Printer Friendly

Racial and ethnic disparities in the use of high-volume hospitals.

Differences in the source of care could contribute to racial and ethnic disparities in health status. This study looks at a major metropolitan area and examines racial and ethnic differences in the use of high-volume hospitals for 17 services for which there is a documented positive volume-outcome relationship. Focusing on the hospitalizations of New York City area residents in the periods 1995-1996 and 2001-2002, we found, after controlling for socioeconomic characteristics, insurance coverage, proximity of residence to a high-volume hospital, and paths to hospitalization, that minority patients were significantly less likely than whites to be treated at high-volume hospitals for most volume-sensitive services. The largest disparities were between blacks and whites for cancer surgeries and cardiovascular procedures.

**********

Although racial and ethnic disparities in health care and treatment outcomes have been extensively documented in the United States, the causes are poorly understood (Institute of Medicine 2002). Disparities persist even when data are adjusted for socioeconomic differences, health insurance status, and other access-related factors. Effective remedial policy depends on an understanding of the causes (Epstein 2004).

Disparities in the process or outcomes of treatment could result from a number of factors including being treated differently, a possibility that has prompted calls for improved "cultural competence" among providers, or receiving care from different providers whose abilities or practice styles vary. Several studies support the former scenario (Schulman et al. 1999; Balsa, McGuire, and Meredith 2005; Green et al. 2007), and there is considerable evidence of racial and ethnic differences in source of care. Minority patients tend to be treated by providers with different training (Bach et al. 2004; Educational Commission for Foreign Medical Graduates 1992) and in different settings than white patients (Lillie-Blanton, Martinez, and Salganicoff 2001; Bradley et al. 2004; Kahn et al. 1994; Birkmeyer et al. 2002; Rothenberg et al. 2004; Barnato et al. 2005; Groeneveld, Laufer, and Garber 2005; Skinner et al. 2005; Hasnain-Wynia et al. 2007; Jha et al. 2008).

A key question is whether the use of different providers has quality implications. Quality shortfalls have been documented among physicians treating more minority patients with regard to credentials, continuity of care, and timely referrals (Bach et al. 2004; Lillie-Blanton, Martinez, and Salganicoff 2001; Hargraves, Cunningham, and Hughes 2001; Mukamel, Murthy, and Weimer 2000). Evidence about racial differences in the quality of the source of hospital care is more mixed (Kahn et al. 1994; Rothenberg et al. 2004; Rask et al. 1994; Leape et al. 1999; Liu et al. 2006), and researchers have only begun to examine whether quality disadvantages may result from the use of different providers by minorities other than blacks (Liu et al. 2006).

This study addresses the extent and causes of racial and ethnic quality disparities in the source of care by examining patterns in the use of high-volume hospitals in the New York metropolitan area for 17 services for which a positive volume-outcome relationship has been documented in previous research. Liu et al. (2006) recently reported that such disparities exist for 10 complex surgical procedures in California, with minorities less likely than whites to use high-volume hospitals and more likely to use low-volume hospitals.

Our study complements the California study in several ways. First, we focus on a racially diverse geographic area in which the entire population lives relatively close to high-volume hospitals. Second, we examine a larger set of services than was examined in the California study, including both surgical and nonsurgical services for which a positive volume-outcome relationship has been documented-that is, patients have better outcomes at higher-volume hospitals. Third, rather than define "high-volume" hospitals as hospitals with the highest 20% of patients, as was done in the California study, we define it in terms of the volume levels used in the empirical studies that have shown better outcomes at higher-volume hospitals. Thus, the number of high-volume hospitals in the New York area varied greatly from service to service (for example, 38 hospitals met the volume threshold for acute myocardial infarction [AMI], while only one hospital was high-volume for gastric cancer surgery and pediatric cardiac surgery). Finally, by studying two different time periods, we can assess the extent to which idiosyncratic factors (e.g., greater awareness of the volume-outcome relationship) might contribute to disparities.

A positive volume-outcome relationship has been documented for a wide array of services, although there is debate about potential explanations (e.g., whether volume results in quality or vice versa, and whether surgeon volume or hospital volume is more important). (For literature reviews, see Halm, Lee, and Chassin 2002; Gandjour, Bannenberg, and Lauterbach 2003.) The robustness of the volume-outcome relationship varies across services. However, in the period covered in this study, volume was identified by the Leapfrog Group and the Agency for Healthcare Research and Quality as a proxy for quality for several surgical procedures and was the basis for some policy decisions in New York state, including the regionalization of AIDS care and coronary bypass surgery.

