2MEASURING THE EFFECTS OF SOCIOECONOMIC STATUS ON HEALTH CARE

2–1. STUDYING THE EFFECTS OF SOCIOECONOMIC STATUS ON HEALTH

Since 1985 there has been a substantial increase in the number of studies about the relationships between socioeconomic factors and health. In an article published in 1999, Nancy E.Adler and Joan M. Ostrove sketched the evolution of the theoretical framework now used in studying disparities in health (Adler and Ostrove, 1999). Before the mid-1980's, socioeconomic status was largely absent in studies on health except as a control variable. Studies focused on poverty and its association with health. The model assumed a threshold effect: the health of people below the poverty level was believed to improve as their income increased and reached the poverty threshold. Above the poverty threshold, the level of health was constant as income increased.

At a 1987 conference sponsored by the Kaiser Family Foundation, leading social scientists from the U.S. and Great Britain presented a number of papers that showed that the effect of socioeconomic factors was much broader than just poverty. In fact, many social and economic factors are related to health. Moreover, there is a gradient effect between socioeconomic status and health: as socioeconomic status increased, health improved. The conference resulted in the 1989 publication of Pathways to Health (Bunker et al., 1989).

The papers were groundbreaking and ushered in an era of profound intellectual and pioneering work to understand the effects of socioeconomic status on health. A reading of Pathways to Health today shows that the 1980 Black Report³ (Black, 1982) stimulated the thinking of many social scientists because it had found that gaps in health had widened since the establishment of the National Health Service in 1948. The Black Report became the underpinning of the belief that health care does not play a very important role in health. Robert and House describe the prevailing views of social scientists during the 1980s and 1990s:

Most research suggests that access to medical care plays a relatively minor role in explaining socioeconomic inequalities in health…socioeconomic differences are seen both in diseases that are amenable to medical treatment and in diseases that are not amenable to medical treatment …with deaths from diseases amenable to treatment representing only a fraction of all deaths in any case (Robert and House, 2000, p. 121).

These conference papers also may have encouraged the development of models that would focus on the effects of socioeconomic status on health without inclusion of race as an independent variable. The likely premise has been that racial differences in morbidity and mortality are reflections of differences in social and economic factors. However, in recent years social scientists have noted that “studies may need to address how class-related experiences of racial/ethnic and gender discrimination may harm health” (Krieger et al., 1997, p. 369).

2–2. REVIEW OF METHODS USED IN STUDYING THE EFFECTS OF SOCIOECONOMIC STATUS ON HEALTH

The following is an overview of the methods that social scientists use in studying the effects of socioeconomic status on health to determine what is applicable to studying the effects of socioeconomic status on health care. Cross-fertilization of knowledge between social scientists and health services research promises to be beneficial all around. The dissemination of information about disparities in the use of Medicare services has helped to change the perception that health insurance by itself assures equal access and use of health care (Robert and House, 2000). An example of the beneficial effects of cross-fertilization of knowledge from social scientists—and one that is central to this paper—is the recent recognition by researchers that socioeconomic status is an important variable in studying disparities in health care, particularly disparities by race and ethnicity.

There is no simple conclusion or overwhelming agreement on the causes of disparities in health, reasons for growing gaps in health, ways to address them, or even how to study the issues. As Robert and House observe, “we still do not well and consensually understand why socioeconomic inequalities in health exist and persist, nor what policies are most likely and necessary to reduce these inequalities” (Robert and House, 2000, p. 115).

Nonetheless, a significant body of knowledge is available from studying the effects of socioeconomic status on health that is useful in studying the effects of socioeconomic status on health care. Four major issues on which a consensus has been reached are discussed next:

1. Is There a Single Best Approach to Measuring or Analyzing Socioeconomic Status?

The field of research about the effects of socioeconomic status on health (sometimes termed health status or health outcomes) is still new. The link between socioeconomic status and health is not well understood. Among social scientists there is a consensus that there are many different pathways connecting socioeconomic status and health. This means that a broad perspective is needed to understand the multiple pathways linking socioeconomic status and health.

This literature addresses two fundamental questions about methods of study: first, among the variables used as measures of socioeconomic status, is there a single best measure? Secondly, are some approaches used to analyze the effects of socioeconomic status better than others? The answers are, in general, “No.” There are inherent imperfections and limitations in all of the measures of socioeconomic status just as there are in measures of race and ethnicity. But, when used thoughtfully, each measure can have its own ring of truth.

