Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
154 INTEGRATING FEDERAL STATISTICS ON CHILDREN interest in other sources of social science evidence regarding how children and youth negotiate the transition to adulthood. REFERENCES Alexander, K.L., A.M. Pallas, and S. Holupka 1987 Consistency and change in educational stratification: recent trends regarding so- cial background and college access. Pp. 161-185 in R. Robinson, ed., Research in Social Stratification and Mobility, Volume 6. Greenwich, Conn.: JAI Press. Bureau of Labor Statistics 1988 BLS Handbook of Methods, Bulletin 2285. Washington, D.C.: U.S. Government Printing Office. Charner, I., and B.S. Fraser 1988 Youth and Work: What We Know, What We Donât Know, What We Need to Know. Washington, D.C.: The William T. Grant Foundation on Work, Family and Citi- zenship. Chase-Lansdale, P.L., F.L. Mott, J. Brooks-Gunn, and D.A. Phillips 1991 Children of the National Longitudinal Survey of Youth: a unique research oppor- tunity. Developmental Psychology 27:918-931. Chubb, J., and T. Moe 1990 Politics, Markets, and Americaâs Schools. Washington, D.C.: The Brookings Institution. Davis, C., and B. Sonnenberg, eds. 1993 Programs and Plans of the National Center for Education Statistics, 1993 Edition. Washington, D.C.: U.S. Government Printing Office. Duncan, G., and J. Morgan 1985 The Panel Study of Income Dynamics. In G.H. Elder, Jr., ed., LifeCourse Dynam- ics: Trajectories and Transitions, 1968-1980. Ithaca, N.Y.: Cornell University Press. Ekstrom, R.B., M.E. Goertz, and D.A. Rock 1988 Education and American Youth. London: Falmer Press. Elder, G.H., Jr. 1985 Life Course Dynamics: Trajectories and Transitions, 1968-1980. Ithaca, N.Y.: Cornell University Press. Hauser, R.M. 1991 Trends in College Entry Among Whites, Blacks, and Hispanics, 1972-1988. Dis- cussion Paper No. 958-91. Institute for Research on Poverty, University of Wis- consin, Madison. Kaufman, P., M. McMillen, and D. Bradby 1992 Dropout Rates in the United States: 1991. Washington, D.C.: National Center for Education Statistics. Kominski, R., and P.M. Siegel 1993 Measuring education in the Current Population Survey. Monthly Labor Review September:34-38. Kominski, Robert 1990 Estimating the national high school dropout rate. Demography 27:303-311. Manser, M., M. Pergamit, and W.B. Peterson 1990 National longitudinal surveys: development and uses. Monthly Labor Review July:32-37.
EDUCATIONAL ATTAINMENT AND THE TRANSITION TO WORK 155 Mizell, H. 1988 A commentary on Ivan Charner and Bryna Shore Fraserâs Youth and Work: What We Know, What We Donât Know, What We Need to Know. Pp. 97-106 in I. Charner and B.S. Fraser, Youth and Work: What We Know, What We Donât Know, What We Need to Know. Washington, D.C.: The William T. Grant Foundation on Work, Family and Citizenship. OERI Goal 2 Work Group 1994 Reaching the Goals: Goal 2âHigh School Completion. Washington, D.C.: Of- fice of Educational Research and Improvement, U.S. Department of Education. Polivka, A.E., and J.M. Rothgeb 1993 Redesigning the CPS questionnaire. Monthly Labor Review September:10-28. Rosenbaum, J.E., T. Kariya, R. Settersten, and T. Maier 1990 Market and network theories of the transition from high school to work: their application to industrialized societies. Annual Review of Sociology 16:263-299. Wilson, W.J. 1987 The Truly Disadvantaged: The Inner City, The Underclass, and Public Policy. Chicago: University of Chicago Press.
156 INTEGRATING FEDERAL STATISTICS ON CHILDREN Monitoring Changes in Health Care for Children and Families Paul Newacheck and Barbara Starfield INTRODUCTION In 1994, when the workshop on integrating federal statistics on children was held, health care reform was a leading item on the agendas of the Clinton administration and the Congress. Although federal legislative progress in this area is now uncertain, the concern of many stakeholders over access, costs, and quality of the health care system is driving change at the state level. Indeed, several states have already moved to implement health care reforms. In addition, powerful economic forces are leading to rapid changes in the organization and delivery of health care. For example, the move throughout the country to managed health care in the private sector is one indication of the kind of reform that may gather strength, with or without federal legislation. These changes and others in the health care system are the result of several factors. Paul Newacheck is at the Institute for Health Policy Studies, University of California at San Francisco. Barbara Starfield is at the Department of Health Policy and Management, Johns Hopkins University. This paper was drafted and presented in 1994, when proposals for health care reform were being put forward and debated by the Clinton administration and in Congress. The authors have revised the text somewhat to reflect changing circumstances. However, the implications for federal statistics on children of eventual changes in the nationâs health care system remain significant. 156
MONITORING CHANGES IN HEALTH CARE 157 First, the high and rapidly growing costs of the health care system have generated concerns about the tax burdens of public programs such as Med- icaid and Medicare, threats to the competitiveness of industry as a result of the high costs of employee health benefits, high administrative costs associ- ated with a disorganized mechanism of paying for services through uncoor- dinated insurance policies, and perverse incentives for the use of unneces- sary technology. Second, access to basic health care services remains problematic for millions of people as a result of poorly distributed personnel and facilities, lack of insurance and inadequate insurance coverage for a substantial pro- portion of the population (especially children and young adults), cost shar- ing that imposes a barrier to seeking needed health services, employment insecurity and the lack of portability of insurance coverage with changes in place of employment, and the refusal of many health care providers to deliver services to individuals covered by public programs such as Medic- aid. Third, there is continuing concern about the availability, quality, and effectiveness of care as reflected by the relatively poor performance of the United States on major health indicators (especially among the young) and evidence of lack of effectiveness of many health services interventions. Fourth, persistent and perhaps growing inequities among subgroups of the population have resulted from systematic differences in access to care and quality of care, insurance mechanisms such as experience rating, in which the sickest and most disadvantaged are less able to obtain insurance, and a premium structure for health insurance that requires the least well off to pay as much as the wealthiest for their health insurance coverage. Health care reforms are likely to bring about profound changes in the manner in which children and their families in this country obtain health care. Universal coverage would give almost 10 million children who are now uninsured health insurance coverage (Snider and Boyce, 1994). Mil- lions more would be likely to see improvements in the breadth and depth of their coverage. Others would see their coverage become more restrictive or their choice of providers become more limited. Under managed competition, the approach favored by the president and many in Congress, there would be an acceleration of the already rapid growth of managed care arrangements, such as health maintenance organi- zations and preferred provider organizations. Managed care offers impor- tant potential for reducing costs and improving access to quality care. Yet many urge caution regarding the rapid adoption of managed care because of the financial incentives to underserve patients in prepaid systems of care (Newacheck et al., 1994; Fox and McManus, 1992). Although little research has been conducted to date on the effects of managed care on children, it is clear that customary doctor-patient relation-
158 INTEGRATING FEDERAL STATISTICS ON CHILDREN ships will be disrupted for many families as they enter organized systems of care. Proponents of managed care argue that access to primary care will be facilitated; however, it is not clear that this would be the case. Access to specialist services would almost certainly be reduced in managed care set- tings. Pediatricians would increasingly be asked to serve as gatekeepers for the service system; at the same time they would experience reduced au- tonomy in choosing courses of treatment for their patients. Beyond these direct effects on pediatric patients and providers, health care reforms carry significant indirect ramifications for families. For ex- ample, some families would see their out-of-pocket expenses for medical care decline, freeing resources for other family needs. Other families would experience increased outlays for medical care. Changes occurring under health care reform would also reverberate in the labor market. Most health care reform proposals would end the practice of imposing preexisting con- dition clauses in insurance policies. These restrictive clauses, common in health insurance plans offered through small employers, can circumscribe job mobility for families with disabled children. The effect of eliminating this practice would be profound, since 4 million American children now suffer from disabling chronic conditions (Benson and Marano, 1994). It is less clear how the myriad special programs serving vulnerable populations of children would be affected by health care reform. The fed- eral government now helps to fund a network of over 500 community and migrant health centers across the United States. Low-income families and their children often seek care in these and other publicly sponsored health centers and clinics. In addition, the federal government currently spends upward of $500 million annually on maternal and child health block grants. The states supplement these grants with their own funds to operate a variety of programs serving low-income and other children with special needs. Federal, state, and local governments also support emergency medical services for children, regionalized systems of prenatal care, regional centers serving children with rare chronic diseases, family planning programs, and other services. The fate of these programs and others that make up the public health system remains uncertain. Monitoring and evaluating the changes that accompany health care re- forms for children, their families, and the health care system that serves them presents a formidable challenge. Fortunately, the United States pos- sesses a system for collecting and disseminating health statistics that is the envy of most other developed countries. Even so, our health statistics programs were not designed for the purpose of monitoring and evaluating health care reform. Consequently, an assessment of their utility for this purpose is in order. In doing so, it is important to keep in mind that changes in the health care system will not affect all populations in the same manner. Childrenâs
