A Review of the National Access-to-Care Surveys

A Review of the National Access-to-Care Surveys By Marc L. Berk and Claudia L. Schur

Chapter Three documents the strategies used over time by Foundation grantees tracking access to health care among Americans. The four national surveys of access to care that are discussed represent the Foundation's flagship investment in analysis related to problems of access. One goal of these surveys has been to gain an understanding of the extent and nature of the problem. An even more important goal has been to focus attention on the problems faced by individuals who cannot get basic health care services.

The chapter avoids reporting findings or trends from the surveys, which have been published in a large number of academic papers, Foundation reports, and monographs. Instead, the chapter examines the evolution of the surveys and the shifting aims and priorities over the past twenty years. It highlights the challenges survey researchers face in studying access to health care, and it explains how approaches to measuring access have also evolved over the past two decades.

This chapter introduces readers to the logic of the Foundation's efforts in this area. It offers a guide to interpreting access-to-care research and puts the surveys supported by the Foundation into the context of other data-collection efforts that have measured service use, access, and expenditures on health care over the years.

The authors, Marc L. Berk and Claudia L. Schur, are the director and deputy director, respectively, of the Center for Health Affairs at Project HOPE. They have directed the Foundation's most recent access-to-care survey and are publishing a series of papers and reports on findings from the 1994 survey. Meanwhile, fieldwork on the latest Foundation-supported access survey was completed as this book was being printed. The 1997 approach to measuring access allows for assessments of access problems in twelve distinct communities along with the nation as a whole. The community focus of the newest survey reflects the growing importance of understanding how emerging market dynamics--which vary from community to community--affect access to care among vulnerable populations.

Even after the extensive debate over the reform of the health care system in the early 1990s, the basic issue of universal health care as a fundamental right of Americans remains unresolved among our nation's leaders and among our citizens. Media reports about people who lack access to medical care may push emotional buttons--the elderly man left unattended for hours in an emergency room, say, or the poor pregnant woman who receives no medical attention until she goes into premature labor and delivers a one-and-a-half-pound infant--but only occasionally do they have an impact on public policy. It has long been known that what policy makers need is not anecdotal evidence like this but hard, reliable information on access to medical care.

In a 1980 foreword to the classic book on access to care, David Rogers and Linda Aiken of The Robert Wood Johnson Foundation wrote of the lack of availability of such information:

Despite the large efforts made during the preceding decade to improve medical care for Americans, there was startlingly little solid information on the extent of the remaining access problem, on who was having difficulty obtaining medical care, or on whether the multitude of programs initiated during the 1960s had been successful. The studies then available were primarily head counts on the numbers of doctors within geographic areas, or simply records on mortality and morbidity within different regions. Although these statistics had some usefulness for overall planning, they did not provide the kind of information needed to design targeted strategies for getting underserved people better health and medical care.¹

Acknowledging the necessity of having objective, systematic, and scientific data available for policy making, the Foundation launched its first national survey of access to care in 1976.

The survey followed a number of other efforts (Table 3.1) aimed at systematically measuring access to and expenditures on health care for population subgroups and for the nation as a whole.² In the early 1930s, the Falk Commission on the Costs of Medical Care--an independent group of representatives from organized medicine, the government, academic social science, and the hospital and insurance industries--sponsored the first survey intended to produce estimates of national health care expenditures. During the next several decades, there were sporadic, small-scale, health-related survey efforts, but none were national in scope; nor were they focused primarily on health expenditures or access to care.

With legislators and health care professionals demanding information to use in making policy decisions, Public Law 652 was passed in 1956 to authorize the regular collection of health-related data "to produce statistics on disease, injury, impairment, disability, and related topics on a uniform basis for the nation as a whole." This survey was later renamed the National Health Interview Survey (NHIS), and it has been conducted annually since 1957. Until recent years, its primary emphasis was on collecting health status information and, to a lesser degree, the use of services. Only in the past decade has a major focus been on insurance status and other measures of access to care.

The next entry in the progression of national health surveys was the 1963 Health Care Use and Expenditure Survey, conducted by the Center for Health Administration Studies at the University of Chicago. It was followed by a 1970 survey of the same name and, subsequently, two expenditure surveys sponsored by the federal government (the 1977 National Medical Care Expenditure Survey and the 1987 National Medical Expenditure Survey). In contrast to the NHIS, these four surveys were designed to provide data that would help policy makers understand the effects of government financing programs on the use of services and expenditures, assess the distributional implications of programs and the availability of health care services, and identify and evaluate problems in access to care for different population subgroups.

