Cover Image

PAPERBACK
$51.75



View/Hide Left Panel

16
Geography and Racial Health Disparities

Amitabh Chandra and Jonathan S. Skinner


During the past several decades, many studies have documented racial, ethnic, gender, and socioeconomic disparities in both medical care treatments and health outcomes.1 There are no easy economic explanations for such differences: African Americans seem less likely to receive invasive treatments even in the Veteran’s Affairs (VA) system, where doctors’ economic incentives are likely to be blunted (Peterson, Wright, Daley, and Thibault, 1994; Whittle, Conigliaro, Good, and Lofgren, 1993). Nor do differences in insurance coverage seem to eliminate racial or ethnic gaps (Carlisle, Leake, and Shapiro, 1997); indeed Ross and Mirowsky (2000) believe public insurance such as Medicare and Medicaid lead to worse health. More recently, racial differences in cardiac surgery were hypothesized to depend on the race of the physician; however, no significant differences were found (Chen et al., 2001). The Institute of Medicine’s (IOM’s) landmark study (Smedley, Stith, and Nelson, 2002) has conducted a comprehensive survey of the evidence and concluded that racial disparities in medical care treatments and outcomes are pervasive; this topic also has been an integral part of the National Research Council’s research agenda (National Research Council, 1997).

Collectively these papers clearly document important racial differences in treatments, intensity of care, and outcomes. In this chapter, we consider a complicating factor that has implications for both statistical inference and policy recommendations regarding racial disparities: the geography of health care and health outcomes, and its relationship to the measurement of racial



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life 16 Geography and Racial Health Disparities Amitabh Chandra and Jonathan S. Skinner During the past several decades, many studies have documented racial, ethnic, gender, and socioeconomic disparities in both medical care treatments and health outcomes.1 There are no easy economic explanations for such differences: African Americans seem less likely to receive invasive treatments even in the Veteran’s Affairs (VA) system, where doctors’ economic incentives are likely to be blunted (Peterson, Wright, Daley, and Thibault, 1994; Whittle, Conigliaro, Good, and Lofgren, 1993). Nor do differences in insurance coverage seem to eliminate racial or ethnic gaps (Carlisle, Leake, and Shapiro, 1997); indeed Ross and Mirowsky (2000) believe public insurance such as Medicare and Medicaid lead to worse health. More recently, racial differences in cardiac surgery were hypothesized to depend on the race of the physician; however, no significant differences were found (Chen et al., 2001). The Institute of Medicine’s (IOM’s) landmark study (Smedley, Stith, and Nelson, 2002) has conducted a comprehensive survey of the evidence and concluded that racial disparities in medical care treatments and outcomes are pervasive; this topic also has been an integral part of the National Research Council’s research agenda (National Research Council, 1997). Collectively these papers clearly document important racial differences in treatments, intensity of care, and outcomes. In this chapter, we consider a complicating factor that has implications for both statistical inference and policy recommendations regarding racial disparities: the geography of health care and health outcomes, and its relationship to the measurement of racial

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life health disparities. This is a broad topic (see for example, Chapter 11, this volume), and so we will organize our contribution along five basic points. (1) There is considerable variation in the utilization of health care, and in outcomes, by region. The phenomenon of “small area variation” in utilization rates has been studied for a number of decades. Most recently the Dartmouth Atlas of Health Care has used nearly 100 percent samples of Medicare enrollees to measure such differences across 306 Hospital Referral Regions (HRRs) in the United States (Wennberg and Cooper, 1999). Even after controlling for differences in underlying health status across regions, there is clear evidence of persistent and large differences in treatment patterns, even in contiguous areas. Much of the current debate is how to interpret such differences—are they “demand” driven by patient preferences, or “supply” driven by physician beliefs and historical patterns of hospital location? In addition to disparities in treatment patterns, there are also substantial variations in health outcomes by region. Recent research has documented race-specific and gender-specific variations at the county or state level in overall mortality rates as well as disease-specific mortality rates (Barnett et al., 2001; Casper et al., 2001).2 (2) People who are African American or Hispanic or belong to other minority groups tend to seek care from different hospitals and from different physicians compared to non-Hispanic whites. It is not surprising that African-American and Hispanic patients tend to see different physicians and are admitted to different hospitals compared to non-Hispanic whites. This is largely the consequence of where people live: there are far fewer African Americans seeking care in eastern Tennessee hospitals than in Mississippi hospitals, and many more Hispanic patients seeking care in hospitals in Florida, Texas, and California than in Maine and New Hampshire. Furthermore, patients of color who live in the same neighborhood as whites may go to different hospitals or (more clearly) see different physicians and in different settings for a variety of reasons, including financial barriers, as well as racial barriers to care (Lillie-Blanton, Martinez, and Salganicoff, 2001). Patients also tend to be seen by physicians of the same race, although one study (Harrison and Thurston, 2001) suggested this matching is in part the consequence of minority physicians being more likely to live near minority neighborhoods. (3) Racial disparities are pronounced in some areas, but are less so (or may not be present) in other areas.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life In most regions of the United States, there are pronounced racial differences in utilization and outcomes. But in other areas, there are no significant racial differences. In some sense, this is welcome news, in that the medical profession is not some monolithic and uniform “system” that treats patients identically regardless of where they live. Such differences, however, are not easily explained, and may rely on one or two surgeons who account for the majority of procedures in their region. In other cases, the differences in racial disparities may arise from spatial “mismatches” of patients and physicians, for example, because of segregation in residential areas or the location of hospital facilities. (4) These three facts create strong statistical interactions between geography and racial identity: one may falsely diagnose geographical variation as racial disparities, and conversely. On average, Hispanic Medicare enrollees account for the same level of expenditures as their non-Hispanic elderly counterparts (Centers for Medicare and Medicaid Services, 2000, Table 4.8). Although this might reassure observers that there are no obvious utilization disparities between Hispanic and non-Hispanic Medicare enrollees, there is one complicating factor: geography. Medicare expenditures on average are substantially higher in Florida, Texas, and California (Wennberg and Cooper, 1999). Because a large fraction of Hispanic Medicare patients live in these three states, the researcher might find that within each state, Hispanic patients experience lower utilization rates than their non-Hispanic counterparts. More generally, in typical regression analysis when minority patients live in regions with systematically different rates of utilization (e.g., African Americans in the south), and the region of residence is not controlled for, one can estimate larger or smaller racial “disparities” that are in fact the consequence of where people live, and not how they are treated or their outcomes within their community. Nor are typical regional measures, such as Metropolitan Statistical Area (MSA), necessarily accurate mirrors of “local” effects. It is important to note here that we do not argue against the existence of racial disparities, nor do we argue that they are necessarily mitigated by geographical variation. If African Americans live in regions with poor hospital quality, then that in itself represents a valid source of racial disparities. Instead, our central thesis is that ignoring geography (or misspecifying it) will cause the analyst to “cry wolf” when true differences are nonexistent, or to falsely conclude that there are no differences when in fact there are substantial differences in the outcomes of interest. Furthermore, as we

