International Journal of Epidemiology © Oxford University Press 1978
Vol. 7, No. 3 Printed in Great Britain
Uniform Basic Data Sets for Health Statistical Systems JANE HMURNAGHAN1
INTRODUCTION
A prime requirement of information systems for epidemiological surveillance and health planning is that the data be not only comprehensive but comparable, so that meaning can be derived from individual observations by grouping them and making comparisons in time, place, and among different sectors of the population. This paper discusses the development of uniform basic data sets in the United States to coordinate its health statistical systems and meet this requirement. Although there are features of the method that are typically American, the larger issues it addresses are the generics one of comparability and communication. DEFINITION AND DESCRIPTION
As the concept has evolved in the United States, a so-called uniform basic data set is a set of minimum specifications for the content of information
Based on a paper presented at the 8th Internationa] Scientific Meeting of the International Epidemiologies! Association, Puerto Rico. September 1977 1 Assistant Professor, Division of Health Care Organization, School of Hygiene and Public Health, Johns Hopkins University, 615 N Wolfe Street, Baltimore, Maryland 21205, USA; currently on part-time assignment to the U S National Center for Health Statistics
systems. It serves two purposes: first, it defines the central core of data needed on a routine basis by the majority of decision-makers about a given facet or dimension of the health services system; and second, it establishes standard measurements, definitions, and classifications for this core. Unlike a taxonomy or classification system — for example, the International Classification of Diseases — a basic data set is selective, not definitive. In practice, most information systems have special requirements and collect additional data elements or detail; a basic data set covers only the data elements that have universal application. Two steps are involved in their design: definition of scope and content; and choice of terms, definitions, classifications, and units of measurement. Of the two, the first step is in many ways the most challenging, for it means dividing up the universe of people, places, and events into manageable segments relevant to public policy, and it requires making explicit and reaching consensus among many categories of users on the underlying goals and objectives of public policy that determine what to include and what to leave out. Thus the initial questions in the design of a data set are what to measure and why. Anyone familiar with data systems will appreciate how difficult it is to elicit from potential users the answers to these questions. So far basic data sets have been developed in the United States for short-stay hospital inpatient 263
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Murnaghan J H [School of Hygiene 8nd Public Health, Johnj Hopkini University, 615 N Wolfe Street, Baltimore, Maryland 21205, U S A ] . Uniform basic data sets for health statistical systems. International Journal of Epidemiology 1978, 7: 2 6 3 - 2 6 9 . The United States approach to coordinating health statistics involves introduction of multipurpose basic data sets describing health status and the health care system. Standard reporting procedures have been used for many years for vital statistics. Recently designated data sets cover health manpower, inpatient facilities, short-stay hospital discharges, and use of ambulatory care, services. A data set for long-term health care is in the design stage. Advantages of this approach in the United States and internationally are: basic comparisons can be made between health care settings and geographic areas while maintaining the variety and flexibility of existing public and private information systems; shared local, regional, and national data systems can be set up; and better coordination can be achieved between government-sponsored general-purpose and administrative data systems. Problem areas are: avoiding undue proliferation, e.g. of disease-specific data sets; adhering to the principle of minimal requirements; linking data sets and coordinating them with census and other social indicators; promoting widespread use; assuring data quality; establishing mechanisms for review and revision; and extending the concept internationally.
264
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
IMPORTANCE OF BASIC DATA SETS The advantages of national statistical guidelines are self-evident in countries like the United States where the health care system is an exceedingly complex mixture of public and private, national, regional, and local organizations, each supported by its own network of information systems. These data sets are the essential means of communication among the many autonomous and semi-autonomous parts of the system. By this means, basic comparisons can be made between health care settings and geographic areas, and regional and national statistics can be compiled, while maintaining the variety and flexibility of existing information systems. But the utility of basic data sets is by no means limited to mixed public and private systems like that of the United States. The same problem of communication exists at the international level, of course, and has long been a concern of the World Health Organization. The First International Conference of National Committees on Vital and Health Statistics in 195 3 set forth as one of its eight objectives: To help to achieve essential uniformity in records, methods, and tabulations for the production of the minimum core of vital and health statistics needed for national and international purposes (8). This mission was reaffirmed and broadened to include statistics about the organization and activities of health services at the Second International Conference in 1973 (8). More recently WHO groups, together with the European Economic Community, and the Organization for Economic Cooperation and Development, have specifically recommended that uniform data sets be defined for the components of health information systems (9).
In 1975, EEC representatives held a conference to discuss collaboration in conducting national health surveys and in compiling routine statistics on hospital discharges (10). Also in this connection a preliminary 'Glossary of Health Care Terminology' has been prepared under the auspices of the WHO Regional Office for Europe (11). Indeed, developmental work in the United States on basic data sets may well be viewed as an extension of international policy and a contribution to the evolution of international statistical systems. A second point of general interest is that standardization is essential to the construction of regionalized data systems within a country, whereby data are shared by national, district, and community agencies in order to expand the flow of information between levels of organization and reduce the burden of collection and analysis. In the United States, a major policy objective of the Department of Health, Education, and Welfare is the development of the federal—state—local Cooperative Health Statistics System (12). The foundation of this system already exists in the form of cooperative arrangements for collecting and analysing vital statistics. A third point is that regardless of how health services are organized, most governments have difficulty coordinating their general-purpose or base-line statistical systems with their administrative or programme-specific data systems. Large national programmes like health insurance, family planning, or communicable disease control require management information systems. Such systems generate considerable data of more general concern. Without some mechanism for coordination, there may well be loss of essential information or unnecessary duplication. Also management information systems are prone to using their own terminology and measures of performance, making evaluation in relation to external factors like population-based measures of need and satisfaction impossible. In practice, intergovernmental coordination is emerging as one of the most important justifications for basic data sets in the United States. For all these reasons, the general outlines, if not the finer details, of the work on uniform basic data sets in the United States has significance beyond its borders. DEVELOPMENT OF BASIC DATA SETS IN THE UNITED STATES The concept was first explored at a meeting of users and producers of short-stay hospital inpatient statistics in 1969, sponsored by the Department of Health, Education, and Welfare and the Johns Hopkins University (1). Hospital services are the
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
services (1, 2), ambulatory or outpatient services (3, 4), health manpower, and health care facilities or institutions (5). A data set covering institutional and non-institutional long-term health care is at the final planning stage (6). To this list should be added the well-established procedures for compiling mortality and other vital statistics which employ the equivalent of basic data sets although the term is not normally used in this connection (7). Primary responsibility for this work rests with the US National Committee on Vital and Health Statistics, the official advisory body to the Department of Health, Education, and Welfare on health statistical policy. It should be noted, however, that the whole process of designing and gaining acceptance for basic data sets depends upon active collaboration of the public and private sector.
