NATIONAL CENTER Series 4 For HEALTH STATISTICS I Number 11

VITAL amd HEALTH STATISTICS DOCUMENTS AND COMMITTEE REPORTS

PROPERTYOF THE PUBLICATIONS8F?ANC~ EDITORIAL LIBMY

use of hospital data for Epidemiologicand Medical-Care Research

A Report of the United States National Committee on Vital and Health Statistics

How hospital data may be used for epidemiologic and medical- care research is discussed with examples of past applications.

U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration

Washington, D. Cl. June 1969 Public Health Service Publication No. 1000-Series 4 -No. 11

For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D. C., 20402- Price 25 cents NATIONAL CENTER FOR HEALTH STATISTICS

THEODORE D. WOOLSEY, Director

PHILIP S.LAWRENCE, SC.D., Associate Director

OSWALD K. SAGEN, PH.D.>, Assistant Director for Health Statistics Development

WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development

ALICE M. WATERHOUSE, M.D., Medical Consultant

JAMES E. KELLY, D. D. S., Dental Advisor

EDWARD E. MINTY, Executive O//icer

MARGERY R. CUNNINGHAM, Information O//icer

OFFICE OF HEALTH STATISTICS ANALYSIS

IWAO M.MORIYAMA,Ph.D.,Di~ector

DEAN E. KRUEGER, Deputy Directo~

Public Health Service Publication No. 1000-Series 4-No. 11

Library of Congress Catalog Card Number 70-600272 FOREWORD

This report, prepared by a Subcommittee of the U.S. National Committee on Vital and Health Statistics, views the development of hospital data in the United States and discusses how they may be used for epidemiologic studies of chronic disease, for disease surveillance purposes, and for medical-care research. The full potential of mor­ bidity and other data in hospital records needs to be exploited for epidemiologic and medical-care research.

Robert L. Berg, M. D., Chairman U.S. National Committee on Vital and Health Statistics

... Ill U.S. NATIONAL COMMITTEE ON VITAL AND HEALTH STATISTICS

Robert L. Berg, M. D., Chairman Charles E. Lewis, M.D. Professor and Chairman, Department of Professor and Chairman, Department of Preventive Medicine and Community Health Preventive Medicine and Community Health The Medical Center University of Kanaas Medical Center The University of Rochester Kansas City, Kansas Rochester, New York Forrest E. Linder, Ph.D. Department of Biostatistics Iwao M. Moriyama, Ph. D., Executive Secretary School of Public Health Director, Office of Health Statistics Analyais University of North Carolina National Center for Health Statistics Chapel Hill, North Carolina Public Heslth Service* Washington, D.C. Robert Parke, Jr. Program Planning Officer Bureau of the Census Donald J. Davids U.S. Department of Commerce Chief, Records and Statistics Section Washington, D.C. Colorado State Department of Public Health John R. Philp, M.D. Denver, Colorado Health Officer County of Orange Health Department Clyde V. Kiaer, Ph.D. Santa Ana, California Senior Member, Technical Staff Donovan J. Thompson, Ph.D. Milbank Memorial Fund Department of Preventive Medicine New York, New York School of Medicine University of Washington Herbert E. Klarman, Ph.D. Seattle, Washington Department of Public Health Administration Theodore D. Woolsey, ex officio School of Hygiene and Public Health Director, National Center for Health Statistics The Johns Hopkins University Public Health Service* Baltimore, Maryland Washington, D.C.

SUBCOMMITTEE ON EPIDEMIOLOGIC USE OF HOSPITAL DATA

Paul M. Denaen, SC.D., Chairman Alfonse T. Masi, M. D., Dr. P.H. Director, Center for Community Health Department of Medicine and Medical Care The University of Tennessee School of Public Health and Medical Memphis, Tennessee School Harvard University Robert W. Miller, M.D. Bostcm, Msssachuaetts “ Chief, Epidemiology Branch National Cancer Institute National Institutes of Health Warren Winkelstein, Jr., M. D., Secretary Public Health Service* Professor, Department of Epidemiology Bethesda, Maryland School of Public Health University of California Robert Ivf. Sigmond Berkeley, California Executive Director Hospital Planning Association of Allegheny County Jacob E. Bearman, Ph.D. Pittsburgh, Pennsylvania Professor, Biometry Division School of Public Health Vergil N. SIee, M.D. College of Medical Sciences Director, Commission on Professional and University of Minnesota Hospital Activities Minneapolis, Minnesota Ann Arbor, Michigan

Alexander D. Langmuir, M.D, Paul F. Wehrle, M.D. Chief, Epidemiology Program Chief Physician National Communicable Disease Center Pediatrics and Communicable Disease Health Services and Mental Health Services Administration, Public Health Service* Los Angeles County General Hospital .4tlanta, Georgia Los Angeles, California

*Depmtmentof Health, Education, nnd Welfare. iv CONTENTS

... Foreword ------m

Introduction ------1

Epidemiologic Research ------1 Retrospective (case-control) Studies ------2 Prospective (cohort) Studies ------3

Epidemiologic Surveillance ------5

Medical-Care Research ------6 Influence of Dia~oses Per Se------6 Influence of Organizational Structure for Providing Medical Care and of Type of Physician ------7 Influence of the Physician's Characteristics ------7 Influence of Various Types of Case Management ------7

Confidentiality ------8

References ------8

v IN THIS REPORT a Subcommittee of the U.S. National Committee on Vital and Health Statistics considers the potential value, the available methods, and the problems of using hospital data fov epidemiolo~”c studies.