Methods

Data

This study focuses on residents who lived in the five boroughs of New York City plus the adjacent Nassau and Westchester counties and were hospitalized in that area for any of 17 studied services in the period 1995-1996 (N = 224,533) or 2001-2002 (N = 202,667). Nearly all hospitalizations of residents of these seven counties occur therein, and many hospitalizations of residents of the two suburban counties occur within the city.1 Indeed, in terms of travel time, residents of Westchester have as much or greater access to tertiary hospitals in upper Manhattan than do residents of parts of the city such as Brooklyn or Staten Island. Even so, as a check on our results, we replicated our analysis for several procedures excluding the two suburban counties and found there was no clear pattern of change in our results. Also, out of concern that utilization patterns following events of September 11, 2001, might affect the analysis, we carefully compared utilization patterns in 2001 and 2002. Although a few hospitals experienced a decline in admissions following 9/11, overall utilization patterns for 2001 showed little effect and were quite consistent with patterns in 2002. Our inclusion of data from 1995-1996 provides additional assurance that results were not due to the events of late 2001. (We will primarily discuss the 20012002 results, bringing in the 1995-1996 results only where they shed additional light.)

Patient discharge records from New York's Statewide Planning and Research Cooperative System (SPARCS) are the primary source of data. We included discharges only from short-term general hospitals, excluding Veterans Health Administration hospitals and long-term care units of short-term hospitals. Patients receiving the 17 studied services were identified by International Classification of Diseases, ninth revision (ICD-9-CM) codes in diagnosis and procedure fields (except AIDS, which was defined using the New York state major diagnostic category fields). All patients were age 18 or older (except for pediatric cardiac surgery, which included only patients under 18). The data include clinical information as well as each patient's demographic characteristics, insurance status, and street address or zip code, which were used to measure patients' proximity to high-volume hospitals and, where different, the hospital used. Socioeconomic characteristics of patients at the census tract level were obtained from the U.S. census for 2000.

Services Studied

Although positive volume-outcome relationships had been found for at least 21 hospital services (Dudley et al. 2000; Halm, Lee, and Chassin 2002; Gandjour, Bannenberg, and Lauterbach 2003), we excluded services for which: a) only one hospital provided the service (pediatric cardiac surgery); b) no hospital in the study area met our volume threshold during our study period (surgery for ruptured cerebral aneurysms and for unruptured cerebral aneurysms); and c) prevalence was too low to estimate reliable multivariate models in either of the pairs of years studied (esophageal cancer surgery). The 17 selected services included: five types of cancer surgery (breast, colorectal, gastric, lung, and pancreatic); six cardiovascular services (AMI admissions, coronary artery bypass graft [CABG] surgery, coronary angioplasty, abdominal aortic aneurysm repair, carotid endarterectomy, and pediatric cardiac surgery); three orthopedic procedures (hip fracture repair, total hip replacement, and total knee replacement); two prostate procedures (open and transurethral [TURP] prostatectomy); and admissions for AIDS. Table 1 shows the prevalence of each service during our study periods.

The robustness of the volume-outcome relationship varies among these services. For certain services, including gastric and pancreatic cancer surgery, abdominal aortic aneurysm repair, pediatric cardiac surgery, TURP, and AIDS care, the evidence of the volume-outcome relationship is strong. The evidence is weaker or less consistent for the other services that we studied. We included the whole set for two reasons. First, it is desirable to examine a variety of procedures and conditions since idiosyncrasies may be associated with any particular service. Second, in the period we studied, evidence about the volume-outcome relationship was increasingly available to physicians and patients and thus could have influenced decisions about source of care for any of the services that we included.

Outcome Measure

The dependent variable was whether patients received care from a hospital that met our operational definition of high volume for the specific service for which they were admitted. A hospital was defined as high volume for a particular service if its total annual discharges (averaged for each two-year period) met or exceeded a volume threshold based on the research literature. We used volume standards published by The Leapfrog Group for CABG, coronary angioplasty, abdominal aortic aneurysm repair, carotid endarterectomy, and pancreatic cancer surgery (Birkmeyer and Dudley 2004). For the other services, we used the median volume threshold associated with better outcomes in the literature review by Halm, Lee, and Chassin (2004). The thresholds used are also shown in Table 1.