DATA SETS FOR HEALTH STATISTICS
TABLE 1 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
Data elements of the United States uniform hospital discbarge data set
Patient identification Date of birth Sex Race Residence Hospital identification Date and hour of admission Date of discharge Attending physician identification Operating physician identification Principal and other diagnoses Principal and other procedures Disposition of patient Expected principal source of payment
now to rapid development of information systems to provide these data for all areas of the country is a jurisdiction^ tug-of-war within the Department of Health, Education, and Welfare over procedures for data collection. Incidentally, the National Committee's guidelines have recently been translated into Spanish by the Pan American Health Organization (13). Although space does not permit discussion of the design problems, definitions, and details of the Uniform Hospital Discharge Data Set and its 14 data elements are shown in Table 1. Patient descriptors arc limited to a unique identification number, age, sex, race, and residence. Admission and discharge dates are recorded, and the hospital and responsible physicians identified by a unique number. Clinical data consist of a listing of principal and other diagnoses and procedures like operations, to be coded in a manner compatible with the ICD. (There are two adaptations in common use in the US, but it is hoped that there will be only one for the 9th Revision.) Principal diagnosis illustrates the importance of precise definition for there is little agreement in practice on what is meant by this term. After considerable debate, principal diagnosis was defined as 'the condition established after study to be chiefly responsible for occasioning the admission of the patient to the hospital for care'. However, the 9th Revision shifts the meaning significantly to refer to the diagnosis which accounts for the largest portion of hospital resources consumed, thereby raising problems of international comparability (14). Disposition of the patient is divided into five categories: discharged to home, left against medical advice, discharged to another health facility, discharged to an organized home care service, and died. The final item, expected principal source of payment, has eight categories covering the varied sources of public and private funds used in the United States to pay for hospital inpatient care. The initial reaction of many medical staffs and epidemiologists is that the data arc too limited to support their work. But it must be remembered that other users like hospital administrators and planning agencies need summary information on all patients and cannot make good use of detailed clinical data, at least on a routine basis. The principal value of this data set for researchers and medical staffs is to help them identify questions that need answering in special studies and surveys and provide them with an index to medical records. On closer inspection it can be seen that even this modest amount of data, if available from all hospitals in an area, is capable of describing and comparing major reasons for use of hospitals, diagnostic-specific lengths of stay, case-fatality rates, hospital catchment areas, sources
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
most expensive component of the United States health services system, currently accounting for 40 per cent of all health expenditures. Public concern about the efficiency, quality, and accessibility of hospital care had already greatly increased the demand for hospital statistics. As a result, many new reporting and data processing systems were being set up. The conference considered it essential at this formative stage to introduce some measure of coordination. The solution proposed was the Uniform Hospital Discharge Data Set describing short-stay hospital inpatients and their episode of care, to be incorporated in all government and voluntary reporting systems and surveys of hospital activity. In considering the content of this data set, the conference set forth some criteria, which have been used since in the design of other data sets: 1 The data items included in the set must be useful to most if not all potential users. In the case of hospital statistics, the major groups considered were hospital medical staffs; hospital administrators; local, state and federal planning, regulatory, and legislative bodies; private and public hospital insurance agencies; and the research community. 2 The items selected must be ones that can be readily collected with reasonable accuracy. 3 Items should not duplicate data available from other sources. 4 Confidentiality of medical information should be protected. The recommendations of the 1969 conference were reviewed and refined by the US National Committee on Vital and Health Statistics and feasibility studies were made. In 1972, the National Committee issued an official version (2). It has been widely endorsed and accepted; the biggest obstacle
265
266
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 2
Development of uniform basic data sets for health information systems in the United States
Component
Areas covered by existing data sets
Potential areas for development
Vital statistics
Morbidity Disability Health problems, attitudes, and practices
Health services resources
Manpower Inpatient facilities
Ambulatory care facilities Supportive services and equipment ? Skills and knowledge
Use of health
Hospital discharges Ambulatory care visits Long-term care
? Mental health services ? Emergency medical services ? Other categorical services
Health expenditures
Sources and uses of funds Uniform cost accounting
Environmental health
Environmental health hazards Occupational health hazards
of funds, hospital utilization rates by population subgroups, and other basic facts upon which hospital management, planning, and evaluation depend. Table 2 shows the progress that has been made so far on uniform basic data sets in the United States and gives some idea of the uncharted terrain. In the area of health status indicators, standard measures of mortality should be supplemented by standard population-based measures of morbidity and disability. Using White's model of the five D's — death, disease, disability, dysfunction, and dissatisfaction (15) — serious considerations should also be given to expanding the concept of health status to include measures of the health problems, symptoms, pain and anxiety that people experience, their attitudes towards health services, and their health beliefs and practices. Since 1957 the National Health Interview Survey has provided standard estimates of morbidity and disability and related use of services, as reported in a continuing sample survey of households (16). A second national survey, the Health and Nutrition Examination Survey, provides professionally determined measurements of health status (17). But the sample sizes of these surveys are not large enough to provide direct estimates for community planners and policymakers, and few state and local agencies have the resources and skills to conduct surveys of such complexity. A more selective set of indicators and less expensive methods of data acquisition are