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life argue next, the policy prescriptions may differ, depending on whether the racial disparities are caused by regional variations instead of by differences in treatment within hospitals or communities. (5) A potentially large part of overall health disparities in the United States may be the consequence of regional differences in treatment and outcomes. Reducing geographic disparities in quality of care will benefit all Americans, but is likely to yield greater benefits to minority patients. The policy implications of racial disparities are different depending on their proximate causes. Racial differences arising within a hospital or even within a physician’s practice may reasonably be ascribed to differences in underlying health status, patient preferences, financial barriers, provider biases, or some combination of these four factors. Here, however, the insights of the regional variation literature is relevant; it is not the case that the rate of therapeutic interventions for whites should be necessarily viewed as the “correct” or “desired” rate (Tu et al., 1997; Wennberg, 1986). This is because the white rate might reflect inappropriate care—whites get too much done to them, as discussed by Schneider et al. (2001). Alternatively, preferences for care may differ by race or gender. When aggregate racial differences in outcomes are the consequence of minority patients being more likely to live in regions where everyone in the region experiences poorer outcomes, then the policy focus should be on disparities in geography—that specific regions be targeted to improve quality of care or reduce “flat of the curve” health care spending.3 Such policies would ensure that disadvantaged racial groups would be the major beneficiaries of quality improvements. THE GEOGRAPHY OF HEALTH CARE The measurement of regional variation in health care utilization is difficult for a variety of reasons. First, a great deal of statistical power is necessary to measure utilization at the local level; even a sample of 50,000 observations quickly loses power when the data are partitioned into separate regions, and used to focus on specific diseases. Small sample sizes and inadequate statistical power can generate spurious “area variation” just because of random noise in measured average rates.4 Second, the problem of migration to hospitals must be considered; Boston hospitals accept referrals from all over New England, and if these patients were counted, it might appear falsely that Boston residents are at elevated risk of hospitalization. Third, one needs a sample that is not subject to selectivity bias. For example, the sample of Medicaid patients, or of managed care patients, is not likely to be representative of the general population; Medicaid patients can

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life become eligible because of serious illness, and managed care patients tend to be healthier than the general population. Finally, the regions should correspond to actual migration patterns of patients rather than artifacts of historical compromises such as state or county boundaries. In this section, we use data from the Dartmouth Atlas of Health Care that go far to avoid these four shortfalls (Wennberg and Cooper, 1996, 1999). The data comprise a nearly 100 percent sample of Medicare enrollees over age 65, often for 2 years, so the sample sizes are as much as 60 million person-years in a given map or graph; this provides considerable power for regional analysis. Second, the Atlas defines one’s location by the zip code of residence, rather than where one actually gets care. So if a patient from the Burlington, Vermont, region is admitted to a Boston hospital, that hospital stay (and any procedures done there) is assigned to Burlington, not Boston. Third, the Medicare data provide nearly 100 percent coverage of the population over age 65 and is the nearest thing to a national database of utilization in the United States. There have been increases in the population of risk-bearing Health Maintenance Organizations (HMOs) in the Medicare population (now referred to as Medicare+Choice), but that ratio never exceeded 12 percent and has fallen as many insurance carriers have dropped the Medicare+Choice option. In some urban regions the ratio of HMO patients in the Medicare population has been higher than the national average, and this has engendered more concern about selection bias.5 The Dartmouth Atlas has divided the United States into 306 Hospital Referral Regions (HRRs). An HRR is the unit of analysis at which health care for the elderly is delivered. Its geographic boundaries are computed by examining the complex pattern of commuting patterns to major referral hospitals.6 HRRs are named for the hospital service area containing the referral hospital or hospitals most often used by residents of the region. The regions sometimes cross state boundaries—an attribute that is by its very nature ruled out by cross-state analysis. Intuitively, one may think of HRRs as representing the geographic level at which “tertiary” services such as cardiac surgery are received. To demonstrate the construction of the HRRs, in Figure 16-1 we detail the construction of the Evansville, Indiana, HRR. This region includes three states: Illinois, Indiana, and Kentucky. In this region, three hospitals provide cardiovascular surgery services: two in Evansville and one in Vincennes, Indiana. The Evansville HRR also demonstrates that the inclusion of simple MSA fixed effects does not account adequately for geography: the U.S. Census’ Evansville-Henderson MSA is actually comprised of three HRRs. This is not a problem in itself. However, if different HRRs have different practice styles, then it blurs the measure of true regional differences in utilization by aggregating up to the state or MSA level. To demonstrate the