MONITORING CHANGES IN HEALTH CARE 159 health needs and provider networks differ from those of adults. Childrenâs illnesses and injuries are diagnosed and treated in the context of rapid growth and developmental processes that have no counterpart in adults. Their physi- ological, cognitive, and emotional maturation makes children uniquely vul- nerable to certain illnesses. Their developmental vulnerabilities require heightened attention to preventive care and early diagnosis and treatment of their disorders (Jameson and Wehr, 1993). Assessment of the current fed- eral health statistics programs and the design of mechanisms for monitoring the effects of changes in the system must take into account these differ- ences. In the next section, a framework is presented for assessing the types of data needed for monitoring the effects of health care reforms on children and families. DATA NEEDED FOR MONITORING CHANGES IN HEALTH CARE Evaluation of the changing health care system requires data that sys- tematically address concerns about equity, access, quality, and costs of care. Timely and reliable data are needed to support the administration and op- erations of the health care system, to measure the performance of providers, ensure quality, and support public health objectives and programs (Gaus et al., 1993). Table 1 provides a framework for considering the data elements that are required, over time, and at the different levels at which health policies and planning are generated. The table also indicates the types of data sources required to obtain the needed information. Examples of the types of data needed for monitoring changes in health care for children, families, and the health care system are presented below in narrative form. â¢ Information is needed to ascertain the state of health and functional status of the population; the adequacy of access for preventive, curative, palliative, and rehabilitative services; barriers to the receipt of care; the equity of use of services across different population groups; and the degree to which services are distributed where they are most needed. Information concerning the burden of cost sharing and the degree to which it interferes with the receipt of needed services; the extent to which the goals of conti- nuity and coordination of care are achieved; and satisfaction of the popula- tion with perceived technical, interpersonal, and organizational aspects of health service are also needed. The indirect costs of illness, such as time lost from work and school and child care costs attributable to illness, are also important. â¢ Information about the services provided is needed to ascertain the
160 TABLE 1 Data Elements Needed for Monitoring and Evaluating Health Care Reform Units of Analysis Domains Indicators Data Sources Populations Access to care Insurance coverage (including covered and uncovered services), Population surveys sources of primary care and specialty services, travel time, appointment waiting time, office waiting time, unmet needs, barriers to care Health care Hospitalizations; number and types of visits to different kinds of Population surveys; utilization physicians and other health professionals; long-term care; use of medications, provider surveys medical equipment, and supplies; perceived quality of services; continuity, coordination of, and comprehensiveness of services; satisfaction with services Health care Total and out-of-pocket personal health care expenditures, including hospital, Population surveys; expenditures physicians, mental health, other professional services, long-term care, administrative medications, medical equipment and supplies. Indirect costs of illness. records Health status Infant and childhood mortality, low birthweight, health and functional status Vital statistics; indicators (including physical, development, behavioral, and emotional), population surveys; restricted activity days, indicators of risks, and resilience to illness surveillance data INTEGRATING FEDERAL STATISTICS ON CHILDREN
Health Health care Presenting problems, diagnosis, case mix; degree to which care is Population surveys; plans and services comprehensive, coordinated, and continuous; adequacy of prenatal infant, provider surveys; providers childhood, and adolescent care; provision of appropriate immunizations and administrative other preventive services; frequency of well-child examinations, quality of care records for individual acute and chronic conditions Effectiveness Reasons for visits, health services interventions (including procedures, tests, Population surveys; of care medications and other therapies (including home health care), referrals, administrative organizational amenities (such as waiting times); types of practitioners records involved in providing services; outcomes of visits Health Health care Work force deployment, by type of professional and type of physician; Systems systems resources primary care resources; specialty care resources; hospitals; ambulatory care surveillance organizations (including HMOs); long-term care facilities, laboratories; MONITORING CHANGES IN HEALTH CARE home health services organizations, public health organizations & services, (e.g., school health facilities) Health care System-level health care expenditures, mode of financing; mode of payments; System surveillance expenditures expenditures by type of service 161
162 INTEGRATING FEDERAL STATISTICS ON CHILDREN extent and type of health care needs being served; the distribution of ser- vices; the extent to which practice patterns are targeted to the needs of the population served; the extent to which they are organized to provide care that is accessible, comprehensive, continuous, and coordinated; and the ex- tent to which the services provided are justified by evidence of effective- ness and appropriateness. â¢ Information about the organization and distribution of resources within the health care system is needed in order to plan and monitor their adequacy and equity with regard to preventive services, primary care services, and consultative and referral services. Of importance are the deployment of practitioners and facilities, the mode of financing and the distribution of payments for services, the availability of primary care (for both physical and mental child health problems) and specialty services for children with chronic health problems, and the availability and deployment of special modes of intervention such as home visiting, medications, and devices where they are needed. Information concerning the costs of care provided and the sources of payments for various aspects of care is required in order to discern the degree of success in maintaining cost control within the health services system. Information is also needed to determine the balance of expenditures between primary care and specialty care as well as between preventive care and care of illness. â¢ Information is also needed to track environmental and social charac- teristics that predispose children and families to illness or interfere with efforts to ameliorate illness. Data collection strategies should be designed to provide baseline infor- mation on all of these characteristics and should be organized to monitor changes over time as health care reforms are implemented. Planning for information retrieval should be coordinated so that data collected at local levels can be aggregated to the regional, state, and federal level, and so that data collected at higher levels can be related to data collected at lower levels. This will permit comparisons to be made across health plans, geo- graphic regions, and states and will enable health planners to identify more successful models for delivery of care from ones that are less successful in meeting the objectives of reform. In the next section, the major existing sources of data are identified and evaluated against the information needs (outlined above) for evaluating changes in health care. CHARACTERISTICS OF CURRENT DATA SOURCES There are approximately 500 health, social, and demographic data projects in the U.S. Department of Health and Human Services alone. Added to