needed for producing small-area statistics on health levels and health needs. Data sets have been designed and are being used to collect statistics on health manpower and inpatient facilities, although both are presently being extensively overhauled. The constantly changing patterns of services and personnel in the United States make it difficult to settle upon uniform groupings or classifications for health personnel and institutions. As the table shows, there is much room for expansion of these data sets to cover the entire spectrum of tangible and intangible resources. The third broad area of health services utilization also poses a problem of categorization. At present working groups have divided patients and services into three classes: short-stay hospital inpatient, ambulatory, and long-term care, but these distinctions are not entirely logical and may require modification. Long-term health care is particularly difficult to delineate. As currently defined, long-term care includes patients in chronic disease, psychiatric, and custodial institutions, or units of hospitals, as well as those living at home and receiving community-based services. The principal reason for treating long-term care separately is that social, economic and behavioural causes and manifestations of illness assume greater importance in the case of the chronically ill and disabled population than in acute and episodic care. To complicate the picture, special data sets are being
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Health status
DATA SETS FOR HEALTH STATISTICS
PROBLEMS IN THE DEVELOPMENT OF DATA SETS On the basis of United States experience, there are a number of problems in design and implementation that must be solved in order to make the method of uniform basic data sets effective. Defining the scope of a data set and avoiding proliferation. As already suggested, it has been difficult to decide upon the optimal number and scope of data sets in the United States, and this will undoubtedly be a major problem in cross-national statistical guidelines because of wide variations among countries in the organization of health services. In the case of the United States, the
difficulty stems from present uncertainty about the direction of health policy; categorical or disease-specific programmes continue to dominate the flow of public and private funds, while some of the more recent legislation seeks to promote comprehensive planning . and organization of services. The idea of data sets has become quite popular in the last year or so. Proliferation of specialized data sets threatens to dilute the concept of broad, multipurpose guidelines for health statistical systems that permit one to sec the whole as well as the parts and to think flexibly about the health care system as one plans for the future. The process of designing data sets has been evolutionary. The time has come to construct a master plan or conceptual framework for their orderly development, a problem that the US National Committee on Vital and Health Statistics is now beginning to address. Adhering to the principle of minimal or basic requirements. With so many parties involved, and computers standing by to run seemingly endless series of cross-tabulations, there is constant pressure on the designers of data sets to expand them beyond the limits of what is feasible for busy health professionals to provide on a routine basis, and most data recipients to digest and put to good use. The long-term care data set is a good example. In principle, public policy in the United States is directed to a much broader definition of the health care for the chronically ill and disabled population than the health services system is presently funded and organized to provide. The draft data set reflects this philosophy by including measures of dysfunction and social and economic dependency as well as disease, by encompassing mental as well as physical illness, and by covering the entire spectrum of curative, rehabilitative, and supportive services, both inpatient and outpatient. As a result, it started out with some 24 major elements compared to 14 in the Uniform Hospital Discharge Data Set, and may well end up being far from minimal. Coordination and linkage. As work progresses on basic data sets in the United States, problems of internal and external consistency and compatibility are intensified. In the interest of efficiency, the data sets should complement not duplicate each other. More important, it should be possible to link them to build profiles of the health needs, resources, and services of geographic areas and population subgroups. To do so, the data sets must use the same concepts, terms, classifications, and identifiers. In turn, these must correspond with the nation-wide census, the universal denominator for health
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
proposed for mental health and emergency services. In the case of mental health, there is a well-defined service sector with its own source of federal funds. On the other hand, many health professionals consider the distinction between physical and mental illness unsound in theory and practice, and many patients with mental disorders are in fact being treated in the offices of general physicians and in general hospitals and nursing homes. Emergency services are also provided at many sites, and the emergency room in the typical American hospital, although specially staffed and equipped, does double duty as a general, walk-in clinic for the people in its neighbourhood, especially after working hours. Thus the question of how many data sets and subsets are needed to describe patients and their use of services is by no means resolved. The final two components of a comprehensive information system as set forth in Table 2 have yet to be tackled systematically from the viewpoint of establishing uniform guidelines. Expenditures, it should be noted, may also be viewed as resources although treated separately here. Preliminary analysis suggests that the formula used for many years by the Federal Government to provide annual statistics on total expenditures and sources and major uses of health funds may be suitable, with some adaptation, for constructing state and local profiles of the flow of funds (18, 19). There is considerable interest in the related problem of establishing standard cost accounting procedures. The natural and man-made hazards to health in the working place, home, and general environment are now recognized to be so many and varied that they would appear to defy description in a minimum data set. But 1974 planning legislation in the United States includes environmental health indicators in the statistical base for national, state, and local planning agencies, so a start must be made on this front soon.
267
268
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
statistics, and if possible, with other social indicators that give wider meaning to health statistics. In addition, there are certain problems common to all the data sets that require common resolution, notably unique numbering systems for individuals and organizations, geographic coding of residence and location, and procedures for protecting privacy and confidentiality. For all these reasons, the oversight of a single body like the US National Committee is essential.
