.5 17:3 139.6 240.1 400.7 697.0 1,229.1 2,156.1 4 13 8.3 19.7 42.9 78.2 142.9 244.5 420.4 707.6 1,249.8 2,135.0 3,914.3 | 14 B:2 19,1 39.1 77. Xk 140.1 250.2 414.6 738.6 1,219.6 2,093.6 4,347.3] 15 7.3 18.4 37.7 76.1 138.7 254.8 415.5 756.5 1,224.0 2,089.9 » 16 L.7 4.3 8.9 18.6 39.4 83.7 170.5 376.3 802.7 1,602.7 3 «0 1.5 3.7 8.5 19.9 38.6 82.7 178.9 385.1 817.0 1,628.9 3,614.7 | 18 1:53 3.5 6.9 18.3 39.4 84.2 184.6 403.3 800.7 1,662.0 4,469.3 1.4 3.6 8.0 17.6 39.5 86.3 185.1 414.2 825.1 1, 0 4, 0| 20 17.6 33.8 64.7 110.2 188.1 280.5 417.7 713.0 1 6 1,358.9 2,131.0 18.8 33.3 67.7 111.8 178.9 308.9 457.8 743.3 1,069.6 1,333.9 1,912.6] 22 17.9 35.0 69.9 120.3 196.3 33L.4 547.4 938.9 1,007.0 1,244.7 1,629.5 13.1 35.6 70.6 119.3 209.3 327.4 555.3 974.4 3 1,225.7 1, 5 5.2 27.9 54.5 92.6 155.8 278.2 504.9 724.6 993.2 1,942.1 6.7 16.0 29.2 55.0 91.9 176.5 296.7 500.6 734.8 1,097.7 1,806.3 4.3 14.0 25.4 55.) 104.2 199.7 379.1 631.3 658.0 1,100.0 1,716.3 | 27 4.5 1 31.3 59.9 112.7 215.4 423.6 594.2 702.4 1,071.6 1,831.9 0.8 2.3 5.6 11.0 19.7 34.8 58.4 111.7 207.2 390.2 887.1 29 0.8 2.1 4.8 9.9 18.8 32.7 57.6 106.2 207.0 393.9 786.5 | 30 0.8 2.0 5.1 10.2 17.6 32.4 58.1 110.1 197.9 370.6 896.1 | 31 0.8 2.1 4.6 9.7 17.8 31.4 56.2 108.8 194.1 374.1 878.7 | 32 0.9 2.5 6.9 15.1 26,7 44.6 73.1 132.9 239.5 433.3 926.2 | 33 0.9 2.7 6.4 12.8 24,2 40.9 67.4 126.5 233.6 453.1 818.8 | 34 0.9 247 7.4 13.6 23.2 42.1 68.8 130.1 231.5 408.4 901.9 | 35 1.2 2.7 6.5 13.1 23.0 38.2 67.8 130.7 224.3 403.4 883.1 36 0.4 0.9 2.1 4.1 8.5 17.2 34.2 78.4 173.3 364.4 890.2 | 37 0.3 0.6 1.8 4.0 8.1 12.2 36.4 75.7 177.6 360.4 800.7 | 38 0.2 0.7 1.6 4.3 8.8 16.1 35.6 81.7 168.4 352.5 971.6 | 39 0.1 0.7 1.5 3.7 7:9 16.9 34.7 79.7 168.1 361.8 941.5 | 40 2.4 9.7 20.4 35.7 61.2 96.7 145.3 253.2 334.9 476.7 751.0 | 41 2.9 7.8 16.0 32.8 61.3 90.6 148.0 240.1 337.3 384.2 546.7 | 42 3.2 6.3 16.3 34.3 48.2 89.2 154.4 272.6 285.9 353.9 465.6 | 43 2.5 7+6 13.8 29.1 53.0 90.1 145.5 260.6 250.4 337.8 494.5 | 44 1.5 4.8 10.5 16.9 28.7 64.4 104.3 194.7 266.5 335.0 698.7 | 45 1.3 3.3 7:2 18.1 32.6 62.5 106.0 172.6 255.6 361.0 594.9 | 46 1.7 3.0 7.0 14.6 26.3 63.5 125.6 169.7 214.4 298.0 515.0 | 47 1.1 4.7 6.9 17.4 35.5 66.6 13.9 159.2 212.4 309.5 551.4 | 48 FwNH NOW 67 APPENDIX | COMPONENTS OF COMPARABILITY RATIOS FOR 15 LEADING CAUSES OF DEATH AND THEIR MAJOR SUBCATEGORIES Data shown in table I are for the most part previously unpublished statistics from a study of the comparability between the Seventh and Eighth Revisions of the International Classifica- tion of Diseases. Provisional estimates of selected comparability ratios and a brief sum- mary of the statistical design of the compara- bility study have been published in Monthly Vital Statistics Report, Vol. 17, No. 8, Supple- ment, October 25, 1968. The set of comparability ratios shown below was computed to assess the degree of discontinu- ity between 1967 and 1968 for the 15 leading causes of death and their subcategories shown in tables 1 and 2. For each of these causes and subcategories a comparability ratio was especially computed for the present report by dividing the estimated number of 1966 deaths assigned to the Eighth Revision category num- bers corresponding to the title shown in the stubs of tables 1 and 2 by the number of 1966 deaths assigned to the nearly comparable cate- gory numbers of the Sixth and Seventh Revi- sions shown in the last column of tables 1 and 2. This particular set of category numbers according to the Sixth and Seventh Revisions was used in the construction of these compara- bility ratios because trend data for 1950-67 were available for the titles corresponding to these category numbers. It turns out that, with few exceptions (which are designated by a footnote 2 in the table below), this set of category num- bers for the Sixth and Seventh Revisions is the same as the set used to compute for these same Eighth Revision titles the above-mentioned published provisional comparability ratios. It will be observed that even for those Eighth Revison titles for which the set of Sixth 68 and Seventh Revision categories used for the present report are identical with the set used for the published provisional ratios, the provisional ratios are sometimes slightly lower than the ratios shown in this report. This difference results from a better computational procedure used to compute the ratios for the present report. The ratios shown below in column (6) of table I were computed by dividing the estimated number of 1966 deaths assigned to the par- ticular cause using the Eighth Revision ICDA (column (2)) by the number of 1966 deaths assigned to the equivalent cause or combination of causes using the Seventh Revision (column (5)). Adjustment for differences in revisions is made by merely multiplying the original figures by the appropriate ratio. Due to the changes in interpretations and coding rules for classifying causes of death from one revision to another, deaths assigned to an Eighth Revision cause or group of causes rarely had been coded to only one Seventh Revision cause. That is, in most instances, some deaths now assigned to a particular ICDA cause had been coded according to the ICD to causes other than the selected comparable causes. Thus, the difference between the total estimated number of deaths assigned to a particular Eighth Revi- sion cause or group of causes (column (2)) and the estimated number of deaths assigned to this particular Eighth Revision cause that had been coded to the selected comparable Seventh Revi- sion cause (column (3)) is simply the estimated number of deaths that had been previously assigned to Seventh Revision causes other than the selected comparable causes and are now coded to the particular Eighth Revision cause. 69 Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision 66 Revision, based on deaths occurring in 19 with those by Seventh Eighth Revision Seventh Revision Estimated Estimated | number of Number of Ratio: Cause of death and category number Susber of Seah Cause of death and category number Ssgshe in 291, a+ 1966 that went to (2) 1) (2) 3) (4) (5) (6) 1. Diseases of heart------ 390-398,402,404,410-429 730,261 719,550 || Diseases Of heart=-mm=m moo om moomoo 400-402,410-443 727,002 1.0045 449 Rheumatic fever-------ommmmm --400-402 456 14,436 Chronic rheumatic heart disease- -410-416 14,556 5,666 Diseases of mitral valve----=-cecmmoooooooooo._ ---410 5,718 1,693 Diseases of aortic valve specified as rheumatic--- 4 1,713 773 Diseases of pulmonary valve and other endocarditis 773 6,304 Other rheumatic heart diSeases- = aeemmmeceommcmonms re ————————— 6,352 570,220 Arteriosclerotic heart disease, including coronary disease 573,191 158,012 Arteriosclerotic heart disease so described---=-cemeeoeeo-- 158,802 411,920 Heart disease specified as involving coronary arteries- 414,101 1288 Angina pectoris without mention of coronary disease 288 3,560 Chronic endocarditis not specified as rheumatic-------coeeemeeo. 3,785 157 Of mitral valve, specified as nonrheumatic-- 166 2,261 Of aortic valve, not specified as rheumatic- 2,311 1,142 Of other valves, not specified as rheumatic-==-===-=cooomomooooo——_______ 1,308 49,175 Other myocardial degeneration-- 49,796 41,464 With arteriosclerosis---------uo-- 41,935 7,711 Without mention Of ArterioSClerOsis----==memmeemccemcmmmmnn———-——————— 7,861 28,271 Other diseases Of heart-=--====memeeemo ooo 430-434 31,042 706 Acute and subacute endocarditis ----430 757 1,051 Acute myocarditis and acute pericarditis, rheumatic = momo ome --431,432 1,065 7,693 Functional disease of heart------eeeeo__ ---433 10,008 18,827 Other and unspecified diseases of heart-----=-mmecoeeocomomoo_________________~ 434 19,212 53.433 Hypertensive heart disease-=====-m mmm _____________ 440-443 54,176 11,345 42 11,606 42,088 Other hypertensive heart disease~---==-mmemmoe cco mmmmeo;————emmmo ooo 440,441,443 42,570 Active rheumatic fever and chronic 2 rheumatic heart disease------ 390-398 17,293 14,724 Rheumatic fever and chronic rheumatic heart disease------c-omcomeomeo- 400-402,410-416 15,012 1.1519 443 Rheumatic fever=—m=--mmmmom ome TTT 400-402 456 14,281 Chronic rheumatic heart disease 410-416 14,556 5,600 Diseases of mitral valve--=-==-=-meemmeeo oo ____________________TTT7TTTCT 410 5,718 1,686 Diseases of aortic valve specified as rheumatice-----=--omcmoooomccooooo oo. __ 411 1,713 759 Diseases of pulmonary valve and other endocarditis, specified as TheUNAL Lem mmm mm om wm mm mm mm mmm mm mmm LT 3,414 773 6,236 Other rheumatic heart diseases=======-=oocommom oo ___________" 412,415,416 6,352 See footnotes at end of table. 0L Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision with those by Seventh Revision, based on deaths occurring in 1966— Con. Eighth Revision Seventh Revision Estimated Estimated | number of Cause of death and category number pense of degehs dn Cause of death and category number 1966 that went to (2) (1) | (2) (3) (4) EE ea — i EESEEIRES SS — is FEY = 1. Diseases of heart—con.: Hypertensive heart disease with or without renal disease-------- 402,404 | 21,350 20,435 Hypertensive heart disease--------=m---emecccccccc ccc ccc cece ———— 9,672 Hypertensive heart disease with arteriolar nephrosclerosis 10,763 Other hypertensive heart disease=========cmemecceeeecc cece cee | Ischemic heart disease---=-=--- 410-413 | 656,691 569,715 Arteriosclerotic heart disease, including coronary disease-----=-=-c-cccceccaoaaoo 420 157,749 Arteriosclerotic heart disease so described-----======-- -420.0 | 411,678 Heart disease specified as involving coronary arteries- --420.1 | 1288 Angina pectoris without mention of coronary disease-=-====--ecececcmccacanaaana- 420.2 Chronic disease 6f endocardium and other myocardial insufficiency--=-=====ceceu-u- 424,428 9,768 8,688 Nonrheumatic chronic endocarditis and other myocardial degeneration--=--=------- 421, 92 | 1,469 Chronic endocarditis not specified as rheumatic--=----comomommmmm mmm ceeeeee 421 | 129 Of mitral valve, specified as nonrheumatic--- --421.0 | 283 Of aortic valve, not specified as rheumatic-==-=-=---cceecmcmcc ccc cccec cme e 421.1 1,057 0f other valves, not specified as rheumatic--===-c-ceecacaoau- 421.2-421.4 6,983 Other myocardial degeneration without mention of arteriosclerosis- 422.0,422.2 236 Other myocardial degeneration with arteriosclerosis-======---cececcmceaeaaaano * All other forms of heart disease========== 420-423,425-427 ,429 | 25,156 23,375 Other diseases of heart---=-------- 654 Acute and subacute endocarditis- 892 Acute myocarditis and acute pericarditis, not specified as | rheumatice==-===memmme emcee 5 4,344 | Functional disease of heart 433 17,485 | Other and unspecified diseases of heart-=-====-eeeeeeemmmc ccc ccc eee 434 | 2. Malignant neoplasms, including neoplasms of lymphatic and hematopoietic tissues=========- 140-209 304,262 301,920 | Malignant neoplasms, including neoplasms of lymphatic and hematopoietic | tissues: 6,744 | Witn neoplasm of buccal cavity and pharynx 143 1,614 of ne e: 2,205 Of other and unspecified parts of buccal cavity-----=-==-=--- 2,782 Of pharynx-----=---e-eemcccccec cece c cece eee eee ————— 145-148 See footnotes at end of table. Number of deaths in 1966 54,176 11,606 42,570 573,191 158,802 414,101 288 1,065 10,008 19,212 303,736 0.3941 21.1457 20.1823 0.8104 21.0017 6,800 0 \Z Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision with those by Seventh Revision, based on deaths occurring in 1966—Con. Eighth Revision Seventh Revision Estimated Estimated | number of 5 Number of Ratio: Cause of death and category number aber 7% Jeaehs i Cause of death and category number deaths in [col. (2) + 1966 that went 1966 col. (5) to (2) (1) (2) 3) (4) (5) (6) 2. Malignant neoplasms, etc.=—con.: 94,559 Malignant neoplasm of digestive organs and peritoneum, not specified as S€CONAArYmmmmmr mcm mmm mec cmemeemereeee—eemm mmm ————————— 150-156A,157-159 95,079 5,461 Of €S0PhagUS = mmm mm mm mm mee eee eee lolol 1 5,505 17,623 Of stomache--ceeececmmm eee -151 17,623 652 Of small intestine, including duodenum- ---152 693 32,616 Of large intestine, except rectum==-=---- --=-153 32,811 4,568 Cecum, appendix, and ascending cOloM=m=s==-mocoommmo ome. 153.0 4,605 1,397 Transverse colon, including hepatic and splenic fleXUresmemmememeammacmem eee eeeeeeee ee mmeeeemmemeeeeee— mmm memmm————————— 153.1 1,472 1,151 Descending colon --153.2 1,151 5,814 Sigmoid colone-emmmcmmmeccaaaa oo --153.3 5,814 78 Multiple parts of large intestine --153.7 78 18,117 Large intestine (including colon), part unspecified- --153.8 18,155 1,491 Intestinal tract, part unspecified > 10,538 Of rectum 154 10,663 6,543 Of biliary passages and of liver (stated to be primary site)- 6,584 2,112 Liver 2,112 4,431 Other and multiple sites of biliary passages-- 4,472 3,149 Of liver not stated whether primary or secondary 3,149 16,286 Of pancreas==--meeemmmemee eo llllL --157 16,360 1,691 Of peritoneum and of unspecified digestive OY BANS == mmm mmm meme eee me ee mmeeeeoo 158,159 1,691 54,841 Malignant neoplasm of respiratory system, not specified as secondary 54,934 2,623 0f larynx 2,623 20,891 Of bronchus and trachea, and of lung specified as Primary === mcm mm- eee em eee eee eeeeemmeeeeeemmemmmmeeee—————mmm————— 20,913 30,519 Of lung, unspecified as to whether primary or secondar 30,565 808 Of other parts of respiratory system 833 27,377 Malignant neoplasm Of breast==mmmmmmmom oom o oom...) 27,533 40,175 Malignant neoplasm of genital OYgaNS====-==-===mmmmmmememomoommcoooommmmmmoommooo 40,378 7,614 Of cervix uterimm-e-eeeececacocooooooo. 