MONITORING CHANGES IN HEALTH CARE 163 these are state and local data collection efforts, projects of private groups, and administrative and clinical databases of health facilities, plans, and insurance agencies. Evaluation of health care reforms will require a change in thinking about the locus of control for health data systems. Since both governmental and private data sources are needed to address the variety of characteristics in Table 1, a public-private partnership in developing the framework and mechanisms for data collection and analysis is a critical goal. In this section, the major federal, state, and private data collection ef- forts are presented and discussed with regard to their current contributions to knowledge as well as changes that are required in order to make them more suitable for evaluation of health care reforms. Table 1 listed the types of data sources for obtaining information from populations, providers and health care plans, and health systems. Table 2 provides brief descriptions of major databases relevant to monitoring changes in health care. And it is organized according to these types of information sources: â¢ Vital statistics and surveillance systems for obtaining population- based information concerning health status. â¢ Population surveys to obtain data concerning health status, access to services, use of services, experiences and satisfaction with various aspects of services, and both direct and indirect out-of-pocket costs of dealing with health problems. â¢ Provider surveys to obtain information on the characteristics of ser- vices used, types of services provided, distribution of facilities, and types of health conditions diagnosed. â¢ Administrative records from health plans and providers for providing information on the adequacy of arrangements for assessing health care needs of populations served, characteristics of care provided (including the bal- ance between primary care and specialty care, as well as between preven- tive, curative, palliative, and rehabilitative aspects of care), the resources expended in providing care, the adequacy of organizational arrangements for achieving comprehensive, continuous, and coordinated care, documenta- tion of effectiveness of care provided, and costs generated in the provision of services. â¢ Health systems data to determine the distribution of facilities and personnel and their type, the sources of funding for the provision of ser- vices, the distribution of expenditures by type and source of care, and total expenditures for providing care to population groups differing in the extent and type of their health care needs. An additional category of data addresses documentation of environmen-
164 INTEGRATING FEDERAL STATISTICS ON CHILDREN TABLE 2 Population-Based Data Sources for Monitoring Childrenâs Health Status Name and Data Source/ Selected Applications of Data Sponsor Methods for Health Care Reform A. Vital Statistics and Surveillance Vital Statistics State vital Life expectancy; causes of death; Cooperative registration infant mortality; prenatal care Program and birthweight; birth rates; (NCHS) teenage and unmarried births; family formation and dissolution; pregnancy outcomes Linked Birth/ Birth and death Infant mortality rates by birth cohort; Infant Death certificates infant mortality rates by birthweight Program (NCHS) National Death Death certificates Facilitates epidemiologic follow-up Index (NCHS) studies; verification of death for individuals under study; most NCHS surveys are linked to NDI Youth Risk Self-administered Assessment of changes in health Behavior questionnaire in practices over time Surveillance schools System (CDC) B. Population Surveys National Health Personal Large-scale data set with ability to Interview Survey interviews establish trends in access, utilization, (NCHS) and health status National Health Personal Health status; chronic conditions; Interview Survey interviews behavioral problems; learning and on Child and developmental problems; birth Family Health characteristics (NCHS)
MONITORING CHANGES IN HEALTH CARE 165 Child/Family Planned Overall Sample Sample Periodicity All births, deaths, All births, deaths, Annual and fetal deaths; and fetal deaths; sample of sample of marriages and marriages and divorces; divorces; termination of termination of pregnancy for pregnancy for selected selected reporting reporting areas areas All U.S. births All U.S. births Annual and infant deaths and infant deaths All deaths All deaths Annual occurring among children Nationally Only adolescents Every 2 representative surveyed; years sample of students comparable but in grades 9-12; unlinked survey state-level data data for adults available 50,000 households Approximately Annual; 30,000 children insurance and under 18 years access surveys plus family added in 1993 members One child from 17,000 children Every 7-8 each NHIS (in 1988) under years; planned household 18 years plus for 1996 if linked data on funding other family available members continued on next page
166 INTEGRATING FEDERAL STATISTICS ON CHILDREN TABLE 2 Continued Name and Data Source/ Selected Applications of Data Sponsor Methods for Health Care Reform National Maternal Follow-back using Factors associated with low birthweight Infant Health state vital records; and infant death; barriers to prenatal Survey (NCHS) interviews of care; effects of maternal risk factors mothers; hospital on pregnancy outcomes medical records; prenatal care providers National Survey Personal Contraception and sterilization; teenage of Family Growth interviews; sexual activity and pregnancy; family (NCHS) telephone planning and unintended pregnancy; interviews adoption; infertility National Health Personal Physical health; clinical findings; and Nutrition interviews; growth and development Examination physical Survey examinations; laboratory tests National Medical Multiple personal Linked data on access, utilization, Expenditure interviews; health status, and expenditures Survey provider record (AHCPR) checks; insurance policies analysis National Personal interviews Longitudinal tracking of health status Longitudinal Survey of Youth- Child Data (DOL and NICHD) Early Childhood Personal interviews Longitudinal tracking of access, Longitudinal with parents and utilization, and health status Survey children; school (NCES) questionnaire; teacher questionnaire
MONITORING CHANGES IN HEALTH CARE 167 Child/Family Planned Overall Sample Sample Periodicity 10,000 live births; 10,000 live births; Every 8 4,000 fetal deaths; 4,000 fetal deaths; years, with 6,000 infant deaths; 6,000 infant deaths longitudinal oversample of follow-up; blacks last conducted in 1988 with 1991 follow-up 8,500 women ages 8,500 women ages Every 5 years 15-44; oversample 15-44; oversample subsequent of blacks of blacks follow-up; last conducted in 1988 with 1990 telephone interviews 30,000 persons ages 14,000 children Periodic; 2 months and older under age 20 currently in 4th examined for year of 6-year NHANES III cycle; continuous examinations to begin in 1996 14,000 households 10,500 children Every 8-10 (in 1987) plus years; last linked data on conducted in other family 1987; planned members for 1996 Children born to Approximately Every 2 years female members 8,500 children of the NLSY A cohort of 23,000 A cohort of 23,000 Will follow kindergarten kindergarten cohort through children children 5th grade; data collection to begin with 1998- 1999 school year continued on next page
168 INTEGRATING FEDERAL STATISTICS ON CHILDREN TABLE 2 Continued Name and Data Source/ Selected Applications of Data Sponsor Methods for Health Care Reform Survey of Income Personal interviews Longitudinal tracking of program and Program participation, health insurance Participation coverage, and health status (Census) Survey of Program Personal interviews Longitudinal tracking of access, Dynamics (Census) utilization, and health status Current Population Personal interviews Health insurance coverage; public Survey (Census) program participation National Sample survey of Prevalence of substance abuse Household Survey population age 12 on Drug Abuse and older (ADAMHA) C. Provider Surveys National Hospital Hospital records; Patient characteristics; length of stay; Discharge Surveys computerized data diagnosis and multiple diagnoses; (NCHS) sources surgical and diagnostic procedures National Survey Abstract forms Patient characteristics; diagnosis and of Ambulatory completed by multiple diagnoses; surgical and Surgery (NCHS) facility staff diagnostic procedures National Encounter forms Characteristics of patientsâ visits to Ambulatory completed by physicians; diagnoses and treatment Medical Care physicians Survey (NCHS) practicing in private offices
MONITORING CHANGES IN HEALTH CARE 169 Child/Family Planned Overall Sample Sample Periodicity 32-month 15,000 children A new panel is longitudinal panel under age 18 introduced every survey; 20,000 in 1990 panel 12 months; households beginning 1996, a 48-month panel of 50,000 households is planned 20,000 households Approximately Will follow over from the 1993 SIPP 15,000 children a 10-year period panel if funding available 57,000 households 33,000 children Monthly; health monthly under age 15 insurance and program participation data collected in March surveys 18,000 households Approximately Biennial 5,000 youth ages 12-17 494 hospitals; Approximately Annual 274,000 discharges 60,000 discharge records for children under 18 years 700 facilities; Unknown (to be Annual; to be 180,000 patient implemented in implemented in records 1994) 1994 3,400 physicians Approximately Annual in office-based 7,500 records for practices; 45,000 children patient visits continued on next page
170 INTEGRATING FEDERAL STATISTICS ON CHILDREN TABLE 2 Continued Name and Data Source/ Selected Applications of Data Sponsor Methods for Health Care Reform National Home Questionnaires Number of patients; functional status of and Hospice completed by home patients; number of discharged patients; Care Survey health agencies characteristics of home health agencies (NCHS) and hospices and hospices and their patients National Hospital Encounter forms Characteristics of patientsâ visits in Ambulatory completed by hospital outpatient departments and Medical Care physicians and emergency departments; diagnoses and Survey (NCHS) other hospital staff treatment National Employer Computer-assisted State-level estimates of private health Health Insurance telephone insurance; detailed national-level Survey (NCHS) interviewing analysis of private health insurance; characteristics of employer spending for health care Healthcare Cost State hospital Hospital utilization; variation in practice and Utilization discharge records patterns Project (AHCPR) from public and private systems D. Administrative Records National Health Health care Characteristics of hospitals, nursing and Provider Inventory facilities; state related care homes, hospices, and home (NCHS) licensing agencies; health agencies; sampling frame for professional provider surveys associations Claims Data Claims submitted Hospitalization; quality of care; variation to private insurance in practice patterns; health care payments; companies for diagnosis and treatments; some data have insured individuals cost sharing and claims submitted to state Medicaid agencies