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life FIGURE 16-1 Construction of the Evansville, Indiana, Hospital Referral Region.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life overwhelming degree to which even adjacent HRRs practice different “styles” of medicine, we now draw on the findings of the Dartmouth Atlas of Health Care. Figure 16-2 demonstrates that there is substantial variation in Medicare payments for services reimbursed on a fee-for-service basis (including non-risk-bearing health maintenance organizations). Even after controlling for age, sex, race, illness patterns, and differences in regional prices, reimbursements per enrollee varied greatly: as noted in the Atlas, even though the average payment was $4,993 per beneficiary, payments ranged from FIGURE 16-2 Geographic variation in illness-adjusted Medicare payments.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life $9,033 in the McAllen, Texas, hospital referral region to $3,074 in Lynchburg, Virginia.7 In Figure 16-3 the Atlas illustrates the enormous geographic variation in a relatively standard procedure—Percutaneous Coronary Intervention (PCI), which includes the use of angioplasty and the placement of stents. PCI is an invasive procedure in which a catheter is inserted into the thigh and guided to the narrowed artery, where a balloon is expanded to clear the blockage and improve blood flow. Percutaneous Transluminal Coronary Angioplasty (PTCA) is often used immediately following a heart attack, or shortly thereafter, or to relieve pain for patients with ischemic heart disease. In 1996 more than 200,000 of these procedures were conducted with an average rate of 7.5 per 1,000 Medicare enrollees. As in previous figures, the data have been standardized for demographic characteristics, and the unit of reporting is a HRR. Note how in Texas, Pennsylvania, and California, the ratio of rates (to the U.S. average) can vary drastically even across adjacent HRRs. FIGURE 16-3 Geographic variation in PTCA rates.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life The same pattern exists for other surgical procedures. Figure 16-4 summarizes the variation in rates (on a log scale) at which 10 common surgical procedures are used relative to the U.S. average (in 1996). Similar results have been documented for the variation in rates at which different diagnostic tests are utilized. Together the ten procedures listed in Figure 16-4 made up 42 percent of Medicare inpatient surgery and accounted for 44 percent of reimbursements for surgical care in 1995-1996. For many of these procedures, regional variation occurs because of fundamental uncertainty about the effectiveness of the procedure and ambiguity about the efficacy of alternatives. For example, variation in rates of radical prostatectomy might be partly attributable to the lack of controlled clinical trials comparing the risks and benefits of surgery, radiation therapy, and watchful waiting. For other procedures, even the best clinical trials are often not sufficient to eliminate variation in procedure rates: physicians vary in how they interpret findings from the carefully controlled settings of clinical trials to decision making for individual patients in other settings. The variation for hip fractures is small because the fracture can be diagnosed easily and virtually all physicians agree on the appropriate treatment therapy. For this procedure, the observed variation more accurately reflects variation in the FIGURE 16-4 Surgical variation for ten common procedures. Each data point represents an observation for a Hospital Referral Region relative to the U.S. average standardized for age-gender-race and illness. SOURCE: Wennberg and Cooper (1996).

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life actual rate of the hip fractures. Similarly, hospitalizations for colectomy reflect variations in the incidence of colorectal cancer, rather than differences in treatment strategies. One might suspect these variations may be in part the consequence of differences in underlying patterns of cardiovascular disease. Certainly one might expect that HRR-level rates of PTCA should be associated with HRR-level rates of heart attacks (acute myocardial infarctions, or AMIs). This is because nearly one-third of heart attack patients are treated with PTCA and community rates of AMI should be correlated with true (diagnosed and undiagnosed) levels of ischemic heart disease. However, the correlation coefficient (weighted by the Medicare population) between PTCA rates and AMI rates is essentially zero (correlation = 0.05, p = 0.35) and not significant, meaning that these variations are unlikely to be explained by differences in cardiovascular health status. The provocative nature of these results has not gone unnoticed, and several hypotheses have been put forward to explain these variations. These include the role of sampling variation, differences in underlying severity, patient preferences, the role of capacity, and the nature of physician learning. Wennberg et al. (2002) demonstrate that higher Medicare spending does not result in more high-quality care, such as flu vaccines, use of beta blockers when appropriate, or better health outcomes. Instead, higher spending is typically associated with more “supply-sensitive services” such as physician visits, specialist consultations, and days in the intensive care unit. Supply-sensitive services are those that are provided in the absence of specific clinical guidelines on frequency of use, and where medical texts provide little guidance. Utilization rates for such services appear to be highly correlated with the supply of resources—the number of physicians, specialists, labs, and beds. As such, there appears to be little support for the notion that costs or inadequate training drive practice variation. Another class of rationalizations is developed by Phelps and Mooney (1993) and Bikhchandani et al. (2002), who suggest that explanations based on the nature of physician learning are most likely to account for much of the empirically observed locality of treatment. In the Phelps-Mooney model, physicians are Bayesian learners who attempt to reach an optimal rate for the application of a particular treatment. Eventually, as physicians sample both their own and their colleagues’ experiences, the two will converge toward an optimal rate. This hypothesis suggests a number of implications: a physician’s propensity to treat converges toward the community norm, and faster if the community is more informed and the doctor is less informed (e.g., younger). Among the implications of this theory is the hypothesis that the provision of more precise medical information in medical studies can enhance the learning of physicians, and thus offer dramatic social efficiency gains. Bikhchandani and colleagues (2002) consider a modi-