7) 7,665 5,697 Of other and unspecified parts of uterus=----- -172-174 5.731 9,145 Of ovary, fallopian tube, and broad 1igament====mmeeeeeccococooccmemeccmee mm mee 175 9,163 850 Of other and unspecified female genital organs- --176 865 15,856 Of PrOSLaAte mmm mm mmm meee eee ee eee eee eee meme mmm meme mm mmm mmm mmm mmm mm 177 15,941 1,013 Of all other and unspecified male genital organs -178,179 1,013 14,073 Malignant neoplasm of urinary organs-==-me==-=s -180,181 14,166 5,827 Of Kidney=emmmeseemm moana nme e er aeeme meme see eee eee mmm mmm —————— 180 5,841 8,246 Of bladder and other Urinary OrganS====mmee-mececcccoccccecccceseeecceceeememm———— 181 8,325 See footnotes at end of table, TL Table I. Components of comparability ratios for 15 leading cau of death and their major subcategories for comparing assignments by Eighth Revision Revision, based on deaths occurring in 1966—Con. with those by Seventh Eighth Revision Seventh Revision S Estimated Estimate: number of Number of Ratio: Cause of death and category number yunber of deaths in Cause of death and category number deaths in |col. (2) + deaths in |col. (5) 1966 1. (5) 1966 that went eo to (2) 1) (2) 3) 4) (5) (6) 2. Malignant neoplasms, etc.=con.: 34,667 Malignant neoplasm of otiier and unspecified Sitegmmmmmemcocccacmacann 156B,165, 132- 199 35,032 4,545 f SkinNeemme==s=eecssscssesssesmec-ecesseseseeessessmssseseco-e---sssssssoosoo +191 4,560 358 Of eyemmm=== meme eemeessseEseEmeSeSeeseSe SESS esEESS mmm sesem-sseesssss-s--seososse 358 7,152 Of brain and other parts of nervous system= 33 1,355 1,008 thyroid gland=e====== meee —————— ——— --194 1,008 1,792 --196 1,792 1,318 connective tiSSUE=e=mmemsssssessmmesssscscssecesecsememe-----osssseosss—sos----= 197 1,318 1,408 other specified sites, not seated fo be secondary==-= --===---195,199A 1,422 17,086 Of unspecified siteseeese=ee== ——————— ——————— cmmmeme- - 1568, 165,198,199B 17,219 13,926 Leukemia and aleukemi@=====-==-cs-eesmeememc-eeeeecesesesmcee--o----ses--—---ooo-- 204 14,012 15,558 Lymphosarcoma and other neoplasms of lymphatic and hematopoietic £iSSUESmmmmmmmmmmmmmmm--eesem-emsemmmmeee—eeesssmeme—-e-—---e-ss-s-—soo- -200- 203,205 15,802 7,548 Lymphosarcoma and reticulosarcoma= 00 7,563 3,398 Hodgkin's disease--=eeeee=== ceememmmessresssssssssss==== 3,412 4,612 Other neoplasms of lymphatic and hematopoietic tissueses=-=-----=--=-=---=- 202,203,205 4,827 Malignant neoplasms of buccal cayLEy and pharynXeeseeeeseceeeeanan= 40-149 7,205 6,638 Malignant neoplasm of buccal cavity aud pharynX=-me===cecocmmmmomm mmm mmm m mn 140-148 6,800 1.0596 126 Of lip====cesceemssmmccssescceccscesmsmsesmsesssssssssseesse—sssmmmsess sees 140 150 1,603 Of LONQUE=====s=memm==m=eeoeec esses emsesmecmemeoeoooo--oo-sssooososossosses 141 1,629 2,194 0f other and unspecified parts "ok buccal cavity --142-144 2,224 2,715 Of PharynXes=seess=ssmeseececssccmmemeesesessssssessoooossooooooosoossssooos 145-148 2,797 Malignant neoplasms of Sigering organs and peritoneum==-==--=-- 150-159 91,848 90,664 Malignant neoplasm of digestive organs and peritoneum, not specified as 2 SECONJAry=m-===mmmmm-=mem==-e-eesssee=s-e-eeeeeesesssssses—ss--—ses 150-156A,157-159 95,079 0.9660 5,461 Of eSOphaguS==e====mmm-=mcceceeccseesscmsmeemsemesesossosssssossoooosoooossssos 150 5:30 17,547 Of stomacheeeemmesceccccccccccccannnnan --151 17,623 652 Of small intestine, including duodenum- --152 69 32,418 Of large intestine, except rectum====-==- ---153 32,811 4,494 Cecum, appendix, and ascending colon---- --153.0 4,605 1,397 Transverse colon, including hepatic and splenic flexures --153.1 1,472 1,151 Descending colon===---e-eeceeemccemnaccccnnn~ --153.2 1,151 5,774 Sigmoid colon==-==- cemmmmmmm———— --153.3 5,814 478 Multiple parts of large intestine--e-ee-=-=---- -=153.7 78 18,078 large intestine (including colon), part unspecified --153.8 18,155 1,446 Intestinal tract, part unspecified------==-=---- --153.9 1,536 10,455 Of rectuUm--=-=-=s=ss=-=-ese-eescssececsssessss=-e=-=—---=e=-=== ----154 10,663 6,398 of biliary passages and ag liver (stated to be primary site)- ----155 5 2,008 Livereeseeeececcccccarennarenn mmeem-ecessssessss-sssessesss--es-se-eo-e--o-- 155.0 2,112 4,390 Other and multiple Sites of biliary passages--- 155,1,135.8 4,472 - Of liver not stated whether primary or secondary’ ---1 3,149 16,248 Of pancreas=-=-===-=-=sss-ss-ceecsme-ceecess===== --157 16,360 1,485 Of peritoneum and of unspecified digestive organs--=--=-==--=e=-=--=-=-=-=--=- 158,159 1,691 See footnotes at end of table. Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision with those by Seventh Revision, based on deaths occurring in 1966—Con. Eighth Revision Seventh Revision Estimated Estimated | number of } Number of Ratio: Cause of death and category number mmber of (deaths in Cause of death and category number deaths in | col. (2) + deaths in | col. (5) 1966 1 5) 1966 that went col. ( to (2) (1) (2) 3) (4) (5) (6) 2. Malignant neoplasms, etc.—con.: Malignant neoplasms of respirasary ‘ 4.55% ou EE 0- ified system 0-163 56,668 54,5 MaLigTANE Jeoplase OF Iespivanony Sister, BO I rnsmsSEEeSRARESREAY 160-164 | 54,934 1.0316 2,397 Of larynxX-===----ecceceece eee ee meee eeeeee—eee———————— ---161 2,623 20,801 Of bronchus and trachea, and of lung specified as primary- 162 20,913 30, 2403 0f lung, unspecified as to whether primary or secondary 30,565 783 Of other parts of respiratory system-=--=-==-ecececece—ea- 833 Malignant neoplasms of breast----- 174 27,293 27,221 Malignant neoplasm of breast----==-mrmemseeemeec ees... —————————— 170 27,533 0.9913 Malignant neoplasms of genital Organs----=-=-ceceemmceacaa== 180-187 40,516 39,969 Malignant neoplasm of genital organs- -171-179 40,378 1.0034 7,562 Of cervix uterim==rer=ressesemensnsssessnne senna seems. ————— 171 7,665 5,680 Of other and unspecified parts of uterus------ -172-174 5,731 9,092 Of ovary, fallopian tube, and broad ligament--- --=175 9,163 835 0f other and unspecified female genital organs- 176 865 15,806 Of prostate--=--ermmemrecmee——ee—— ee ————————————— --=177 15,941 994 0f all other and unspecified male genital organs--===-==--ececcecceccaceccana 178,179 1,013 Malignant neoplasms of urinary OrganS===========e=eeemam—=— 188,189 14,408 13,987 Malignant neoplasm of urinary organ: --180,181 14,166 1.0171 5,787 Of kidney---==--e-eeemeeemeaaaa—- 180 5,841 8,200 0f bladder and other urinary organs eee eee eee emeeceeeeeecea——— 181 8,325 Malignant neoplasms of all other and unspecified sites----170-173,190-199 35,774 31,296 Malignant neoplasm of other and unspecified sites-------===--coecooaa- 156B,165,190-199 35,032 21.0212 4,498 Of skin==w--eceeececmrcccmcc cece cee eee m eee e eee seme eee e ee ——————— 190,191 4,56 358 Of eye--==s==mecmce ecm m mmm meme meme mee —————— 358 71,137 Of brain and other parts of nervous system --193 1,355 1,008 Of thyroid gland--======ec-ceemccecceeacaaa-" --194 1,008 1,698 of bon --196 1,792 1,318 Of connective tisSue=====ceceececcece cca ee meee cee eee ceeee—ee—————— 197 1,318 1,367 Of other specified sites, not stated to be secondary==-======ceecececcececaaa" 195,199A 1,422 13,912 Of unspecified sites-==-==-=memmmmmm meee 156B,165,198,, 1998 17,219 Leukemig=-=-==-ccocommcannaan 204-207 13,976 13,834 Leukemia and aleukemia=--=v-=eemmecreececcce eee — eee eee ————————————————— 204 14,012 0.9974 Other neoplasms of lymphatic and he- 2 matopoietic tissues--200-203,208,209 16,602 15,478 Lymphosarcoma and other neoplasms of lymphatic and hematopoietic tissues---200-203,205 15,802 1.0506 7,497 Lymphosarcoma and reticulosarcoma--=-==-==--memeeeeccceecceeeecce meee cme —————— 200 1,563 3,398 Hodgkin's disease-======c=cmmmme mec eee m ee meee 201 3,412 4,583 Other neoplasms of lymphatic and hematopoietic tissues-- --202,203,205 4,827 See footnotes at end of table. €L vei Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision Revision, based on deaths occurring in 1966—Con. with those by Seventh Eighth Revision Seventh Revision Estimated Estimated |number of y Number of Ratio: Cause of death and category number Yhge gt ogthe sin Cause of death and category number deaths in |col. (2) + 1966 that went 1965 cal, 5) to (2) 1) (2) 3) 4) (5) (6) 3. Cerebrovascular diseases=---430-438 202,894 200,325|| Vascular lesions affecting central Nervous SySteme-m=mmmmm-eeeecoommeeeommmmeenn 330-334 204,841 0.9905 8,287 Subarachnoid hemorrhage=-== === mmm eee emcee eee eeeeeeeememan 330 8,531 109,400 Cerebral hemorrhage-=-==-=--- -331 111,446 62,937 Cerebral embolism and thrombosis--==eeeeceececccmaaaaaas --332 63,943 19,701 Other vascular lesions affecting central nervous System--------=--oeeeccemmenn 333,334 20,921 4. AccidentsSe-e-ececcecaaaans E800-E949 108,683 106,455 || ACCidenmts =m mmm mmm mmm ene eee eee eee E800-E962 113,563 0.9570 1,000 Railway accidents E800-E802 1,027 52,456 Motor vehicle accidents=-- --E810-E835 53,041 51,373 Motor vehicle traffic accidents E810-E825 51,933 i 1,757 Motor vehicle traffic accident involving colliston with railway train-------- E810 1,800 8,612 Motor vehicle traffic accident to pedestriane--e-eeeeceeecececicmaaoanaon -E812 8,675 22,297 Other motor vehicle traffic accidents involving collision- E811,E813-E819 22,641 14,202 Motor vehicle noncollision traffic accidentSeseeeeeeeecccecacccacacaaans E820-E824 14,291 4,505 Motor vehicle traffic accident of unspecified nature==e=eeeececcecmeccceceaoasd E825 4,526 1,083 Motor vehicle nontraffic accidentSe=emmemeceemcecccocececececoccocemaaeaas E830-E835 1,108 | 2921 other road vehicle accidentse-- E840-E845 292 | 1,630 Water transport accidents=-- E850-E858 1,630 1,486 Aircraft accidents==mmeeeee- E860-E866 1,510 1,806 Accidental poisoning by solid and liquid substances --E870-E888 2,283 | 1,523 Accidental poisoning by gases and Vapors mmm——eeeeee- --E890-E895 1,648 | 16,906 Accidental fallS===emeeeeccencacan E900-E904 20,066 5,312 Fall from one level to another- E900-E902 5,772 ! 4,839 Fall on same level=mmmemmmmm moomoo eee eee emo 5,593 | 6,755 Unspecified falls 8,701 1,432 Blow from falling or projected object or missile-===-==-=essmomcmccococmmocomeoos E910 1,459 1,989 Accident caused by machinery=-e--e--e-a- --E912 2,070 | 1,025 Accident caused by electric current --E914 1,025 7,813 Accident caused by fire and explosion of combustible material---=---eoceocomcaaoo E916 8,084 | 337 Accident caused by hot substance, corrosive liquid, steam, and radiation- E917,E918 409 2,189 Accident caused by firearm====mmm mmm eo cme om emma E919 2,558 | 1,672 Inhalation and ingestion of food or other abject causing obstruction or suffocation 1,831 5,431 Accidental drowning 5,687 | 430 Excessive heat and insolation 531 1,129 Complications due to nontherapeutic medical and surgical procedures, therapeutic | misadventure, and late complications of therapeutic procedures=------=--w-= E940-E959 1,411 5,909 All other accidentS===eeeececee-a E911,E913,E915,E920,E923-E928,E930,E932-E936,E960-E962 7,001 SL Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision with those by Seventh Revision, based on deaths occurring in 1966— Con. Eighth Revision Seventh Revision Estimated Estimated | number of 3 Number of Ratio: Cause of death and category number Qmogr of Sen Cause of death and category number deaths in | col. (2) + that went 1966 col, (5) to (2) 1) (2) (3) (4) (5) (6) 4. Accidents—con.: Motor vehicle accidents---E810-E823 52,622 52,250 Motor vehicle accidents=== =m como eee oo E810-E835 53,041 0.9921 51,243 Motor vehicle traffic accidents E810-E825 51,933 1,757 Motor vehicle traffic accident involving collision with railway train--------- E810 1,800 8,570 Motor vehicle traffic accident to pedestrian-----s---ccccmmmmmcmmoeeoo E812 8,675 22,274 Other motor vehicle traffic accidents involving collision- -E811,E813-E819 22,641 14,158 Motor vehicle noncollision traffic accidents-------cocccommmmmomoomaooaoo E820-E824 14,291 4,484 Motor vehicle traffic accident of unspecified nature--------comemmomoccmceanaoo E825 4.526 1,007 Motor vehicle nontraffic accidents------ceomo mmm m mea E830-E835 1,108 All other accidents-------- E800-E807 ,E825-E949 55,983 53,771 Other accidents---- -E800-E802,E840-E962 60,522 0.9250 1,000 Railway accidents--====m= memo occ ee emma n E800-E802 1,027 292 Other road vehicle accidents- --E840-E845 292 1,630 Water transport accidents---- --E850-E858 1,630 1,486 Aircraft accidents --E860-E866 1,510 1,806 Accidental poisoning by solid and liquid substances- --E870-E888 2,283 1,523 Accidental poisoning by gases and vapors------------ --E890-E895 1,648 16,878 Accidental falls--------ccommommnononn -E900-E904 20,066 5,312 Fall from one level to another-- --E900-E902 5,772 4,839 Fall on same level--=--mm momo ome eee ee eee eee meee emma E903 5,393 6,727 Unspecified falls-----oommmmmmmmmmm mimeo --E904 8,701 1,405 Blow from falling or projected object or missile------=-=-coommcmmmommmaaaaaao E910 1,459 1,936 Accident caused by machinery----------ccoceceoonan --E912 2,070 1,025 Accident caused by electric current---------eccmecoocaoooo --E914 1,025 7,788 Accident caused by fire and explosion of combustible material----------=--c--o-- E916 8,084 337 Accident caused by hot substance, corrosive liquid, steam, and radiation---E917,E918 409 2,189 Accident caused by firearm-------commmmm mo ee E919 2,558 1,672 Inhalation and ingestion of food or other object causing obstruction or suffocation-=-=--- momo mm me oe ee meme 1,831 5,431 Accidental drowning- 5,687 430 Excessive heat and insolation-------m-eeeccceccc cece meee 531 1,129 Complications due to nontherapeutic medical and surgical procedures, therapeutic misadventure, and late complications of therapeutic procedures------------ E940-E959 1,411 5,814 All other accidents----------- E911,E913,E915,E920,E923-E928,E930,E932-E936, Eoco- E962 7,001 5. Influenza and 2 pneumonia---------- 470-474 ,480-486 66,413 61,058 || Influenza and pneumonia, except pneumonia of newborn------------o-comommmmnoo 480-493 63,615 1.0440 2,683 PE SS FF tata 480-483 2,830 58,375 Pneumonia, except pneumonia of newborn- --490-493 60,785 8,565 Lobar pneumonia--------=----ccocooo ----490 8,864 31,949 Bronchopneumonia----=--=------ -491 33,276 5,409 Primary atypical pneumonia------ -492 5,729 12,452 Pneumonia, other and unspecified====-cccmcmm momma 493 12,916 See footnotes at end of table. 