MONITORING CHANGES IN HEALTH CARE 171 Child/Family Planned Overall Sample Sample Periodicity 1,500 home health Approximately 700 Annual agencies and records for children hospices; 7,000 under 18 years current patients; 7,000 discharged patients 525 hospitals; Approximately Annual 70,000 patient visits 15,000 records for children under age 15 Stratified national Not applicable Annual; to be sample of implemented in establishments to 1994 include all sizes of businesses, both private and public sector; 100,000 establishments screened to conduct 51,000 interviews Hospital discharge Depends on state Continuous data records from 14 collection effort state data systems funded through 1994 All licensed/certified Not applicable Periodic; facilities in covered recently categories expanded to include additional providers Depends on system Depends on system Depends on system continued on next page
172 INTEGRATING FEDERAL STATISTICS ON CHILDREN TABLE 2 Continued Name and Data Source/ Selected Applications of Data Sponsor Methods for Health Care Reform State Hospital Approximately 20 Resources expended; medical practice; Discharge Data states variations Tape to Tape Medicaid Hospitalization and ambulatory care Medicaid Files enrollment, claims, visits; quality of care; variation in (HCFA) and provider-level practice patterns data from selected states E. System Surveillance Area Resource Compilation of Health personnel, health facilities, File other data utilization data; relationships between sources services provided and area resources available ADAMHA Alcohol, Drug, and Mental Health Administration AHCPR Agency for Health Care Policy and Research Census Bureau of the Census CDC Centers for Disease Control and Prevention DOL Department of Labor tal and social characteristics that have an impact on health and responsive- ness to medical interventions. These data sources are generally maintained by public health and human services agencies at the federal, state, regional, and local levels. Since the relevance and usefulness of these data sources extend beyond personal health services, they are not separately reviewed in this section. Below, we describe the principal sources of data for monitoring changes in health care. Vital Statistics and Surveillance Data Vital statistics provide the means to describe changes in the rates and causes of death and some major causes of morbidity in the population. Many but not all sources of vital statistics contain information that enable
MONITORING CHANGES IN HEALTH CARE 173 Child/Family Planned Overall Sample Sample Periodicity All hospitals in Data on Continuous selected states hospitalizations only; individuals not traceable Enrollment and Enrollees in 5 Continuous claims forms and states provider data from 5 states Data derived from Same data Continually multiple surveys specifically updated and administrative relevant for data systems children, e.g., births to teenagers; no individual-level data HCFA Health Care Financing Administration NICHD National Institute for Child Health and Human Development NCES National Center for Education Statistics NCHS National Center for Health Statistics analysts to detect differences (primarily concerning race and selected socio- economic characteristics) among major population subgroups in order to identify possible sources of inequity in the provision of services. Greater specificity in relating differences in rates of preventable mortality and mor- bidity to differences in the use of medical services depends on linking these vital statistics to other sources of information. In some cases, these other sources of information are vital statistics themselves. For example, linked birth-death records provide a means to examine the relationship between infant mortality (from death certificates) and prenatal care (often contained on birth records). In other instances, linkages must be made with other types of data. For example, when linked with the Area Resource File (a system surveillance dataset), vital statistics can provide information on the relationship between preventable mortality or morbidity and the level of resources available in the same geographic area. It should be noted, how-
174 INTEGRATING FEDERAL STATISTICS ON CHILDREN ever, that relationships that are uncovered by these types of analyses are âstructure-outcomeâ or ecological relationships. That is, they provide only circumstantial evidence of a relationship, since information to assess the âprocessâ by which the availability of resources influences the outcome is unavailable (Starfield, 1973). As noted below, one of the most marked deficits in existing information systems is the general absence of linkages that make it possible to determine the specific impact of different types of health resources on health. The major contribution of vital statistics to date has been to facilitate epidemiologic studies of the influence of some important sociodemographic characteristics on mortality and morbidity and the elucidation of time trends that are ecologically associated with the availability of system resources. With the establishment of a means of linking most surveys sponsored by the National Center for Health Statistics to the death certificates through the National Death Index, it is now possible to explore the relationship between an even wider variety of individual characteristics and death from different causes. Since most vital statistics are available in comparable form in other industrialized nations, international comparisons carried out at specific times or across time have been useful in identifying possible relationships be- tween major health system characteristics and vital events. If linkages between vital statistics and sources of information on practice characteris- tics were developed, it would be possible for the first time to evaluate the direct impact of health services on major aspects of health. Surveillance data, such as immunization surveys, case registries, the Pregnancy Risk Assessment Monitoring System, the Behavioral Risk Factor Surveillance Study, and the Youth Risk Behavior Surveillance System, are generally most useful when certain types of problems arise and are of great concern. They are helpful in pointing to directions for changes in public policy, including the need for new funding and more targeted delivery of specific services. Since they are limited to information on the specific issue of concern, they have limited long-term applicability to overall health sys- tem reform. As the development of data systems for health care reforms progresses, it might be possible to adopt data elements from these surveil- lance activities for routine incorporation into ongoing surveys, reserving the full surveillance surveys for detection of long-term trends by administration every few years. Population Surveys Almost all national population-based health surveys focus heavily on health status and utilization of health services. Household surveys com- monly obtain information on number of hospitalizations and physician vis-
MONITORING CHANGES IN HEALTH CARE 175 its, and some collect information on use of other professional services. Most also obtain information on self-reported health status, and at least one survey provides physical examination data. Some also address insurance coverage and enrollment in or receipt of services from special programs that are government-administered or administered privately. Less frequently, household surveys are fielded to collect information on total and out-of- pocket expenses for health services. Despite the apparent duplication of information obtained by different agencies in their survey efforts, few surveys attempt to obtain information concerning the types of health services that are sought (especially with regard to the distinction between primary and consultative services), the reasons for individual visits to physicians and other health personnel, the characteristics of those visits as reported by individuals, and the perceived outcomes of those visits. Most surveys are cross-sectional, making it im- possible to track health status or experiences with the health system over time to determine changes that might be related to use of services. Further- more, the validity of specific questions concerning health status is largely untested. When the National Health Interview Survey was first initiated, major efforts were made to determine the validity of information about reported doctor visits and hospitalizations. For the most part, however, questions concerning reported chronic conditions among children are of unknown accuracy, as are questions on acute disability and limitations of activity. Although a major methodological study comparing household sur- vey data and medical record data on chronic conditions was recently con- ducted for adults, children were excluded from the study (Edwards et al., 1994). Until methodological studies begin to include children, survey re- sults will remain of unknown accuracy. Health surveys in the era of health care reform must be able to discern the impact of health services on perceived and documented health status. Health status, as determined from surveys, is composed of many different characteristics reflecting diagnosed conditions (âcondition listsâ), acute dis- ability (as determined by restrictions on usual activities due to acute or chronic conditions), and long-term limitations of activity due to chronic conditions. Only infrequently is the presence of symptoms ascertained. Mental health symptoms, in the rare instances in which they are elicited, are generally treated as separate and distinct from physical symptoms and con- ditions, despite the recognition among mental health professionals that they are interrelated (Leginski et al., 1989). Indeed, even when data on mental and physical health are collected in the same survey, it can be impossible to assess joint occurrence. This was the case with the 1988 National Health Interview Survey on Child Health. Useful questions were asked about physical, emotional, developmental, and learning problems, but different reference or