The surveys conducted after 1965 and into the 1970s attempted to measure changes in access brought about by major governmental expansion into health care: the Medicare and Medicaid programs. As federal expenditures climbed and government became a major purchaser of health care, costs became the primary focus and access was relegated to a secondary consideration. The emphasis on cost necessitated several design changes between the 1970 Health Care Use and Expenditure Survey and the 1977 National Medical Care Expenditure Survey. Multiple interviews and shorter recall periods--requiring respondents to remember events over a three- to four-month period rather than a year--facilitated the collection of detailed cost and payment data. In addition, components were added to verify the charge and use data of medical providers and to obtain information from health insurers on health benefits not available from the household; these numbers were used to support estimates of national health care expenditures and sources of payment by a variety of population subgroups.

When the first Robert Wood Johnson Foundation access survey was undertaken in 1976, however, little of this groundwork had been laid. In addition to asserting the need for hard empirical evidence, the Foundation, along with the National Center for Health Services Research, supported what remains the seminal work in establishing well-defined and widely accepted means of measuring access to care. These measures of access were developed by Ron Andersen and Lu Ann Aday as part of a framework for studying access to health care. The measures were tested on the 1976 access-to-care survey. While this was the fifth in a series of health care surveys undertaken by the University of Chicago, the first four emphasized families' health care experiences and costs, whereas the 1976 survey concerned individuals' access to the health care system. A major focus of the study, as voiced by the Foundation, was to "carefully describe the exact subgroups of our population who continue to have difficulty in getting medical care." Thus, there were twin themes of measuring access and charting the equity of access across different population subgroups.

The first two were intended to reflect potential access to medical care, while the last two indicators measure realized access. Behind this framework was the assumption that public policy could be used to affect realized access through changes in potential access. Suppose, for example, that policy makers learn that people who live in areas with few health professionals have fewer health care visits than those who live in areas where health care providers are plentiful. Within the Aday-Andersen framework, the number of health care visits made by different population subgroups is an indicator of realized access; in order to increase the number of visits for those in specific geographic areas, policy makers might try to increase the supply of health care providers in those areas. So, too, altering a part of the health care delivery system, for example, reorganizing the financing and delivery of services, might affect realized access to care.

This broad framework has supported much of the policy research related to access to care over the last twenty years. During this period, access to a wide range of services has improved, and for substantial portions of the population. Much of the overall improvement in the health status of Americans is attributable to increased access to care: more widespread use of Pap smears and pelvic exams has contributed to a decline in mortality from cervical cancer, and the expansion of 911 emergency medical systems has increased survival from out-of-hospital cardiac arrest, stroke, and trauma.

The original framework developed by Aday and Andersen for studying access to care remains in wide use today. Health services researchers continue to be concerned about the organization of health care systems and the availability of resources; about how personal characteristics affect interactions with the health care system; about the level of use of health care services; and about consumers' satisfaction with their care. Within these broad categories, however, emphases have shifted over time, and there has been a move toward measuring the effects of access on health outcomes.

The themes explored in the four Foundation-supported access surveys (1976, 1982, 1986, and 1994; see Table 3.2) have evolved to reflect society's most pressing problems. The methodological approaches have varied in keeping with goals, constraints on resources, and changes in the way data are collected. The first of the surveys relied on in-person interviews with eight thousand people. A number of groups were oversampled: persons with episodes of illness, nonurban Southern blacks, and Hispanics living in the Southwest. Substantively, the survey emphasized attributes of the respondents' usual source of care, including convenience and consumer satisfaction, illness-related symptoms, and the use of health care services.

While the study originally intended to provide baseline data for an evaluation of a specific health initiative, the larger goal was to develop empirical indicators of access that could be used to monitor changes in the study population. Such data could also inform policy makers about the population subgroups who faced the greatest barriers to access, and the data could provide a basis for analyzing how well existing programs succeeded in improving access.