Research about socioeconomic status and health began to gain momentum in the mid-1980s. While much has been learned, frequently lacking in research is a clear conceptualization of what is being studied and why a particular measure of socioeconomic status is used. In fact, measures and methods are at times chosen because of data availability rather than because of theoretical premises. For example, in the U.K. occupation is used more frequently in studies about disparities in health. In the U.S. income and education are used more frequently. These choices are due, in part, to the type of social and economic information collected.

Recently, the use of composite measures has gained attention. Different composite measures of deprivation relating to material and social disadvantage have been developed for studying the effects of socioeconomic status on the individual and area levels (Pampalon and Raymond, 2000). Composite indices are generally constructed by combining information (often from a national census) about factors such as income, employment, communications, transportation, support, education, owned home, and living space. Peter Townsend (Townsend, 1987) and Morris and Carstairs (Morris and Carstairs, 1991) in the U.K. introduced composite indices for area-level analyses based on four factors. Three factors in both of the indices are unemployment, lack of a car, and overcrowded housing. For the fourth factor the Townsend index uses home ownership while the Carstairs index uses lower social class.

A different formulation of a composite index (named CAPSES) has been developed based on the theory that socioeconomic status is a function of three domains of capital: material capital (such as incomes, homes, and stocks); human capital (such as education, skills, and abilities); and social capital (such as membership in social networks).⁴ A recent pilot study testing CAPSES against individual and other composite indices of socioeconomic status showed considerable consistency across the various socioeconomic status measures (Oakes and Rossi, forthcoming).

Composite indices for area-level analyses have been used in different ways. They may be particularly useful as area-wide planning tools. For example, in the 1960s, the Planning Department in Baltimore City designed a composite index for census tract areas based on several social and economic factors. The index was used to rank census tracts from the most advantaged to the most deprived. For an experimental program set up in Baltimore in the 1970s, these rankings were used to establish a health program for children and youth in census tract areas that were most deprived.⁵

A Quebec study provides some insight into the potential difficulties in interpreting results of area rankings from a composite index of deprivation. Comparisons between area rankings and factors used in the index showed that areas deprived socially were not necessarily deprived materially and vice versa. Thus, the Quebec study provides a cautionary note that “lumping” socioeconomic status measures together can be confounding because the index does not necessarily provide a measure of area-level socioeconomic status that can be readily interpreted.

In their review of methods used in studying socioeconomic status and health, Robert and House conclude that questions about which measures and methods to use “remain unanswered and perhaps unanswerable in a generic sense” (Robert and House, 2000, p. 8). Moreover, there are many remaining methodological problems relating to studying the effects of socioeconomic status on health. These problems include:

1.

The lack of precision and reliability of various measures as well as difficulties in generating measures of socioeconomic status;

2.

Unresolved questions about how to measure the effects of socioeconomic status over the life course that would reflect change in social and economic factors from birth to old age;

3.

Some measures of socioeconomic status that are useful for studying their effects within some races and ethnic groups may not be useful for other races and ethnic groups, a methodological issue that also applies to gender;

4.

A lack of understanding about why the relationships between socioeconomic status and health are stronger for men than for women;

5.

Difficulties with classifying married women, the unemployed, and retired persons in a household;

6.

Difficulty of including mental and other psychosocial factors that affect health in measures of socioeconomic status; and

7.

The intertwining of race, ethnicity, and socioeconomic status, and how to assess the separate effects.

The list of unresolved conceptual and measurement questions is long although the viewpoint of experts such as Krieger, Williams, and Moss is clear about certain issues: “we underscore the issue is not whether one measure is ‘right' or another ‘wrong'…rather, numerous studies suggest that measures at each level, over time, may be informative, separately and in combination” (Krieger et al., 1997, p. 349). They add that “the utility of socioeconomic indices for public health research remains unclear…. One concern is that combining measures of income and education into one index…can conflate pathways and obscure each component's distinct—and conceivably different—contribution to specified health outcomes” (Krieger et al., 1997, p. 366).

This overview of methods used to study the effects of socioeconomic status indicates that there is no one right measure. The choice of a “right” measure depends upon the study. Table 2–1 briefly summarizes the advantages and disadvantages of using specific measures of socioeconomic status.