Assuring data quality. A recent study by the National Academy of Sciences of the reliability of data abstracted from hospital medical records (20) illustrates how much time and effort must be devoted to evaluating and improving the quality of the basic data being collected. Comparison of a sample of original abstracts prepared by hospital personnel with those of an independent team of coders employed in the study showed a high degree of agreement in the areas of patient age, sex, dates of admission and discharge, and expected principal source of payment. But in the key areas of principal diagnosis and principal procedure, agreement in the 4-digit codes assigned was only 65 per cent and 73 per cent respectively. Often the medical record itself was not clear in regard to the patient's principal and other diagnoses and procedures. Since hospitals have had considerable experience preparing diagnostic indexes and they are widely used, it is clear that much more work must be done on validity and reliability and caution exercised in the introduction of new and unfamiliar data items and concepts. Establishing mechanisms for review and revision. Changes in health policy in the United States are constantly shifting priorities for information. Furthermore, some of the data elements and
Extending the concept internationally. From the start United States committees have recognized the obvious benefits of coordinating their efforts with those of other countries, and it should be added that they have relied heavily on experience and practices in other countries, especially the United Kingdom and Canada, to help design the hospital, ambulatory, and long-term care data sets. It is encouraging, therefore, that there is growing interest on the part of international groups not only in common measures and classifications of diseases and health problems but also in common approaches to describing the ways in which health resources arc allocated, organized, and used to address these problems. If definitional differences can be resolved, a uniform core of international data linking measures of the population's health problems with measures of the distribution of health resources, use of services, and related expenditures offers exciting new possibilities for cross-national comparisons. Work on uniform basic data sets in the United States, it is hoped, will contribute to this important international movement. REFERENCES (1) Hospital Discharge Data: Report of the Conference on Hospital Discharge Abstract Systems, Airlic House, June, 1969. Eds Mumaghan J H and White K L. Lippincott, Philadelphia. 1970. (Reprinted from Medical Care 8(4): Supplement, 1970.) (2) Uniform Hospital Abstract: Minimum Basic Data Set. US Department of Health, Education, and Welfare, US National Committee on Vital and Health Statistic!. DHEW Publication No (HRA) 7 5 - 1 4 5 1 . Government Printing Office, Washington. 1972. (3) Ambulatory Care Data: Report of the Conference on Ambulatory Care Records, Chicago, April, 1972. Ed Mumaghan J H. Lippincott, Philadelphia. 1973. (Reprinted from Medical Care 11(2): Supplement, 1973.) (4) Ambulatory Medical Care Records: Uniform Minimum Basic Data Set. US Department of Health, Education, and Welfare, US National Committee on Vital and Health Statistics. DHEW Publication No (HRA) 7 5 - 1 4 5 3 . Government Printing Office,' Washington. 1974. (5) The Health Manpower Inventory and Health Facilities
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Promoting widespread use. It is not enough, of course, to promulgate a series of well-tested guidelines; development of statistical standards must go hand-in-hand with development of data systems like the Cooperative Health Statistics System in the United States. Aside from the political problems .of establishing shared data systems, American experience has shown that promoting uniformity extends to the way in which data are collected and analysed. Attention must be devoted to common procedures as well as common terminology. Also experience shows that potential recipients like planners and administrators require much direct guidance and technical assistance to make effective use of even the small amount of data contained in the basic data sets discussed here.
concepts are necessarily new and experimental. Only widespread use can demonstrate fully the strengths and weaknesses of a basic data set. To assure that these statistical guidelines are dynamic — that they do not suffer from hardening of the bureaucratic arteries — there must be a regular schedule for review and revision. Most of the data sets discussed here are already being re-examined by the US National Committee, but with more experience, a longer cycle may be feasible as in the case of data sets for vital events.
DATA SETS FOR HEALTH STATISTICS
(13)
(14)
(15) (16)
(17)
(18) (19)
(20)
DHEW Publication No(HRA) 77-1456. Government Printing Office, Washington. 1977. El Resumen Uniforme de Egreso de Hospitalcs: Un Conjunto Minimo dc Datos Basicos. Pan American Health Organization. PAHO Document HS/Seric RM/No 16/1975. Washington, 1975. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, 1975 Revision, Vol 1, pp 737-74O. World Health Organization, Geneva, 1977. White K L. Epidemiologic intelligence requirements for planning personal health services. Ada socio-medica Scandinavia 2: 143, 1972. Health Interview Survey Procedures, 1957-1974. US Department of Health, Education, and Welfare, National Center for Health Statistics. DHEW Publication No (HRA) 7 5 - 1 3 1 1 . Government Printing Office, Washington. 1975. Plan and Operation of the Health and Nutrition Examination Survey, United States — 1971 — 1973. US Department of Health, Education, and Welfare, National Center for Health Statistics. DHEW Publication No (HRA) 76—1310. Government Printing Office, Washington. 1976. Gibson R M and Mueller M S. National health expenditures, fiscal year 1976. Social Security Bulletin 40(4): 3, 1977. Final Report on the Development of the Community Funds Flow Data System. Wirnowski J P, Project Manager, Arthur Young & Co, 1025 Connecticut Avenue, Washington, D C 20036, 1975. Reliability of Hospital Discharge Abstracts. National Academy of Sciences, Institute of Medicine. The Institute, Washington. 1977.
(revised version received 6 March 1978)
Infectious Disease Newsletter This is intended as an informal method of communication between members of the International Epidemiological Association interested in the field of infectious diseases. Portions may be published in the International Journal of Epidemiology. The newsletter is not intended to compete with official surveillance reports such as the WHO Weekly Epidemiological Record or the U S Morbidity and Mortality Weekly Reports, and other similar sources of formal data. Rather it is to
supplement these with informal reports, comments, news of meetings etc, so that a closer association can be formed of members of IEA interested in this area. If you would like to contribute to the newsletter and/or be placed on the mailing list write to Dr AS Evans, Professor of Epidemiology, Yale University School of Medicine, 333 Cedar Street, New Haven, Ct 06510, USA. This is a pilot project so your support, comments and suggestions are welcome both as to substance and to format.
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Minimum Data Set are being revised by the US National Committee on Vital and Health Statistics. Mimeographed copies of 1974 versions are available from the Division of Health Manpower and Facilities Statistics, National Center for Health Statistics, Hyattsville, Maryland 20782. (6) Long-Term Care Data: Report of the Conference on Long-Term Health Care Data, Tucson, May, 1975. Ed Mumaghan J H. Lippincott, Philadelphia. 1976. (Reprinted from Medical Care 14(5): Supplement, 1976.) (7) The 1968 Revision of the Standard Certificates of Live Births, Death, Fetal Death, Marriage and Divorce or Annulment. US Department of Health, Education, and Welfare, National Center for Health Statistics. Public Health Service Publication No 1000, Series 4, No 8. Government Printing Office, Washington. 1968. (8) New Approaches in Health Statistics. World Health Organization. Technical Report Series 559. Geneva, v 1974. (9) Some WHO Activities in the Field of Vital and Health Statistics and Related Fields During 1975. WHO Document/HS/Nat Com/76.352. World Health Organization, Geneva, 1976. (10) Armitage P. National health survey systems in the European Community. International Journal of Epidemiology 5t 321, 1976. (11) Hogarth J. Glossary of Health Care Terminology. World Health Organization Regional Office for Europe, Copenhagen. 1975. (12) The Cooperative Health Statistics System: Jts Mission and Program. US Department of Health, Education, and Welfare, National Center for Health Statistics.