The 1982 survey continued this general approach, focusing on whether the previous trends for the major indicators of access were continuing, which groups appeared to have the greatest problems, and the severity of certain problems that might be exacerbated by the political and economic climate.³ Emphasis on this latter objective was facilitated through an oversample of families with incomes below 150 percent of the federal poverty level. Attention continued to be focused on the usual source of care (including the site of care and waiting times); the use of hospital, physician, and preventive services; and insurance coverage or eligibility for public programs. A new area of concern--prompted by increasing health care cost consciousness combined with an economic recession--emphasized barriers encountered in obtaining care for a family member with a serious illness. Methodologically, the 1982 survey undertook one of the first highly visible applications of random-digit dialing techniques, using telephone interviews to reach its approximately sixty-six hundred respondents at a cost substantially lower than that of in-person interviews.

In their analysis of survey data, Aday, Andersen, and Gretchen Fleming wrote of "the special access problems created by long-term chronic illness,"⁴suggesting that even people with usually adequate third-party coverage were unable to pay for the long-term health care needs posed by chronic illness. With these findings in mind, the 1986 survey (consisting of telephone interviews with approximately ten thousand people) included an oversample of Americans with chronic and serious illness. The content of the survey continued to focus on the availability of a usual source of care; use of services; self-reports of health status, including the presence of serious health conditions; problems in paying for care; and satisfaction. In contrast to the findings from the two earlier surveys, results from 1986 indicated an overall decline in the use of medical care and increasing disparities in the use of services between the American mainstream and members of vulnerable population subgroups such as the poor, the uninsured, and minorities.⁵

Although a secondary goal of the 1994 survey was to monitor trends in access, the Foundation was primarily interested in taking a fresh look at both the methodology and the policy objectives of the survey. We were particularly interested in increasing the survey's analytic capabilities (larger sample sizes of populations of interest) without increasing data collection costs.

This was accomplished with a methodological approach employed in other data-collection efforts but never before for an access survey: using the National Health Interview Survey to target respondents who reported access problems. Some of these individuals were contacted for a follow-up interview that focused additional attention on their problems in gaining access to health care. Most interviews were conducted by telephone (in order to keep costs down); these were supplemented with in-person interviews of people without telephones or people who were functionally unable to be interviewed by telephone. The 1994 survey comprised three components: (1) a national probability sample, (2) a sample of persons who reported access barriers or who met other specific criteria suggesting low access to care, and (3) a sample of persons with one of two specific chronic conditions for which well-accepted standards of care exist.

Given the Foundation's continued interest in the organization of medical care and its effect on access, data on the usual source of care were collected, although it was deemphasized somewhat. The survey also examined predisposing characteristics, including age, race, sex, and attitudes about the efficacy of medicine. The ability to assess the role of health status in the use of services was emphasized through a new symptom-response sequence developed with separate funding from the Foundation. A focus on financial barriers to care was expanded through a series of questions on paid sick leave from work.

Over the last two decades, seemingly small changes in the emphasis of the access surveys are part of a larger pattern in the study of access to health care. While it remains important to analyze changes over time, social science research cannot be held hostage to old ideas. It must adapt to changing circumstances, even though comparability of particular questions over time is sacrificed. Researchers have become increasingly sophisticated in their approach to data collection. They are better able to target populations of interest and select appropriate methods to reach them. They have adopted a healthy skepticism about soliciting the opinions of our health care consumers ("Don't tell us what you think; let us observe what you do"). Faced with increasing budgetary pressures, our society is reluctant to increase access for its own sake. Instead, it demands to know how increases in resources and in the use of health care services affect health outcomes.

The most recent survey went beyond counting the instances in which people felt that they could not get care; it provided a context giving their reason for seeking care. In addition, the survey tracked the condition for which they sought care, whether they were eventually treated, and how they felt their health was affected by the treatment or lack of treatment.

Although collection of in-depth information on the circumstances surrounding unmet need was motivated by policy considerations, it was made possible by more sophisticated data-collection methods than had been previously used. In any nationally representative sample, only a small number of the respondents encounter serious access problems, so having a sufficient number of "cases" for analytic purposes is difficult. Increasing the sample size is costly and inefficient; for every person with access problems added to the sample, many other persons with no problems are added, and data collection costs rise prohibitively.