9L Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories Revision, based on deaths occurring in 1966— Con. for comparing assignments by Eighth Revision with those by Seventh Eighth Revision Seventh Revision Estimated Estimated | number of Number of Ratio: Cause of death and category number Sumbee of deaths Cause of death and category number deaths in | col. (2) + 1966 | that went 1966 col. (3) to (2) (1) (2) 3) (4) (5) (6) 5. Influenza and pneumonia— con. : Influenza--------cccmmmmeaaoo 470-474 2,709 2,676 Influenza = = mmm mmm ee eee ee emcee 480-483 2,830 0.9572 Pneumonia-------cocommmmaaoo 480-486 63,704 58,375 Pneumonia, except pneumonia of newborn------- eee ooo ___ 490-493 60,785 ’1.0480 8,565 Lobar pneumonia-------=ccommmemaaano ---490 ,8 31,949 Bronchopneumonia---===--=- --491 33,276 5,409 Primary atypical pneumonia------- --492 5,729 12,452 Pneumonia, other and unspecified---==cmcmoommm oom __ ---493 12,916 6. Certain ca ses of mortality in early 2 infancy-760-769.2,769.4-772,774-778 47,425 46,706 || certain diseases of early infancy---==-=c momo eee 760-776 51,644 0.9183 191 Pneumonia Of MeWbOIM= === momo coo eee eee ____ 763 2,477 5 Diarrhea of newborn-==-----ecoooommme_ 250 46,510 Remainder of Certain diseases of early infancy. 48,917 7. Diabetes mellitus------eoeeuoa_ 250 34,496 33,870 || Diabetes mellitus-=== === om mmm. 260 34,597 0.9971 8. Arteriosclerosis--------meeooooo 440 34,873 32,083 || General arteriosclerosis === =m mmm mmm om oom _______ 450 38,907 0.8963 5 BOlirie, Wpysems 2d 490-493 29,84 is, chronic and lified, and Emph ithout mention of thma======-cccccmmceeo - 9,841 28,602 thma, B hitis ronic and unqua e an physema without mention o Reka A enlists, o meme mii Tiel. RN 241,501,502,527.1 29,740 1.0034 4,089 AS EMA = = = mmm me ee meen 241 4,324 4,89 Bronchitis, chronic and unqualified----- 5,164 19,619 Emphysema without mention of bronchitis 20,252 Chronic and unqualified bronchitis--=-=----cmmeeeaeo- 490,491 5,484 4,851 Bronchitis, chronic and unqualified-----meeemmeecscecceeceemec am ———————— 501,502 5,164 1.0620 Emphysema-=--==cccmmmmmmao__ 492 21,350 19,555 Emphysema without mention of bronchitis-=====em coo. 527.1 20,252 1.0542 Asthma--====-=ccmmmm mee 493 3,007 2,623 AStAMA === mm meee ee ee mmm ————— 241 4,324 0.6954 10. Cirrhosis of liver-----m-----o- 571 26,839 26,203 || Cirrhosis Of LiVer====- =m comm moomoo oe een meme eee 581 26,692 1.0055 16,962 Without mention of alcOhOliSm==m mm mmm mm mmm m ome. 581.0 17,320 9,241 With alcoholism- 9,372 See footnotes at end of table. LL Table I. Components of comparability ratios for 15 leading causes of death and their major subcategories for comparing assignments by Eighth Revision with those by Seventh Revision, based on deaths occurring in 1966— Con. Eighth Revision Seventh Revision Estimated Estimated | number of Number of Ratio: Cause of death and category number Suber of Cen” Cause of death and category number eaths in aly qr 1966 that went : to (2) (1 (2) 3) (4) (5) ! (6) | T 11. Suicide--===m=mmmmmmmmmmnan E950-E959| 20,158 20,046 || Suicides===mmmmmmmmm mmm m FR E963,E970-E979 21,281 0.9472 5,192 Suicide by poisoning--============-a- 5,588 2,825 Suicide by hanging and strangulation: 974 2,863 10,072 Suicide by firearm and explosive---- 976 10,407 1,957 Suicide by all other means========= moomoo meee 2,423 12. Congenital anomalies--------- 740-759 18,529 17,453 || Congenital malformations====== === mmm oe eee 18,158 1.0204 os | — 2 1,123 Spina bifida and meningocele=== === =m mmo eee n 751 1,151 1,655 Congenital hydrocephalus and other congenital malformations of nervous system and | SEeTSe OrgaNnS==========-eeeoccomcacma———an -—- 152,753 1,756 | 8,819 Congenital malformations of circulatory system====-=====ecomcomo cece 754 9,020 | 5,853 Other congenital malformations========c= omen 750,755-759 6,231 13. Homicide====-sm=emeceeanaan E960-E978 11,570 11,395 || Homicides==m==m mmm ooo ee ee eee meee E964 ,E980-E985 11,606 | 0.9969 6,773 Assault by firearm and eXploSive----=-=- om mmm E981 6,855 2,300 Assault by cutting and piercing instruments-====mmme ceo como een E982 2,330 2,023 Assault by other means-----------ccoooooo -E964 ,E980,E983 2,122 298 Injury by intervention of police=====m momo mmo eee E984 298 1 EXeCULiOn= mmm mmm mm mm ee ee ee ee ee ee eee --E985 1 14. Nephritis and nephrosis------ 580-584 10,227 9,955 || Nephritis and nephrosis=-= = mmo m comme eee 590-594 11,540 0.8862 462 Acute nephritis=m= mmm momo ee ee een 590 485 593 Nephritis with edema, including nephrosis--==-=c oom mmm 591 679 8,900 Chronic and unspecified nephritis and other renal sclerosis-- --592-594 10,376 15. Peptic ulcer-------=-=-====n- 531-533| 10,172 9,429 || Ulcer of stomach and dUOdENUm===========m=======mmmommm mm m———ee————————————— 540,541 10,321 0.9856 4,896 Ulcer of stomach: meee 540 5,599 4,533 Ulcer of duodenume mm mmo oe ee eee eee eee ee eee 541 4,722 "There were no deaths in the sample assigned to the Seventh Revision title Angina pectoris without mention of coronary disease (ICD No. 420.2). NCHS nosologists state, how- ever, that with a possible few exceptions these deaths would be assigned by the Eighth Revision to Angina pectoris (ICDA No. 413), The exceptions, if any, resulted from the dropping of the preference in the Seventh Revision of angina pectoris over cardiovascular diseases. For reasons described in appendix I, the set of Seventh Revision category numbers used to construct this ratio is different from the set used to construct the previously published provisional comparability ratio for this same Eighth Revision title. n computing the above-mentioned published comparability ratio for Malignant neoplasms of digestive organs and peritoneum (ICDA Nos. 150-159), the Seventh Revision title Malignant neoplasm of liver not stated whether primary or secondary (ICD No. 156A) was not included in the set of titles most nearly comparable to Malignant neoplasms of diges- tive organs and peritoneum. Category number 156A was thus excluded because deaths assigned to this category by the Seventh Revision were removed by the Eighth Revision from under the title for the system of the body attacked by the neoplasm (in this case, from under digestive organs and peritoneum) and placed under the following subsection of "Section II. Neoplasms': Malignant neoplasm of other and unspecified sites (ICDA Nos. 190-199). More specifically, these deaths were transferred to the new Eighth Revision ti- tle: Malignant neoplasm of liver, unspecified (ICDA No. 197.8). “There were no deaths in the sample assigned to the Seventh Revision title Malignant neoplasm of multiple parts of large intestine (ICD No. 153.7). NCHS nosologists state, however, that these deaths would be distributed by the Eighth Revision over the four-digit categories under Malignant neoplasm of large intestine, except rectum (ICDA No.153). APPENDIX II TECHNICAL NOTES Death Statistics Tabulations of deaths used in this report are based on information obtained from copies of death certificates. These copies were received from the registration offices of all States, certain cities, and the District of Columbia. The sta- tistical information on these records was edited, classified, transferred to a tape for computer processing, and tabulated in the National Center for Health Statistics (NCHS). The rates shown in this report are based on deaths tabulated by place of occurrence, that is, all deaths occurring in the death-registration States from 1900 to 1932, and all deaths occurring in the continental United States there- after, with Alaska added in 1959 and Hawaii in 1960. Deaths among armed forces overseas and U.S. nationals living abroad are excluded for all years. Race The category “white” includes, in addition to persons reported as “white,” persons reported to be Mexican or Puerto Rican. The categories “races other than white” or “all other” consist of persons reported as Negro, American Indian, Chinese, and Japanese; other numerically small racial groups; and persons of mixed white and other races. Populaticn Bases Rates were computed on the bases of popu- lation statistics made available by the U.S. Bureau of the Census. Rates for decennial years are based on the populations enumerated in censuses of those years, which are taken as of April 1. Rates for all other years are based on 78 midyear (July 1) estimates. Sources of the popu- lations used, published by the Bureau of the Census, are given below. Vital Statistics Rates in the United States, 1900-1940, Washington, U.S. Government Printing Office, 1943. Current Population Reports, Series P-25: No. 98. “Estimates of the population of the United States and of the components of change, by age, color, and sex: 1940 to 1950,” 1954. No. 265. “Estimates of the population of the United States, by age, color, and sex: July 1, 1950 to 1962,” 1963. (Used only for data years 1961 and 1962.) No. 276. “Estimates of the population of the United States, by age, color, and sex: July 1, 1963,” 1963. No. 310. “Estimates of the population of the United States and compo- nents of change, by age, color, and sex: 1950 to 1960,” 1965. No. 321. “Estimates of the population of the United States, by age, color, and sex: July 1, 1960 to 1965,” 1965. (Used only for data years 1964 and 1965.) No. 352. “Estimates of the population of the United States, by age, color, and sex: July 1, 1966,” 1966. No. 385. “Estimates of the population of the United States, by age, color, and sex: July 1, 1964 to 1967,” 1968. (Used only for data year 1967.) No. 416. “Estimates of the population of the United States, by age, color, and sex: July 1, 1968,” 1969. “Estimates of the population of the United States, by age, color, and sex: July 1, 1969,” 1970. No. 441. The population estimates by color used for 1962 and 1963 exclude New Jersey. Birth, death, and fetal death records of the State of New Jersey did not contain the race item in the beginning of 1962, and the certificate revision without this item was used for most of 1962 as well as for 1963. Therefore the National Center for Health Statistics estimated a population base by color for these years which excluded New Jersey. The estimates for 1963 are shown in table 6-5, Part A, Volume II, of Vital Statistics of the United States, 1963. Those for 1962 are shown in the comparable report for that year. Rates All rates are shown per 100,000 population. In many cases the rates are shown beyond the last significant figure, not because they can be interpreted with that degree of accuracy, but merely for convenience in computation and publication. 79 APPENDIX III THE UNITED STATES STANDARD CERTIFICATE OF DEATH, 1968 REVISION Standard certificates of death issued by the National Center for Health Statistics and its predecessor offices have served for many years as the principal means of attaining uniformity in the content of documents used to collect infor- mation on deaths. They have been modified in each State to the extent necessitated by the par- ticular needs of the State or by special provisions of State vital statistics laws. The certificates of most States, however, conform closely in con- tent and arrangement to the standard certificates. 80 The most recent revision of the standard certificate of death is shown on page 81. It was prepared in close collaboration with State health officers and registrars; Federal agencies concerned with vital statistics; national, State, and county medical societies; and others working in the fields of public health, social welfare, demography, and insurance. It was recommended to the States for adoption as of January 1, 1968. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE —PUBLIC HEALTH SERVICE— NATIONAL CENTER FOR HEALTH STATISTICS REV. 1-68 PHS—797-1 FORM APPROVED BUDGET BUREAU NO. 68-R1901 TYPE, OR PRINT IN PERMANENT INK SEE HANDBOOK FOR INSTRUCTIONS DECEASED USUAL RESIDENCE WHERE DECEASED LIVED. IF DEATH OCCURRED IN INSTITUTION, GIVE — LOCAL FILE NUMBER - (PHYSICIAN, MEDICAL EXAMINER OR CORONER) U.S. STANDARD CERTIFICATE OF DEATH r US. GOVERNMENT PRINTING OFFICE 1967 OF —241-659 1 STATE FILE NUMBER DECEASED — NAME FIRST MIDDLE LAST SEX 2 3 DATE OF DEATH ( MONTH, DAY, YEAR) RACE WHITE, NEGRO, AMERICAN INDIAN, ETC. (SPECIFY) 4 AGE — Last BIRTHDAY (YEARS) UNDER | DAY HOURS MIN 5¢ UNDER | YEAR | DAYS 5b DATE OF BIRTH (MONTH, DAY, YEAR ) le COUNTY OF DEATH CITY, TOWN, OR LOCATION OF DEATH INSIDE CITY LIMITS SPECIFY YES OR NO HOSPITAL OR O 6 a HER INSTITUTION —NAME (If NOT IN EITHER, GIVE STREET AND NUMBER ) ™. Te 4 STATE OF BIRTH (if NOT IN u.S.A., NAME [CITIZEN OF WHAT COUNTRY MARRIED, NEVER MARRIED, SURVIVING SPOUSE (IF WIFE, GIVE MAIDEN NAME ) COUNTRY) WIDOWED, DIVORCED ( speciry) 8 | 10 "n SOCIAL SECURITY NUMBER USUAL OCCUPATION (GIVE KIND OF WORK DONE DURING MOST OF WORKING LIFE, EVEN IF RETIRED ) KIND OF BUSINESS OR INDUSTRY RESIDENCE BEFORE "” 13 13 ADMISSION. RESIDENCE — STATE COUNTY CITY, TOWN, OR LOCATION INSIDE CITY Limits [STREET AND NUMBER pp (SPECIFY YES OR NO) 14b 14¢ 14d 14¢ FATHER — NAME FIRST MIDDLE LAST MOTHER — MAIDEN NAME FIRST MIDDLE Last PARENTS 16 INFORMANT —NAME MAILING ADDRESS (STREET OR R.F.D. NO., CITY OR TOWN, STATE, ZIP) y . N 17 o PART |. ° | DEATH WAS CAUSED BY [ENTER ONLY ONE CAUSE PER LINE FOR (a), (b), AND (c)] Mangler Roig J > 1" TMMEDIATE CAUSE o FH 3 (a) > - BUF YS, OF AS A CONSEQUENCE OF CONDITIONS, IF ANY, WHICH GAVE RISE 10 (b) INTE husk to, DUE TO, OR AS A CONSEQUENCE OF LYING CAUSE LAST EIT. © PART Il. OTHER SIGNIFICANT CONDITIONS: CONDITIONS CONTRIBUTING TO DEATH BUT NOT RELATED TO CAUSE GIVEN IN PART | (a) AUTOPSY IF YES WERE FINDINGS CON- (YES OR NO) | SIDERED IN DETERMINING CAUSE OF DEAT! 1% 19% ACCIDENT, SUICIDE, HOMICIDE, [DATE OF INJURY (MONTH, DAY, Year) [HOUR HOW INJURY OCCURRED ( ENTER NATURE OF INJURY IN PART | OR PART II, ITEM 18) OR UNDETERMINED (speciFy) 200. 