269
Vol. 7, No. 3 Printed in Great Britain
Uniform Basic Data Sets for Health Statistical Systems JANE HMURNAGHAN1
INTRODUCTION
A prime requirement of information systems for epidemiological surveillance and health planning is that the data be not only comprehensive but comparable, so that meaning can be derived from individual observations by grouping them and making comparisons in time, place, and among different sectors of the population. This paper discusses the development of uniform basic data sets in the United States to coordinate its health statistical systems and meet this requirement. Although there are features of the method that are typically American, the larger issues it addresses are the generics one of comparability and communication. DEFINITION AND DESCRIPTION
As the concept has evolved in the United States, a so-called uniform basic data set is a set of minimum specifications for the content of information
Based on a paper presented at the 8th Internationa] Scientific Meeting of the International Epidemiologies! Association, Puerto Rico. September 1977 1 Assistant Professor, Division of Health Care Organization, School of Hygiene and Public Health, Johns Hopkins University, 615 N Wolfe Street, Baltimore, Maryland 21205, USA; currently on part-time assignment to the U S National Center for Health Statistics
systems. It serves two purposes: first, it defines the central core of data needed on a routine basis by the majority of decision-makers about a given facet or dimension of the health services system; and second, it establishes standard measurements, definitions, and classifications for this core. Unlike a taxonomy or classification system — for example, the International Classification of Diseases — a basic data set is selective, not definitive. In practice, most information systems have special requirements and collect additional data elements or detail; a basic data set covers only the data elements that have universal application. Two steps are involved in their design: definition of scope and content; and choice of terms, definitions, classifications, and units of measurement. Of the two, the first step is in many ways the most challenging, for it means dividing up the universe of people, places, and events into manageable segments relevant to public policy, and it requires making explicit and reaching consensus among many categories of users on the underlying goals and objectives of public policy that determine what to include and what to leave out. Thus the initial questions in the design of a data set are what to measure and why. Anyone familiar with data systems will appreciate how difficult it is to elicit from potential users the answers to these questions. So far basic data sets have been developed in the United States for short-stay hospital inpatient 263
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Murnaghan J H [School of Hygiene 8nd Public Health, Johnj Hopkini University, 615 N Wolfe Street, Baltimore, Maryland 21205, U S A ] . Uniform basic data sets for health statistical systems. International Journal of Epidemiology 1978, 7: 2 6 3 - 2 6 9 . The United States approach to coordinating health statistics involves introduction of multipurpose basic data sets describing health status and the health care system. Standard reporting procedures have been used for many years for vital statistics. Recently designated data sets cover health manpower, inpatient facilities, short-stay hospital discharges, and use of ambulatory care, services. A data set for long-term health care is in the design stage. Advantages of this approach in the United States and internationally are: basic comparisons can be made between health care settings and geographic areas while maintaining the variety and flexibility of existing public and private information systems; shared local, regional, and national data systems can be set up; and better coordination can be achieved between government-sponsored general-purpose and administrative data systems. Problem areas are: avoiding undue proliferation, e.g. of disease-specific data sets; adhering to the principle of minimal requirements; linking data sets and coordinating them with census and other social indicators; promoting widespread use; assuring data quality; establishing mechanisms for review and revision; and extending the concept internationally.
264
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
IMPORTANCE OF BASIC DATA SETS The advantages of national statistical guidelines are self-evident in countries like the United States where the health care system is an exceedingly complex mixture of public and private, national, regional, and local organizations, each supported by its own network of information systems. These data sets are the essential means of communication among the many autonomous and semi-autonomous parts of the system. By this means, basic comparisons can be made between health care settings and geographic areas, and regional and national statistics can be compiled, while maintaining the variety and flexibility of existing information systems. But the utility of basic data sets is by no means limited to mixed public and private systems like that of the United States. The same problem of communication exists at the international level, of course, and has long been a concern of the World Health Organization. The First International Conference of National Committees on Vital and Health Statistics in 195 3 set forth as one of its eight objectives: To help to achieve essential uniformity in records, methods, and tabulations for the production of the minimum core of vital and health statistics needed for national and international purposes (8). This mission was reaffirmed and broadened to include statistics about the organization and activities of health services at the Second International Conference in 1973 (8). More recently WHO groups, together with the European Economic Community, and the Organization for Economic Cooperation and Development, have specifically recommended that uniform data sets be defined for the components of health information systems (9).
In 1975, EEC representatives held a conference to discuss collaboration in conducting national health surveys and in compiling routine statistics on hospital discharges (10). Also in this connection a preliminary 'Glossary of Health Care Terminology' has been prepared under the auspices of the WHO Regional Office for Europe (11). Indeed, developmental work in the United States on basic data sets may well be viewed as an extension of international policy and a contribution to the evolution of international statistical systems. A second point of general interest is that standardization is essential to the construction of regionalized data systems within a country, whereby data are shared by national, district, and community agencies in order to expand the flow of information between levels of organization and reduce the burden of collection and analysis. In the United States, a major policy objective of the Department of Health, Education, and Welfare is the development of the federal—state—local Cooperative Health Statistics System (12). The foundation of this system already exists in the form of cooperative arrangements for collecting and analysing vital statistics. A third point is that regardless of how health services are organized, most governments have difficulty coordinating their general-purpose or base-line statistical systems with their administrative or programme-specific data systems. Large national programmes like health insurance, family planning, or communicable disease control require management information systems. Such systems generate considerable data of more general concern. Without some mechanism for coordination, there may well be loss of essential information or unnecessary duplication. Also management information systems are prone to using their own terminology and measures of performance, making evaluation in relation to external factors like population-based measures of need and satisfaction impossible. In practice, intergovernmental coordination is emerging as one of the most important justifications for basic data sets in the United States. For all these reasons, the general outlines, if not the finer details, of the work on uniform basic data sets in the United States has significance beyond its borders. DEVELOPMENT OF BASIC DATA SETS IN THE UNITED STATES The concept was first explored at a meeting of users and producers of short-stay hospital inpatient statistics in 1969, sponsored by the Department of Health, Education, and Welfare and the Johns Hopkins University (1). Hospital services are the
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
services (1, 2), ambulatory or outpatient services (3, 4), health manpower, and health care facilities or institutions (5). A data set covering institutional and non-institutional long-term health care is at the final planning stage (6). To this list should be added the well-established procedures for compiling mortality and other vital statistics which employ the equivalent of basic data sets although the term is not normally used in this connection (7). Primary responsibility for this work rests with the US National Committee on Vital and Health Statistics, the official advisory body to the Department of Health, Education, and Welfare on health statistical policy. It should be noted, however, that the whole process of designing and gaining acceptance for basic data sets depends upon active collaboration of the public and private sector.