176 INTEGRATING FEDERAL STATISTICS ON CHILDREN recall periods precluded assessment of joint occurrence of physical and nonphysical problems. No systematic efforts have yet been made, in the conceptualization of child health status in the national health surveys, to put these various char- acteristics together in a composite profile or index of child health, or even to examine the relationships among the various components of health status. Since it is unlikely that health care reform would be expected to improve any of the specific components of health status as obtained in existing surveys, summary measures that would allow for the detection of changes in the composite of health characteristics would be more useful. Fortu- nately, researchers in child (as well as adult) health status are now develop- ing methods that will make such an approach feasible (Starfield et al., 1992). Monitoring health care reforms will also require more in-depth infor- mation on access to care, use of health services, and satisfaction with care than has previously been available in national surveys. The new Family Resources Supplement to the National Health Interview Survey represents a major step forward in this area. This module contains a number of useful questions on usual sources of care, characteristics of the source of care and caregiver, satisfaction with services, and delays in obtaining care. Since most reform efforts explicitly recognize that primary care is a central com- ponent, efforts to develop assessments of the adequacy of primary care will also be critically important. Although the Family Resource Supplement includes questions designed to distinguish primary care services from spe- cialty care services, other national surveys do not. Therefore, the design and implementation of standardized methods to determine childrenâs expo- sure to and experiences with primary care are of high priority. Provider Surveys Historically, survey data from practitioners and facilities have been used to provide information about the characteristics of hospitalizations and ambulatory care visits, primarily those to office-based physicians. New surveys sponsored by the National Center for Health Statistics have ex- panded the range of facilities and services studied to include ambulatory surgery centers, home and hospice care providers, and ambulatory care pro- vided by hospitals. However, since the information obtained typically con- cerns individual encounters, it is not possible to link hospitalizations or visits made by individuals from one time to another or from one provider to another. Therefore, elucidation of differences in practice patterns is not possible. These types of data are primarily useful in providing information about the challenges of practice, e.g., information on the distribution and types of problems encountered in practices of different specialists, and a
MONITORING CHANGES IN HEALTH CARE 177 major contribution of such surveys, particularly the National Ambulatory Medical Care Survey, has been to elicit the nature of the problem, as per- ceived by the patient, that is responsible for initiating the visit. (However, the problem is generally recorded by the physician and thus may not accu- rately represent the problem just as the patient views it.) Refinements of these datasets so that they contain coded patient identifiers to enable link- ages between events, and the encouragement of the use of coding systems to facilitate the analysis of patientsâ presenting problems and their relation- ships with the ensuing processes of care, will contribute in a major way to improving their utility for exploring differences in practice patterns, for clinical research, and for provider education. Another concern pertains specifically to children. Due to limited funds, the sponsoring agency for datasets such as the National Ambulatory Medi- cal Care Survey often makes use of supplementary funds provided by other agencies to enrich the dataset. Since these agencies generally heavily rep- resent adult interests rather than those of children, data obtained from this and other surveys have generally contained more information about care provided for adult problems than for childrenâs problems. Better balance between adult and childrenâs interests is warranted as the era of health care reform emerges. Administrative Records Although there are some notable attempts to compile and use adminis- trative data concerning the provision of health services (for example, the databases maintained by Systemetrics/Medstat and the United Health Care Corporation), there are some major problems in using existing administra- tive record data for health care reforms. One notable problem concerns the identification of the âdenominator.â Data are available for beneficiaries (i.e., those covered by health plans), but they exclude family members (of- ten children) who are not covered. The identity of these nonbeneficiaries is generally unknown. Second, the most commonly used datasets are restricted to beneficiaries of corporate plans, and their generalizability to persons insured individually or in small groups is questionable. In other types of plans, ownership determines control over accessibility and dissemination of the data. Consequently, many administrative datasets are not available for surveillance of costs, access and quality. None of the administrative data- bases are currently able to provide adequate denominators, or to focus on per-capita total expenditures and utilization (rather than prices and indi- vidual claimsâGaus et al., 1993). Third, the completeness and quality of data are uneven as a result of unstandardized procedures for collecting them at the original source in health plans. Fourth, the data lack elements that would be critical in accomplishing the aims of health care reform. In
178 INTEGRATING FEDERAL STATISTICS ON CHILDREN particular, the data sources rarely contain information on presenting prob- lems and baseline health status, against which improvements resulting from health services interventions could be judged. The usefulness of existing administrative data is also compromised by inadequacies in current coding systems for ambulatory care diagnoses and problems. The International Classification of Diseases, which was devel- oped primarily to code causes of death, is not well suited to classifying the problems in ambulatory care, especially in primary care, in which so many of the health problems cannot be resolved to formal diagnoses. Practitio- ners in many other countries have adopted the International Classification of Health Problems in Primary Care (ICHPPC) to facilitate the recording and analysis of information about the diagnoses that are managed in pri- mary care, and the International Classification of Primary Care (ICPC) builds on the ICHPPC but adds the capability to code and classify present- ing problems as well. To date, however, administrative databases in the United States have not attempted to adopt or adapt these primary-care- oriented classification systems. The inadequacy of systems of classification and coding has been par- ticularly problematic in the area of mental health problems. The develop- ment of the Diagnostic and Statistical Manual (DSM) has been a major advance for the practice of psychiatry, since it specifies criteria for the full range of diagnoses that occur in mental health services. It has not, how- ever, been useful in primary care. Work currently under way will lead to the development of DSM-PC, an adaptation suitable for primary care. Com- panion efforts will also lead to the development of a version for children, which should be rapidly tested, promulgated, and considered for use, along with the ICPC, in clinical and administrative data. Administrative data sources also lack information on the disposition of visits, particularly with regard to referrals for consultation (secondary care) and highly complex services (tertiary care). They also lack information on the type of professionals who deliver the servicesâinformation that is es- sential in understanding the appropriate roles for different types of primary care providers and specialists. The adoption and promulgation of a national standardized ambulatory care minimum dataset, with carefully defined elements are long overdue. First proposed in 1980, such a dataset is still under discussion. Responsi- bility for its delayed adoption is due at least in part to controversies over the importance and necessity of certain data elements, issues concerning the confidentiality of data about individual patients, individual providers, and individual health care plans (which may regard such information as propri- etary), and intensive lobbying by special-interest groups for inclusion of data elements of particular interest to them. Clearer specification of the goals of health care reform, with regard to its impact on effectiveness and