200 20 M.| 20d INJURY AT WORK PLACE OF INJURY AT HOME, FARM, STREET, FACTORY, LOCATION ( STREET OR R.F.D. NO., CITY OR TOWN, STATE) (SPECIFY YES OR NO) OFFICE BLDG. ETC. (SPECIFY) \ 20¢. 21. 20g CERTIFICATION — MONTH DAY YEAR | MONTH DAY YEAR AND LAST SAW Him/HER ALIVE ON || DID/DID NOT VIEW THE DEATH OCCURRED AT THE PLACE, ON THE PHYSICIAN: To MONTH DAY YEAR BODY AFTER DEATH. (HOUR) DATE, AND, TO THE BEST | ATTENDED THE OF MY KNOWLEDGE, DUE 2a. DECEASED FROM [2m 2c 214d 210 M. TO THE CAUSE(S) STATED. CERTIFICATION —MEDICAL EXAMINER OR CORONER: ON THE BASIS OF THE HOUR OF DEATH THE DECEDENT WAS PRONOUNCED DEAD EXAMINATION OF THE BODY AND/OR THE INVESTIGATION, IN MY OPINION, ITH DAY YEAR HOUR «In DEATH OCCURRED ON THE DATE AND DUE TO THE CAUSES) STATED 2% M226 Mm. CERTIFIER— NAME (Type OR PRINT) SIGNATURE DEGREE OR TITLE DATE SIGNED (MONTH, DAY, YEAR) 2c MAILING ADDRESS — CERTIFIER STREET OR R.F.D. NO. CITY OR TOWN STATE re \, 23d BURIAL, CREMATION, REMOVAL CEMETERY OR CREMATORY — NAME LOCATION CITY OR TOWN STATE (SPECIFY ) 240. 2b 24¢ DATE ( MONTH, DAY, YEAR) FUNERAL HOME — NAME AND ADDRESS ( STREET OR R.F.D. NO., CITY OR TOWN, STATE, ZIP) ud 250. FUNERAL DIRECTOR — SIGNATURE \ 25 ge [rersran=sienarere 2 i EIVED BY LOCAL REGISTRAR DATE REC 26b. 81 APPENDIX IV BRIEF SUMMARY OF STATISTICAL DESIGN General Plan for Deaths at All Ages The strata, corresponding to the causes or groups of causes of death according to the Seventh Revision, into which the total number of deaths in 1966 are divided for the purposes of this report, are designated by Lj; the total number of deaths in 1966, by N; and the number of deaths in the general or /ith stratum, by N,,. Therefore, ‘ (1) N=XN =N_+N_+...+N_ ...+N . wh 1 2 h t Similarly, the strata according to the Eighth Revision are designated by L'; and the number of deaths in the general or A'th stratum, by Ni. Thus, according to the Eighth Revision, the total number of deaths in 1966 may be represented as follows: . (2) N=ZN,=N +N, +... +N... +N’. n W ! 1 2 ’ ! h L Let the number in the stratum according to the Eighth Revision to which the Ath stratum according to the Seventh Revision is most nearly comparable be designated by N',. Then the equation for the comparability ratio (designated by R,,) that is to be estimated is: R,=-> (3) Let the estimate of R,, be denoted by r.,; and the estimate of Ny be denoted by x. Inas- much as N, is known, the problem of obtaining an estimate of R,, reduces to the problem of 82 obtaining an estimate of N'r. This estimate 7, may be defined as follows: x' ' r= _". (4) N, To obtain the estimate x},, L random sub- samples were drawn by the computer from the L strata into which the total number of deaths in 1966 were divided according to classification by the Seventh Revision. These L random sub- samples were then classified according to the Eighth Revision. The size of the random subsample drawn from the Ath stratum of N is denoted by n,; and the size of the stratified random sample drawn from all deaths in 1966, classified according to the Seventh Revision is: (5) L n=22n,. YN h The number of deaths in the stratum according to the Eighth Revision to which the hth stratum of the stratified random sample n, according to the Seventh Revision, is most nearly com- parable, is designated by nly and the total number of deaths at all ages drawn in the strat- ified random sample, classified according to the Eighth Revision, is designated by n'. Thus, n' may be represented as follows: L n= ns (6) h where L' designates the strata in the sample according to the Eighth Revision. Estimator of the numerator.—Inasmuch as the denominators of these comparability ratios (defined in equation 3) are the total counts of deaths in the stratum in 1966 according to the Seventh Revision, the only values estimated from the stratified random sample (n) are, as stated above, the numerators—the numbers of deaths that would be assigned to each of the strata if all deaths in 1966 were classified by the Eighth Revision. The estimator of these numera- tors is represented as follows: N x s Nx x', = Pr aX H (7) h n, j#h n, where x, is the number of deaths in the Ath stratum of the sample—the stratum according to the Seventh Revision selected as most nearly comparable to the h'th stratum of the Eighth Revision, and where x; represents the number of deaths in any except the Ath stratum of the stratified random sample that were coded to the hth stratum by the Eighth Revision. The total number of strata other than the Ath that con- tained deaths that were assigned to the A'th stratum by the Eighth Revision is represented by J ™h In equation 7 x, =% x j=3 the value 1 if the death it represents was assigned by the Eighth Revision to stratum N' Ws and takes on the value 0 if the death it Pn. sents was not assigned to stratum N' n. , where x, takes on h'* J Similarly, in equation 7 x=Z x, where x, takes =1 on the value 1 if the death it represents was assigned to stratum Nis and takes on the value 0 if the death it represents was not assigned to stratum N,. Variance of the numerator.— The variance of x's, the estimated total of deaths that are coded by the Eighth Revision to the category numbers comprising Ni, 1! is: nN 72 2 3, s” (8) J 1 - 2 =f) Nf) EN Gf) ny, where f, is the sampling fraction (n n/N, ), and /; is the sampling fraction (n; IN). For a variate, such as in this study, that takes on only the value 0 or 1, 82 = NP, Q,/(N, = 1 ) (9) where P, is the proportion of deaths in N, that would , on the value 1, that is, that would be assigned to N' yr if coded by the Eighth Revision, and where Q)- 1- P. . The sample variance for stratum A may be expressed as follows: =m,p,q,)(n, -1). (10) Similarly, the sample variance for the general term among all the strata in the stratified random sample that include deaths assigned to stratum 4’ by the Eighth Revision may be written: =(n,p,9,)/(n, - 1). (11) Substituting the right-hand members of equa- tions 10 and 11 for Ss? and S23, respectively, in equation 8 gives the following as the estimate 02 from the sample of xp : (12) , N (N, -n,) J N,(N, -n)) nt n, -1 Png, * = n -1 v,9,)- Inasmuch asp, = (x /n,);q, = 1 -pnip= (x,/n,); and q; = 1-p,, we have (13) ’ 2 2 N,(N, -n.) x, x, 5, = - + h n, 7, ~1 n(n, -1) Standard error of the estimate of the ratio. —Inasmuch as the variance of a constant times a random variable is the constant squared 83 times the variance of the random variable, it follows that 2 2 1 2 (14) S$ ’ = , nd ’ "h N, *h Taking the square root of the above equation gives the estimated standard error of r.,, denoted s_ : hn Ya Sr EN sli) =3 "h N Xn x, N ' h h Confidence interval for the ratio.—Once the standard error of the estimate of the compara- bility ratio (denoted s, ) is known, confidence h intervals for R,, the true value of the compara- bility ratio as defined in equation 3 may be computed. The required degree of confidence that a range will cover the true value of R, has been specified for this study to be 95 percent. Employing the usual notation in the table for the normal distribution? it may be stated that for given percentage p (equal to 5 percent in the present study) the p% value Ap of the normal distribution is defined by the condition: p (16) Probability (In -R | >Ap s,,) = 5 h With the b5-percent value Ap of the normal distribution equal to 1.9600 this gives: I’ ry = 1.9600 5, CONTENTS Page PLELBCE v'vnnvt sumennn se nunme te isuaoiat sabaonesnmmmenns russ nasyrrsnss 11 Introduction... «oot ee eee 1 Background of HANES ...ccvuicisnsnsitststsnmonvnrarnsemmnnnns sansa 2 Experimental Design and Data Collection ............ccciiiiinininnnnnn.. 3 Experimental DESIGN o.ovvunmss vunununtnsanasssannsnssivammsnesrnmn 3 Data COlCCHON saunssvnvnnisrnimniinisavsstsns romans unrsmnnssapus 4 Findings of the San Antonio Study ........... cco inininnnn.. 5 Examination ROLES «uous runs sen ts vonnann ss anannssnssassisninsmnsvmns 5 Number of Contacts Made by HER’s ...........ciiiiiiinrinnnrnnnnn. 7 IMPlEMEntation ou. seesuns sonnets aamaiuts svnunnsronussss rasamnys ne 8 BEfErenees vuvuas svvinonisnomeis savnnnnss prusmmrsnnnsnms rs nvenay sens 9 List of Detailed Tables ..........iuiuiniiii init ii ininnan.n. 10 Appendix 1. Technical Notes on Methods ..couvirvvsnsisirsnsisinnmmnns os 19 General QUalIfICAIONS «s+ rvnuansssmsnntsssimensnrnrnnens nunmans ss 19 Appendix II. Forms and Questionnaires ..............ouuiiuiunnenennennn.. 22 Appointment Form .......... i eee 22 Exit Interview Form... o.oo ee 23 A STUDY OF THE EFFECT OF REMUNERATION UPON RESPONSE IN THE HEALTH AND NUTRITION EXAMINATION SURVEY E. Earl Bryant, Mary Grace Kovar, and Henry Miller * INTRODUCTION The National Health Survey Act of 1956 pro- vided for the establishment and continuation of a National Health Survey to obtain information about the health of the United States population. The re- sponsibility is placed with the National Center for Health Statistics (NCHS), a research-oriented sta- tistical organization within the Department of Health, Education, and Welfare. Three separate and distinct kinds of programs are employed by NCHS in meeting the objectives of the Act—house- hold health interview surveys, surveys of health resources, and a health examination survey. rr The overall plan of the Health Examination Survey (HES) has been to conduct successive examination cycles for specific age segments of the civilian, noninstitutionalized U.S. population and, bymeans of medical and dental examinations, tests, and measurements, to characterize certain health aspects of the specified population. Between 1959 and 1970, three cycles were completed. The first cycle was a survey of adults aged 18 through 79 years; the second, of children aged 6 through 11 years; and the third, of youths aged 12 through 17 years. Numerous methodological and ana- lytical reports based on those three surveys have been published. aAt the present time, Mr. Bryant is Chicf, Statistical Methods Staff, Office of Data Systems; Ms. Kovar is a senior statistician, Division of Analysis; and Mr. Miller is Chief, Health Examination Field Operations Branch, Division of Operations, National Center for Health Statistics. The fourth cycle of HES, which began in April 1971, was expanded to include a newly assigned responsibility for measuring and monitoring the nutritional status of the U.S. population. This cycle, referred to as the Health and Nutrition Examination Survey (HANES), was planned to serve a dual purpose. The first purpose was to measure the nutritional status of the U.S. popu- lation 1-74 years of age, and the second was to collect data on the health status and health care needs of the population 25-74 years of age. The success of these surveys depends upon voluntary participation of individuals selected in the sample, For the first three cycles the partici- pation was excellent; the examination response rates were 87, 96, and 90 percent, respectively. For HANES, however, it was apparent early in the survey that response rates were much lower than expected from experience in the previous cycles. After extensive efforts to improve interviewer techniques and to increase publicity and commu- nity involvement, the response rates remained low; only 64 percent of the 5,641 sample persons selected for the first 15 sites were examined (the rates at different sites ranged from 46 to 86 per- cent), Thus, other measures were required to improve the response rate. It was hypothesized that response rates might be increased if an honorarium were paid to indi- viduals who participated in the survey. Very little data from controlled experiments relating to this problem were available to support this hypothesis. Remuneration had been used extensively in mail surveys, but the amount of the honorarium had generally been small and the response rates so low that the results were not relevant, In house- hold interview surveys the results of paying re- spondents had been mixed. For example, Dohren- wend reported no difference in response rates when an honorarium of $5 was offered in 163 households in New York City. > Because of the lack of conclusive evidence from previous studies, the decision to test the ef- fects of remuneration in HANES was made. Al- though most surveys conducted by the Federal Government are based on unpaid, voluntary par- ticipation, it was reasoned that remuneration for participating in HANES could be justified because full participation in the survey requires several hours of the respondent's time and for many adults this means time lost from work, theneed to pay a babysitter, or other inconveniences. Also, the cost of remuneration would result in some offsetting economies if the number of contacts required to obtain response could be reduced. However, even if the unit costs of the survey were increased by a $10 honorarium—the amount proposed for the study, the cost would be small compared with the importance of the total program if remuneration should increase the response rate to a satisfactory level. Necessary clearances were submitted and plans were developed in November 1971 to institute a study of the effect of remuneration upon re- sponse, The earliest possible date that the study could be started was January 1972; at this time operations would be starting at three sites— Tucson, Arizona, West Palm Beach, Florida, and San Antonio, Texas. The last site was selected for two primary reasons—a sample size of about 600 persons as compared with 350 and 500 at the other two, and the fact that the San Antonio population was expected to be more typical of future HANES survey sites, particularly with respect to income