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life fication of this model and demonstrate that it is possible for physicians to fall into a localized “cascade” because of the difficulty in experimenting with alternative treatment choices. The message of Figures 16-2, 16-3, and 16-4 is that the practice or “intensity” of medicine varies tremendously across space. But there are also large differences within states and even within cities. Fisher, Wennberg, Stukel, and Sharp (1994) construct cohorts of Medicare beneficiaries on the basis of initial hospitalization for AMI, stroke, gastro-intestinal bleeding, hip fracture, or surgery for breast, colon, or lung cancer. They find substantial differences in the intensity with which beneficiaries were treated (as measured by readmission rates) even across similar teaching hospitals in the Boston area. Specifically, there is substantial variation across readmission rates for Massachusetts General Hospital, Brigham and Women’s Hospital, Beth Israel, and Boston University Medical Center. Most interestingly, there is no relationship between mortality (both 30 day and over the entire study period) and the intensity of hospitalization. Clearly, racial differences in migration patterns to hospitals of patients within Boston could have first-order effects on utilization rates, although in this case, probably not with respect to outcomes. RACIAL DIFFERENCES IN WHERE (AND FROM WHOM) HEALTH CARE IS PROVIDED A variety of studies have documented the large differences in insurance status and presence of regular providers (versus emergency room visits) among African Americans, Hispanics, and non-Hispanic whites (e.g., Lillie-Blanton, Martinez, and Salganicoff, 2001). In addition, simple differences in where people live will lead to minority patients being seen at different hospitals, and by different providers, from whites. This is not terribly surprising; clearly, hospitals in Washington, DC, will be more likely utilized by African Americans and Hispanics than those in Minot, South Dakota. To capture this difference, we use a nearly 100 percent sample of Medicare fee-for-service patients who were admitted for a heart attack, or AMI, in 1998-1999; these data come from the National Bureau of Economic Research Medicare claims panel developed by McClellan and Staiger (1999). There were a total of 468,663 admissions in those two years to 4,737 hospitals. Nonblack admissions totaled 439,350, while black admissions were 29,313. We use a Lorenz curve approach to characterize the extent to which black and nonblack AMI patients tend to be admitted to different hospitals, as shown in Figure 16-5. The 4,737 hospitals were sorted according to the total number of black AMI Medicare patients admitted during 1998-1999, starting with the lowest number (to the left) and ranging to the right of the graph showing hospitals with the largest number

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life requisite intervention. However, it is important to note that the analysis also controls for the hospital that the patient was treated at (through the inclusion of a hospital fixed effect). As noted earlier, regions vary substantially with regard to the degree of racial differences in utilization; controlling for hospital effects could have further attenuated such differences. These studies taken together suggest an additional focus for improving quality of care among the black elderly population. If African Americans are more likely to be seen at low-quality hospitals, public policies that attempt to improve hospital quality would disproportionately benefit African Americans. This conclusion remains consistent with one of the salient conclusions of the IOM report: Significantly, minority access to better quality facilities is often limited by the geographic distribution of care facilities and patterns of residential segregation, which results in higher quality facilities being less accessible (Smedley et al., 2002, Chapter 3, p. 114). GEOGRAPHY AND RACIAL DISPARITIES: POLICY IMPLICATIONS AND CONCLUSIONS Most studies in the literature on health disparities find dramatic differences in utilization by race, but are generally vague on the question of whether differences are driven by demand (e.g., blacks do not want the more intensive care) or supply (e.g., physicians treat blacks with otherwise identical characteristics differently) or perhaps that blacks and whites differ by unmeasured health characteristics (e.g., Johnson, Lee, Cook, Rouan, and Goldman, 1993) or respond to different nonmedical incentives such as insurance coverage. Hence most studies do not provide strong policy prescriptions on how one goes about fixing the problem. It is often useful to characterize such differences into three general categories: Preferences, or the underlying demand “function” by patients. “Supply” or physician, health professional, and hospital behavior. Implicit and explicit “prices,” or differences in insurance coverage, travel time, and other factors without explicit prices such as location of residence that are likely to affect behavior. A massive body of literature in social science and medicine may be classified under the first two categories. Indeed, the recent IOM report provides a detailed literature review of these two categories. The report concludes that while a small number of studies demonstrate that minority patients are more likely to refuse care, these differences in refusal rates are insufficiently large to explain a significant share of the observed disparities. A smaller subset of studies has also considered the impact of insurance coverage

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life and travel time on utilization. We group location in this third categorization; in theory, an individual could travel 300 miles to a different hospital, but the costs of travel broadly defined (including the potential for adverse outcomes during the travel) are too high to make it feasible. Most health care is local, and we believe that this third category is critical in evaluating evidence for racial disparities and developing policies to reduce such disparities. We would suggest that, if possible, racial disparities be decomposed into their proximate causes, for example, with respect to “across hospital” variation (i.e., patients are more likely to be admitted to hospitals with perhaps less aggressive treatment protocol) and “within hospital” variation (i.e., black, Hispanic, and non-Hispanic white patients are treated differently within a hospital). Both variations can lead to lower utilization rates for minority populations; the difference, however, lies in the policy implications. The latter type of variation clearly involves the internal workings of specific hospitals or provider groups, and further inquiry into causes of such differences (financial barriers, preferences of patients, or provider behavior) is clearly warranted. The former type of variation, however, relates less to race per se and more to geographical variations in treatment patterns of all patients. The research on regional variations, health outcomes, and shared decision making provides illuminating lessons particularly with theses types of variations. For example, a cursory examination of the medical and social science literature on racial disparities in outcomes reveals that for nearly every study, the white treatment rate is seen as the “gold standard” against which to evaluate black outcomes. This may or may not be the right approach: For economists interested in the study of the racial wage gap, for example, it makes sense to view white wages or white test scores as the standard against which black outcomes should be measured (Chandra, 2000, 2002). Increases in incomes, wealth, or test scores are viewed as being desirable, and decreases in these measures are viewed unanimously as being adverse events. However, with medical outcomes there are at least two reasons why the above logic may not translate over. First, a number of recent studies suggest that “more is not necessarily better.” Simply put, the fact that whites have higher rates of PCI or bypass surgery following AMI does not necessarily mean that blacks should have the same rate (Schwartz et al., 1999).20 This is because it is entirely possible that the white rate of PTCA is a consequence of aggressive medicine and is therefore not the desired benchmark. In the technical jargon of economics, if physicians are operating in a region of negative marginal product on the production function, then scaling back on intensity could actually improve outcomes. Similar issues are considered in asking whether some regions that practice more intensive health care are in fact practicing “flat of the curve” medicine with no observable benefit in terms of better health outcomes (Skinner, Fisher, and Weinberg, 2001).