To overcome this problem, the 1994 survey used the National Health Interview Survey as a giant screener to identify people having low access. These people fell into one of the following categories:

With 120,000 respondents to the NHIS annually, we were able to select people who reported access problems in the NHIS for follow-up interviews, thereby increasing the sample size in a cost-efficient manner and concentrating interview time on those with the most serious access problems.

The public opinion polls have been awash in news of consumer dissatisfaction with the health care system; political careers have been made and lost gambling on just how widespread consumer dissatisfaction is, and pundits argue over the interpretation of the polling results. While the majority of people report dissatisfaction with the general state of health care, as individuals they are quite satisfied with the care they receive.⁶This presents something of an anomaly for policy makers: do they attempt to fix the health care system about which people complain or do they protect the status quo because most individuals are happy with their own care? As one of the present authors noted in a recent Health Affairs commentary, "the problem is not in the polls themselves, but rather in the unrealistic expectations that we have placed on the polling process. Polls are only one predictor of how people are likely to behave and, indeed, polls are probably not the most accurate."⁷

It is better, instead, to find out what people do rather than what they say. With the growing role of economics in the study of health policy, analysts are using the methods of what economists call "revealed preference": deducing an individual's preferences from observed choices or behavior. Thus, researchers collect data on people's health insurance coverage, their demographic characteristics and health status, and their medical care use and expenditures, and they make inferences about the underlying relationships and causal factors. In fact, from the perspective of trying to understand and explain why and how individuals use health care, opinion questions have never proved to be particularly good predictors of health care use.

Researchers have also begun to think more objectively about what can be learned from questions about satisfaction. Responses to these questions may tell more about the site of care or the expectations of the patient than about the quality of the care delivered. Some differences in satisfaction have been traced to variations in the supply of health resources; in these cases, individuals are responding to different circumstances surrounding the delivery of care. In reviewing findings on satisfaction over time, Howard Freeman and Martin Shapiro remark that "it is a reasonable hypothesis that the ambiance of physicians' offices and the validation of parking receipts may be equally as strong determinants of satisfaction as correct diagnoses and appropriate treatment."⁸

Thus, within the access surveys themselves, there has been a moderate deemphasis on subjective measures of access such as satisfaction with care and overall views about the American medical system, and an increasing emphasis on observable behavior. Certain questions were included in the 1994 survey to discover attitudes toward the efficacy of medicine; others elicited reactions to health care reform. In addition, the 1994 survey included a sequence in which respondents were asked whether they ever experienced a particular symptom and, if so, whether they sought medical care. With independent ratings of the seriousness of the symptoms from physicians, a more objective assessment could be made of care-seeking behavior.

Researchers who monitor trends are always faced with a dilemma: how to improve a survey and still make it comparable to surveys that have been done in the past? As more resources are invested in gathering data with which to make health policy decisions, more is learned about the difficulty of gathering reliable data. Seemingly small changes in survey design--minor alterations to the way the respondents are chosen, revisions in the wording of a question, and even changes in the ways interviewers are trained--can fundamentally affect the responses people give.

The designers of the Foundation-supported access surveys, as well as the designers of other major health surveys, must decide whether it is more important to measure changes over time or to obtain the most accurate estimate possible of how things stand at a single point in time. An example of this type of decision and how it affects analysts' ability to provide policy-relevant information is a change in the Current Population Survey (CPS), which, among other things, is used to make estimates of health insurance coverage and, in particular, to monitor growth or decline in the size of the uninsured population. In the late 1980s, the issue of whether CPS estimates of the uninsured population were "annual" or "point-in-time" was the subject of some controversy.⁹Because of the attention focused on the issue, wording changes were made; while these may have increased the reliability of the estimates since 1989, they have made the longitudinal study of patterns of coverage more difficult. To avoid these difficulties, many surveys, such as the National Health Interview Survey, place emphasis on maintaining continuity and make changes only where there is strong evidence that the current procedure is not working well; even then, changes are closely monitored to ensure the reliability of data.

When the 1994 access survey was designed, eight years had elapsed since the last survey. We convened a technical advisory panel of leading experts in survey methods and health policy. There was a consensus that while monitoring trends in access to care was important, preserving the ability to measure such trends should not come at the expense of accurately enumerating the problems in access to care that were being experienced in the 1990s. Accordingly, when better ways were found to ask questions, or to select our sample, we used them, even though doing so complicated our ability to do trend analysis.