and age distributions, than that of either Tucson or West Palm Beach. This report describes the design and findings of that study. In addition, since remuneration was instituted in the succeeding part of the HANES survey, the report includes a comparison of re- sponse rates in the National survey before and after implementation of remuneration, BACKGROUND OF HANES Reports on the background, sample design, general plan, and operation of HANES %and all data collection forms of HANES 7 have been published, The sample design and procedures used in San Antonio were the same as those described in the reports except for the changes made specifically for the remuneration study procedures described in this report. So that the reader can understand how the remuneration study relates to the national survey, a brief description of the HANES sample design and survey procedures is presented. HANES was similar to the three previous cycles of HES, in both general survey methodology and design. The examinations took place in spe- cially built and equipped mobile examination cen- ters consisting of three interconnected trailers. The staff of an examination center included physi- cians, dentists, nurses, laboratory and health technicians, and dietary interviewers. The sample was based on a highly stratified, multistage prob- ability design which made it possible to produce National and regional estimates by various socio- economic and demographic characteristics. The sample consisted of approximately 30,000 persons from 65 primary sampling units (PSU's), i.e., counties or groups of contiguous counties through- out the United States. The persons selected for the examination were chosen to make up a representa- tive sample of the total population with oversam- pling of groups of persons with a high risk of mal- nutrition, In keeping with the dual purpose con- cept of HANES, a subset of persons aged 25-74 years received, in addition to the nutrition exam- ination, a more detailed examination designed primarily to detect certain chronic diseases and to permit an assessment of unmet medical needs through comparing examination findings for sev- eral target conditions with the individual's self- perceived health needs and behavior. The first contact with a sample household was made by a Bureau of the Census interviewer, At that time, a brief interview was conducted to de- termine the age and sex of each household member and to collect other demographic and socioeco- nomic information required for the survey, If no one was found at home after repeated calls, or if the household members refused to be interviewed, the interviewer tried to determine household com- position from neighbors. The primary purpose for the data collected in this interview was to provide a framework for selecting a subsample of house- hold members to receive the examination. The next contact was made by a member of the HANES staff, referred to as a Health Exami- nation Representative (HER). The purpose of this visit was to administer a medical history question- naire to the sample persons and to make appoint- ments for the sample persons to be examined at the centrally located examination center. Intensive efforts were made during the 3-6 week duration of the survey in an area (the length of time depended on the number of people to be examined) to maxi- mize the response rate, Call-backs were madeto those who broke appointments as well as to those who had not made appointments at the time of the first visit by the HER. EXPERIMENTAL DESIGN AND DATA COLLECTION Experimental Design The design for the study was superimposed upon the "within PSU" sample design of HANES for the San Antonio Standard Metropolitan Statisti- cal Area. As such, that portion of the#ANES de- sign needs to be briefly described so that the experimental design can be understood, Enumeration Districts (ED's) in each PSU were divided into segments of an expected six housing units each. Then a systematic sample of segments was selected, The number of segments selected for any particular PSU was based on a predetermined sample size of between 300 and 600 sample persons. The size was set by the PSU's population and the number of persons living in the ED's with median family income of less than $3,000. The ED's that fell into the sample as a re- sult of the segment selection were then coded into two economic classes—median family income of less than $3,000 per year and $3,000 or more per year according to 1960 Bureau of the Census clas- sifications. All sample segments in the low income ED's were retained in the sample. For those sam- ple segments in the higher income ED's, the seg- ments were divided into eight random subsamples and one of the subsamples was chosen to remain in the sample. The expected result of this sampling plan was that about a fourth of the sample persons would have family incomes of less than $3,000, The initial sample in San Antonio consisted of 651 households; of these, 631 were interviewed by Bureau of Census interviewers, The 2,010 persons in the initial sample were listed by age and sex Table A. Subsampling rates used in HANES Sampling Age Sex Toile 1-5 years-==---=====-== Both 1/2 6-19 years---=-=====--- Both 1/4 20-44 years-==-======== Male 1/4 20-44 years--========= Female 1/2 45-64 years----------- Both 1/4 65-74 years=---=====-- Both 1/1 (information about the age and sex of the members of the 20 noninterviewed households was obtained from neighbors), and a systematic sample of 747 "eligible" persons was selected using the HANES sampling rates shown in table A, The final sam- ple of 603 persons was determined by systemati- cally deleting 144 persons from the eligible sam- ple. This subsampling was necessary because a maximum of about 600 persons could be examined at any one site (maximum and minimum limits of 600 and 300 were set as part of the design). The final 603 sample persons came from 402 house- holds in 138 segments, The first step in the ex- perimental design was to classify the 138 segments by segment size (number of occupied households in segment) and by median family income, using the information that had been collected by the household interviewers. The segments were then sorted into seven size-income classes as shown in table B. Table B. Distribution of segments by seg- ment size and median family income Median annual family income for segment Number of occu- pied households in segment Less $4,000 Total than or $4,000 | more Total---=-=--= 138 44 94 cooly srl) | {hn Jem 39 15 24 4 or more------- 42 15 27 Segments were randomly paired within each cell, One segment of each pair was then randomly selected to have all of the sample persons in that segment told about the $10 remuneration, The other segment of the pair was selected to have none of the sample persons told. Note, however, that all persons who were examined received $10. The difference was that persons in the "Not told" segments did not know about the remuneration until they were at the examination center while those in the "Told" segments knew in advance of the exam- ination. The decision to classify everyone in a particular segment as either "Told or "Not told" was made because it was felt that there might be communication between households within a seg- ment and the 'Not told" sample person would learn of the payment from a neighbor. The pairs of segments were then randomly assigned to the HER's so that each interviewer's assignment consisted of a representative subsam- ple of the segments, An attempt was thus made to control three variables—income, interviewer, and segment size. Income was selected as a control variable because it was believed that an offer of $10 would influence persons with low income more than it would those with higher incomes. Interviewer as- signments were selected because some inter- viewers are more successful than others in ob- taining response in surveys. In HANES wherethe function of the HER's is to interview sample per- sons and to persuade them to come in for an ex- amination, the interviewer's effect may be even more important than in a survey where theinter- viewer's function is only to obtain an interview, Segment size was selected as a control because of the possible interaction of the sample persons within segments and because the size of interview assignments had to be regulated as some of the interviewers could work for only 2 weeks in San Antonio before they had to report to another HANES examination site. Assigning too many sample persons to these interviewers would have made it impossible for them to complete their as- signments, Data Collection The design and purpose of the study was thor - oughly explained to the HER's before the HANES interviewing began, They were told that they must conduct the survey according to regular HANES procedures, except for the changes required for the study. The major difference between their usual routine and the experimental procedure was that they must tell all sample persons in the experimental segments about the remuneration. Under no circumstances were they to tell those in the control group about the $10 unless a person in a control segment had heard about remunera- tion and asked. Then, of course, he was told and that fact was recorded. To assure a standard approach in the offer of remuneration, a statement was prepared and made part of the interviewer's introduction to the house- hold. The statement read: ''"The United States Pub- lic Health Service is conducting a study on the health of the American people. The people chosen for the study are part of a carefully selected scientific sample, representative of all people in the United States. For the study to accurately picture the health of theNation, we need your help. Today, I will ask some questions about your health and related matters. Then I would like tomake an appointment for you to receive a free health exam- ination at our special examination center. As an expression of appreciation for your help in this important survey, and as compensation for your time and inconvenience, you will receive a fee of $10 after the examination, Also, we will send any significant findings of the examination to the phy- sician and dentist that you may want to designate." This statement was either read or paraphrased for each sample person in the experimental seg- ments. If more than one family member was in the sample, the interviewer emphasized that each sample person would receive $10. For those in the control segments, the statement excluded the sen- tence about remuneration. After the sample person had been examined, each was asked to complete an exit interview form. The primary purpose of the exit interview was to determine whether the sample person knew about the remuneration before coming to the examination center, A facsimile of the form is in appendix II. In any experiment of this kind, itis inevitable that the design will not be followed exactly and problems will occur, One of the problems thatdid arise in this study resulted from the need to have interpreters accompany interviewers to approxi- mately 10 percent of the households where noone could speak English. Some training was given to all the interpreters but they could not be randomly as- signed and, consequently, the results are probably contaminated to some extent by interpreter ef- fects. A second problem arose because some of the HER's were not able to complete their assignments before leaving San Antonio. The goal had been to have the assigned interviewer complete at least the first contact with a sample person, attempt to make the examination appointment, and to offer remuneration if the sample person was in an ex- perimental segment, At the end of the fourth week of the survey, four of the six interviewers had de- parted without completing the first contact with 109 sample persons; 50 in the experimental seg- ments and 59 in the control segments, These sample persons were randomly reassigned to the two remaining interviewers, Also, it was necessary to hire additional tem- porary interviewers near the end of the study to followup on persons who had broken appointments or who for other reasons had not been examined. However, these interviewers were well-trained, experienced interviewers and their assignments included similar proportions of sample persons from both experimental and control segments, FINDINGS OF THE SAN ANTONIO STUDY Examination Rates Telling a sample person that he would be given $10 after being examined had a positive ef- fect on the response rate in San Antonio. Among the 303 persons in the experimental segments who were contacted by the HER's, 82 percent were ex- amined; among the 292 persons in the control seg- ments who were contacted by the HER's, 70 per- cent were examined, (Eight persons whom the HER's were never able to contact are excluded from this analysis of examination rates.) The dif- ference of 12 percentage points was statistically significant and was large enough to have an im- portant implication for future HANES procedures. The differences reported here are probably conservative since some persons were not told about remuneration even though they should have been and a few were told even though they should not have been. According to the records kept by interviewers, there were 10 errors of not telling people who should have been told and 4 errors of telling people who should not have been, According to answers given by those sample persons who filled out the Exit Interview questionnaire, as many as 20 percent of the experimental group may not have known or understood about remuneration, while 14 percent of those inthe control group may have known, It is difficult, however to evaluate the sample persons' responses to the Exit Interview because interpreters were not available and because there is