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life For example, in a widely publicized study, actors of different races and gender described identical symptoms in videos that were then shown to physicians, who were then asked whether they would prescribe cardiac catheterization (Schulman et al., 1999).21 The results of the study indicated that for the actors who were white males, black males, and white females, prescribed catheterization rates of about 91 percent were identical. For the two actresses who were African American, prescribed catheterization rates were 79 percent. (These findings were reported to the media in a quite different way; see Schwartz et al., 1999.) The researchers suggested such differences were evidence of provider discrimination, but what is not known is whether the 91 percent rate is too high or the 79 percent rate is too low (or both) (Schwartz et al., 1999). This question also has been confronted in studies of geographical variation; we don’t know which rate is correct (Tu et al., 1997; Wennberg, 1986). This point also constitutes the central thesis of a recent paper by Schneider et al. (2001). In this important paper, the authors use RAND criteria to classify Coronary Artery Bypass Grafting (CABG) and PTCA procedures on a sample of Medicare beneficiaries who had undergone coronary angiography. The sample was drawn from more than 170 hospitals, and each beneficiary’s treatment was classified as being appropriate, uncertain, or inappropriate.22 The authors found that there was substantial cross-state variation in the inappropriate use of both bypass surgery (CABG) and PTCA; for PTCA inappropriate rates were 24 percent in California, 14 percent in Pennsylvania, 8 percent in Georgia, and 12 percent in Alabama. These regional differences clearly have implications for the percentage of Hispanics and African Americans receiving inappropriate care. Furthermore, they find almost all of the measured racial gap in PTCA revascularization is explained by the higher rate of inappropriate care for whites as well as higher rate of PTCA that is viewed as being of “uncertain” legitimacy. By contrast, they found lower rates of CABG use where appropriate among African-American patients. The null hypothesis in the racial disparities literature always appears to be that there should be no differences in utilization. This is reasonable for procedures where nearly 100 percent of patients should be in favor of such treatments (immunization, eye exams for diabetics) or where 100 percent of patients should be against treatment (inappropriate PTCA as mentioned earlier). It is not unreasonable, however, that preferences for certain types of care may differ across patients, even for demonstrably effective elective surgical procedures (where appropriate) such as hip replacements. It is highly unlikely that observed differences in utilization can be attributed solely to preferences, however. Preferences for a kidney transplant were slightly lower among African-American men and women, but these differences in preferences could explain only a fraction of overall racial differ-

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life ences in transplant rates (Ayanian, Cleary, Weissman, and Epstein, 1999). When seriously ill patients were asked about preferences for life-sustaining technology, preferences among African Americans were stronger for more intensive care (Hopp and Duffy, 2000). Still, it should be kept in mind that the null hypothesis is not exact equality across racial or ethnic groups, but instead that rates of procedures (by race) match with informed preferences for that procedure. In summary, this chapter has argued that local area variations need to be taken seriously in considering racial disparities in health care. This is true for two reasons: First, statistical pitfalls can trip up otherwise careful and valid empirical research documenting the existence and prevalence of disparities. Second, the policy solutions to racial disparities that occur because African Americans and Hispanics tend to live in different places from non-Hispanic whites are quite different from the more obvious sources of racial differences in treatment within a hospital or provider group. A potentially important, but not well understood, source of racial disparities cannot be solved by equal access to health care at the local level, or by universal health insurance for everyone. Instead, the disparities that occur when hospital or provider quality is worse in regions with a larger percentage of African Americans can be solved only by addressing the problem of geographic disparities in health care. Furthermore, reducing geographic disparities is likely to have a first-order impact on improving racial disparities in health care and health outcomes. ACKNOWLEDGMENTS We have benefited from conversations with Elliott Fisher, Douglas Staiger, Kate Baicker, and Jack Wennberg, and this chapter draws on our work with these individuals. We are grateful to the National Institute of Aging for generous support and to Angus Deaton, Christopher Jencks, Jim Smith, Richard Suzman, two anonymous reviewers, and other participants at the National Research Council Workshop on Ethnic Disparities for useful comments. Chandra also acknowledges support from the Nelson A. Rockefeller Center through the Rockefeller Faculty Fellowship program. All errors are our own. Annex to Chapter 16 DEFINITIONS USED IN THE DARTMOUTH ATLAS OF HEALTH CARE23 Hospital Service Areas Hospital Service Areas (HSAs) represent local health care markets for community-based inpatient care. The definitions of HSAs used in the 1996