Increasing standardization of diagnostic procedures improves the reli­ ability of hospital diagnoses. Increasing utilization of hospitals veduces the difference between the medical expedience of the hospitalized and the geneval populations. Both trends enhance the potential of hospital data fov epidemiolo~”c study. Hospitil vecovds provide access to large num km of cases of specific diseases which would be difficult to identify in general population suvveys.

Techniques ave descvibed, pvoblems are identified, and examples ave given for several types of hospital studies: ~etvospective (usually case- control) studies of specific diseases or conditions; prospective (coho~t) studies with either analytical or clinical- tvial objectives; disease sur­ veillance intended to identify changes in levels of incidence; and med­ ical- care veseavch which is concerned with relationships between utilization and the chavacte-vistics of physicians, hospitals, and case management as well as the specific diagnosis.

Opportunities fov hypothesis testing ave pointed out, along with ~equive­ ments, in terms of procedures for vecovding data and the selection of conhol groups.

vi USE OF HOSPITAL DATA FOR EPIDEMIOLOGIC AND MEDICAL-CARE RESEARCH

INTRODUCTION tional studies of data from chronic disease and psychiatric hospitals and from nursing homes At the suggestion of the Epidemiology Section will sometimes be necessary because short-term of the American Public Health Association made general hospitals are utilized primarily by pa­ during its 1964 meeting in New York City, the tients with acute illnesses and with surgical prob­ Subcommittee on Epidemiologic Use of Hospital lems. Data was constituted by the U.S. National Com­ The three areas in which the information in mittee on Vital and Health Statistics. The task of hospital records can contribute are epidemiologic this Subcommittee was to make a study of the uses research, disease surveillance, and medical-care of diagnostic and other data on hospital patients, research. Each of these areas will be briefly re- including statistics needed for epidemiologic and viewed with respect to the role of the hospital medical-care research, studies of current ther­ record. apeutic practices, and for health surveillance. With the increased standardization of diag­ EPIDEMIOLOGIC RESEARCH nostic tests and criteria, hospital data have be- come more reliable in recent years. The fact that While data on hospital patients have, in the the general hospital is being regarded more and past, provided a resource for testing hypotheses more as the hub of a complex of facilities provid­ on disease causation, it is anticipated that this ing medical care—inpatient and outpatient, pre­ resource will be increasingly useful in the future. ventive and curative—increases the possibility y This is due to the increasing availability of hos­ that hospital data will provide a useful indicator pital care through the extension of private and of the medical experience of the general popula­ governmental support as well as to the increased tion. This opens up the possibility of utilizing di­ recognition of the hospital as a technological agnostic information in hospital records for pur­ center. Masil has pointed out the potential uses poses other than the treatment of patients. Indeed of hospital data for epidemiologic research as the subcommittee believes that the full potential well as their limitations. While some problems of information in hospital records has yet to be attend the use of hospital data for epidemiologic exploited. purposes, the availability of these data as well The major advantage presented by such data as their comprehensive nature suggest that their lies in the ready accessibility of large numbers of use be fully exploited. cases of specific illnesses. To exploit this advan­ The two approaches to the epidemiologic test­ tage to the full, it is essential that rigorous scien­ ing of hypotheses are (1) the retrospective or tific procedures be followed. case-control method and (2) the prospective or In this report hospital refers to the short- cohort method. In general, findings of retrospec­ term general hospital. It is recognized that addi­ tive studies will need to be extended by the appli-

1 cation of more rigorous designs based on either c. High degree of diagnostic accuracy in community population samples or on prospective defining case and control populations. studies of carefully defined hospital patients and On the other hand, this study also illustrated controls. Prospective studies may be regarded as some of the problems which have been cited in more definitive testing of hypotheses. connection with the use of hospital populations.