Explanatory Variables

Patients' race and ethnicity were coded into five mutually exclusive categories from the SPARCS database as follows: Spanish or Hispanic origin (hereafter Hispanic), the non-Hispanic racial categories of white, black, Asian or Pacific Islanders (hereafter Asian), and other. Missing data on patient race and ethnicity (approximately 5.5%) was imputed using a random method technique (Little and Rubin 1987) based on distribution of race and ethnicity in that patient's census tract. (2)

We also included other patient characteristics, many of which have been shown in previous research to be associated with choice of hospital, though not necessarily with racial and ethnic disparities (McGuirk and Porell 1984; Adams et al. 1991; Chernew, Scanlon, and Hayward 1998; Gregory et al. 2000; Wan-Tzu, Porell, and Adams 2004; Basu 2005). These include age, gender, insurance (Medicare, Medicaid, private insurance, self-pay), whether insurance involved managed care, whether the admission was scheduled, and whether the patient had been transferred from another hospital. (3) The descriptive statistics are shown in Table 2. The complexity of the patient's condition was measured by an index of comorbidities that were not consequences of the procedure being performed (Comorbidity Software 2003; Elixhauser et al. 1998).

The proximity of each patient's residence to the hospital used, the nearest high-volume hospital for the service, and the nearest hospital (regardless of volume) that provided the service were calculated using Maptitude Geographic Information System, a combination of software and geographic data. Including both measures of distance suggests that the impact of distance to the nearest high-volume hospital is measured conditional on the proximity of other facilities. This takes into account the propensity of residents in communities or neighborhoods with no local hospital to travel farther, whatever the attributes of the hospital at which they are eventually treated. Conversely, the propensity of a nearby low-volume facility to draw patients away from a high-volume hospital also depends on how much farther patients would need to travel to get to a high-volume facility. If an address was missing (19.6% of the sample in 2001-2002), the centroid of the patient's zip code was used for proximity analyses. (For reasons of confidentiality, patients' addresses were not present in the SPARCS database for AIDS discharges; excluding those cases, address information was missing in fewer than 10% of cases.)

Socioeconomic status was measured by average household income and average educational attainment for residents over age 25 in the patient's census tract. To account for possible differences in patients' receiving word-of-mouth information about available hospitals (National Survey on Consumer Experiences 2001), we included a measure of the prevalence of each service in each census tract.

Statistical Analyses

We first calculated the raw percentage of patients from each racial and ethnic category who used a high-volume hospital for each service studied. To assess the distinctive influence of race and ethnicity on these differences, we then estimated multivariate logistic regression models to determine the magnitude of racial and ethnic disparities controlling for all of the factors described previously, including insurance status and proximity to a high-volume hospital. Separate models were estimated for each service. All analyses used SAS version 8.02 (The SAS System 2001). Statistical significance was set at the p < .05 level.

We present the results of the multivariate analysis as regression-adjusted probabilities of using a high-volume facility, rather than as odds ratios or relative risks. There is active but inconclusive debate over the most appropriate method for presenting such results (McNutt et al. 2003; Cook 2002). Because the percentage of patients who used high-volume hospitals varies greatly across services (from single digits to the mid-90th percentiles), comparisons across procedures in terms of either odds ratios or relative risks ratios could be misleading. By reporting predicted probabilities, we allow the reader to decide whether disparities are considered most meaningfully in terms of absolute differences in use of high-volume facilities or the relative rates for minority compared to white patients. (4)

Results

Basic Differences in the Use of High-Volume Hospitals

There were many large racial and ethnic differences in the use of high-volume hospitals for the volume-sensitive services studied, (5) and the results we report are consistent when both unadjusted and regression-adjusted results are Considered. As shown in Table 3, black patients were significantly less likely than white patients to use a high-volume hospital for all but one service examined in 2001-2002; the exception was admissions for AIDS, where the opposite pattern was found. Hispanics were less likely than whites to use high-volume hospitals in 15 of the 17 services, but more likely for AIDS. (It is likely that the data for the 17 services somewhat exaggerate the use of high-volume hospitals among Hispanics, because several public hospitals that were not high volume on most services coded few patients as Hispanic, despite location in heavily Hispanic neighborhoods. (6)) Asians used high-volume hospitals less frequently than whites for 13 services. The differences were most pronounced among cancer surgeries and cardiovascular procedures, averaging over 20 percentage points between blacks and whites for these 11 services. For orthopedic and prostate procedures, the racial disparities between blacks and whites were smaller but still statistically significant. Differences between whites and Hispanics and Asians were generally smaller than the black-white difference.