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life edition of the Atlas were retained in the 1999 edition. HSAs were originally defined in three steps using 1993 provider files and 1992-1993 utilization data. First, all acute care hospitals in the 50 states and the District of Columbia were identified from the American Hospital Association Annual Survey of Hospitals and the Medicare Provider of Services files and assigned to a location within a town or city. The list of towns or cities with at least one acute care hospital (N = 3,953) defined the maximum number of possible HSAs. Second, all 1992 and 1993 acute care hospitalizations of the Medicare population were analyzed according to zip code to determine the proportion of residents’ hospital stays that occurred in each of the 3,953 candidate HSAs. Zip codes were initially assigned to the HSA where the greatest proportion (plurality) of residents were hospitalized. Approximately 500 of the candidate HSAs did not qualify as independent HSAs because the plurality of patients resident in those HSAs were hospitalized in other HSAs. The third step required visual examination of the zip codes used to define each HSA. Maps of zip code boundaries were made using files obtained from Geographic Data Technologies, and each HAS’s component zip codes were examined. To achieve contiguity of the component zip codes for each HSA, “island” zip codes were reassigned to the enclosing HSA, and/or HSAs were grouped into larger HSAs. This process resulted in the identification of 3,436 HSAs, ranging in total 1996 population from 604 (Turtle Lake, North Dakota) to 3,067,356 (Houston) in the 1999 edition of the Atlas. Intuitively, one may think of HSAs as representing the geographic level at which “front end” services such as diagnoses are received. Hospital Referral Region Hospital Service Areas make clear the patterns of use of local hospitals. A significant proportion of care, however, is provided by referral hospitals that serve a larger region. Hospital Referral Regions were defined in the Atlas by documenting where patients were referred for major cardiovascular surgical procedures and for neurosurgery. Each Hospital Service Area was examined to determine where most of its residents went for these services. The result was the aggregation of the 3,436 HSAs into 306 HRRs. Each HRR had at least one city where both major cardiovascular surgical procedures and neurosurgery were performed. Maps were used to make sure that the small number of “orphan” hospital service areas—those surrounded by HSAs allocated to a different HRR—were reassigned, in almost all cases, to ensure geographic contiguity. HRRs were pooled with neighbors if their populations were less than 120,000 or if less than 65 percent of their residents’ hospitalizations occurred within the region. HRR were named for the HSA containing the referral hospital or hospitals used most often by residents of the region. The regions sometimes cross state bound-