MONITORING CHANGES IN HEALTH CARE 179 efficiency of care, may help to specify those data elements that are most important. Administrative databases frequently do not provide identifiers that en- able the linkage of family members, and sometimes do not even contain identifiers that permit linkage of different health care events experienced by individuals. For example, hospital discharge data provide counts of hospi- talizations and the characteristics associated with them, but they do not provide data on the number of individual people experiencing those hospi- talizations or the characteristics of people who have more than one hospital- ization. Despite the many limitations in administrative data, the potential of these types of data for contributing to knowledge and policy is amply dem- onstrated by the data systems associated with the Medicare program. For example, it is now possible to link the Medicare Current Beneficiary Survey with Health Care Financing Administration claims payment files and with the National Death Index, Social Security records, and the Area Resource File (Wunderlich, 1992). System Surveillance Data The Area Resource File is the most commonly used source of informa- tion on system characteristics. It is a compilation of health-related data for each of the 3,070 counties in the country. Since the dataset contains census data on the size and characteristics of the population, rates can be calcu- lated and compared over time and across areas. Data are derived from many sources, including the American Medical Association, the American Osteopathic Association, and the American Hospital Association; also in- cluded are data on health maintenance organizations in each county. Health care reform is likely to pose additional challenges to system surveillance. Increasing formalization of health services delivery arrange- ments will make it possible and necessary to collect data on deployment of personnel by type and with regard to the characteristics of the populations served. Ensuring adequacy and equity of services will require attention to the balance between specialists and primary care practitioners in individual health plans; currently, such information is not available. Comparisons of health care risks, both for financial planning as well as for evaluation of effectiveness of care, in different populations will require improved case- mix measures applied in the same way in all areas. Data sources such as the Area Resource File could be uniquely valuable in health care reform research by producing knowledge about the relation- ships between the structures, processes, and outcomes of health services and by exploring the trade-offs between access, quality, and costs of care. Since these types of data sources provide critical information about many
180 INTEGRATING FEDERAL STATISTICS ON CHILDREN aspects of the structure of services and some important aspects of health status at the population level, they can play an important role in efforts to understand the impact of system characteristics on provider practice charac- teristics, on peopleâs use of services, and on the impact of health services on health outcomes and expenditures. CONSIDERATIONS IN DESIGNING A STRATEGY FOR MONITORING CHANGES IN HEALTH CARE FOR CHILDREN AND FAMILIES As demonstrated in the previous section, the nation possesses a rich set of vital statistics and surveillance data, survey data, administrative record data, and systems surveillance data that can serve as building blocks for a system to monitor health care reforms. Since these databases were not originally designed to monitor the performance of the health care system and its impact on health, modifications are needed. Below, we outline what we believe to be essential attributes of an effective system for monitoring health care reforms and summarize, in a general fashion, what changes need to be made to create a monitoring system with these attributes. The at- tributes include comprehensiveness, timeliness, ability to meet descriptive and analytical needs, ability to assess change, ability to measure short-term and long-term effects, provision of adequate geographic detail, capacity to assess effects on vulnerable populations, flexibility to address emerging issues, and capacity to integrate information within and across data bases. Comprehensiveness The framework presented in Table 1 suggests several domains of data needed for a comprehensive assessment of the impact of health care reforms on children and families. These domains include access, utilization, expen- ditures, health status, health resources, quality, and effectiveness of care. Existing data sets are generally designed to provide information on one or two of these domains, and in rare cases three or four. Consequently, assess- ing effects of reforms across multiple domains requires integrating data- bases (as discussed below). Since both public and private databases are needed to assess the effects of reform across these domains, a public-private partnership will be required. The effects of health care reforms are also felt at multiple levels. At the population level, health care reforms affect individuals, families, and communities. Within the health care provider community, health care re- forms have effects at the level of the individual practitioner or facility, the health plan level, and at the system level. In monitoring reforms, it is important to describe and document changes in each of these areas. How-
MONITORING CHANGES IN HEALTH CARE 181 ever, existing databases (with the exception of systems surveillance data) are designed for the most part to provide information at the level of the individual patient or consumer. They are generally less suitable in their current form for assessing outcomes aggregated at the level of families, communities, and systems of care. Additional efforts will be required to assess outcomes at these aggregate levels. For example, there is a need for family-level identifiers in all household-based health surveys. In some surveys, such as the National Health Interview Survey, it is currently very difficult to identify records for parents of sample children, making family- level analysis all but impossible. Other surveys, such as the Current Popu- lation Survey, include family-level records and identifiers. Yet even in this dataset, it can be difficult to identify relationships among family members, especially in multigenerational households. Much more work is required to assess effects at the community level. As a starting point, geographic identifiers are needed in national surveys to link systems surveillance data, such as that available in the Area Resource File, to population data. In the past, strict interpretation of confidentiality provisions have precluded such linkages. Although confidentiality must be respected, a review of policy in this area is called for. Timeliness In gauging the impact of health care reforms, policy makers, research- ers, advocacy groups, and others will require timely information. Speedy acquisition, processing, and release of data are also essential if the monitor- ing system is to provide an early warning of problems or unintended conse- quences of reform. Hence, the implementation of reforms will increase pressure on sponsoring agencies to speed the collection, processing, and release of data. The pressure will be greatest for the larger, more complex survey efforts, such as the National Medical Expenditure Survey. This survey involves multiple personal interviews, diaries, and verification of expenditures through record checks. Partly as a consequence of the enor- mity of this undertaking, the 1977 and 1987 editions of this survey were largely outdated by the time processing was completed and data were re- leased. Trade-offs are always present between precision of data and production time. These competing needs must, of course, be balanced. One approach to addressing this issue is to develop a monitoring system that combines quick turnaround descriptive data for surveillance purposes with more pre- cise but necessarily more slowly produced data for analytic and evaluative purposes. Taking this approach, data from large-scale population and pro- vider surveys would be used for in-depth analysis and evaluation of health care reform. A complementary tracking system, based on rapid turnaround
182 INTEGRATING FEDERAL STATISTICS ON CHILDREN data sources, would be created to provide an early warning of problems and unintended consequences. The tracking system should be designed to pro- vide rapid release information on a continuous basis as reform is imple- mented. Information for the tracking system could be derived from a com- bination of administrative data, public health surveillance data, and telephone survey data. Administrative data would be derived from health plans, claims payment systems and other sources. Public health surveillance data would come from existing and new surveillance efforts organized by the Centers for Disease Control and Prevention, as well as state and local public health agencies. Telephone surveys would provide immediate data on access bar- riers, satisfaction with care, and other indicators not available from admin- istrative and public health surveillance systems. A capacity to conduct rapid turnaround telephone surveys is now being developed at the National Center for Health Statistics (NCHS, 1993). However, new funding would be required to implement such a program. Meeting Descriptive and Analytical Needs It is necessary for the monitoring system to provide information for both descriptive and analytical purposes. The objective of descriptive stud- ies is to describe a phenomenon of interest, whereas the purpose of an analytical study is to explain the phenomenon (Aday, 1989; Moore, 1993). Descriptive data are needed to profile the status of children, families, and communities as well as providers and systems of care under health care reforms. Data are also needed for analytical purposes, such as unraveling relationships among variables or explaining trends. For example, the fed- eral and state maternal and child health agencies may be interested in as- sessing whether and how patterns of care change for chronically ill children after health care reform is implemented. A descriptive profile of utilization patterns could be created from population and provider surveys and admin- istrative records. If chronically ill childrenâs utilization patterns were found to change substantially after implementation of reform, there would be a clear need to move beyond description of the changes to an explanation of whether the changes were caused by reforms or other factors, and which specific factors were responsible for them. Data for both descriptive and analytical purposes are essential, but da- tabases must be designed carefully if they are to meet both needs. Typi- cally, more depth is needed within a topic area for analytical studies, whereas descriptive studies are often facilitated by fewer, carefully chosen mea- sures. Hence, trade-offs can exist in meeting descriptive and analytical needs. One of the objectives in designing a monitoring system is balancing these needs.