DATA SETS FOR HEALTH STATISTICS
TABLE 1 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
Data elements of the United States uniform hospital discbarge data set
Patient identification Date of birth Sex Race Residence Hospital identification Date and hour of admission Date of discharge Attending physician identification Operating physician identification Principal and other diagnoses Principal and other procedures Disposition of patient Expected principal source of payment
now to rapid development of information systems to provide these data for all areas of the country is a jurisdiction^ tug-of-war within the Department of Health, Education, and Welfare over procedures for data collection. Incidentally, the National Committee's guidelines have recently been translated into Spanish by the Pan American Health Organization (13). Although space does not permit discussion of the design problems, definitions, and details of the Uniform Hospital Discharge Data Set and its 14 data elements are shown in Table 1. Patient descriptors arc limited to a unique identification number, age, sex, race, and residence. Admission and discharge dates are recorded, and the hospital and responsible physicians identified by a unique number. Clinical data consist of a listing of principal and other diagnoses and procedures like operations, to be coded in a manner compatible with the ICD. (There are two adaptations in common use in the US, but it is hoped that there will be only one for the 9th Revision.) Principal diagnosis illustrates the importance of precise definition for there is little agreement in practice on what is meant by this term. After considerable debate, principal diagnosis was defined as 'the condition established after study to be chiefly responsible for occasioning the admission of the patient to the hospital for care'. However, the 9th Revision shifts the meaning significantly to refer to the diagnosis which accounts for the largest portion of hospital resources consumed, thereby raising problems of international comparability (14). Disposition of the patient is divided into five categories: discharged to home, left against medical advice, discharged to another health facility, discharged to an organized home care service, and died. The final item, expected principal source of payment, has eight categories covering the varied sources of public and private funds used in the United States to pay for hospital inpatient care. The initial reaction of many medical staffs and epidemiologists is that the data arc too limited to support their work. But it must be remembered that other users like hospital administrators and planning agencies need summary information on all patients and cannot make good use of detailed clinical data, at least on a routine basis. The principal value of this data set for researchers and medical staffs is to help them identify questions that need answering in special studies and surveys and provide them with an index to medical records. On closer inspection it can be seen that even this modest amount of data, if available from all hospitals in an area, is capable of describing and comparing major reasons for use of hospitals, diagnostic-specific lengths of stay, case-fatality rates, hospital catchment areas, sources
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
most expensive component of the United States health services system, currently accounting for 40 per cent of all health expenditures. Public concern about the efficiency, quality, and accessibility of hospital care had already greatly increased the demand for hospital statistics. As a result, many new reporting and data processing systems were being set up. The conference considered it essential at this formative stage to introduce some measure of coordination. The solution proposed was the Uniform Hospital Discharge Data Set describing short-stay hospital inpatients and their episode of care, to be incorporated in all government and voluntary reporting systems and surveys of hospital activity. In considering the content of this data set, the conference set forth some criteria, which have been used since in the design of other data sets: 1 The data items included in the set must be useful to most if not all potential users. In the case of hospital statistics, the major groups considered were hospital medical staffs; hospital administrators; local, state and federal planning, regulatory, and legislative bodies; private and public hospital insurance agencies; and the research community. 2 The items selected must be ones that can be readily collected with reasonable accuracy. 3 Items should not duplicate data available from other sources. 4 Confidentiality of medical information should be protected. The recommendations of the 1969 conference were reviewed and refined by the US National Committee on Vital and Health Statistics and feasibility studies were made. In 1972, the National Committee issued an official version (2). It has been widely endorsed and accepted; the biggest obstacle
265
266
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 2
Development of uniform basic data sets for health information systems in the United States
Component
Areas covered by existing data sets
Potential areas for development
Vital statistics
Morbidity Disability Health problems, attitudes, and practices
Health services resources
Manpower Inpatient facilities
Ambulatory care facilities Supportive services and equipment ? Skills and knowledge
Use of health
Hospital discharges Ambulatory care visits Long-term care
? Mental health services ? Emergency medical services ? Other categorical services
Health expenditures
Sources and uses of funds Uniform cost accounting
Environmental health
Environmental health hazards Occupational health hazards
of funds, hospital utilization rates by population subgroups, and other basic facts upon which hospital management, planning, and evaluation depend. Table 2 shows the progress that has been made so far on uniform basic data sets in the United States and gives some idea of the uncharted terrain. In the area of health status indicators, standard measures of mortality should be supplemented by standard population-based measures of morbidity and disability. Using White's model of the five D's — death, disease, disability, dysfunction, and dissatisfaction (15) — serious considerations should also be given to expanding the concept of health status to include measures of the health problems, symptoms, pain and anxiety that people experience, their attitudes towards health services, and their health beliefs and practices. Since 1957 the National Health Interview Survey has provided standard estimates of morbidity and disability and related use of services, as reported in a continuing sample survey of households (16). A second national survey, the Health and Nutrition Examination Survey, provides professionally determined measurements of health status (17). But the sample sizes of these surveys are not large enough to provide direct estimates for community planners and policymakers, and few state and local agencies have the resources and skills to conduct surveys of such complexity. A more selective set of indicators and less expensive methods of data acquisition are