As can be seen from the data from the mid1990s (Table 4), these disparities were not new and thus cannot be attributed to idiosyncratic factors or to greater dissemination in the white population of the then-developing evidence about the volume-outcome relationship. In the 1995-1996 period, blacks were significantly less likely than whites to use a high-volume hospital for the same 16 services, with AIDS again the exception. The differences were rather stable over our two time periods, though there was some growth in the black-white disparity associated with cancer surgery. Hispanics were less likely than whites to use high-volume hospitals for 12 services during 1995-1996, and Asians were less likely than whites on 14 services. Clearly the general pattern of racial and ethnic disparities was not peculiar to the 2001-2002 period.

Isolating the Effects of Race and Ethnicity

Our multivariate regression found that a number of factors other than race and ethnicity were related to the use of high-volume hospitals for volume-sensitive services. (7) Table 5 illustrates this with results for four of the most common services--coronary artery bypass grafts, breast cancer surgery, total knee replacement, and transurethral resection of the prostate. (These were selected because their prevalence made the regression estimates more reliable and because they represented one example from each of our four categories of surgical procedures.)

In the entire set of services studied, the most consistent predictors other than race and ethnicity were: 1) residential proximity to a high-volume facility (greater distance significantly reduced patients' use of high-volume hospitals for 14 of 17 services in 2001-2002); 2) patient age (older patients were significantly less likely to be treated at high-volume hospitals for 11 services); and 3) socioeconomic characteristics of the patient's neighborhood (those from neighborhoods with higher income and education levels were significantly more likely to use high-volume hospitals for more than half of the services studied). Patient insurance also displayed some significant effects. For about half of the services, patients with Medicare or commercial insurance were significantly more likely to use a high-volume hospital, compared to Medicaid or self-paying patients. Managed care insurance was associated with greater use of high-volume hospitals for cancer surgery and cardiovascular procedures, although findings were mixed among orthopedic and prostate surgeries. Patients who had scheduled admissions or were transferred from another hospital were more likely to be treated at a high-volume hospital for cancer surgery and cardiovascular procedures. Neighborhood prevalence of the services was associated with greater use of high-volume hospitals for five services (two cancer, two orthopedic, and AIDS services) in the period 2001-2002.

Notwithstanding the importance of these various factors, we found that in both pairs of years the racial and ethnic disparities in the use of high-volume hospitals were essentially unchanged when we controlled for these other factors (Tables 3 and 4). (8) The small number of substantive differences that disappeared were for services for which few patients used a high-volume hospital (gastric cancer surgery and hip fracture repair) or for which most patients used a high-volume hospital (angioplasty, AIDS). Substantial black-white differences remained for the other 13 services; one or two differences with whites disappeared for Hispanics and Asians.

To check further whether the differences persisted when there was a common primary payer, we compared Medicare enrollees who had CABG, breast cancer surgery, total knee replacement, or TURPs with patients who did not have Medicare (data not shown). (This is not the same as a comparison by age, since 16% of patients age 65 and older from New York City did not have Medicare as a payer [Gray et al 2006].) We found that the racial and ethnic disparities persisted, even with a common payer. Although white-black disparities were smaller among Medicare patients than among other payer types for three of the four procedures (TURPs being the exception), disparities between whites and Hispanics and Asians tended to be larger among Medicare patients.

In sum, for a majority of services, accounting for sociodemographics, insurance, proximity, and path to hospitalization had little impact on racial and ethnic disparities. The introduction of controls eliminated such disparities in a few cases and substantially reduced them in a few more. In some cases (e.g., repair of abdominal aortic aneurysms), the measured disparities were actually larger when controls were introduced. Altogether, averaging across all procedures/conditions, the other predictor variables accounted for only about 25% of the apparent magnitude of the racial and ethnic disparities in the unadjusted data.