Retrospective (case-control) Studies a. Hospital detected cases may not be in­ clusive and are subject to selection ac­ Retrospective (case-control) studies can pro­ cording to factors such as socioeconomic ceed along either of two lines —namely, one in status, severity of disease, and psycho- which cases and controls are both drawn from the logical and culturally determined attitudes hospital population or one in which cases are of the hospital patient. drawn from a hospital population and controls from nonhospital populations or the community b. Because of the selection referred to, the at large. population base furnishing the cases can- Cases and contvols jrom hospital popula­ not be precisely defined. tions. —Several studies in which both cases and controls were derived from hospital populations c. Most hospital records are not designed are available for comment. These studies usually with research uses in mind. The records involve diseases which cause the majority of pa­ are frequently incomplete and the ap­ tients with the diseases to be hospitalized and proaches to data collection are not stand­ which can be diagnosed using fairly standard and ardized. Further, there is considerable reproducible criteria, e.g., lung cancer,2 sar­ variability between hospitals in diagnostic coidosis,3 and Hashimoto’s disease.4 accuracy and reproducibility. However, The usual method of conducting a retrospec­ such difficulties may be minimized in a tive study involves identifying a sample of patients research institution like the Roswell Park with the diagnosis and subsequently obtaining rel­ Memorial Institute, where the record sys­ evant information regarding the characteristics of tem and the clinical activities may be these patients. A control sample is obtained from subjected to the control of the research patients with other diagnoses, and the frequency staff. In other situations, research uses of the characteristics under study is compared in may be expected to promote standardiza­ the two groups. A classic example of such a study tion and improvement in quality. was reported by Levin and others2 and was con­ The most serious criticism of the type of cerned with the relationship between cancer of study outlined above is that the hospital admission certain sites and tobacco smoking. The investiga­ rates for each of the designated disease groups tors routinely obtained a history of tobacco usage (cases and controls) may be different and may be from all patients admitted to the Roswell Park related to the independent variable under study. Memorial Institute, a cancer research hospital, Berkson5 has discussed this problem and has de­ beginning in 1938 and showed an association be- lineated the conditions under which valid compari­ tween cigarette smoking and cancer of the lung sons may be made. If unequal admission rates and between pipe smoking and cancer of the lip. are unrelated to the independent variable under This study clearly demonstrated the following ad- study, then comparisons of the diagnostic groups vantages of using data on hospitalized patients for can be made without bias. The problem is to de­ retrospective epidemiologic study: termine whether or not the independent variable affects the admission rate. This determination is a. Availability of a large number of cases of a particular disease and of suitable con­ not always possible, but it must always be at- tempted. trols for comparison analysis. Cases from hospital populations and controls b. Possibility of collecting information on from nonhospitulized populations, -There are independent variables of interest such as several ways of guarding against the possibility y smoking from cases and controls. of error from unrepresentativeness of controls,

2 Among them is the selection of hospital case tive study in that data are collected according to controls from a wide variety of diagnoses so that a defined protocol on a defined cohort. In the pro­ possible hidden factors of bias related to certain spective study the investigator begins with a hy­ diagnostic categories will be diluted. Another ap­ pothesis and designs a protocol to collect sys­ proach is to draw the control sample from the tematically the history, physical examination data, general population usually by one of two methods. and laboratory data necessary tb test the hypoth­ The first is to obtain a matched sample by con- esis. Thus, the fullest control can be exercised trolling on a number of standard variables such in collecting the required information. However, as age, race, and economic status. The second, there may be selection in the hospitalized segment and preferable approach, is to obtain a random of the total cases in the community depending upon sample of the population and then to match it to the hospital utilization pattern for any disease. the sample. The advantage of the latter is that one Examples of some prospective controlled epide­ can get frequency estimates of the independent miologic studies conducted in hospitals will il­ variables of interest in the population at large lustrate some of these points. and compare them with both the matched control A study of viral hepatitis in a group of Boston sample and with the case group. Both of these hospitals highlights the transition from a retro­ methods were used in a study of pregnancy wast­ spective to a prospective analysis of hospital data age and coronary artery disease in females by as well as the increased power of the prospective Winkelstein and others.6 These investigators ob­ study design. After a retrospective review of hos­ tained a case population of women who had sur­ pitalized hepatitis patients, Grady and Chal- vived a myocardial infarction and compared their mers7’8 designed a prospective study in 1963, in­ pregnancy patterns with a matched sample of volving 10 Boston hospitals, to determine whether women drawn by canvassing houses adjacent to hepatitis patients had different types of exposures the residences of the cases and matching on age, than matched controls. Biweekly visits were made race, and marital status. A second control was to each of the participating hospitals. Included in obtained from a random sample of the population the study were patients over 15 years of age with surveyed for other purposes, which also provided a working diagnosis of hepatitis. These patients complete information on pregnancy experience. were given a standard questionnaire which in­ These findings indicated that there was about cluded questions about personal contact with other twice as much pregnancy wastage among the cases hepatitis patients, possible anicteric cases, trav­ of myocardial infarction as among either matched el, eating habits and places, parenteral exposures, controls or controls drawn from the population and transfusions. The same questionnaire was sample, A difficult problem to overcome in this given to paired controls. The controls were type of study is the difference in recall for the matched by age, race, sex, marital status, hospital characteristic under consideration between hos­ pay status, and date of admission. They were pre­ pitalized patients and population controls. Fur­ viously healthy persons admitted for incidental thermore, it is difficult to hide from the inter- surgery, e.g., fractures and appendicitis. viewer the identity of the cases as contrasted with The findings from the study showed that hep­ the controls. atitis patients who had not recently been trans- The advantage of this type of retrospective fused had a significantly larger consumption of design is apparent when dealing with diseases of raw clams and oysters than the controls. These low prevalence in the population. Under such cir­ cases had also received more injections from cumstances it is usually difficult, if not impos ­ physicians. Some other factors such as dental sible, to identify cases from a population sample; procedures and exposure to certain insects showed therefore, reliance must be on hospital data for no differences. Thus, some of the exposures the initial recruitment of the case population. previously suspected of leading to hepatitis were confirmed in this study, and others such as ex­ Prospective (cohort) Studies posure to cockroaches and dental procedures, which were also suspected on clinical inference, Analytic studies. —A prospective hospital- were clearly excluded. based analytic study differs from the retrospec ­