internal evidence that the ques- tions were not always understood. For example, when answering the question: "Before coming for the examination, were you told that you would receive payment as compensation for your time if you came?" one person answered ''No" and then explained how he knew that he would re- ceive $10, The possible effects of this type of error should be kept in mind when interpreting the re- sults in this report since all response rates were computed according to the original assignment of the segments. Tables 1-6 provide a comparison of the re- sponse rates for the experimental and control groups for a number of subsets of the population according to age, sex, income group, and number of sample persons in a household. One notable observation is that the observed response rate was almost uniformly higher when renumeration was offered than when it was not; in only 3 of the 41 different (but not always mutually exclusive) subclasses shown in the tables was the observed difference zero or negative, Although some of the positive differences were small and consequently of little practical importance, more than half of them were 10 percentage points or larger. To provide a more objective evaluation of a large number of positive response differences, consider the six mutually exclusive age-sex classes shown in table 1 where five of the differ- ences are positive and one is zero, The prob- ability of occurrence of this event by chance alone, assuming that there is no difference between the response rates of the control and experimental groups regardless of age or sex, is ,¢ (%)¢ or about 9 in 100 trials, In table 2, all of the six in- come-age classes show positive differences, an event which would occur by chance alone about 2 times in 100 trials. Even stronger evidence of the existence of dif- ferences among population subgroups is the fact that most of the differences observed in the exper - imental study were also observed for succeeding stands of HANES where remuneration was routine- ly offered. This is discussed in the section ""Imple- mentation," Because of the limited sample size for the experiment and the resulting small number of per - sons in subclasses of the sample, itis not possible to draw firm conclusions for most of the population subgroups when considered separately. Neverthe- less, some knowledge about relationships can be gained by examining differences among the sub- groups. The following analysis is based primarily on the normal deviate test, using the 0.051evel of significance. The test statistics are not exact since sampling errors were approximated using the pro- cedure described in appendix I. The test is prob- ably conservative, however erring in terms of not rejecting the null hypothesis of nodifferenceinre- sponse rates when in fact there is a difference. Tables 1-3 show the number of sample per- sons, the proportion examined, and the difference between the "Told" and '"Not told" groups for per- sons classified by age, sex, and family income group, These three variables are particularly im- portant in the analysis of HANES data on nutrition, Reliable measures of the nutrition status of chil- dren, women in the childbearing years, and low- income persons are needed to design and evaluate programs aimed at improving the nutritional lev- els of these high-risk groups. The examination rate for persons 1-19 years of age who were not told about remuneration was relatively high—=83 percent (table 1.) The differ- ence between that and the rate of 90 percent for persons in the same age group who were told was not significant, However, women inthe childbear- ing ages (20-44 years) did show a significant dif- ference in response rates with 90 percent re- sponding in the "Told" group as compared with 65 percent of those who were not told about re- muneration, With respect toincome (table 2), the examina- tion rate for persons ina family with an annual in- come of $4,000 or more was significantly higher in the '"Told'"" group (85 percent) than in the "Not told" group (72 percent), However, for persons ina fam- ily with an income of under $4,000 the difference 6 between 78 percent in the ''Told" group and 67 percent in the "Not told" group was not statistically significant, The lack of a significant difference be- tween the two groups in this income class as contrasted with the higher income group may be due to the fact that there were only214 persons in the lower income category as compared with 344 in the higher income category. With a smaller number of persons in the category, a difference must be larger before it can be detected by a sta- tistical test, The lowest examination rate for any age-sex class was that for women aged 45-74 (table 1). Being told that they would receive $10 after the examination had no detectable effect onthe exam- ination rates; only 56 percent of those told about remuneration were examined compared with 52 percent of those not told. There may be many reasons why more of the older women did not respond. They may include fear or reluctance to be examined by a strange physician, fear of having certain physical condi- tions diagnosed, general bad health and already under rather intensive medical care, and reluc- tance to travel long distances ina taxi, Also, over half of these women (58 percent) were the only sample person in their household. In addition tothe three demographic variables of age, sex, and family income, which have been considered so far inthis analysis, there is another variable, number of sample persons in the house- hold, which may help to explain differencesinre- sponse rates, Because of the way the HANES sample is drawn—first a sample of segments and thena sam- ple of persons listed in the households in those segments-—it is possible tohave one, two, or more sample persons in the same household. It seemed possible that the number of sample persons inthe household might also have a positive influence on the examination or response rates, First, some sample persons might be less apprehensive about the trip to the examination center and the exam- ination if they were in the company of another sam- ple person from the same household, Second, the combined or total amount of remuneration avail- able to a household with two or more sample per- sons might also have a positive effect on response. The response rates by those variables are given in tables 4-6 and summarized intable C, Among those persons not told about remuner - ation, 65 percent were examined when there was Table C. Proportion of persons examined by remuneration status, according to num- ber of sample persons in the household Remuneration status Number of sample per- sons in Not Dif- household Total Told 0 fer- told ence Total----- 76 .82 | .70| l.12 One-=-========- .67 .68 .65 .03 Two or more--- .82 .90 74 1,16 Difference---- Lis] 1.22 | .09 ces lsignificant at the 5-percent level. only one sample person from a household and 74 percent when there were two or more sample per - sons from the same household (table 4), The prob- ability of a difference of this size or larger oc- curring by chance is about 0.75; thus the evidence of a real difference is relatively weak. Among those told about remuneration, 68 percent were examined when there was only one sample person from a household but 90 percent when there were two or more sample persons from the same house- hold, a statistically significant difference of 22 percentage points. The difference between the "Told" and 'Not told" groups was not significant when there was only one sample person from a household but was statistically significant when there were two or more sample persons from the same household. The differential response rates according to the number of sample persons in the household of- fers a possible explanation for the failure to detect a difference in rates for persons in low income households. As shown in table 6, differences be- tween the "Told" and "Not told" groups were small (5 to 6 percentage points) and not signifi- cant when there was only one sample person, regardless of income class, For households with two or more sample persons, the estimated differ- ences for both income classes were substantial, being 24 and 18 percentage points for the less than $4,000 and $4,000 or more classes, respectively. However, using the standard normal deviate test, even these large differences are not statistically significant. In the study, 47 percent of the sample per- sons from households with family incomes of under $4,000 were one-sample-person house- holds compared with 36 percent of those with family incomes of $4,000 or more. Thus, the amount of remuneration per household was more likely to be $10 per household in low income than in high income households. Number of Contacts Made by HER'’s The purpose of the remuneration study was to determine whether response rates would be changed by paying the sample persons to come to the examination center. The examination rates were improved, but it could have been possible, however, that other factors, such as more inten- sive followup among the "Told" group, may have accounted for the difference. As shown in table7, this was clearly not true; the average number of contacts per sample person for eachage, sex, and income category was almost identical for each group. Also, thereisno evidence from the data re- corded at the time of the study thatdifferent sur- vey procedures were used for sample persons as- signed to the two groups. The final point to be investigated was whether there were any economies in terms of fewer con- tacts per examined person, If so, these factors would offset, at least to some extent, the cost of remuneration, Tables 8-10 provide some evidence that people are more cooperative and that less effort is required to obtain response when re- muneration is offered. Table 8 compares the proportion of persons making appointments at the first contactby HER's according to age, sex, and family income, Although none of the differences are statistically signifi- cant, the appointment rate was largest for each age, sex, and income class in the group told about remuneration, Table 9 shows that a larger propor - tion of the "Told" group kept their appointments than of the "Not told" group for each of the sub- classes. Possibly the strongest evidence that remu- neration influences cooperation and thus reduces the number of contacts required to elicit response is shown in table 10. Only 2.1 sample person con- tacts per examined person were required for the "Told" group as compared with 2.5 such contacts per examined person for the "Not told" group, This difference of 0.4 is statistically significant, Note also that the savings is apparent for each age, sex, and income class. One interesting point which is not shown in the table is the amount of effort spent in trying to per- suade women aged 45 years or older tocome in for an examination, As discussed earlier, this group had a very low examination rate regardless ofre- muneration status, but it was not due tolack of ef- fort, This group received 3.5 sample person con- tacts per examined person, in contrast tothe 2.25 sample person contacts per examined person for the entire stand. Neither remuneration nor inten- sive effort had much effect in attracting these women, IMPLEMENTATION The findings of this study were considered sufficient to include remuneration as a routine procedure in the national survey. Remuneration of $10 per person examined was initiated simultane- ously at the twenty-first (Avoyelles, Louisiana) and twenty-second (San Francisco, California) stands in the sequence of operations to cover the 65 stands scheduled for the survey. When the first 35S stands of the national survey had been com- pleted, (excluding the San Antonio stand), 6,035 persons had been offered remuneration and 77.5 percent of them had been examined. This com- pares with 68.1 percent of the 7,335 persons in- terviewed when remuneration was not offered, Examination rates have been higher for each of the age-sex classes (table 11) and for each of the age- income classes (table 12) since remuneration has been a routine procedure in HANES, However, the inference from the San Antonio data that remuner - ation would be more effective with two or more sample persons in the household than with one has not been substantiated in the national survey (table 13), It is not possible to assess just how much of this rather substantial response difference in the national survey can be attributed to remuneration since other factors not related to remuneration were also involved, Interviewer training continued throughout the survey, additional interviewers were added to the staff so that more intensive con- tacts were possible, and all survey procedures thought to affect response rates wereimproved as much as possible, The preremuneration stands included anum- ber of large metropolitan areas where, on the basis of experience in previous health examination surveys, examination rates were expected to be low, Data on population size are available, how- ever, to compare the examination rates with and without remuneration according to population size of the areas surveyed (table 14). Regardless of the size of the population, examination rates were higher with remuneration than without; re- gardless of whether remuneration was offered or not, response rates were lowest in the areas with one million or more people, Provisional response data for the 65 HANES stands show that of the 28,043 persons in the total sample, 20,749 or 74.0 percent were examined, During the last 30 stands, those in which remun- eration was offered to all sample persons, the response rate was 76.4 percent. The overall re- sponse rate at the 45 stands where remuneration was offered in HANES (excluding San Antonio) was therefore 76.8 percent as compared to 68,1 percent for the 19 stands where remuneration was not offered. REFERENCES tional Center for Health Statistics: Origin, program and operation of the U.S. National Health Survey. Vital and Health Statistics. PHS Pub. No. 1000-Series 1-No. 1. Public Health Service. Washington. U.S. Government Printing Office, Aug. 1963. 