Commentary

Several points are notable about the patterns of racial and ethnic disparities that we have documented in the use of high-volume New York area hospitals for volume-sensitive services. The first concerns the magnitude of the differences. For several services, whites were more than twice as likely as blacks to use a high-volume hospital, and for several others there was a 20 to 30 percentage-point difference in such usage. The magnitude of the disparities with whites was smaller for Hispanics and Asians than for blacks, but even at these reduced levels, the impact of race and ethnicity for these groups still was larger than that associated with lack of insurance or an unplanned admission.

Second, racial and ethnic disparities in use of high-volume hospitals were found not only for services for which admissions were largely unplanned (e.g., AMI and hip fracture repair), but also for services that generally are planned in advance (e.g., the cancer surgical procedures and hip and knee replacement surgery). Thus, it is probably not surprising that the disparities are not a simple function of proximity. As Table 2 shows, whites on average live farther from high-volume hospitals for the services studied than do the minority groups. This is partially a function of whites' disproportionately living in the two suburban counties; however, even within the city of New York, blacks on average live closer to high-volume hospitals than whites (data not shown). Of course, distance does not translate directly into convenience in a metropolitan area such as New York, but lack of proximity is generally not an advantage regarding ease of access.

Third, the disparities we found are not a simple function of insurance status. It is true that they are exacerbated by the disproportionate numbers of minorities among Medicaid and self-pay patients, since their use of high-volume hospitals tends to be lower than among Medicare and privately insured patients. However, within every insurance grouping--private insurance, Medicare, Medicaid, and self-pay--whites were more likely than minorities to use a high-volume hospital.

Fourth, we found racial and ethnic disparities both for services provided at a large number of hospitals (12 studied services were provided at more than 75 hospitals in our study area in 2001-2002) and services provided at only a few hospitals. Among the former, breast cancer surgery--for which whites were twice as likely as blacks to use a high-volume hospital in 2001-2002--was performed at 91 hospitals, of which 10 met the volume threshold. Similarly, lung cancer surgery--for which whites were more than 33 percentage points more likely than blacks to use a high-volume hospital in 2001-2002--was provided at 80 hospitals, of which 25 met the volume threshold.

At the other end, only 21 hospitals performed CABG surgery in 2001-2002 (with nine meeting the threshold), but there was a substantial white-black difference (72% vs. 52%) in the use of high-volume hospitals. Pediatric cardiac surgery was provided at only 13 hospitals in 2001-2002, and 43% of whites compared with 25% of blacks used the single hospital that met the volume threshold.

In a similar fashion, significant disparities were seen both among services for which large numbers of hospitals met the volume threshold (e.g., AMI and TURP) and among services in which only one hospital met the threshold (gastric cancer and pediatric cardiac surgery).

As noted previously, services for which disparities were relatively small, either in absolute terms or when adjusted for potential confounding factors, tended to have very high or very low use of high-volume hospitals. The former included angioplasty and admissions for AIDS, which were also the two services that had the highest proportion of high-volume hospitals among the facilities that provided the services (51 high-volume hospitals of the 101 hospitals that treated AIDS patients and 18 of the 30 that did angioplasties). The regression-adjusted disparities were particularly small for gastric cancer surgery and hip fracture repair. Overall use of high-volume hospitals was particularly low for these two services (fewer than 20% of patients used a high-volume hospital), perhaps because very few service-providing hospitals were above the volume threshold--only one of the 80 that treated gastric cancer and six of the 92 that performed hip fracture repair.

If proximity and insurance status do not account for the racial and ethnic differences in the use of hospitals, what might do so? One possible explanation lies with factors associated with consumer/patient information and choice. Evidence linking hospital volume to outcomes was already well-recognized in the clinical literature by the mid-1990s, and statistics on hospital volume in New York were becoming available to the public. Minority patients might have been less likely to be aware of these data or to appreciate their importance, or they may have been more concerned about other hospital attributes when deciding (or discussing with their doctor) where to be treated.

We have no information on these considerations for the New York metropolitan area, but a statewide survey found that more than half the residents of New York state saw hospital quality as a concern and that minority respondents were most likely to be concerned (Boscarino and Adams 2004). Our own tabulations of data from a Kaiser Family Foundation survey (2000) suggest that most Americans view hospital quality as variable and that minority respondents are no different in this regard (Table 6). Even so, consumers concerned about quality might not view volume as a proxy measure. However, data from the Kaiser survey suggests that by 2000, virtually all of the public recognized that volume could predict quality (for both hospitals and physicians), and African Americans and Latinos were just as likely to make this linkage as were white Anglos.