3 As a byproduct of the smdy, it was found technology for the conduct of studies that have the that there were quite distinctive clinical features greatest potential for improving our knowledge for cases of hepatitis transmitted through the for reducing illness, disability, and mortality, and parenteral and oral routes. Rigorously controlled for improving the prognosis of the patients. analytic prospective epidemiologic studies offer Clinical tvials. —A particular class of pro­ opportunities such as this to describe more pre­ spective study is that which deals with the testing cisely the pattern of disease because it is neces­ of the relative efficacy of certain forms of ther­ sary to define and categorize cases prior to the apy, including drugs and vaccines. As with retro­ study. spective studies, concern must be with the prob­ In another analytic prospective hospital- lems of selecting patients and the possibilities of based study, the National Center for Radiological spurious association. As in many experimental Health of the Public Health Serviceg is conducting studies, however, the risk of shortcomings using a cooperative multihospital followup of some hospital patients is diminished wherever treat­ 40,000 adult patients with a history of thyrotoxi ­ ments can be randomly assigned. The generali­ cosis, About three-fourths of these patients had zation of conclusions is, of course, limited to the radioactive iodine treatment for thyrotoxicosis, nature of the hospital population in terms of dis­ and the remainder had surgery. Both groups are eases, their severity, and other selective factors. being followed biannually according to a specified However, for the purposes of certain treatments, protocol. The main objective was to determine if there may be little reason to suspect that there there were more complications (e.g., blood dys ­ will be a significant difference in effect between crasias or thyroid cancer among the radioactive nonhospital cases and hospital patients. For iodine treated patients) than would be expected example, if aspirin can be shown to be effective from the experience among surgically treated in treating rheumatoid arthritis in hospital pa­ controls which might be attributable to radiation tients, is there any reason a priori for thinking exposure. Since no treatments were imposed on that the nonhospital sample would respond dif­ either the case or control groups, this was an ferently? This is analogous to the dilemma of observational analytic study. A systematic follow- using volunteers in a vaccine field trial. Volun­ up is being made to determine any differences in teers of certain age, race, and sex with a partic­ outcome. ular socioeconomic or educational background As with retrospective studies, concern in may possibly react differently to a vaccine than prospective studies must be with the selection of would members of the entire community. Never­ patients and the possibilities of spurious associa­ theless, if the vaccine shows a high degree of tions. Generalization of the conclusions is, of efficacy in the test group, it should also prove to course, strictly limited by the sample of patients be effective when applied to the whole community. coming to the hospital in terms of the severity of An example of a study in this area is the one disease and other selective factors. conducted by Chalmers and others 10 on veterans The difference between an epidemiologic of the Korean war, where hepatitis patients were study and a clinical study sometimes becomes randomly assigned to one of four hospital wards hazy. If the major purpose of an investigation is identical in all aspects of management except for to test a hypothesis using case and control popu­ a single factor of treatment. Systematic evalua­ lations or to derive comparative prevalence and tion of the patients over the short term of several 1 incidence rates in a defined population, it may be months revealed no differences in the outcome, said that the study is epidemiologic. On the other regardless of the treatment regimen. A 10-year hand, a descriptive study of diseases not involving followup of the majority of these patients showed any hypotheses or a control population is a clini­ no adverse effects, which could be attributed cal study. With increasing collaboration between either to type of exercise or to dietary regimen. 11 clinicians and epidemiologists, the difference be- More recently, Garceau and others12 in a co­ comes unimportant. What is important is to em- ordinated study of 10 Boston teaching hospitals ploy epidemiologic methods that are effectively comprising the Boston Inter-hospital Liver Group consistent with statistical theory and clinical tested the efficacy of surgery versus conserva-