2 National Center for Health Statistics: Plan and initial program of the Health Examination Survey. Vital and Health Statistics. PHS Pub. No. 1000-Series 1-No. 4. Public Health Service. Washington. U.S. Government Printing Office, July 1965. 3 watiomsl Center for Health Statistics: Plan, operation, and response results of a program of children’s examinations. Vital and Health Statistics. PHS Pub. No. 1000-Series 1-No. 5. Public Health Service. Washington. U.S. Government Printing Office, Oct. 1967. _ _ : National Center for Health Statistics: Plan and operation of a health examination survey of U.S. youths, 12-17 years of age. Vital and Health Statistics. PHS Pub. No. 1000-Series 1-No. 8. Public Health Service. Washington. U.S. Government Printing Office, Sept. 1969. 5 Dohrenwend, B. S.: An Experimental Study of Payments to Respondents. Public Opinion Quarterly 34:621, 1970. 6 . : National Center for Health Statistics: Plan and operation of the Health and Nutrition Examination Survey, United States, 1971-1973. Vital and Health Statistics. Series 1-No. 10a. DHEW Pub. No. (HSM) 73-1310. Washington. U.S. Government Printing Office, Feb. 1973. ? Smtond Center for Health Statistics: Plan and operation of the Health and Nutrition Examination Survey, United States, 1971-1973. Vital and Health Statistics. Series 1-No. 10b. DHEW Pub. No. (HSM) 73-1310. Washington. U.S. Government Printing Office, Feb. 1973. Table 1: 10. il. 12, 13. 14. LIST OF DETAILED TABLES Number of sample persons and proportion examined by remuneration status, according to sex and age: HANES Remuneration Study======-=-w-=w-w- memes ee sees ——————— Number of sample persons and proportion examined by remuneration status, according to family income and age: HANES Remuneration Study=-===--- mmm meee ————— —————— Number of sample persons and proportion examined by remuneration status, according to family income and sex: HANES Remuneration Study----==-eeeemceccaaoa- ww Number of sample persons and proportion examinedby remuneration status, according to number of sample persons in household and age: HANES Remuneration Study------ Number of sample persons and proportion examined by remuneration status, according to number of sample persons in household and sex: HANES Remuneration Study------ Number of sample persons and proportion examined by remuneration status, according to number of sample persons in household and family income: HANES Remuneration Study ===== em em meee ee remem mmm ————————— Number of sample persons and average number of HER contacts per person by remu- neration status, according to age, sex, and family income: HANES Remuneration Study = mmm mm mmm mmm mmm ee ee eee eee Number of sample persons and proportion making appointment at first contact by remuneration status, according to age, sex, and family income: HANES Remunera- tion Study=-=-== cm mmo ee ee ence meen Number of sample persons and proportion examined after only one HER contact by remuneration status, according to age, sex, and family income: HANES Remunera- tion Study==-=- === mmm em em ee ee emer emma Number of examined persons and total sample person HER contacts per examined person by remuneration status, according to age, sex, and family income: HANES Remuneration Study=-===ce= cc comme ee meme ——— Number of sample persons and proportion examined by whether remuneration was offered, according to sex and age: First 35 HANES stands (excluding San Antonio)-- Number of sample persons and proportion examined by whether remuneration was offered, according to family income and age: First 35 HANES stands (excluding San Antonio) === =m comme ee ee emcee mmm Number of sample persons and proportion examined by whether remuneration was offered, according to number of sample persons in household and age: First 35 HANES stands (excluding San Antonio) ====== ceo comme eee eee Number of sample persons and proportion examined by whether remuneration was offered, according to population of areas in survey: First 35 HANES stands (ex- cluding San Antonio) ==--= momo moe meme eem Page 11 11 12 12 13 13 14 14 15 15 16 17 18 18 Table 1. Number of sample persons and proportion examined by remuneration status, ac- cording to sex and age: HANES Remuneration Study Told Not told Sex and age of sample person Difference Nupbee Proportion Nother Proportion sample examined Ske examined Both sexes, 1-74 years------- 303 .82 292 .70 Laz 1-19 years=======m===-mmceceeo----- 119 .90 110 .83 .07 20-44 years-----===-==--cee-e----- 85 .86 85 .67 1,19 45-74 yearS~=======mmmmmmmmcaoa——- 99 .69 97 +39 .10 Male, 1-74 years===-======-=-- 129 .88 123 74 14 1-19 years----=====-mmmmmcccoooo-= 57 +93 55 .80 +13 20-44 year§-=======mmmm-m--oeeo---- 27 .78 23 74 .04 45-74 years-=======mmmm-ccacoo-—-- 45 .84 45 .67 .17 Female, 1-74 years=====----- 174 .78 169 .67 .11 1-19 yearS=========--mmeeeecceo—--- 62 .85 55 .85 .00 20-44 yearS=======m-mmmm--eoooo--- 58 .90 62 +65 1,25 45-74 yearS-=======mmmmmmmcomoeo——- 54 .56 52 +52 .04 significant at the 5-percent level. Table 2. Number of sample persons and proportion examined by remuneration status, ac- cording to family income and age: HANES Remuneration Study Told Not told Family income and age of sample Difference person Nurber Proportion Furber Proportion sample examined sample examined All incomes, 1-74 years----- 303 .82 292 .70 1,12 1-19 years----=======-m----------- 119 .90 110 .83 .07 20-44 years 85 .86 85 .67 1,19 45-74 years 99 .69 97 .59 .10 Under $4,000, 1-74 years---- 115 .78 99 .67 .11 1-19 years--========---c---------- 44 +93 35 .83 .10 20-44 years=========m---e--------- 25 .84 21 .67 .17 45-74 years-=====mm-mm-me-——ee—o== 46 .60 43 «53 .07 $4,000 or more, 1-74 years-- 171 .85 173 .72 1,15 1-19 years---=======m-m---o------- 68 .88 68 .82 .06 20-44 years---=====mmm=m--caoo----- 56 .88 58 .67 .21 45-74 years-======mem=m-mmmemo————== 47 77 47 .62 .15 Unknown income, 1-74 years-- 17 .76 20 .75 .01 !significant at the 5-percent level. 11 Table 3. HANES Remuneration Study Number of sample persons and proportion examined by remuneration status, ac- cording to family income and sex: Told Not told Family income and sex of sample Difference person Raber Proportion Baber Proportion sample examined sample examined All incomes, both sexes----- 303 .82 292 .70 1,12 Male-==cccccmcmcccc ccc ee 129 .88 123 74 14 Female===ccccccccaanccccccccccaea a 174 .78 169 .67 .11 Under $4,000, both sexes---- 115 .78 99 .67 +11 Male=-weccccac cca caccccccceeeeam 47 .87 38 .68 .19 Female~=ccccccnanccaccccacaanaanaa 68 «72 61 .66 .06 $4,000 or more, both sexes-- 171 .85 173 72 }.A3 Male=ewecccccc cc cc ccc ccna aaa 74 .89 78 77 .12 Femaleseecnncnannccaacancacananaaa 97 .81 95 .67 14 Unknown income-====-=ceccaa-- 17 .76 20 73 .01 lsignificant at the 5-percent level. Table 4. Number of sample persons and proportion examined by remuneration status, ac- cording to number of sample persons in household and age: HANES Remuneration Study Told Not told Number of sample persons in household and age of sample Difference person Raber Proportion Nuber Proportion sample examined sample examined All households, 1-74 years-- 303 .82 292 .70 1.12 1-19 years-==-e-ecccccmmcccncnacaan 119 .90 110 .83 .07 20-44 years=mmmmmmemmmcececee——aaa 85 .86 85 .67 1,19 45-74 yearS-=---=m=meememceceaaao- 99 .69 97 .59 .10 One sample person, 1-74 yearS===mmemeece mca 114 .68 125 .65 .03 1-19 years=====-e-cmcccmccccaneaa—- 40 13 42 .74 .04 20-44 yearS-==----ccccccccccaaaaa- 22 .68 30 .60 .08 45-74 yearS=====-mmecccccccacnna—— 52 .62 53 .60 .02 Two or more sample persons, 1-74 yearse==mmmcememeeea-- 189 .90 167 vA 1.16 1-19 years===-=----cececccmncanna- 79 .96 68 . 88 .08 20-44 years====-----eeeecmemaaaa 63 .92 55 .71 .21 45-74 years===-=-emeccccmccm——a——— 47 77 44 «57 .20 significant at the 5-percent level. 12 Table 5. Number of sample persons and proportion examined by remuneration status, ac- cording to number of sample persons in household and sex: HANES Remuneration Study Told Not told Number of sample persons in household and sex of sample Difference person Ruder Proportion Rupise Proportion sample examined sample examined All households, both sexes-- 303 .82 292 .70 1 2 Maleeececacaaa sessecssacccccacaca- 129 .88 123 «74 14 Female=ecacaacacacanaaa- smececmemaa- 174 +78 169 .67 11 One sample person, both SE€XESemsuncanncaannnnn —mm———— 114 .68 125 .65 .03 Maleesemceccccacana- seeccseccceca- 42 .83 52 + 71 .12 Femalee=ccecccccccncanacaan cececaa- 72 .60 73 .60 .00 Two or more sample persons, 1 both sexesee=ee-a- cmmceeca- 189 .90 167 74 +16 Malesceecccncanaccacacccccacccaaaa 87 .90 71 +76 .14 Femaleeweeacaaa cesses ccccccccacaa- 102 .90 96 .73 17 Significant at the 5-percent level, Table 6. Number of sample persons and proportion examined by remuneration status, ac- cording to number of sample persons in household and family income: HANES Remunera- tion Study Told Not told Number of sample persons in 3 Difference household and family income Faiise Proportion Ll Proportion sample examined sample examined All households, 2 all incomes’ me=ececcaccaca- 286 .82 272 .70 +12 Under $4,000-=-c=cccaccacccanaaaan 115 .78 99 .67 ,- 11 $4,000 OF MOre===emcememceccaceaax 171 .85 173 +72 +13 One sample person, all incomes==eecccccccacaacaaan 109 .67 117 .67 .00 Under $4,000-==ececcccacacacacanan 48 .60 53 .66 -.06 $4,000 OF mMOre====mmecaccacaaan “-- 61 +72 64 .67 .05 Two or more sample persons, 5 all incomes==eecccacccaann - 177 .92 155 «72 .20 Under $4,000===ccccccccacax cemcaaa 67 + JL 46 .67 .24 $4,000 or more===e==- cseccccccccaa- 110 .92 109 74 .18 Excludes 37 persons with unknown income. Significant at the 5-percent level. Table 7. Number of sample persons and average number of HER contacts per person by remuneration status, according to age, sex, and family income: HANES Remuneration Study Told Not told Age; ex; 20d family income Number Contacts Number Contacts in per in per sample person sample person Total, 1-74 years------=---ceecceecea-- 303 1.7 292 1.7 Age 1-19 YRALSrm mmm mm mmm mmm mm om mmm mm rm on on mo mm 119 1.7 110 1.7 20-44 yeargemrmmmmmmmmmm seme ————————— 85 1.8 85 1.8 U5 Th JE AIL Gomme mimo mm om am a 99 1.7 97 1.7 ex Male---=-==-mmmeceec mmc 129 1.6 123 1.5 Female---===mmrmrecccc ccc; cee mm em 174 1.8 169 1.9 Family income Under $4,000-----=--cmmmmmmmmm mmo emmo 115 1.6 99 1.7 $4,000 or more----=----m--mmmmmce meee mmm —— mmm 171 1.7 173 1.8 UNKNOWN Y miomio = ii n d i.mkod of h t Fbl, t SR e r 17 2.1 20 1.6 Table 8. Number of sample persons and proportion making appointment at first contact by remuneration status, according to age, sex, and family income: HANES Remuneration Study Told Not told Proportion Proportion Age and sex of sample person . and family income Number who made Number who made Difference in | oo ear | in | ments at ments a men sample first sample first contact contact Total, 1-74 years----------- 303 .66 292 .61 .05 Age 1-19 years-----------=----cc------ 119 .69 110 .63 .06 20-44 years------==-=--m---eooaaa- 85 .67 85 .66 .01 45-74 yearS--------cc-cmcmecnnaaa- 99 .63 97 .56 .07 Sex Male-====cccmccmc remem meme mm 129 .73 123 +21 «02 Female------=mememccm ccm meee ee 174 BL 169 .54 .07 Family income Under $4,000-----==--==mmcmmeccen-n 115 .73 99 +70 .08 $4,000 or more----===mmmmme——————— 171 +653 173 .56 .09 Unknown---====m=mm=m--e——————————— 17 .35 20 .65 -.30 14 Table 9. Number of sample persons and proportion examined after only one HER contact by remuneration status, according to age, sex, and family income: HANES Remunera- tion Study Told Not told Age and sex of sample person : ‘ : and family income Number Bropoytion Number Proposition Difference a after one sam Le after one sump contact P contact Total, 1-74 yearS======e==a-- 303 +53 292 44 .09 Age 1-19 yearse==cecccecccccccccccccan- 119 .59 110 47 .12 20-44 yearSeee-eecececccccacacanan 85 .49 85 +42 .07 45-74 yearSeeeece-cecececcncncnan= 99 + 51 97 42 .09 Sex Male--emmmcccacaccccancaracccacann 129 +39 123 . 54 .05 Female-==ececccacaccccccaacccacnan 174 49 169 +37 +12 Family income Under $4,000==eccacccacacax cmmeee- L115 35 99 45 .10 $4,000 Or more==e=e-ececccaceca= -- 171 . 54 173 43 11 Unknownee=eeeaeaca- memes cesaseaa- 17 I. 20 . 