2. What is the Relationship between Socioeconomic Status and Health?

The direction of the relationship between socioeconomic status and health is a fundamental issue in understanding pathways leading to disparities in health. While some economists believe that health drives socioeconomic status—because poor health has a negative effect on job opportunities and socioeconomic position (social drift)—most social scientists believe the direction of the relationship is the other way around. Among those who have studied disparities in health, there is a consensus that biological and genetic differences account for a relatively small proportion of the disparities in health. Supporting that belief is a study of the effects of six risk factors—smoking, alcohol consumption, systolic blood pressure, cholesterol level, body mass index, and diabetes. The study showed that these six factors together accounted for only 31 percent of the difference in mortality between Blacks and Whites. Income accounted for 38 percent of the difference in mortality, while the remaining 31 percent of excess mortality among Blacks was unexplained (Otten et al., 1990).

Among those who have studied disparities in health care, there is also a consensus that biological, genetic, and health status differences account for very little of the persistent disparities by race in health care. For example, one study found that Black veterans with coronary artery disease were 64 percent less likely than White veterans to undergo coronary artery bypass graft (CABG) and balloon angioplasty (Peterson et al., 1994). Several other studies in the literature have found disparities by race in the use of revascularization procedures (Ayanian et al., 1993; Udvarhelyi et al., 1992; Wenneker and Epstein, 1988; Whittle et al., 1993). However, because certain diseases such as hypertension, diabetes, and osteoporosis are not uniformly distributed in the population, such differences must be recognized because they can lie at the crux of the credibility of studies about disparities in access, utilization, and quality of health care. For example, differences in amputations of all or part of the lower limb must be examined in light of differences in diabetes (Gornick et al., 1996).

3. Should Socioeconomic Status Be Used as a Primary Independent Variable to Analyze Health Outcomes?

Social scientists ceased using socioeconomic status as a control variable when they recognized that health was affected not only by poverty, but also by a much broader set of variables including income, education, and occupation. Thus, if the intent is to understand factors that affect disparities in access, utilization, or quality of care, socioeconomic status should not be used as a control variable.

This is critical to studying the effects of socioeconomic status on health care, especially in relatively new areas of research. For example, suppose it were found that on average highly educated people rate health plans better than less educated people. It could be hypothesized that this consistent pattern biases the ratings, and therefore controlling for education across plans is warranted. However, better-educated members of a plan may get better health care if their interactions with the plan are more successful. For example, they may experience less waiting time for appointments or they may be more successful getting referrals to specialists than less educated members of the plan (Fiscella et al., 2000).

4. Why Does Research on Disparities Require a Clear Conceptualization?

Ameliorating disparities in health care requires a conceptual framework that evolves from hypothesis testing, especially those hypotheses that can help pinpoint potential agents of change. For example, a framework might first evolve from formulating hypotheses about how individuals and the health care delivery system interact in terms of behaviors of individuals, providers, and institutions. This would be followed by testing how these interactions are associated with access, utilization, and quality of care. Behaviors have been shown to be factors associated with disparities in the use of preventive services because these services are often self initiated (Gornick et al., 2001; Lemon et al., 2001).

As an example, elderly women with higher incomes and supplementary insurance are more likely to obtain mammograms than lower income women and women without additional coverage. Under Medicare, mammography requires a co-payment, which suggests that the co-payment may lead to the disparities associated with income. In every insurance category—Medicare only, Medicaid, and private supplementary coverage—mammography use rises with income (Blustein, 1995). Yet, there are even greater disparities in the use of flu shots, which are “free.” These facts do not rule out the effect of income, but they do suggest that there are likely to be multiple pathways leading to disparities in utilization.

2–3. REVIEW OF METHODS USED IN STUDYING DISPARITIES IN HEALTH CARE

Disparities in health care have been studied for many years. For example, before the advent of Medicare it was known that the elderly who were minorities and who were poor received inpatient hospital care at a much lower rate than Whites and more advantaged persons. Early studies focused primarily on known “barriers to care.” Lack of health insurance and a regular source of health care were identifiable obstructions to obtaining health care. When these barriers were removed and the elderly and the poor could enter the health care system, it was expected that there would be equal access to covered services and that the use of any particular service would reflect need. In the past decade, disparities in Medicare have led to the awareness that there are other barriers to health care that are related to race, ethnicity, and socioeconomic status.