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life aries. Intuitively, one may think of HRRs as representing the geographic level at which “back end” services such as invasive surgery are received. ENDNOTES 1.   For a partial list of references, see Alter et al. (1999); Blustein and Weitzman (1995); Chen et al. (2001); Gornick et al. (1996); Peterson et al. (1997); Rathore et al. (2000); and references therein. 2.   Also see Skinner et al. (2001) for measures of morbidity (i.e., heart attacks, stroke, gastrointestinal bleeding, colon cancer, lung cancer) across HRRs as developed in the Dartmouth Atlas of Health Care (Wennberg and Cooper, 1999). 3.   The “flat of the curve” refers to a region where the marginal health intervention has zero impact on outcomes. For economists, this corresponds to the region of zero marginal product. This notion is formalized by Skinner and colleagues (2001); and Wennberg et al. (2002). 4.   It is possible that much of the observed variation reflects random deviations from identical practice patterns across communities (Diehr, Cain, Kreuter, and Rosenkranz, 1992). While this possibility must be considered for smaller samples, the very large samples in the Medicare claims data preclude this explanation; also see McPherson, Strong, Epstein, and Jones (1981). 5.   In statistical analysis, controlling implicitly for selection using the percentage of HMO enrollees in the area has not affected empirical estimates. Beginning in 2000, HMOs were expected to report hospital procedures to the Centers for Medicare and Medicaid Services, suggesting better data on managed care enrollees in the future. 6.   For further details on the construction methods, see http://www.dartmouthatlas.org/99US/toc8.php. 7.   Illness has been controlled for by using age-sex-race-specific mortality and hospitalization rates for five conditions: hip fracture, cancer of the colon or lung treated surgically, gastrointestinal hemorrhage, acute myocardial infarction, or stroke. These conditions were chosen because hospitalization for them is a proxy for the incidence of disease. The cost of living indices were computed by using nonmedical regional price measures. Doing so avoids contaminating the analysis with physician workforce or hospital market conditions. 8.   Values of the index over 60 are considered high. It means that 60 percent of the members of one group would need to move to a different neighborhood in order for the two groups to be equally distributed. 9.   The isolation index measures the extent to which minority members are exposed only to each other, and is calculated as the minority-weighted average of the minority proportion in each area. 10.   States are used instead of HRRs to increase statistical power. 11.   Population weights are for the state-specific African-American and non-African American population for both men and women, not just men alone. 12.   We are grateful to Melinda Pitts for pointing out this correlation to us. 13.   Furthermore, the means for beta-blocker use differ substantially between the two studies—56 percent versus 72 percent, suggesting different criteria may have been used to determine appropriateness. 14.   When Washington, DC, is included in the sample, the observed (unweighted) negative correlation disappears. This is because DC is an “outlier”—the population is 61 percent African American, but exhibits 93 percent beta-blocker use. 15.   Jenks et al. (2000) rank states on the basis of whether interventions that are known to be correct were administered for conditions such as AMI, heart failure, stroke, pneumonia, screening for breast cancer for women aged over 53, and eye exams and lipid profiles for diabetics.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life 16.   Thus, percentage African American may be a better proxy for SES than income or other indirect measures of economic well-being. This interpretation is one that will be consistent with the results presented in this chapter. However, it would not be a proxy for income or social inequality, as is demonstrated conclusively by Deaton and Lubotsky (2001). 17.   More technically, one might think of this error as being identical to measurement error in a covariate (geography, in our example). For a subset of observations, the wrong state has been included (Kentucky instead of Ohio). In general, measurement error biases the coefficient toward zero, implying that the researcher is prone to incorrectly concluding that geography does not matter. 18.   For example, if 15 percent of the residents in HRR A seek care in the more aggressive HRR B, then because HRR measures of utilization are based on residence, the measured level of utilization for HRR A would be higher than is the true level of utilization in its local hospitals. 19.   Opioids refer to codeine, morphine, and other drugs whose effects are mediated by specific receptors in the central and peripheral nervous systems. They are used for severe pain management in cancer patients. 20.   CABG is surgery in which a vein is harvested from the leg, or an artery is harvested from the internal mammary artery, to bypass the coronary artery that has narrowed because of the buildup of atherosclerotic plaque. 21.   Cardiac catheterization (or an angiogram) is a nonsurgical procedure performed under X-ray guidance in a cardiac catheterization lab to aid in the diagnosis of coronary artery disease. 22.   The RAND appropriateness criteria for CABG and PTCA are discussed by Leape et al. (1999). These criteria are not based on the cost of the procedure, and classify a procedure as being appropriate or inappropriate based on the expected health benefit (quality of life or longevity) versus the expected health costs (probability of death or disability). The criteria are constructed for nearly 3,000 clinical scenarios or indications. 23.   We have duplicated the definitions used by Wennberg and Cooper (1998, 1999). For further details on the construction methods, see http://www.dartmouthatlas.org/99US/toc8.php. REFERENCES Alter, D.A., et al. (1999). Effects of socioeconomic status on access to invasive cardiac procedures and on mortality after acute myocardial infarction. New England Journal of Medicine, 341(19), 1359-1368. Ayanian, J.Z., Cleary, P.D., Weissman, J.S., and Epstein, A.M. (1999). The effect of patients’ preferences on racial differences in access to renal transplantation. New England Journal of Medicine, 341(22), 1359-1368. Baicker, K. (2002). The government subsidization of hospital care and health care outcomes. Unpublished, Dartmouth College. Balsa, A.I., and McGuire, T.G. (2002). Testing for statistical discrimination: An application to health care disparities. Unpublished, Department of Health Care Policy, Harvard Medical School. Barnett, E., Casper, M., Halverson, J., et al. (2001). Men and heart disease: An atlas of racial and ethnic disparities in mortality. Atlanta: Centers for Disease Control and Prevention. Bikhchandani, S., Chandra, A., Goldman, D., and Welch, I. (2002). The economics of Iatroepidemics and Quakeries: Physician learning, informational cascades and geographic variation in medical practice. Department of Economics, Dartmouth College, working paper.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life Blinder, A.S. (1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources, 8(4), 436-455. Blustein, J., and Weitzman, B.C. (1995). Access to hospitals with high-technology cardiac services: How is race important? American Journal of Public Health, 85, 345-351. Carlisle, D.M., Leake, B.D., and Shapiro, M.R. (1997). Racial and ethnic disparities in the use of cardiovascular procedures: Associations with type of health insurance. American Journal of Public Health, 87, 263-267. Casper, M., Barnett, E., Halverson, J., et al. (2001). Women and heart disease: An atlas of racial and ethnic disparities in mortality. Atlanta: Centers for Disease Control and Prevention. Centers for Medicare and Medicaid Services. (2000). Health and health care of the elderly population: Data from the 1996 Current Beneficiary Survey. Washington, DC: U.S. Department of Health and Human Services. Chandra, A. (2000). Labor market dropouts and the racial wage gap. American Economic Review, 90(2), 333-338. Chandra, A. (2002). Is the convergence in the racial wage gap illusory? Unpublished, Department of Economics, Dartmouth College. Chen, J., Rathore, S.S., Radford, M.J., Wang, Y., and Krumholz, H.M. (2001). Racial differences in the use of cardiac catheterization after acute myocardial infarction. New