needed for producing small-area statistics on health levels and health needs. Data sets have been designed and are being used to collect statistics on health manpower and inpatient facilities, although both are presently being extensively overhauled. The constantly changing patterns of services and personnel in the United States make it difficult to settle upon uniform groupings or classifications for health personnel and institutions. As the table shows, there is much room for expansion of these data sets to cover the entire spectrum of tangible and intangible resources. The third broad area of health services utilization also poses a problem of categorization. At present working groups have divided patients and services into three classes: short-stay hospital inpatient, ambulatory, and long-term care, but these distinctions are not entirely logical and may require modification. Long-term health care is particularly difficult to delineate. As currently defined, long-term care includes patients in chronic disease, psychiatric, and custodial institutions, or units of hospitals, as well as those living at home and receiving community-based services. The principal reason for treating long-term care separately is that social, economic and behavioural causes and manifestations of illness assume greater importance in the case of the chronically ill and disabled population than in acute and episodic care. To complicate the picture, special data sets are being
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Health status
DATA SETS FOR HEALTH STATISTICS
PROBLEMS IN THE DEVELOPMENT OF DATA SETS On the basis of United States experience, there are a number of problems in design and implementation that must be solved in order to make the method of uniform basic data sets effective. Defining the scope of a data set and avoiding proliferation. As already suggested, it has been difficult to decide upon the optimal number and scope of data sets in the United States, and this will undoubtedly be a major problem in cross-national statistical guidelines because of wide variations among countries in the organization of health services. In the case of the United States, the
difficulty stems from present uncertainty about the direction of health policy; categorical or disease-specific programmes continue to dominate the flow of public and private funds, while some of the more recent legislation seeks to promote comprehensive planning . and organization of services. The idea of data sets has become quite popular in the last year or so. Proliferation of specialized data sets threatens to dilute the concept of broad, multipurpose guidelines for health statistical systems that permit one to sec the whole as well as the parts and to think flexibly about the health care system as one plans for the future. The process of designing data sets has been evolutionary. The time has come to construct a master plan or conceptual framework for their orderly development, a problem that the US National Committee on Vital and Health Statistics is now beginning to address. Adhering to the principle of minimal or basic requirements. With so many parties involved, and computers standing by to run seemingly endless series of cross-tabulations, there is constant pressure on the designers of data sets to expand them beyond the limits of what is feasible for busy health professionals to provide on a routine basis, and most data recipients to digest and put to good use. The long-term care data set is a good example. In principle, public policy in the United States is directed to a much broader definition of the health care for the chronically ill and disabled population than the health services system is presently funded and organized to provide. The draft data set reflects this philosophy by including measures of dysfunction and social and economic dependency as well as disease, by encompassing mental as well as physical illness, and by covering the entire spectrum of curative, rehabilitative, and supportive services, both inpatient and outpatient. As a result, it started out with some 24 major elements compared to 14 in the Uniform Hospital Discharge Data Set, and may well end up being far from minimal. Coordination and linkage. As work progresses on basic data sets in the United States, problems of internal and external consistency and compatibility are intensified. In the interest of efficiency, the data sets should complement not duplicate each other. More important, it should be possible to link them to build profiles of the health needs, resources, and services of geographic areas and population subgroups. To do so, the data sets must use the same concepts, terms, classifications, and identifiers. In turn, these must correspond with the nation-wide census, the universal denominator for health
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
proposed for mental health and emergency services. In the case of mental health, there is a well-defined service sector with its own source of federal funds. On the other hand, many health professionals consider the distinction between physical and mental illness unsound in theory and practice, and many patients with mental disorders are in fact being treated in the offices of general physicians and in general hospitals and nursing homes. Emergency services are also provided at many sites, and the emergency room in the typical American hospital, although specially staffed and equipped, does double duty as a general, walk-in clinic for the people in its neighbourhood, especially after working hours. Thus the question of how many data sets and subsets are needed to describe patients and their use of services is by no means resolved. The final two components of a comprehensive information system as set forth in Table 2 have yet to be tackled systematically from the viewpoint of establishing uniform guidelines. Expenditures, it should be noted, may also be viewed as resources although treated separately here. Preliminary analysis suggests that the formula used for many years by the Federal Government to provide annual statistics on total expenditures and sources and major uses of health funds may be suitable, with some adaptation, for constructing state and local profiles of the flow of funds (18, 19). There is considerable interest in the related problem of establishing standard cost accounting procedures. The natural and man-made hazards to health in the working place, home, and general environment are now recognized to be so many and varied that they would appear to defy description in a minimum data set. But 1974 planning legislation in the United States includes environmental health indicators in the statistical base for national, state, and local planning agencies, so a start must be made on this front soon.
267
268
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
statistics, and if possible, with other social indicators that give wider meaning to health statistics. In addition, there are certain problems common to all the data sets that require common resolution, notably unique numbering systems for individuals and organizations, geographic coding of residence and location, and procedures for protecting privacy and confidentiality. For all these reasons, the oversight of a single body like the US National Committee is essential.