Two examples illustrate this point. Previous access surveys used symptom/response questions to measure whether people are receiving care for specific symptoms they have experienced. The Foundation had already invested considerable resources to improve symptom/response questions. In the course of developing the 1994 survey, a greater effort was made to develop lists of symptoms for which there was a strong medical consensus about appropriate medical care. These questions (developed by Shapiro and Freeman) were substituted for those used previously. While this change makes comparisons with earlier surveys more difficult, it does provide a more valid indicator of differences in obtaining medical care across population subgroups.

A second example is our use of in-person interviews. The 1982 and 1986 access surveys used telephone interviews exclusively. Some studies have indicated that telephone surveys may exclude poor or disadvantaged people who do not have a telephone; these people are disproportionately likely to have access problems. In order to reduce the bias that might be associated with relying solely on telephone interviews, the 1994 survey conducted in-person interviews with those who had no phone or who had a disability that prevented them from using a telephone.

"Coming full circle" might best describe the evolution of the four access surveys' outlook on financial barriers to care. Health services researchers often make a distinction between financial factors that inhibit access (such as insurance coverage and income) and cultural or attitudinal factors (such as inability to speak English or lack of trust in physicians). While the first survey oversampled members of disadvantaged groups (Southern blacks and Hispanics in the Southwest), the focus was not primarily on their cultural differences but on their economic differences, and how the latter disparity influenced the use of medical care.

Most research indicates that financial barriers to care are better predictors of the use of services than variables that reflect attitudes about the delivery or the efficacy of medicine, or even indicators of cultural or language differences. Yet we wanted to account fully for all possible dimensions of access: a lack of health insurance and low income are probably the most important financial barriers faced by those seeking services, but we hoped to measure other economic obstacles to care. For example, an employee may have good health insurance but not have adequate sick leave to visit a doctor without loss of pay. Similarly, a child might be covered under a parent's insurance policy, but the parent cannot afford to take time off from work without financial consequences. Thus, we asked a series of questions about sick leave as well as the travel cost associated with seeking medical care.

In developing these questions, we wanted to get a better idea of the real consequences of taking time off from work. Many employers do not officially allow employees to take time off from work to care for sick children, for example, but in practice supervisors may grant such leave when it is requested. Conversely, although an employee may have a large bank of sick leave, absence from the office can adversely affect how he or she is evaluated or compensated. The 1994 access survey asks questions specifically designed to find out about informal mechanisms that might have financial consequences for those seeking care.

An important factor influencing the use of health services is the need for medical care, which the access surveys have attempted to measure in a variety of ways. Without reviewing medical records, however, measuring need remains a tricky endeavor; some people are simply not very good at reporting on their health in a way that allows for systematic comparisons with what other people have reported. With this rather substantial caveat in mind, the most often-used measure of health is a self-reported health status measure with four possible levels: excellent, good, fair, or poor.

In order to better study those with the highest level of need for health care services, both the 1976 and the 1986 surveys oversampled sick people. The 1976 survey placed special emphasis on people with episodes of illness, while adults and children with chronic conditions or severe illnesses were overrepresented in the 1986 study. In the 1994 survey, with the added power provided by the NHIS, two specific chronic conditions were selected for special investigation: asthma and ischemic heart disease. People who reported having either of these conditions in the NHIS were selected for follow-up interviews.

Focusing on specific health conditions allows analysts to establish a baseline of need and devote attention to assessing differences in access among population subgroups. The two chronic conditions were chosen with the following factors in mind: high incidence rates to provide sufficient numbers of cases for analysis; empirical evidence suggesting that household respondents can self-report accurately; and well-accepted standards for care of the conditions. For each of the conditions, a physician helped design questions to elicit from survey respondents a brief natural history of the disease, past treatment, symptoms, and functional status. Comparison across population subgroups in terms of treatment and health care use will be used to evaluate access to care.

The central role of the federal government in the financing of health care spurred the growth of health services research as we know it today. With the advent of the Medicare and Medicaid programs in the 1960s came the imperative to study their impact--a major thrust of research on access to care in the early 1970s was assessing the changes brought about by these programs--and to prepare for national health insurance.