4 tive management of portal hypertension. Patients pirin was studied in a systematically controlled with liver cirrhosis who fulfilled certain physical fashion, and, in turn, the effect of other medica­ and laboratory criteria were invited to partici­ tions will be compared with it. pate in the study knowing that they would have a The same theory that applies to inpatients random chance of being selected for surgery or would apply to outpatient populations whether they for medical management. This study was impor­ be individuals attending a clinic or occupational tant because the criteria for eligibility for surgery groups enrolled in hospitalization programs. were defined prior to the study, and eligible volun­ Another possibility not developed in this report teers were randomly assigned to this category. because of the paucity of published reports is the Although patients who were randomly picked for potential of using Blue Cross records and similar surgery did better than those who were not eligible hospital insurance programs as a source of cases for surgery because of medical contraindications, and controls. It is hoped that eventually these rec­ they did no better than the comparison group—the ords can be used prospectively to determine patients randomly assigned to conservative man­ comparative morbidity in a study of disease asso­ agement. If a rigorous study method can be applied ciations or possibly for confirming high-risk to a procedure as major as portacaval shunt sur­ morbidity groups identified through independent gery in a disease as variable and potentially life sources. threatening as advanced liver cirrhosis, then this approach can also be applied to most therapeutic EPIDEMIOLOGIC SURVEILLANCE questions in medicine. The most important factor is a dedication to discover the true answer to a The existing system of epidemiologic sur­ therapeutic dilemma. As long as the dilemma veillance based primarily on the reporting of noti­ exists and the patient is frankly informed of his fiable diseases by physicians is not satisfactory or her risks and the alternatives, then there is because of the gross incompleteness of reporting little or no infringement of personal or profes­ and the limited nature of coverage of disease sional ethics in conducting such studies. Rather problems. The diagnostic data in hospital records the investigator should try to discover the answer should provide a particularly useful source of in- to such questions as soon as possible in order to formation for detection of sudden changes in mor­ provide all patients with the preferred therapy bidity. and to spare many a needless procedure. The key to this problem is to develop case- Currently a long-term, collaborative inter- finding methods which would provide the needed hospital drug trial is being started to test the data and at the same time overcome the problem efficacy of various medications on the treatment of time lag inherent in discharge data. In some and survival of patients with coronary heart dis - instances admission diagnoses could alert the ease. Is More than 8,400 patients in 50 clinics health official to the need for thorough investiga­ across the country will be involved. tion. Another possibility would be to assign there­ Certain coronary patients will receive one of sponsibility to an epidemiologic officer16’17 within four drugs or a placebo to evaluate their ability the hospital who could experiment with the most to prevent recurrence of myocardial infarction. effective means of finding cases of public health This prototype of a drug-trial study has unlimited concern. potential for determining the effects of drugs, and In the use of hospital records, care must be it represents a fertile field of mutual endeavor taken to avoid erroneous conclusions due to errors for epidemiologists, statisticians, and clinicians. or incompleteness in the observations recorded. Mainland and Sutcliffe 14’15 have provided In surveillance as in the hypothesis-testing use of magnificent examples of experimental studies of hospital records, the observer must not confound arthritis among hospital outpatients. In these differences of etiologic significance with results studies the Cooperating Clinics Committee of the arising from the fact that patients choose hospi­ American Rheumatism Association is engaged in tals or are referred to hospitals in nonrandom testing the efficacy of various medications on the fashion. Undue concern with this bias has, in re- course of rheumatoid arthritis. The effect of as­ cent years, resulted in the underutilization of hos -

5 pital records as a source of information of great teristics of the patient as age and sex. Therefore, potential value. interpretation of utilization patterns must also Problems are also to be expected in the es­ take into account diagnoses and associated disa­ timation of the incidence and prevalence of various bilities. Such diagnostic information is usually diseases. Estimates based on hospital admissions more accessible and reliable when obtained from will be less valid for chronic diseases than for hospital records than from other sources. The illnesses characterized by a single acute attack uses of these data are discussed below under that requires hospitalization and results in im­ four headings, each of which represents a ma­ munity of long duration. Furthermore, the use of jor factor influencing patterns of hospital uti­ in-hospital data will cause the loss of large num­ lization. bers of acute and minor diseases which are diag­