England Journal of Medicine, 344(19), 1443-1449. Cooper-Patrick, L., Gallo, J.J., Gonzales, J.J., Vu, H.T., Powe, N.R., Nelson, C., and Ford, D.E. (1999). Race, gender and partnership in the patient-physician relationship. Journal of the American Medical Association, 282(6), 583-589. Deaton, A., and Lubotsky, D. (2001). Mortality, inequality and race in American cities and states. (NBER Working Paper No. 8370). Cambridge, MA: National Bureau of Economic Research. Diehr, P., Cain, K.C., Kreuter, W., and Rosenkranz, S. (1992). Can small area analysis detect variation in surgery rates? The power of small area variations analysis. Medical Care, 30(6), 484-502. Exner, D.V., Dries, D.L., Domanski, M.J., and Cohn, J.N. (2001). Lesser response to angiotensin-converting-enzyme inhibitor therapy in black as compared with white patients with left ventricular dysfunction. New England Journal of Medicine, 344, 1351-1357. Fisher, E.S., and Skinner, J. (2001). Geography and the debate over Medicare reform. Health Affairs, 21(1). Fisher, E.S., Wennberg, J.E., Stukel, T.A., and Sharp, S. (1994). Hospital readmission rates for cohorts of Medicare beneficiaries in Boston and New Haven. New England Journal of Medicine, 331, 989-995. Gornick, M.E., Eggers, P.W., Reilly, T.W., et al. (1996). Effects of race and income on mortality and use of services among Medicare beneficiaries. New England Journal of Medicine, 335, 791-799. Harrison, M., and Thurston, N.K. (2001). Racial matching among African-American and Hispanic physicians and patients: Causes and consequences. Journal of Human Resources, 37(2), 410-428. Hopp, F.P., and Duffy, S.A. (2000). Racial variations in end-of-life care. Journal of the American Geriatrics Society, 48(6), 658-663. Jencks, S.F., Cuerdon, T., Burwen, D.R., Fleming, B., Houck, P.M., et al. (2000). Quality of medical care delivered to Medicare beneficiaries: A profile at state and national levels. Journal of the American Medical Association, 284(13), 1670-1676. Johnson, P.A., Lee, T.H., Cook, E.F., Rouan, G.W., and Goldman, L. (1993). Effect of race on the presentation and management of patients with acute chest pain. Annals of International Medicine, 118, 593-601.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life Kennedy, B.P., Kawachi, K.I., and Prothrow-Stith, D. (1996). Income distribution and mortality: Cross-sectional ecological study of the Robin Hood Index in the United States. British Medical Journal, 312, 1004-1007. King, G. (1997). A solution to the ecological inference problem. Princeton, NJ: Princeton University Press. Leape, L.L., Hilborne, L.H., Bell, R., Kamberg, C., and Brook, R.H. (1999). Underuse of cardiac procedures: Do women, ethnic minorities, and the uninsured fail to receive needed revascularization? Annals of International Medicine, 130(3), 183-192. Lillie-Blanton, M., Martinez, R.M., and Salganicoff, A. (2001). Site of medical care: Do racial and ethnic differences persist? Yale Journal of Health Policy, Law, and Ethics, 1(1), 1-17. Massey, D.G. (2001). Residential segregation and neighborhood conditions in US metropolitan areas. In N.J. Smelser, W.J. Wilson, and F. Mitchell (Eds.), America becoming: Racial trends and their consequences (Vol. I). Washington, DC: National Academy Press. McClellan, M., and Staiger, D. (1999, revised 2000). The quality of health care providers. (NBER Working Paper #7327). Cambridge, MA: National Bureau of Economic Research. McNeil, B.J. (2001, November). Shattuck Lecture: Hidden barriers to improvement in the quality of care. New England Journal of Medicine, 345(22). McPherson K, Strong, P.M., Epstein, A, and Jones, L. (1981). Regional variations in the use of common surgical procedures: Within and between England and Wales, Canada and the United States of America. Social Science and Medicine, 15A, 273-288. Morrison, R.S., Wallerstein, S., Natale, D.K., Senzel, R.S., and Huang, L. (2000). We don’t carry that—failure of pharmacies in predominately non-white neighborhoods to stock opioid analgesics. New England Journal of Medicine, 342(14), 1023-1026. National Research Council. (1997). Racial and ethnic differences in the health of older Americans. L.G. Martin and B.J. Soldo (Eds.), Committee on Population, Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review, 14(3), 693-709. Oaxaca, R. (1975). Estimation of union/nonunion wage differentials within occupational/ regional subgroups. Journal of Human Resources, 10(4), 529-537. Oliver, M.L., and Shapiro, T.M. (1995). Black wealth/white wealth: A new perspective on racial inequality. New York: Routledge. Peterson, E.D., Wright, S.M., Daley, J., and Thibault, G.E. (1994). Racial variation in cardiac procedure use and survival following AMI in the Department of Veterans Affairs. Journal of the American Medical Association, 271, 1175-1180. Peterson, E.D., Shaw, L.K., DeLong, E.R., Pryor, D.B., Califf, R.M., and Mark, D.B. (1997). Racial variation in the use of coronary-revascularization procedures: Are the differences real? Do they matter? New England Journal of Medicine, 336, 480-486. Phelps, C.E., and Mooney, C. (1993). Variations in medical practice use: Causes and consequences. In R.J. Arnauld, R.F. Rich, and W. White (Eds.), Competitive approaches to health care reform. Washington, DC: The Urban Institute Press. Rathore, S.S., Berger, A.K., Weinfurth, K.P., et al. (2000). Race, sex, poverty, and the medical treatment of acute myocardial infarction in the elderly. Circulation, 102(6), 642-648. Ross, C.E., and Mirowsky, J. (2000). Does medical insurance contribute to socioeconomic differentials in health? The Millbank Quarterly, 72(2), 291-321. Schneider, E.C., Leape, L.L., Weissman, J.S., Piana, R.N., Gatsonis, C., and Epstein, A.M. (2001). Racial differences in cardiac revascularization rates: Does “overuse” explain higher rates among white patients? Annals of Internal Medicine, 135(5), 328-337.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life Schulman, K.A., et al. (1999). The effect of race and sex on physicians’ recommendations for cardiac catheterization. New England Journal of Medicine, 340(8), 618-626. Schwartz, L.M., et al. (1999). Misunderstandings about the effects of race and sex on physicians’ referrals for cardiac catheterization. New England Journal of Medicine, 341(4), 279-283. Skinner, J., Fisher, E., and Wennberg, J.E. (2001). The efficiency of Medicare. (NBER Working Paper 8395). Cambridge, MA: National Bureau of Economic Research. Skinner, J., Staiger, D., Chandra, A., Lee, J., and McClellan, M. (2002). Racial differences in hospital quality for the treatment of acute myocardial infarction: Evidence from the Medicare population (Unpublished). Dartmouth Medical School. Smedley, B.D., Stith, A.Y., and Nelson, A.R. (2002). Unequal treatment: Confronting racial and ethnic disparities in health care. Washington, DC: National Academies Press. Smith, J.P., and Kington, R.S. (1997). Race, socioeconomic status, and health in late life. In L.G. Martin and B.J. Soldo (Eds.), Racial and ethnic differences in the health of older Americans. Committee on Population, Commission on Behavioral and Social Sciences and Education, National Research Council. Washington, DC: National Academy Press. Susser, M. (1994a). The logic in ecological. The logic of analysis. American Journal of Public Health, 84(5), 825-829. Susser, M. (1994b). The logic in ecological. The logic of design. American Journal of Public Health, 84(5), 830-835. Tu, J.V., Naylor, C.D., Kumar, D., DeBuono, B.A., McNeil, B.J., and Hannan, E.L. (1997). Coronary artery bypass graft surgery in Ontario and New York State: Which rate is right? Annals of Internal Medicine, 126(1), 13-19. Wennberg, J. (1986). Which rate is right? New England Journal of Medicine, 314(5), 310-311. Wennberg, J.E., and Cooper, M. (Eds.). (1996). Dartmouth atlas of health care. Chicago: Dartmouth Medical School and American Hospital Association. Wennberg, J.E., and Cooper, M. (Eds.). (1999). Dartmouth atlas of health care 1999. Chicago: Dartmouth Medical School and American Hospital Association. Wennberg, J.E., Fisher, E., and Skinner, J. (2002). Geography and the debate over Medicare reform. Health Affairs, February, 1-19. Whittle, J., Conigliaro, J., Good, C.B., and Lofgren, R.P. (1993). Racial differences in the use of invasive cardiovascular procedures in the Department of Veterans Affairs medical system. New England Journal of Medicine, 329, 621-627.

OCR for page 604
Critical Perspectives on Racial and Ethnic Differences in Health in Late Life This page intentionally left blank.