But the moment passed, and in the 1980s the nation's attention turned to supply-side economics and tax cutting. The health care sector was faced with the persistence of escalating health care costs and the intransigence of the various health care players in being the first to change behavior. There was a widespread realization as well that nudging coinsurance rates and deductibles had little effect on the cost of health care. The real players in the health care system were the providers, and it was their behavior that needed to be modified. The move in 1983 from cost-based hospital reimbursement under Medicare to paying a fixed amount based on diagnosis was the beginning of the evolution toward altering financial incentives for providers.

Then came the effort in the 1990s to reform the nation's health care system. By the time the reform effort ended, it had become increasingly clear that changes in the health care system do not come exclusively--or even, perhaps, primarily--from policy makers at the national level. The delivery of health care has been changing at a mind-altering pace ever since.

Health care experts and policy makers must understand not only these changes but also the way they affect how researchers formulate questions and collect data. It is no longer sufficient to ask respondents if they are covered by a private insurance policy; in order to understand their use of health care, we must attempt to ascertain whether it is a health maintenance organization or a preferred provider organization, whether the physician is paid on a fee-for-service or a capitated basis, and so on. And we must deal with analytic complexities, imposed not only by the myriad insurance arrangements but also by the inability of both consumers and providers to understand their arrangements. We may hypothesize, for instance, that a specific payment arrangement (for example, a salary bonus to physicians at the end of the year if expenditures on treatment have been lower than expected) may influence the way in which physicians prescribe care. Yet does it make sense to even study this hypothesis when recent studies show that many physicians do not understand how their treatment decisions affect their pay? If a physician doesn't know that he or she is subject to a salary withhold, does it affect how he or she provides care?

At the same time, the political climate continues to change--with dramatic potential for altering access to care under public programs. Every new Congress has its own views on what to do with the federally sponsored programs that make up the safety net. In order to monitor changes and develop an understanding of the effects of local public programs on access to care, the 1994 survey developed a methodology for studying the role of community-based and local health programs in promoting access. By gathering information about county-based and other local health care programs in the communities from which we drew respondents, it is possible to investigate the effect of these programs on access to care.

Although much of what we learn when we study access to health care is disheartening, this may be a glass-is-half-empty phenomenon. By and large, most studies of access find that the majority of Americans are able to obtain the health care they need and are satisfied with it. One of the first findings from the 1994 access survey, published in Health Affairs, noted that "approximately 6 percent of Americans are unable to obtain the medical care or surgical care they believe they need."¹⁰ This was similar to the proportion of persons reporting unmet need in both the 1982 and 1986 surveys, showing a surprising constancy over time. Looking at the number from the glass-is-half-full perspective, in each of the survey years 94 percent of those surveyed were essentially saying that they had no trouble getting the medical or surgical care they needed. For Americans as a whole, physician visit rates--in particular, the proportion of persons who are able to get into the system for at least one visit--have also been relatively steady over time, while hospitalization rates have tended slightly downward, in part because of changes in the financial incentives facing those institutions. A clear majority of Americans queried in public opinion polls express satisfaction with their own health care; this is true for persons in traditional fee-for-service arrangements as well as those in HMOs or other managed care plans.

There is some evidence that certain types of health care use may be declining. Preliminary trend estimates from the access-to-care surveys (part of ongoing analyses) indicate that, on a per capita basis, the number of physician visits is declining somewhat. It should be stressed that this may be "just what the doctor ordered"--that the focus should not be on maintaining a given number of visits but on the appropriateness of care. In an increasingly managed care environment, fewer visits may mean that costs are under control and that incentives to provide increasing amounts of care have been neutralized. This notion applies to the downturn in hospital admission rates as well. Researchers are increasingly making the distinction between overall hospitalizations and those for a special class of conditions sensitive to ambulatory care: if individuals receive appropriate ambulatory treatment for the condition, they should not require hospitalization. In examining this specific class of conditions, we can more clearly assess whether changes in the use of services are indicative of an access problem.