Assuring data quality. A recent study by the National Academy of Sciences of the reliability of data abstracted from hospital medical records (20) illustrates how much time and effort must be devoted to evaluating and improving the quality of the basic data being collected. Comparison of a sample of original abstracts prepared by hospital personnel with those of an independent team of coders employed in the study showed a high degree of agreement in the areas of patient age, sex, dates of admission and discharge, and expected principal source of payment. But in the key areas of principal diagnosis and principal procedure, agreement in the 4-digit codes assigned was only 65 per cent and 73 per cent respectively. Often the medical record itself was not clear in regard to the patient's principal and other diagnoses and procedures. Since hospitals have had considerable experience preparing diagnostic indexes and they are widely used, it is clear that much more work must be done on validity and reliability and caution exercised in the introduction of new and unfamiliar data items and concepts. Establishing mechanisms for review and revision. Changes in health policy in the United States are constantly shifting priorities for information. Furthermore, some of the data elements and
Extending the concept internationally. From the start United States committees have recognized the obvious benefits of coordinating their efforts with those of other countries, and it should be added that they have relied heavily on experience and practices in other countries, especially the United Kingdom and Canada, to help design the hospital, ambulatory, and long-term care data sets. It is encouraging, therefore, that there is growing interest on the part of international groups not only in common measures and classifications of diseases and health problems but also in common approaches to describing the ways in which health resources arc allocated, organized, and used to address these problems. If definitional differences can be resolved, a uniform core of international data linking measures of the population's health problems with measures of the distribution of health resources, use of services, and related expenditures offers exciting new possibilities for cross-national comparisons. Work on uniform basic data sets in the United States, it is hoped, will contribute to this important international movement. REFERENCES (1) Hospital Discharge Data: Report of the Conference on Hospital Discharge Abstract Systems, Airlic House, June, 1969. Eds Mumaghan J H and White K L. Lippincott, Philadelphia. 1970. (Reprinted from Medical Care 8(4): Supplement, 1970.) (2) Uniform Hospital Abstract: Minimum Basic Data Set. US Department of Health, Education, and Welfare, US National Committee on Vital and Health Statistic!. DHEW Publication No (HRA) 7 5 - 1 4 5 1 . Government Printing Office, Washington. 1972. (3) Ambulatory Care Data: Report of the Conference on Ambulatory Care Records, Chicago, April, 1972. Ed Mumaghan J H. Lippincott, Philadelphia. 1973. (Reprinted from Medical Care 11(2): Supplement, 1973.) (4) Ambulatory Medical Care Records: Uniform Minimum Basic Data Set. US Department of Health, Education, and Welfare, US National Committee on Vital and Health Statistics. DHEW Publication No (HRA) 7 5 - 1 4 5 3 . Government Printing Office,' Washington. 1974. (5) The Health Manpower Inventory and Health Facilities
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Promoting widespread use. It is not enough, of course, to promulgate a series of well-tested guidelines; development of statistical standards must go hand-in-hand with development of data systems like the Cooperative Health Statistics System in the United States. Aside from the political problems .of establishing shared data systems, American experience has shown that promoting uniformity extends to the way in which data are collected and analysed. Attention must be devoted to common procedures as well as common terminology. Also experience shows that potential recipients like planners and administrators require much direct guidance and technical assistance to make effective use of even the small amount of data contained in the basic data sets discussed here.
concepts are necessarily new and experimental. Only widespread use can demonstrate fully the strengths and weaknesses of a basic data set. To assure that these statistical guidelines are dynamic — that they do not suffer from hardening of the bureaucratic arteries — there must be a regular schedule for review and revision. Most of the data sets discussed here are already being re-examined by the US National Committee, but with more experience, a longer cycle may be feasible as in the case of data sets for vital events.
DATA SETS FOR HEALTH STATISTICS
(13)
(14)
(15) (16)
(17)
(18) (19)
(20)
DHEW Publication No(HRA) 77-1456. Government Printing Office, Washington. 1977. El Resumen Uniforme de Egreso de Hospitalcs: Un Conjunto Minimo dc Datos Basicos. Pan American Health Organization. PAHO Document HS/Seric RM/No 16/1975. Washington, 1975. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, 1975 Revision, Vol 1, pp 737-74O. World Health Organization, Geneva, 1977. White K L. Epidemiologic intelligence requirements for planning personal health services. Ada socio-medica Scandinavia 2: 143, 1972. Health Interview Survey Procedures, 1957-1974. US Department of Health, Education, and Welfare, National Center for Health Statistics. DHEW Publication No (HRA) 7 5 - 1 3 1 1 . Government Printing Office, Washington. 1975. Plan and Operation of the Health and Nutrition Examination Survey, United States — 1971 — 1973. US Department of Health, Education, and Welfare, National Center for Health Statistics. DHEW Publication No (HRA) 76—1310. Government Printing Office, Washington. 1976. Gibson R M and Mueller M S. National health expenditures, fiscal year 1976. Social Security Bulletin 40(4): 3, 1977. Final Report on the Development of the Community Funds Flow Data System. Wirnowski J P, Project Manager, Arthur Young & Co, 1025 Connecticut Avenue, Washington, D C 20036, 1975. Reliability of Hospital Discharge Abstracts. National Academy of Sciences, Institute of Medicine. The Institute, Washington. 1977.
(revised version received 6 March 1978)
Infectious Disease Newsletter This is intended as an informal method of communication between members of the International Epidemiological Association interested in the field of infectious diseases. Portions may be published in the International Journal of Epidemiology. The newsletter is not intended to compete with official surveillance reports such as the WHO Weekly Epidemiological Record or the U S Morbidity and Mortality Weekly Reports, and other similar sources of formal data. Rather it is to
supplement these with informal reports, comments, news of meetings etc, so that a closer association can be formed of members of IEA interested in this area. If you would like to contribute to the newsletter and/or be placed on the mailing list write to Dr AS Evans, Professor of Epidemiology, Yale University School of Medicine, 333 Cedar Street, New Haven, Ct 06510, USA. This is a pilot project so your support, comments and suggestions are welcome both as to substance and to format.
Downloaded from http://ije.oxfordjournals.org/ at University of California, Santa Barbara on June 27, 2015
Minimum Data Set are being revised by the US National Committee on Vital and Health Statistics. Mimeographed copies of 1974 versions are available from the Division of Health Manpower and Facilities Statistics, National Center for Health Statistics, Hyattsville, Maryland 20782. (6) Long-Term Care Data: Report of the Conference on Long-Term Health Care Data, Tucson, May, 1975. Ed Mumaghan J H. Lippincott, Philadelphia. 1976. (Reprinted from Medical Care 14(5): Supplement, 1976.) (7) The 1968 Revision of the Standard Certificates of Live Births, Death, Fetal Death, Marriage and Divorce or Annulment. US Department of Health, Education, and Welfare, National Center for Health Statistics. Public Health Service Publication No 1000, Series 4, No 8. Government Printing Office, Washington. 1968. (8) New Approaches in Health Statistics. World Health Organization. Technical Report Series 559. Geneva, v 1974. (9) Some WHO Activities in the Field of Vital and Health Statistics and Related Fields During 1975. WHO Document/HS/Nat Com/76.352. World Health Organization, Geneva, 1976. (10) Armitage P. National health survey systems in the European Community. International Journal of Epidemiology 5t 321, 1976. (11) Hogarth J. Glossary of Health Care Terminology. World Health Organization Regional Office for Europe, Copenhagen. 1975. (12) The Cooperative Health Statistics System: Jts Mission and Program. US Department of Health, Education, and Welfare, National Center for Health Statistics.
269