We do not know how great a role medical care plays in explaining disparities by race and socioeconomic status in health and health care. What is known is that patterns of health care utilization among the healthiest elderly differ from those of the least healthy. Moreover, the patterns of health care use among the healthiest are those that experts recommend, specifically a concentration on prevention and health promotion. In the Medicare program, three distinct patterns have become evident. Compared to Whites and beneficiaries (White or Black) of higher socioeconomic status, Blacks and beneficiaries (White or Black) of lower socioeconomic status use fewer preventive and health promotion services such as influenza immunization and mammography. They also use fewer diagnostic tests such as colonoscopy and undergo fewer common surgical procedures such as CABG. In addition, they use more of the types of procedures that are associated with poor management of chronic disease such as excisional debridement and amputations of part or all of the lower limb (Gornick, 2000).

Three principal approaches are used to study the effects of race, ethnicity, and socioeconomic status on access, utilization, and quality of care. The first approach uses information about health care collected in nationally representative household surveys. The second draws from administrative databases from such sources as the Medicare and Medicaid programs, the Veterans Administration, and hospital discharges. The third is based on clinical data collected by sources such as medical records and disease registries.

The detailed data collected about health care in surveys of nationally representative households provide a rich source of information to study the effects of race, ethnicity, and socioeconomic status (income and education) on potential access. This is the dimension of access to care that is measured by characteristics of individuals and of the health care system (Aday et al., 1984). Measures of potential access most often used include health insurance coverage and a regular source of care. Other measures of potential access include availability of resources such as physicians-to-population ratios, hospital beds per capita, out-of-pocket costs of services, and waiting and travel time. Household surveys such as the Medical Expenditure Panel Survey (MEPS), sponsored by AHRQ, and the Medicare Current Beneficiary Survey (MCBS), sponsored by the Centers for Medicare and Medicaid Services (CMS), focus on collecting different measures of potential access. A special strength of survey data is that utilization rates are not subject to inaccuracies created by multiple payers. However, health care services that are less common, such as heart procedures, cannot be analyzed using survey data because of small cell sizes. In addition, the extent to which self reporting biases estimates of disparities in health care is not well understood.

A second approach is through information about “realized access” that is available in administrative databases. Realized access is measured by the actual use of services (Aday et al., 1984). Assessing the effects of socioeconomic status on realized access requires information about the use of different types of services such as those for health promotion and disease prevention, referral (including diagnostic tests and surgery), pain management, mental health, aftercare and rehabilitation, and long term care. The sample size of household surveys is generally large enough to generate utilization rates for frequently used services, such as influenza immunization and mammography, but not for less frequently used services such as CABG surgery. Data sources that collect information for a large number of people such as administrative data, surveys of hospital discharges, or statewide hospital discharge systems are needed to generate utilization rates for the majority of medical and surgical services. Both aspects—potential access and realized access—are essential dimensions in assessing access to care.⁶

The major strength of administrative data is the size of the files, which is often large enough to develop population-based utilization rates for many different types of services. A major limitation of administrative data is the inadequate information about race and ethnicity and the lack of clinical information about the need for certain services such as a particular heart procedure. In addition, administrative data typically do not contain enough detail to assess appropriateness or effectiveness. For example, administrative data capture whether a certain test was performed, but not the results of the test.

A critical factor in the use of administrative data is whether reliable information is available to generate denominator data that correspond to the numerator data. In general, denominators can be generated using Medicare, Medicaid, and VA administrative data. Over time, programmatic changes such as the growth of managed care enrollment can threaten the reliability of administrative data, although analysts have devised ways of adjusting numerators and denominators for enrollment in health maintenance organizations (HMOs). However, in many cases, the relevant denominators for research on topics such as the number of persons by race for whom a procedure is indicated can only be determined by using selected patient-based studies.

To address the absence of data on socioeconomic status, Medicare data were linked in 1995 to U.S. census data on a ZIP code basis to study the effects of race and socioeconomic status on health care. This approach is derived from studies that validated the use of aggregate data on socioeconomic status from the census as a proxy for the socioeconomic status of an individual. This is based on the understanding that the proxy measure of socioeconomic status reflects both the characteristics of the individual and the area where the individual lives (Geronimus et al., 1993; Geronimus et al., 1995; Krieger, 1992). The match was incomplete for 4 percent of White beneficiaries and 6 percent of Black beneficiaries because of unmatched ZIP codes or missing income data on the census files. These beneficiaries were excluded from the study. The MCBS was used to validate this approach (Gornick et al., 1996). It was intended that the ZIP code analyses would be refined in future studies by using census tracts aggregations. However, that approach was abandoned for methodological reasons, including the fact that about 30 percent of addresses in the U.S. do not have a census tract.