MONITORING CHANGES IN HEALTH CARE 183 Capacity to Assess Change Assessment of change is fundamental to monitoring the impact of health care reform. Ideally, a monitoring system should have the capacity to provide information on changes occurring for children, families, and com- munities as well as providers and systems of care across multiple domains (access, utilization, expenditures, health status, health resources, quality, and effectiveness of care). Two analytical approaches can be taken to measure change. First, data collected on a periodic basis can be analyzed and compared over time. For example, health insurance data from the Current Population Survey are fre- quently compared over time to assess changes in the number and proportion of uninsured persons. For such comparisons to be meaningful, at least certain items in the questionnaire must remain constant, since even small changes in questionnaire wording can have significant effects on survey responses. An example of this problem can be seen in the 1981 and 1988 Child Health Supplements to the National Health Interview Survey. Both supplements contained a checklist of child health conditions. Although the wording for many condition entries remained the same, the wording for others changedâprecluding comparisons of prevalence over time. Although it is sometimes possible to estimate the impact of wording changes when questionnaires are changed, doing so is difficult and requires special atten- tion. The second approach consists of assessing changes over time within a panel of subjects (e.g., children, families, practitioners). In this case, the same subjects are followed with measures of interest collected at regular intervals. Panels can be created from samples obtained from vital statistics (e.g., birth records), administrative records (e.g., claims records ), or survey samples. The panel approach is used most often in the survey context. Although a powerful tool for assessing and analyzing change, panel surveys have been used only occasionally in child health. The most well- known panel surveys, including the Survey of Income and Program Partici- pation and the National Longitudinal Survey of Youth, have limited health content. The few examples in the child health field include the National Survey of Family Growth and the National Maternal and Infant Health Survey, both sponsored by the National Center for Health Statistics. With the exception of these surveys, which have very specific purposes and lim- ited target populations, there are no ongoing panel surveys of childrenâs health. Two promising possibilities for panel surveys are now being planned that could be of enormous value in monitoring health care reforms. First, the National Center for Health Statistics is considering conducting a child and family survey as part of the National Health Interview Survey. This
184 INTEGRATING FEDERAL STATISTICS ON CHILDREN comprehensive survey of child and family health would be modeled after the very successful 1988 National Health Interview Survey on Child Health. Like its predecessor, the new survey could collect in-depth information on pregnancy and birth, injuries, impairments, acute and chronic conditions, developmental and learning and behavioral problems, use of health ser- vices, and participation in health programs such as the Supplemental Food Program for Women, Infants, and Children, Medicaid, and Healthy Start. In addition, the planned survey would include several indicators of family functioning. If a panel or follow-up component was added, the survey could be an extremely valuable tool for assessing the effects of health care reforms on children and families. However, it is not clear at present whether even the baseline for this important survey will be fielded. The costs of this survey must be met externally by agencies interested in the survey results, and so far only limited funding has been committed. Unless additional funding becomes available soon, survey planning will be abandoned. The second promising panel survey is the Survey of Program Dynam- ics, now being fielded by the Bureau of the Census. This survey includes a panel of 20,000 households first assembled in 1993 as part of the Survey of Income and Program Participation. The households will be followed over a 10-year period, with household interviews conducted annually or more fre- quently. As envisioned by its planners, the survey would focus on provid- ing an information base for policy makers and researchers interested in welfare reform and health care reform issues. Initial funding has been provided by federal agencies interested in welfare reform, but so far no funding has been provided for the health care reform component. Capacity to Measure Short- and Long-Term Effects Health care reforms are likely to bring about a myriad of changes in the organization, financing, and delivery of care. Some of these changes will begin occurring during implementation or shortly after reforms are imple- mented; others will occur much later. For example, changes in access and utilization of care could be expected to occur soon after implementation of a reform. In contrast, changes in child health status, if they occur, are likely to become apparent only years later. Hence, the monitoring system must be capable of capturing changes occurring over the short and the long term. As reforms are implemented at the state level, there are likely to be policy adjustments in eligibility, benefits, cost-sharing, and other compo- nents especially during the early years of reform and perhaps on an ongoing basis, if past experience with Medicaid and Medicare can serve as a guide. Consequently, the âeffectsâ of reform are likely to change over time, as policy makers fine-tune various elements of the health care reform plan.
MONITORING CHANGES IN HEALTH CARE 185 The monitoring system should then be designed with recognition of the temporal and ongoing nature of change. This means a monitoring system based on a simple âbefore and afterâ design will provide only partial and perhaps misleading conclusions concerning the effects of reform. Instead, the monitoring system should be conceived as an ongoing activity with continuous collection of outcome data. Provision of Adequate Geographic Detail Most major federal health data bases were originally designed to pro- vide national or regional level estimates. State-level data will be critical in evaluating health care reforms, since reform is likely to be implemented differently in each state (and perhaps only in selected states). Indeed, as weâve said, even without national health care reform, states are already beginning to implement their own reform agendas. Consequently, state- level data represent a key component of the monitoring system. All states maintain vital records and many maintain data collection systems for hospi- tal discharge information. Claims payment systems are also maintained by every state Medicaid agency, although not all lend themselves to analysis. A few states have also developed household survey databases (Hawaii, Rhode Island, Puerto Rico), and all states participate in the behavioral risk factor surveys sponsored by the Centers for Disease Control and Prevention. However, no state currently has a data system in place that is capable of assessing the broad range of effects shown in Table 1. Moreover, none of the major federal health surveys currently has the capacity to assess the effects of state-level reform efforts on children and families. Two survey design considerations are particularly relevant to producing state-level estimates from national surveys. First, the sample design, par- ticularly the selection of primary sampling units, must be developed in a manner that is consistent with state-level estimation. In the past, major health surveys, such as the National Health Interview Survey, have selected primary sampling units that overlap state boundaries or have excluded some states entirely. This is now changing at least in some surveys. For ex- ample, the 1995 redesign of the National Health Interview Survey will in- clude only primary sampling units that do not cross state boundaries, and all states will be represented for the first time. Second, there must be a suffi- cient number of cases at the state level to permit accurate estimation. Even in large household surveys, such as the National Health Interview Survey, there are insufficient numbers of observations in all but the largest states to permit useful state-level estimates. To a degree, this problem can be cir- cumvented by combining multiple years of survey data. However, this strategy will have limited applicability to small states. Other approaches,