Looking beyond outpatient and hospital inpatient services, we observe phenomena that remind us, as researchers, to be ever vigilant in defining access to care. Some of the greatest barriers to care are registered when the definition of health care includes a wider range of services, such as dental care and eyeglasses. Casting the net wider, 16.1 percent of respondents to the 1994 access survey--representing more than forty-one million people--were unable to obtain at least one service they believed they needed. This is substantially larger than the 6 percent figure cited above. Since there are no comparison data about these other services from past surveys, the finding by itself provides no evidence as to whether access to care has either improved or deteriorated. Yet it calls attention to the need to be cautious in defining access. As is true with all research, the answer one gets depends on the question asked. While prescription drugs and dental care consume far less of the health care dollar than inpatient services or ambulatory physician services, they are critical for the nation's health, have serious implications for quality of life, and should be examined more systematically.

In addition to traditional services, alternative medicine is seen as important to health for an increasing number of people. Estimates from the 1994 survey indicate that nearly 10 percent of the U.S. population saw a professional in 1994 for at least one of the following four therapies: chiropractic, relaxation techniques, therapeutic massage, or acupuncture. Some may argue, perhaps justifiably, that insuring access to therapeutic massage is not part of the nation's health policy agenda, but the use of these services and their health-related outcomes should be monitored.

Since their inception, the access-to-care surveys have focused not only on the level of access achieved by the majority but also on equity in access across population subgroups. Vulnerable populations--whether Southern blacks in the 1976 survey or individuals with chronic illness in the 1986 survey--have always been a major focus of the Foundation's interest. It is thus disturbing to find that access for vulnerable subpopulations may be deteriorating. In our early analyses of the 1994 data, we are giving particular attention to the uninsured--not focusing on the numbers of people without insurance but on their use of health care services. Less emphasis is being placed on overall trends in the use of services and more on the differentials across population subgroups. Although declining physician visits might indicate that care is being managed more efficiently, what does it mean if visits are declining at different rates for those with and without insurance coverage? While analyses are still under way, preliminary results confirm our suspicions that access for the most vulnerable groups in our country has declined. Despite expansions in Medicaid eligibility at both the federal and state levels, there are those who are still excluded from our health care system. Although the safety net may be larger, it is farther from the ground and the fall is potentially more threatening. Studies of access to care must focus greater attention on holes that have appeared in the safety net, targeting specific vulnerable subgroups of the population for further study and exploring the policy solutions to problems of access. The 1994 access survey reflects our belief, as well as the Foundation's, that providing policy makers with appropriate and reliable information increases the likelihood that effective programs will be designed and implemented.

One of the secondary benefits of the 1994 design was increasing dialogue between the public and private sectors about the responsibilities of foundations and the federal government in monitoring changes in access to care. Since the 1994 access survey was conducted by a federal agency, by law the project was coordinated with other federal data-collection activities. We have learned the importance of ongoing and close collaboration between The Robert Wood Johnson Foundation and the federal government, even if the Foundation's future access surveys are not directly linked to those conducted by the government. Although the federal government does not sponsor any health care surveys whose primary objective is to study access, there are a number of federal surveys that do allow analysis of access issues. These include the National Health Interview Survey, the Medicare Current Beneficiary Survey, and the Medical Expenditure Panel Survey (which will be conducted annually and replace the National Medical Expenditure Survey). The resources of private foundations such as The Robert Wood Johnson Foundation should augment rather than duplicate such efforts.

Future access surveys should also be designed with awareness of other Foundation-supported activities intended to measure trends in access to care, such as those of the Center for Studying Health System Change. The Center's first major activity is the design and implementation of the Community Tracking Study, a longitudinal study designed to measure the effects of various health systems on a number of outcomes, including those related to access.

The Community Tracking Study, as well as federal surveys, will follow the access-to-care experiences of the nation overall, as well as in twelve communities across the country. These surveys, however, are of limited value in analyzing the specific access barriers experienced by some of the nation's most vulnerable groups. Still, they can help the Foundation understand which groups are experiencing problems getting care, even though there may be too few surveyed persons in each group to permit more detailed analysis on why they are having problems or on the health-related outcomes that result.

Future access surveys can better inform policy makers if they carefully target those groups of people who continue to have trouble obtaining adequate health care. The tracking mechanisms already in place can be used to help the Foundation identify which groups need to be targeted at any given time. Because of the long delays required in procurement and approvals, the federal government is poorly suited to sponsor these types of small, targeted efforts; in contrast, the Foundation is in a unique position to develop quickly surveys focused on the identified populations.