A third approach is through patient-based studies to analyze treatments and quality of health care vis-à-vis patient need for medical and surgical care. The strength of patient-based studies is that they generally draw upon data sources containing clinical information, such as hospital medical records. One limitation in patient-based studies is that they are not likely to be nationally representative. Moreover, they do not reflect the population at risk of needing the treatment. Nonetheless, they are a rich source of information for analyzing quality of care. A number of patient-based studies have used a database established from the linkage of information available in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program with information available in the Medicare administrative data system. The SEER/Medicare database has also been linked to U.S. census data on a ZIP code basis. This has resulted in a unique source of information for studying the effects of race and socioeconomic status on disparities in the incidence and treatment of cancer, the second leading cause of death in the U.S..

These three approaches to studying disparities in health care have provided a wealth of information about the relationships among race, ethnicity, and socioeconomic status. It is important to note, however, that results are likely to differ somewhat according to the data source. For example, Medicare utilization data from the administrative data system reflect the experience of beneficiaries receiving services in the fee-for-service sector, whereas utilization data from the MCBS reflect the experience of all beneficiaries. Therefore, analysts need to be aware of the design features as well as the limitations and strengths of their data sources. The following tables illustrate the types of data available from household surveys and administrative data.

Table 2–2 also illustrates the gradient effect: as income increases, ambulatory visits and mammography rates increase. The gradient effect is in the opposite direction for emergency department visits and for amputations. The Black-to-White ratio of the rates for each service is shown at the bottom of Table 2–2, unadjusted and adjusted for income. The Black-to-White ratio for ambulatory visits— when adjusted for differences in income—rose from 0.89 to 0.93. Similarly, for mammography, the Black-to-White ratio rose from 0.66 to 0.75. For emergency department visits, the Black-to-White ratio improved slightly, declining from 1.45 to 1.37. Amputations had the same pattern, declining from 3.64 to 3.30.

The lower bank of data shows generally similar patterns, but the percentages using the preventive services are generally lower. The rates differ because the number of women in the income and education groups changes somewhat. For example, among White women, the proportion of those in the higher income group (28.8 percent) was only about half the proportion of those in the higher education group (65.9 percent). Among Black women, the shifts are greater: 7.2 percent were in the higher income group while 36.0 percent were in the higher education groups.

Table 2–4 shows that the total utilization rate for each service for White women exceeded the rate for Hispanic women. However, it can be observed that the rates of use of mammograms and Pap smears were greater for higher income Hispanic women than for higher income White or Black women. As Table 2–4 illustrates, when the Hispanic-to-White ratios are adjusted for income, the ratio for mammograms increased from 0.83 to 0.91 and the ratio for Pap smears increased from 0.94 to 1.07. This illustrates the sizeable effect of differences in income distributions between White and Hispanic women. Table 2–4 also provides a comparison in utilization patterns between Hispanic and Black women. Among those with high income, the utilization rate for Hispanic women far exceeds the rate for Black women.

The next two tables provide examples of two patient-based studies. Table 2–5 is from a patient-based study analyzing rates of resection of patients diagnosed with resectable non-small-cell lung cancer, by race and income. Patients were newly diagnosed during the period 1985–93. The study used the SEER/Medicare database linked to U.S. census data on a ZIP code basis.

Table 2–6 is from a patient-based study analyzing rates of different procedures following acute myocardial infarction. This study used data from the Veterans Administration. The percentage of patients with each of the procedures shown was higher for White patients than Black patients.

Tables 2–2 through 2–6 illustrate the ways that household surveys, administrative databases, and patient records can be used to analyze patterns of health care for various types of services by race, ethnicity, and socioeconomic status. Clearly, these data sources have provided substantial evidence that vulnerable subpopulations receive different health care than more advantaged subpopulations. Such descriptive analyses are valuable in identifying disparities in health care and are needed to raise concerns about unequal access and utilization of health care. But the question remains: why do disparities in health care exist? The lack of knowledge about why disparities exist —even among insured populations—indicates that ongoing monitoring of health care disparities should be joined by research that focuses on analyses to understand the pathways that lead to disparities in health care and the testing of initiatives to effect a change. Table 2–7 briefly summarizes the approaches used by researchers to examine disparities.