nosed and treated in physicians’ offices or in Influence of Diagnoses Per Se outpatient departments. It is recognized that the reporting of diseases For this purpose discharge data by diagnostic and preliminary y diagnoses by one hospital or by categories are useful. Data for a specific hospital a few hospitals in a large metropolitan area would may be of value to the hospital itself in evaluating be of limited value because patients choose or are its functions and in servicing the community. referred to hospitals in nonrandom fashion. Bias However, these data are usually of limited value introduced in this way could lead to differences for any one hospital in the community since the which would suggest spurious etiologic relation- population served by a given hospital is ordinarily ships. However, undue concern with such possi­ difficult to delineate. If all hospitals in the com­ bilities in recent years has led to hospital records munity are included, discharge data for residents being underutilized as a valuable source of in- of that community and the population estimates formation. These risks would be greatly reduced for the total community may be used to establish if all hospitals in a metropolitan area or other discharge rates. defined geographic area would join in a common Diagnostic data for the country as a whole computer service for broad administrative and are available from the Hospital Discharge Survey fiscal purposes; this would make possible the sys­ of the National Center for Health Statistics, which tematic collection of admission and discharge is a probability sample of U.S. hospitals .Is!lg For diagnoses that would have maximal value for smaller areas such as States and metropolitan epidemic surveillance. areas, corresponding data are not routinely tabu­ lated. There are, however, collection mechanisms MEDICAL-CARE RESEARCH in existence which do have the potential for pro­ The use of hospital data for purposes of viding such data for a sizable and definable popu­ medical-care research is comparable to its use lation subgroup. Among such mechanisms are the for epidemiologic purposes. However, rather than Blue Cross and/or commercial hospital insurance to identify the characteristics of patients with programs and certain group practice programs specific diseases, the goal in this case is to iden­ such as the Health Insurance Plan of Greater tif y the characteristics of patients who seek vari­ New York20”22 and the Kaiser- Permanence Pro- ous care and to determine the effect of organiza­ gram in California. 23 In addit ion, the Professional tional and professional characteristics of the Activity Study at Ann Arbor Michigan, insofar as it hospital on receipt of care. Such information sug­ serves hospitals in several areas, might provide gests hypotheses as to why certain patients do or similar information. Data potentially available do not receive care and why such care may be from social programs such as Medicare and Med­ costly. In this case the hospital also provides a icaid may also be used for this purpose. In Alle­ resource for testing hypotheses. gheny County, Pennsylvania, the Blue Cross has a Obtaining medical care is greatly influenced program which provides a continuous “patient reg­ by the disease process, including symptoms and ister” and r-elated statistical tabulations for each severity, as well as by such demographic charac­ hospital.