50 -.15 Table 10. Number of examined persons and total sample person HER contacts per examined person by remuneration status, according to age, sex, and family income: HANES Remu- neration Study Told Not told Age and sex of sample persons v . . Number Contacts Number Contacts Difference and family income oF per of per examined examined examined examined persons person persons person Total, all examined y PErSONS-====mmceceecaa= 248 24) 205 2.5 Ab Age 1-19 years==e=eeeeemcececceeeeaa= 107 1.9 91 2+) .2 20-44 yearSeeeeemmmeamccccana= 73 2.1 57 2.6 «3 45-74 yearSe-m=ececceccccacaaa- 68 2.5 57 2.9 4 Sex TIL rere seme rm me i ee 113 1.8 91 2.1 3 Female-==acameecaccecccccccann 135 2.3 114 2.8 5 Family Income Under $0,000» ssenmmenanmmnans 90 2.1 66 2.6 15 $4,000 Or MOrE========-aaaaa== 145 2.1 124 2.5 Loy UnKNOWN==eecccasaccnacacaenan= 13 2.1 15 2.1 .0 Stott it Significant at the S5-percent level. Table 11. Number of sample persons and proportion examined by whether remuneration was offered, according to sex and age: First 35 HANES stands (excluding San Antonio) Offered Not offered remuneration remuneration Sex and age of sample person Difference Number Proportion Fiber Proportion sample examined sample examined Both sexes, 1-74 years----- 6,035 .78 1,335 .68 .10 1-19 years==---=ccccmccmmnnemae 2,068 .86 | 2,471 +77 .09 20-44 years=--=--m-mmcmcm mmm 1,959 «27 2,390 .65 .12 45-74 yearg=--m-mmcmmmmme eee 2,008 .70 2,474 .62 .08 Male, 1-74 years-------==-- 2,548 .78 | 3,070 .69 .09 1-19 years------=-cc-cccmcmmccea 1,021 .86 1,228 77 .09 20-44 yearS=-------cceecccnmannaa- 584 .73 718 .62 11 45-74 years==------c-mmmcceeeaaooo 943 74 1,124 .66 .08 Female, 1-74 years-----=---- 3,487 77 | 4,265 .67 .10 1-19 years=-=eecceccmcmccceecaa 1,047 .86 1,243 «27 .09 20-44 yearS==----=---mmeeecmcennn 1,375 .78 1,672 .66 .12 45-74 years==-----ememeeeecneoonon 1,065 .67 1,350 39 .08 'See appendix 1, page 20 for procedures to determine sampling errors. Table 12. Number of sample persons and proportion examined by whether remuneration was offered, according to family income Antonio) and age: First 35 HANES stands (excluding San Offered Not offered remuneration remuneration Family income and age § 1 of sample person Difference Number P 3 Number . in roportion in Proportion sample examined sample examined All incomes, 1-74 yearse-==-=- 6,035 .78 73335 .68 .09 1-19 yearse-eeeececccaccccccancan= 2,068 .86 2,471 227 .09 20-44 yearS=memeemmmmecscesececaa= 1,959 .77 2,390 «65 «12 45-74 yearSeeeeeemceeeeccccescaa= 2,008 70 2,474 .62 .08 Under $4,000, 1-74 years=-=- 1,408 .81 1,455 .68 +12 1-19 yearSemeeecaccsccccccccncaa= 437 .90 384 77 14 20-44 yearSeemmeneseseecececese-= 291 +82 309 .61 11 45-74 yearSeeeeeeeeeeseesceescna- 680 .74 762 +67 07 $4,000 or more, l=74 yeaArSe==m=mmmmmmemnaa= —————— 4,326 .78 5,406 .69 .08 1-19 yearSee=escacacacaccccacana= 1,558 .85 1,976 +13 .07 20-44 yearSee=eeeseemsecea== aceea- 1,585 .76 1,929 .66 .10 45-74 yearSememmeeecccceccccceaena= 1,183 .70 1,501 .62 .08 $4,000-$9,999, 1-74 years-=- 2,341 .76 2,917 .70 .07 1-19 years=e===== ceeesceccccacaa- 835 .84 1,092 .78 .06 20-44 yearSeeeeeeemsececccemcea-- 785 .76 950 +67 .09 45-74 yearSememmeemecmccccceceaa= 721 .68 875 +63 .05 $10,000 or more, 1-74 YEArS==mmmmmecmmemmaneaa== 1,985 .79 2,489 . 69 .10 1-19 yearSe=eeeceecceccceccencana 723 .85 884 «77 .08 20-44 yearS=eeeeeeececccscaceeaa= 800 .76 979 .66 .10 45-74 yearS==em-ee-esccecmcccnana=~- 462 wld 626 .60 .13 Unknown income, 1-74 years-=--m=mmcmmmmmm———-- 301 .62 474 . 54 .09 ‘see appendix I, page 20 for procedures to determine sampling errors. Table 13. Number of sample persons and proportion examined by whether remuneration was offered, according to number of sample stands (excluding San Antonio) persons in household and age: First 35 HANES Offered Not offered remuneration remuneration Number of sample persons in household and age of sample Difference! person Number | Proportion | Number | Proportion in examined in examined sample sample All numbers, 1-74 years--- 6,035 .78 7.335 .68 .10 1-19 years-----------cmcmmomean- 2,068 .86 | 2,471 77 .09 20-44 years-------m-mmmmmme————- 1,959 771 2,390 .65 .12 45-74 years------------mmmm—e——- 2,008 .70 | 2,474 .62 .08 One sample person, 1-74 yearsS-=-====mmemm-—m—————— 2,564 15 3,237 .67 .08 1-19 years---------c-cccemmmmm———- 636 .86 788 .80 .06 20-44 years-----------eecmcmaa—— 747 .74 919 .64 .10 45-74 years-------c--emmemeena—- 1,181 +71 1,530 .62 .09 Two or more sample per- sons, 1-74 years--------- 3,464 .79 | 4,090 .69 +10 1-19 years---====---cememee ea 1,430 .85 1,682 .76 .09 20-44 years-----=----m-mmme———a- 1,208 .78 1,465 .66 12 45-74 years----==-mmmmmemmeee——— 826 .69 943 .62 .07 Unknown number of sample PersonsS------=------e=-==- 7 .57 8 «12 45 See appendix I, page 20 for procedures to determine sampling errors. Table 14. Number of sample persons and proportion examined by whether remuneration was offered, according to population of areas in survey: First 35 HANES stands (exclud- ing San Antonio) Offered Not offered remuneration remuneration Population of areas in survey Difference’ Ratoer Proportion Raber Proportion sample examined sample examined Total--==----eeemcmmccc mm 6,035 .78 7,335 .68 .09 One million or more------------- 2,631 .70 | 2,929 .62 .08 Other urbanized areas----------- 436 +16 1,212 yx .06 Urban placeS-=-=-===-emeeeceeeeaa- 755 .85 780 «73 +12 Rural--=-=--ccccmm mmm emcee mmm 2,213 .84 | 1,914 +73 +11 ‘see appendix I, page 20 for procedures to determine sampling errors. APPENDIX | TECHNICAL NOTES ON METHODS General Qualifications The Remuneration Study was based on a sample of 603 persons who had been selected from the San An- tonio SMSA as part of the National Health and Nutrition Examination Survey (HANES) sample. The purpose of the experiment was to determine if an offer of $10 would influence one's willingness to participate in HANES. The analysis presented in the report is based largely on normal deviate tests of hypotheses. The es- timator for sampling errors required for the analysis assumes a two-stage, stratified cluster design, whereby a simple random sample of segments of about six house- holds each was selected independently from seven size- income classes; within segments, a random sample of people was selected. These assumptions deviate some- what from the actual design. Poststratification was used rather than the assumed prestratification, households rather than persons were chosen at the second stage of selection, and differential sampling rates were used in sample selection within age, sex, and income classes. The effect of these assumptions probably results in an underestimate of variance. On the other hand, there is a component of vari- ance due to interviewers that is not fully reflected in the variance estimates and because of the way the study was carried out, it is not possible to obtain an accurate estimate of the interviewer variance. However, since each interviewer was initially assigned a random sam- ple of segments which had been randomly paired by experimental procedure, the interviewer's effect on the difference in response rates for the two experi- mental groups should be minimized. Estimation Procedure Let P; =response rate for experimental proce- dure A, Py =response rate for control procedure B. = number of sample persons in stratum i who are assigned to procedure A. Gy n, =total number of sample persons as- signed to procedure A. p;, = response rate in ith stratum among peo- I ple assigned to procedure ng, ng and py are defined similarly for pro- cedure B. Sampling Errors Variance estimator for response rates, —Assum- ing the n’ are fixed constants as they would be for prestratification, 2 2 7 ("Bi 2 Op) = z Ops 8 -1\ng Bi The variance of Pp; (and Py ) have two com- ponents—between segments and within segments-—as follows: 0, =— += Pai 2 m 2 Ail ’ ’ 2 Sais Mai PLiQaij : z = Ee r— ” i=1 Ang "aij Ai where 2 2 a (PRij = Pai) i=1 my = 1 AiB The undefined terms in the equation are: n,. = number of sample persons assigned to procedure Ain ith stratumin jth segment, PL; = response rate in jth segment of ith stratum among persons assigned to procedure A, Qr =1- PL Aij = number of segments assigned to pro- cedure A in ith stratum, Ai The variance estimator for py has the same form, Variance of difference, D' =P; - Pp; .—~In gen- eral, the variance of the difference between two ran- dom variables is: For this study, the covariance term in the above equation is considered to be zero. For most variables presented in the report, the assumption is probably close to the truth, The response rates for the two ex- perimental groups should be nearly independent since segments (clusters of households) were randomly as- signed to the two procedures, There may be some in- teraction between the two groups, however, due pri- marily to interviewer effects, since interviewer assignments included both experimental and control households, Neither the magnitude nor direction of the difference is known, but our speculation is that the ef- fect is relatively small and positively correlated. If this is true, then ap = on, + on is an overestimate of the variance. A ? Variance estimator for estimates of number of visits (or contacts).—lL et xy = average number of contacts per person in procedure A; stratum i , El =average number of contacts per person in procedure A, stratum i, segment j . 20 X jk = number of contacts for kth person in segment j, stratum i, procedure A, Then the variance estimator for xy is approx- imately Yr Y! 2 Y! , mar (Xa; - Xai) 1 [ma man Kae - Xay) Og, = z + = > z ~ Tomy (my - 1) Ap im mp (gc) The variance estimator for Xg is the same as X,, except that estimates for the procedure B sample replace those for procedure A. Presentation of variances (San Antonio study and HANES).—Because of scarce resources it was not fea- sible to compute variances for every statistic shown in this report. Instead, sampling variances were com- puted only for a few key statistics as indicated in tables I and II. Sampling errors for other estimated response rates shown in the report were approximated by use of the "design effects'' (DEFF'S) shown in the tables. To determine the approximate sampling variance of an estimated response rate for either the Remuner- ation Study or for the entire HANES, the following for- mula can be used: 2 ’ ’ (DEFF)"P' (1-P') n » n response rate and 3 I sample size for the class in which the response rate applies. The variance estimator of the difference between re- sponse rates is as follows: 2 2 _ (DEFF), Py (1-P}) (DEFF)g Py (1-Pg) ¥ o Pa-Pg + na n Table I. Sampling errors and design effects for response rates, by remuneration status and se- lected population characteristics Told Not told Characteristic Response Sanpling Design [Response Sampling Design rate of Tate effect rate of Tate effect Total, 1-74 years------=-cceeeac-- .82 | .035 | 1.6 | .70 | .038 | 1.4 1-19 years--------cc-emcmmmemmcccceeeeo .90 .069 2.5 33 .047 L:3 20-44 years------cccmmm mcm eeemeeeeem .86 .068 1.8 .67 .066 1.3 Female, 1-74 years----------ccceccceaa-- .78 .047 1.5 .67 .047 1.3 Female, 20-44 years--------ccecccccacaaoa- .90 .071 1.8 .65 .073 1.2 Two or more sample persons in the household: 1-19 years--=----c--emmmmmceeeeeoo .96 .086 3.9 .88 .076 1.9 20-44 years--------cccmcmmcmeeeeeao .92 .089 2.6 i! .073 1.2 45-74 years-----ccecccmccccmceeeeean «77 .087 1.4 +97 .079 1.1 Female, 1-74 years------coccceccccaa-- .90 .071 2.4 .73 .068 1.5 Family income under $4,000------ccceea-- .78 .058 1.5 .67 .066 1.4 Table II. Sampling errors and design effects for number of contacts per examined person, by re- muneration status and selected population characteristics Told Not told Characteristic Contacts Contacts per Sampling | Design per Sampling | Design examined error effect | examined error effect person person Total, 1-74 years-----------c-c--- 2.1 | .101 | 1.8 | 2.5] .101 | 1.8 1-19 years-=-=---cmmemee eee ee 1.9 .139 1.6 2.1 .129 1.4 20-44 years-=------ececcccmeeeeeeee 2.1 .180 1.5 2,6 .146 1.4 Female, 1-74 years-------ccccmccmaccaaao 2.3 .136 1.6 2.8 .123 1.5 Family income under $4,000----------o--- 2.1 .161 2.1 2.6 142 1.6 21 APPENDIX I FORMS AND QUESTIONNAIRES Appointment Form SAMPLE NO. SEGMENT SERIAL COLUMN APPOINTMENT FOR (name) DATE AND TIME OF APPOINTMENT o TIME z IN TIME PERSON OUT [CONTACTED DAY OF WEEK DATE HER DATE TIME RESULT CALL NAME AND ADDRESS REMARKS TELEPHONE: TRANSPORTATION: [IBY PHS AGENT TAXI []BY SELF ADDITIONAL INFORMATION RECORD OF CALLS 22 Exit Interview Form DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Health and Nutrition Examination Survey Sample Number NAME: (CAST) (FTRST) MT Now that you've finished your health examination, we would appreciate some of your opinions about it. This will help us learn how we can improve the survey. 1. Were there any parts of the examination which you did not like for any reason? O Yes [J No If Yes, which parts? 2. Do you feel that the examination was too long? [J Yes 0 No 3. Did you take time off from work to come? J Yes [J No 4. Did you have any problems, worries, or reluctance about coming for this examination? [J Yes [J No If yes, what were they? 5. Before coming for the examination, did you know that you would receive payment as compensation for your time if you came? [J Yes OJ No If Yes: A. By Whom? [J One of our representatives [J A neighbor or friend [J Somebody else - Who? B. If you had not known you would be compensated, would you have come for the examination? [J Yes J No 6. Is there anything else you would like to tell us about the examination? # U. S. GOVERNMENT PRINTING OFFICE : 1975 210-91/¢ 23 VITAL AND HEALTH STATISTICS PUBLICATION SERIES Originally Public Health Service Publication No. 1000 Series 1. Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data, Series 2. Data evaluation and methods reseavch.—Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Series 3. Analytical studies.—Reports presenting analytical or interpretive studies basedon vital and health statistics, carrying the analysis further than the expository types of reports in the other series. Series 4. Documents and committee reports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Series 10. Data from the Health Interview Survev.—Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Series 11. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons, Series 12. Data from the Institutional Population Surveys. — Statistics relating to the health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Series 13. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Series 14. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Series 20. Data on mortalily.—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Series 21, Data on natality, marriage, and divorce.—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility, Series 22, Data from the National Natality and Mortality Surveys,— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Scientific and Technical Information Branch National Center for Health Statistics Public Health Service, HRA Rockville, Md. 20852 U.C. BERKELEY LIBRARIES WIR (021206098