2–4. METHODOLOGICAL ISSUES IN STUDIES OF HEALTH CARE DISPARITIES

In addition to the advantages and disadvantages of specific measures of socioeconomic status, certain other data issues must also be considered. The following presents five issues common in studies of health care disparities.

1. Availability of data on socioeconomic status and other factors that affect disparities in health care. Surveys that generate information about use of health care generally contain only limited information about socioeconomic status. The two measures of socioeconomic status generally collected are income and education. These measures of socioeconomic status may be useful indicators of social and economic status for some subgroups of the population, but are often relatively insensitive for other subgroups, especially for Blacks. In part, this is due to sample size. Other measures of socioeconomic status, such as wealth, would very likely be useful indicators of social and economic status. However, wealth can be extremely difficult to capture using surveys alone since people are generally unwilling to provide that information in household surveys. Moreover, recent studies indicate that lifestyle factors such as nutrition, exercise, obesity, and behavioral characteristics such as smoking cessation are also associated with disparities in health care. This suggests that the role of socioeconomic status will be difficult to disentangle from lifestyle and behavior factors especially because information about lifestyle and behaviors is generally unavailable.

2. Using census data for measures of socioeconomic status. Databases that lack information on socioeconomic status have been linked with U.S. census data at the census tract or ZIP code area level to assign an individual in the database the median income and educational attainment corresponding to his or her area of residence. For analytic purposes, individuals are often distributed into quartiles. If specifications for the quartiles are based on the income of the total population, then the distribution for Blacks will be uneven given the substantial differences between Blacks and Whites in income. Table 2–5 illustrates this problem. The study had a total of 10,124 White patients and 860 Black patients; 52 percent of Black patients fell into the lowest income quartile. Evidently, the three highest quartiles of patients were grouped together to overcome the problem of small cell size. Experience with this approach has shown that the problem can be avoided if income quartiles are specified separately for Blacks and for Whites. However, researchers are often limited to using databases in which certain variables, such as income, are put into a pre-specified grouping. Therefore, the “raw” data are no longer available to alter the groupings.

3. Small cell sizes even with large samples. Except for preventive services, utilization rates may be relatively low. Even with large databases, cell sizes may be too small to analyze rates by age, sex, race, ethnicity and socioeconomic status. Table 2–6 illustrates this problem. This study had 29,119 White patients and 4,522 Black patients. This study was published in 1994, a time when socioeconomic status had not yet been commonly used in studying disparities in health care. Had socioeconomic status been included in this study, sample size would have been sufficient. But had the data also been presented by age and sex, cell sizes for Black patients would have been too small.

4. Differences by race and ethnicity in risk factors. Linking health to health care requires that differences in risk factors be recognized. To make a creditable case that disparities exist in health care, reference needs to be made to differences in risk factors. For example, the rate of amputations of all or part of the lower limb for Black Medicare beneficiaries is substantially higher than the rate for White beneficiaries. In this example, it is important to show that among elderly Blacks, diabetes (frequently the underlying cause of limb amputations) was 1.7 times the rate for elderly Whites. However, as shown in Table 2–2, the amputation rate in 1993 for Blacks was 3.64 times the rate for Whites, far greater than expected based on the difference in diabetes rates (Gornick et al., 1996).

5. Data for persons in managed care plans. Data are generally not available to study the effects of race and socioeconomic status on utilization in managed care plans. Policy papers have discussed the inadequacy of current information from health plans to assess disparities by race and socioeconomic status.

2–5. CONCLUSION

Incorporating knowledge from the social sciences about methods for studying socioeconomic status will help to put the NHDR on a sounder scientific footing and expand the perspective of its audiences. The examples in this paper illustrate the insights that can be gained about racial and ethnic disparities in health care when measures of socioeconomic status are included. Disparities in health care between Blacks and Whites and between Hispanics and Whites were generally reduced even with adjustment by a single measure of socioeconomic status such as income. It is important to recognize that examples in this paper show that substantial disparities in health care also occur within the White population. As income or education increased among Whites, the gradient effect was notable in several instances: the use of preventive and diagnostic services increased while the use of procedures associated with poor outcomes of care (such as lower limb amputation) decreased as income increased.