6 In many of these medical-care programs, the 2. Type of hospital from which patient was diagnostic data become available as part of a sys­ discharged tem designed to pay claims or for some other ad­ a. AHA accredited or nonaccredited ministrative reason. In others it will be necessary b. Teaching or nonteaching or university to code a sample of the records. In such cases it or nonuniversity is desirable that the diagnostic tabulations have c. Auspices (voluntary, proprietary, gov­ some meaning in terms of the organization’s own ernmental) purposes in order to justify the sizable expense d. Type of outpatient department organi ­ represented by such procedures. 24 zation (speciality clinics only, general medical or speciality clinic, and other) Influence of Organizational Structure for Various cross-classifications of these and Providing Medical Care and of Type of other variables may also be desired. Such classi­ Physician fications require that the necessary information be available on the discharge report and appro­ It has been shown that hospitalization rates priately coded. for members of group practice programs are lower than those for comparable groups serviced Influence of the Physician’s Characteristics by other programs or for groups seeking care in- dependently. 20-22 The differences, however, are The following classification of the physician’s not uniform for all diagnostic categories. For professional characteristics is suggested: example, the difference in hospitalization rates between individuals served by group and by non- 1. Physician specialty (general practitioner, group programs is greater for patients with ton­ internist, pediatrician) sillectomies and adenoidectomies than for those 2. Years since graduation with neoplasms. Such knowledge is of considerable 3. Type of staff appointment (none, courtesy, importance in planning a framework by which active, consultant) medical services are made available to the popu­ Again some degree of cross-classification lation. Moreover, since the population is given a may be desired. choice of several different kinds of medical-care In this connection, it would be helpful if each programs (as is the case in the Medicaid program physician in a community was assigned a code in New York City), it is of more than passing in­ number, as was done in Allegheny County, Penn­ terest to examine the utilization patterns by diag­ sylvania. The code number would identify his type nosis among the several plans. Differences in patterns after appropriate adjustments for popu­ of practice, specialty, board certification, age, and other factors. It would permit grouping of lation characteristics may yield clues as to dif­ data from various locations—either by physician ferences in medical practice which may, in turn, or by physician characteristics. It would also per­ suggest ways of examining the quality of medical mit comparisons of the behavior of the same phy­ care in different settings. sician in different settings. Discharge rates by diagnoses are required for the study of this problem, and they need to be Influence of Various Types of Case classified as follows: Management 1. Type of practice of physician discharging The concept of progressive patient care has the patient given rise to a variety of different methods of a. Solo case management. Since the diagnosis is clearly b. Partnership one of the determinants of case management, one c. Group practice may raise the question of whether subsequent ex­ d. Hospital-based perience of a patient with a particular diagnosis varies with different methods of case manage­ and the patient. In some States much of the infer. ment. Subsequent experience may be measuredin mation, particularly the diagnosis, obtained aml many ways, such as readmission rates during a recorded in the hospital records is of a privileged given period after initial discharge,length of time and confidential nature .25 Making such informa­ required to return to work, and subsequent use of tion available for epidemiologic purposes is con­ physician’s or nurse’s services. These data will sidered by some to be a violation of trust. Thus, not necessarily be available from a routine re- it is sometimes necessary to obtain the permis ­ porting system. It may be necessary to do a fol­ sion of every patient whose record is to be used. lowup on the patient and to obtain information by It is obvious that this can be a cumbersome and direct interview. These data may then be corre­ difficult process. lated with the presence or absence and use ofi In some instances, special legislation has been passed permitting the use of hospital data for 1. Formal intensive care unit in the hospital such purposes as medical audits. New York State 2. Home care program has recentIy enacted legislation which gives the 3. Nursing home affiliation by type of affili­ State health officer access to individual records. ation Edgar S. Dunn26 has stated that it is neces­ 4. Formal discharge planning program. sary to distinguish clearly between intelligence It will be essential in carrying out such stud­ data about individuals as individuals, such as med­ ies to have detailed knowledge of the criteria for ical records a doctor keeps to trace changes in the admission to a particular kind of case-manage­ wellbeing of his patients, and statistical data that ment program. A control or comparison group is relate to groups of individuals. Statistical data also needed. In general, the requirements for are concerned with aggregates, averages, per­ more definitive study here are similar to those centages, and so forth that describe relationships in the discussion of hypotheses testing under characteristic of groups of, for example, hospital Prospective Studies. cases. No personal information about the indi­ vidual needs to be available to anyone other than those engaged in the research. Respect for con­ CONFIDENTIALITY fidentiality of medical information has been char­ A problem which has been raised recently in acteristic of epidemiologic studies. The need for some of the epidemiologic studies utilizing hos­ the protection of an individual’s privacy is well pital data has been that of confidentiality and priv­ appreciated and understood by individual re- ileged communications. This issue involves the searchers, and specific steps are usually taken relationship between the hospital staff physician to provide the necessary protection. lMusi, A. T.: Potential uses and limitations of hospital 15Cooperating Clinics Committee of the American Rheu­ dutu in epidemiologic researoh. Am. J. Pub. HeaZth 55(5):658- matism Association: A seven-day variability study of 499 OW1 May 1965. patients with peripheral rheumatoid arthritis. Arthrctis Rheu­ mat. VIII(2) :302-335, Apr. 1965. ‘Lovin, hf. L., Goldstein, H., and Gerhardt, P. R.: Cancer 16 and tobacco emoking. J.A.M .A. 143(4):336-338, hiay 1950. Fuerst, H. T., and Lichtman, H. “S.: Directions in hospi­ ‘lBuck, A. A.: Epidemiologic investigations of sarcoidosis. tal epidemiology. New York J.Med. 66(10):1202-1205, hfay .bn.J.Hg:7. 74(2):137-202, Sept. 1961. 1966. 17 lhf,lqi A. T., and others: Hashimoto’s disease, a clinico­ Lichtman, H. S., and Stuart, G. M.: The Community Ori­ pntholo~i;til study with matched controls. Lancet 1 (7377 ):123- ented Hospital Epidemiologist. Paper prepared for presenta­ tion at the 93d annual meeting of the Epidemiology Section of lW, Jun. 1966. the American Public Health Association, Oct. 21, 1965, Chi­ ‘iBerkson, J.: Limitations of the application of four-fold cago, 111. table analysis to hospital data. Biometrics 2:47-53, 1946. 18Sirken M. G.: The hospital discharge survey. Pub. Hea2th ‘iWink@lstein, W., Jr., Stenchever, M., and Lilienfeld, Rep. 82(1):9-16, Jan. 1967. A. M.: Occurrence of pregriancy, abortion and artificial meno­ 19 pauso among women with coronary artery disease, a prelimi­ National Center for Health “Statistics: Patients dis­ nary study. J. C/won .DLs. 7(4):273-286, Apr. 1958. charged from short-stay hospitals, United States, October- December 1964, Vital and Health Statistics. PHS Pub. No. 7Grndy,. G. F., Chalmers, T. C., and the Boston Inter-Hos­ 1000J3eries 13-No. 1. Public Health ‘Service. Washington. pital Liver Group: Viral hepatitis in a group of Boston hospi­ U.S. Government Printing Office, Oct. 1966. tals-I. A retrospective study of 1675 patients. New EngZund ,J. Mcd, 272(1:3):657-661, Apr. 1965. 20Densen, P. hi., and others: Prepaid medical care and hospital utilization. American Hospital Association. Chicago 8@ndY, G. F., Chalmers! T. C., and the Boston Inter-Hos­ Hospital Monograph ‘Series No. 3, 1958. p. 55. pital Liver Group: viral hepatitis in a group of Boston hospi­ 21 tuls-H. A prospective controlled epidemiologic study. New Densen, P. M., Jones, E. W., Balamuth, E., and ‘Shapiro, Entiltmd J.Mcd. 27’2(1 3):662-666, Apr. 1965. S.: Prepaid medical care and hospital utilization in a dual 0? choice situation. Am. J. Pub. HeaZth 50(1 1):1710-1 726, Nov. hnongcm, E. L., Thoma, G. E., and Tompkins, E. A.: In­ 1960. rirlonce of leukemia following treatment of hyperthyroidism, 22 u ptwliminnry report of the Cooperative Thyrotoxicosis Ther­ Densen, P. M., Shapiro, “S., Jones, E. W’., and Baldinger, ups Follow-up ‘Study. J.A.M .A. 205(1 2):855-862, “Sept. 1968. 1.: Prepaid medical care and hospital utilization, Comparison l%hulmers, T. C., and others: The treatment of acute in­ of a group practice and a self-insurance situation. HospitaZe fm-tious heputitis, controlled studies of the effects of diet, 36(22):62, iiov. 1962. r,,st,, [lnd physical reconditioning on the acute course of the 23\veissman A .: A morbidity study of the Permanence disuaso and on the incidence of relapses and residual ab­ Health Plan population. Pe~manente Found.,41.BuU. X(1-4):12- normuliticw. ,J.Clin.invest. 34(7):1163-1235, July 1955. Part II. 26, Aug. 1952. 11 24 Chulmers, T. C., and Nefzinger, M. D.: The treatment of Densen, P. M., and Shapiro, “S.: Methodological probleme acwto infections hepatitis, 10-year followup study. Gastro­ in the study of special population groups, health insurance rnfmdogy 42:479, 1962. (abstract) and industrial groups. Ann. New York Acad.Sc. 107(2):490- 12Gnrcewr, A. J., the Boston Inter-Hospital Liver Group, 505, May 1’963. 2~ayt. E., and Hayt} and others: A controlled trial of prophylactic portacaval-shunt J.: Legal Aspects of Medical Rec­ surgery. Now EngZand J.Med. 270(10):496-500, Mar. 1964. ords. Berwyn, 111. Physicians’ Record Company, 1964. pp. 74, 1:3National Heart Institute. The Co Tonary Drug Project. 75, 93. PHS Pub. No. 169+5. National Institutes of Health. Washing 26U .’S. Congress, House Committee on Government Opera­ ton. LI.s. Government Printing Office, 1967. tions: The computer and invasion of privacy. Hearings before 14 a subcommittee of the Committee on Government Operations, Mainland, D., and Sutcliffe, M. I.: Aepirin in rheuma­ House of Representatives, 89th Congress, July 26-28, 1966. toid arthritis, n seven-day double-blind trial-preliminary re- Washington. U.S. Government Printing Office, 1966. p. 92. port. J3u11.EJteumat. Dis. 16:388-391, NOV. 1965.