186 INTEGRATING FEDERAL STATISTICS ON CHILDREN such as supplementing household interviews with telephone interviews, need to be considered. Over the long term, federal efforts should also be directed toward im- proving capacity at the state and local levels to conduct surveys, make better use of claims data and other administrative databases, and develop standards and conventions for sharing data between levels of government. The National Center for Health Statistics, with its long history of cooperat- ing with states on the development of uniform standards for vital statistics, may be well suited to leading this effort. With health care reforms in place, there will be a much greater need for coordinated information at the local, state, and national levels. As indicated earlier, data collection and retrieval need to be coordinated so that data collected at the local level can be aggregated to the state and federal level, and so that data collected at higher levels can be related to data collected at local levels. Doing so will permit comparison of effects across communi- ties, states, and the nation. Capacity to Assess Outcomes for Vulnerable Populations The monitoring system should have a built-in capacity for assessing outcomes for vulnerable child populations, including children living in poor and near-poor families, minority children, children in out-of-home place- ments, and children with disabilities and other ongoing health problems. These vulnerable populations are likely to be affected more signifi- cantly by health care reform than other populations. Consequently, there is a heightened need to closely monitor their welfare as reforms are imple- mented. Yet it is difficult to identify these children in vital records, admin- istrative records, and public health surveillance systems. In surveys, sample sizes for these populations are often too limited for meaningful analysis. In the past, sample size problems have plagued analysis of health survey data for minority populations. During recent years, survey designers have im- proved capabilities in this area through oversampling of minority popula- tions. For example, the National Health Interview Survey currently oversamples blacks and will begin oversampling Hispanics in 1995. However, difficul- ties remain in analyzing other vulnerable populations. Two populations of particular concern are children residing in institu- tional settings and homeless children. Virtually all ongoing health surveys exclude children residing in institutions. Even the 1990 decennial census data on the institutionalized population lacks the specificity needed to esti- mate the number of children residing in institutions for health-related rea- sons. Even less information is available on homeless children and families. This population is routinely excluded from national household surveys and other data systems, even though local surveys have demonstrated significant
MONITORING CHANGES IN HEALTH CARE 187 unmet health needs exist for homeless children and families (Wood et al., 1990a, 1990b). Emerging Issues It is not possible to accurately predict the direction or timing of health care reform, or its effects, especially unintended consequences. As a result, flexibility in data collection will be necessary. During recent years, many ongoing surveys have adopted a modularized approach, whereby a core questionnaire is supplemented with modular questionnaires on topics of current interest. This provides an effective means for maintaining flexibil- ity in data collection. Unfortunately, the fielding of a âsupplementâ often depends on the availability of special funding for it. As a result, supplements generally reflect the interests of specific groups that have the resources to offer. For this reason, children have only infrequently been the subject of these added modules. To adequately monitor the effects of health care reform on chil- dren and families, an in-depth child and family supplement should be con- ducted at least every five years and preferably every three years as part of the National Health Interview Survey. Doing so would provide the flexibil- ity needed to address emerging issues for children and families. Integrating Data Collection Efforts Over time, federal databases have become increasingly specialized and categorical in nature. The result is that few databases contain data across the multiple domains relevant to health care reform (access, utilization, expenditures, health status, health resources, effectiveness, and quality of care). Consequently, the monitoring system will be required to draw on multiple data sources. Its capacity for assessing the effects of reforms will be enhanced to the extent that data from the different databases can be integrated. Integrating or linking data from various sources offers the added advantage of increasing the utility of existing data bases at a relatively modest cost (AHCPR, 1991). Integrating data can occur at several levels. At the most basic level, databases should share common definitions and terminology whenever pos- sible; unfortunately, this happens less frequently than is desirable. For example, the National Health Interview Survey, the National Medical Ex- penditure Survey, the Current Population Survey, and the Survey of Income and Program Participation use different questions for assessing health insur- ance coverage. This makes it very difficult to compare results across sur- veys, or even to develop agreed-on estimates of the size of the insured and uninsured populations. Even commonly used terms such as access have
188 INTEGRATING FEDERAL STATISTICS ON CHILDREN imprecise meanings and varying representation in data collection efforts. The National Committee on Vital and Health Statistics should perhaps be given the responsibility for coordinating efforts to define important terms so that they can be measured in consistent ways. Linking data across databases can yield significant benefits but presents additional challenges. Recently, much attention has been focused on link- ing administrative records such as claims payment data to national survey data. For example, data on utilization and expenditures from Medicare administrative files are linked to population-based survey data in the Medi- care Current Beneficiary Survey. Similarly, the National Death Index now permits linkage of death certificates to most of the population surveys spon- sored by the National Center for Health Statistics. The challenge ahead lies in integrating databases across federal agencies as well as between public and private data collection organizations. Linkages of administrative data with survey data could provide a pow- erful mechanism to explore the relationship between services provided and resulting health status in the geographic area served by health services orga- nizations and plans. Accomplishing this level of integration requires close cooperation among the sponsors of the databases. This has not always been possible in the past because of confidentiality issues, as well as concerns such as funding and turf. Health care reform will mean that new efforts toward integrating databases will be essential. CONCLUSION A carefully designed data collection and analysis strategy is central to assessing the impact of the changing health care system on children and families. In previous sections of this paper we presented a framework for assessing data needs, reviewed existing databases, and articulated the at- tributes of an effective monitoring system. All of this needs to be consid- ered within an overall strategy for monitoring health care reforms. Too often data collection and analysis efforts are divorced, resulting in missed opportunities and inefficient use of resources. Data collection strategies must be developed in conjunction with a data analysis strategy. The data collection and analysis strategy should be driven by a clear understanding of the key health care reform issues. The specifics of that strategy and the monitoring system that will support it should be developed using a process that takes into account the needs and views of policy makers, the scientific community, child health advocates, and others concerned with the well- being of children. The strategy of planning for data to monitor and assess changes in health care faces four types of challenges:
MONITORING CHANGES IN HEALTH CARE 189 â¢ Concept development. As the country moves toward primary care and toward outcome and functional status assessment, concerted attention will be needed to clarify the meaning of these concepts, especially as they relate to children, so that approaches to measurement can be developed. â¢ Standardized measurement. At present, the many different data sys- tems and data sets use different criteria to measure the same concept. Im- portant goals of health care reform, such as improving access and quality of care, are currently being measured in many different and incompatible ways. Once the concepts of access, quality, primary care, and health and func- tional status are clarified, attention must be devoted to developing consis- tent or at least compatible ways of measuring them. â¢ Personal identifiers with maintenance of confidentiality. Since per- son-focused (rather than disease-focused) health care requires the tracking of events over time and places for each individual in the population, a method of assigning unique personal identifiers that maintains confidential- ity will need to be developed and instituted. â¢ Responsibility and accountability. Timely and accurate information is needed to ensure that changes consequent to health care reforms proceed in the anticipated direction. In an ideal world, a system for monitoring health care reforms would be designed and implemented without regard to budget constraints. Realisti- cally, limited additional funds are likely to be available for the monitoring system. Consequently, there is a need for careful consideration of the marginal costs and marginal benefits of adding components to the monitor- ing system. As much as possible, the system should rely on existing data bases, modifying them when necessary but keeping in mind their original purposes. Fortunately, the nation already possesses a rich set of health databases that can be used for monitoring health care reforms. With careful planning and coordination, the vast number of current surveys might even be reduced, thus releasing resources that could better be applied to obtain- ing data of better quality with more rapid availability. Such planning will be required to develop a system capable of accurately and economically assessing the effects of health care reform in a timely fashion. REFERENCES Aday, L. 1989 Designing and Conducting Health Surveys. San Francisco: Jossey-Bass Inc. Agency for Health Care Policy and Research (AHCPR) 1991 Report to Congress: The Feasibility of Linking Research-Related Data Bases to Federal and Non-Federal Medical Administrative Data Bases. AHCPR Pub. No. 91-0003. Rockville, Md.: AHCPR.