Research has shown that there are a myriad number of social and economic factors that can influence health and health care. It follows that future analyses of disparities in health care that are better able to measure and adjust for socioeconomic differences are likely to reduce racial and ethnic disparities even further. The major lesson learned from this review of research is that knowledge about disparities in health care increases when we are able to disentangle the separate effects of race, ethnicity, and socioeconomic status. In the example showing White, Black, and Hispanic rates of mammography, flu shots, and Pap smears by income groups, the rates for Hispanic women in the higher income groups differed substantially not only from White women but from Black women as well. Thus, studies of disparities in health care that aggregate data for all minority persons and present an overall measure of access and utilization are likely to obscure the fact that barriers to health care can differ for population subgroups.

The NHDR provides a major opportunity to focus attention on disparities in health care in the U.S., especially in the use of preventive and health promotion services. The vast amount of information available in U.S. data systems—as well as the experience gained in analyzing data collected in household surveys, administrative data, and medical records—can serve as a foundation for the NHDR. The challenge is to provide useful information on whether or not the health care received by vulnerable subgroups continues to differ from the health care received by persons who are more economically and socially advantaged.

Disparities in health care are likely to be more meaningful to Congress and the nation if the NHDR provides information that indicates disparities matter in terms of health outcomes. For example, rates of colonoscopy and sigmoidoscopy for Black Medicare beneficiaries have been consistently lower than rates for White beneficiaries. These differences are more likely to capture the attention of policy experts, the health care community, and the nation if they are juxtaposed against information showing that Black persons aged 65 or older have more advanced stages of cancer at the time of diagnosis and higher colon cancer death rates than White persons their age.

By depicting the types of disparities that occur in health care by race, ethnicity, and social status, the NHDR can serve a vital function not only in reporting disparities in health care, but in stimulating questions about why disparities exist. Thus, the report can serve as a foundation for conceptualizing a framework for testing hypotheses about pathways that lead to disparities in health and health care and ways of effecting a change.