000

~ “, .s. GOVERNMENT PRINTING oFFIC= 19 S9—3~2049/!54

9 OUTLINE OF REPORT SERIES FOR YITAL AND HEALTH STATISTICS

Public Health Service Publication No. 1000

SeTies 1, P~o,gvams and collection pvoceduves.— 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.

Sevies 2. Data evaluation and methods ~esea?’ch. —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.

Sevies 3. 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.

Series 4. Documents and committee Yeports. — Final reports of major committees concerned with vital and health statistics. and documents such as recommended model vital registration laws and revised hi rth and death certificates.

Sevies 10. Data fYom t}ze Health Inte%,zjzew SuYveY. —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 Sur~e~. — Data from direct examination, testing, and measure­ ment of national samples of the 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 psychological characteristics; and (2) .malysis of relationships among the various measurements without reference to an explicit finite universe of persons.

Sevies 12. Data fvom the Institutional Population %~veys. — Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, ~nd personal care received, based on national samples of establishments providing these services and samples of the residents or patients.

SeVies 13. Data fvorn the Hospital Dischavge Suwey. —Statistids relating to discharged patients in short~stay hospitals, based on a sample of patient records in a national sanlPle of hospitals.

Series 14. Data on health ~esouvces: rnanpowev and facilities. —Statistics on ihe numkrs, geographic” distri­ bution, and characteristics of health reso~rces including physicians, dentists, nurses, other health manpower occupations, hospitals, nursing homes, and outpatient and other inpatient facilities.

Series 20. Data on mortality. -Various statistics on mortality other than as included in annual or monthly reports —special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses.

SeYies 21. Data on natality, rnarviage, and diuo?’ce. —Various statistics on natality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility.

Series 22. Data fro”rn the National Natality and Mortality Su)’ueys. —Statistics cm 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, medical experience in the last year of life, characteristics of pregnancy, <>tc.

For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201