Reference List

• Aday, L., G.Fleming, and R.Andersen. 1984. Access to Medical Care in the US: Who Has It, Who Doesn't? Chicago: Pluribus Press Inc; and the University of Chicago.
• Adler, N., and J.Ostrove. 1999. Socioeconomic status and health: what we know and don't know. Pp. 3–15. In Socioeconomic Status and Health in Industrialized Nations: Annals of the New York Academy of Sciences . Vol. 896. New York: New York Academy of Sciences. [PubMed: 10681884]
• Ayanian, J.Z., I.S.Udvarhelyi, C.A.Gatsonis, C.L.Pashos, and A.M. Epstein. 1993. Racial differences in the use of revascularization procedures after coronary angioplasty. JAMA 269 (20):2642–46. [PubMed: 8487447]
• Bach, P.B., L.D.Cramer, J.L.Warren, and C.B.Begg. 1999. Racial differences in the treatment of early-stage lung cancer. N Engl J Med 341 (16):1198–205. [PubMed: 10519898]
• Black, D. 1982. Inequalities in Health: The Black Report . UK: Pelican.
• Blustein, J. 1995. Medicare coverage, supplementary insurance, and the use of mammography by older women. N Engl J Med 332 (17):1138–43. [PubMed: 7700288]
• Bunker, J., Gomby D., and B.Kehrer. 1989. Pathways to Health: the Role of Social Factors . Menlo Park CA: The Henry J.Kaiser Family Foundation.
• Fiscella, K., P.Franks, M.Gold, and C.Clancy. 2000. Inequality in health: addressing Socioeconomic, racial and ethnic disparities in health care. JAMA 283 (19):2579–82. [PubMed: 10815125]
• Geronimus, A.T., J.Bound, and L.Niedert. 1993. On The Validity of Using Census Geocode Characteristics to Proxy Economic Status (Research Reports No. 93–269) . Ann Arbor, MI: Population Studies Center, University of Michigan.
• ——. 1995. On the Validity of Using Census Geocode Characteristics to Proxy Individual Socioeconomic Characteristics. Technical Working Paper 189 . Cambridge, MA: National Bureau of Economic Research.
• Gornick, M. 2000. Vulnerable Populations and Medicare Services; Why Do Disparities Exist . New York: The Century Foundation Press.
• Gornick, M., P.Egger, and G.Riley. 2001. Understanding disparities in the use of Medicare services. Yale Journal of Health Policy, Law, and Ethics 1:133–58. [PubMed: 12669323]
• Gornick, M.E., P.W.Eggers, T.W.Reilly, R.M.Mentnech, L.K. Fitterman, L.E.Kucken, and B.C.Vladeck. 1996. Effects of race and income on mortality and use of services among Medicare beneficiaries. N Engl J Med 335 (11):791–99. [PubMed: 8703185]
• HCFA. 1995. 1995 Report to Congress: Monitoring the impact of Medicare Physician Payment Reform on Utilization and Access, HCFA Publication #03378 . Baltimore MD: Health Care Financing Administration.
• Krieger, N. 1992. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health 82 (5):703–10. [PMC free article: PMC1694121] [PubMed: 1566949]
• Krieger, N., D.Williams, and N.Moss. 1997. Measuring social class in US public health research. Annu Rev Public Health 18:341–78. [PubMed: 9143723]
• Lemon, S., J.Zapka, E.Puleo, R.Luckmann, and L.Chasan-Taber. 2001. Colorectal cancer screening participation: comparison with mammography and prostate-specific antigen screening. AJPH 91 (8):1264–72. [PMC free article: PMC1446758] [PubMed: 11499116]
• Morris, R. and V.Carstairs. 1991. Which deprivation? A comparison of selected deprivation indexes. J Public Health Med 13 (4):318–26. [PubMed: 1764290]
• Oakes, M. and P.Rossi. Forthcoming. The measurement of SES in health research: current practice and steps toward a new approach. Soc Sci Med. [PubMed: 12560010]
• Otten, M.W. Jr., S.M.Teutsch, D.F.Williamson, and J.S.Marks. 1990. The effects of known risk factors on the excess mortality of Black adults in the United States. JAMA 263 (6):845–50. [PubMed: 2296146]
• Pampalon, R. and G.Raymond. 2000. A deprivation index for health and welfare planning in Quebec. Chronic Dis Can 21 (3):104–13. [PubMed: 11082346]
• Peterson, E.D., S.M.Wright, J.Daley, and G.E.Thibault. 1994. Racial variation in cardiac procedure use and survival following AMI in Dept. of VA. JAMA 271 (15):1175–80. [PubMed: 8151875]
• Robert, S., and J.House. 2000. Socioeconomic inequalities in health: integrating individual-, community- and societal- level theory and research. Handbook of Social Studies in Health and Medicine . G. Albrecht, editor; , R.Fitzpatrick, editor; , and S.Scrimshaw, editor. . Thousand Oaks CA: Sage Publications.
• Townsend, P. 1987. Deprivation. J Soc Policy 16 (2):125–46.
• Udvarhelyi, I.S. Gatsonis C., A.M.Epstein, C.I.Pashos, J.P.Newhouse, and B.J.McNeil. 1992. Acute myocardial infarction in the Medicare population: process of care and clinical outcomes. JAMA 268 (18):2530– 36. [PubMed: 1404820]
• Wenneker, M.B. and A.M.Epstein. 1988. Racial inequalities in the use of procedures for patients with ischemic heart disease in Massachusetts. JAMA 261 (2):253–57. [PubMed: 2521191]
• Whittle, J., J.Good C.B.Conigliaro, and R.P.Lofgren. 1993. Racial differences in the use of invasive cardiovascular procedures in Veterans Affairs medical system. N Engl J Med 329 (9):621–27. [PubMed: 8341338]

Footnotes

1

Instead of socioeconomic status, some social scientists use the concept of socioeconomic position, which they believe takes into account more of the social and economic factors that influence health.

2

Researchers use the expression “disparities in health care” while social scientists tend to refer to disparities in health as “inequalities in health.”

3

The report is commonly referred to as the Black Report after Sir Douglas Black, chair of the Research Working Group, Department of Health and Social Security, U.K..

4

CAPSES is an acronym formed from the words capital and socioeconomic status (SES).

5

From personal participation in the Baltimore City Health Department study.

6

The term “access” is commonly used to refer to “potential access” to health care, which can be indicated by, for example, having health insurance coverage or a usual source of care. For ease of discussion, this paper conforms to the commonly accepted practice of using the phrase “access and utilization” to mean potential and realized access.