Even consumers who understand that volume has relevance for quality might not have access to the pertinent information at the time it is needed. Surveys of both New York state and the United States suggest that consumers from minority communities report less exposure than whites to information about physician or hospital performance (Boscarino and Adams 2004). Although such differences are modest (see Table 6), they could account for some variation in the influence of volume on hospital choice.

It is possible that some consumers who value and have access to information regarding volume might nevertheless give more emphasis to other characteristics of health care providers, in addition to their proximity (which we've controlled for statistically in the regression-adjusted results presented earlier). Most consequentially, some consumers may prefer health care providers with whom they are familiar, based on personal experience or word-of-mouth from family and friends. These experiences might seem particularly salient for minority patients, who typically are less likely than white patients to assume that medical professionals are generically trustworthy (Schnittker and Liang 2006).

The Kaiser survey data (Table 6) suggest that many Americans prefer a surgeon with whom they, their family, or their friends have had positive experiences over one with a good statistical profile, and this preference is more common among African Americans and Latinos than whites. But preferences for hospitals look a bit different--although African Americans again prefer familiar hospitals to a greater extent than do other consumers, Latinos actually give less emphasis to familiarity than do non-Latino whites. Consequently, although this aspect of preferences might contribute to the disparities in high-volume hospital use between whites and African Americans, it cannot explain the equally large disparities that we observed between whites and Latino patients.

We believe that the most likely explanation for the racial and ethnic differences we have found rests less with factors pertaining to patients' choice of hospitals than with differences in the paths by which patients reach particular hospitals, beginning with whatever physician is first seen and the referral process that subsequently transpires (Burns and Wholey 1992; Wan-Tzu, Porell, and Adams 2004). That process would likely be affected by factors such as where the initial physician has hospital privileges or to which specialists he or she customarily makes referrals. Information about referring physicians was not available in the SPARCS data. The role of physicians' referral patterns in creating racial and ethnic differences in the hospitals used merits further investigation.

Importance of Our Findings

The size of the disparities that we have documented have clinical relevance and policy importance. We applied the volume-related mortality differences demonstrated in previous research, as reported by Halm, Lee, and Chassin (2002), to the disparities that we uncovered. In doing so, we estimated that the racial and ethnic disparities that we found would be associated with 5% to 20% higher service-specific mortality rates among minority patients than among white patients for common services, and 10% to 60% higher mortality rates for infrequent procedures (Table 7). (9) The differential use rates in these calculations were based on the regression-adjusted results, and therefore controlled for socioeconomic status, insurance coverage, and proximity. These differentials were large, both in an absolute sense and in comparison to findings from reviews of racial and ethnic disparities in clinical outcomes (Institute of Medicine 2002).

A second way of assessing the impact of the racial and ethnic disparities we found is to compare the impact of minority status with the effects of other variables included in the regression analyses. For example, our analysis of cardiovascular procedures and conditions in 2001-2002 shows that, on average, being black has almost three times the inhibiting effect on reaching a high-volume hospital as does lacking health insurance (compared to patients with commercial insurance), and one-and-a-half times the inhibiting effect as having an unscheduled admission to the hospital (compared to scheduled admissions). Among Hispanics and Asians, the magnitude of the disparities is smaller than those identified for blacks. Even at these reduced levels, the impact of race and ethnicity for these groups is still larger than those associated with lack of insurance or an unplanned admission.

Public Policy and Racial and Ethnic Disparities

Our findings suggest that simply having more high-volume hospitals will not necessarily reduce racial and ethnic disparities. We observed procedures and conditions (TURP and AMI) in which between 35 and 50 hospitals exceeded the volume threshold in 2001-2002, yet the magnitude of disparities involving blacks (either adjusted or unadjusted) were comparable to other procedures in the same clinical grouping that involved fewer hospitals. Examining hip replacement and pancreatic cancer surgery, two procedures for which the number of facilities exceeding the volume threshold increased substantially between 1995-1996 and 2001-2002, we see that the unadjusted racial disparities actually increased over this period (the regression-adjusted disparities did decline slightly, in both cases, but remain large and statistically significant).

Consolidation of care at specific hospitals exhibits more substantial, but somewhat inconsistent, effects on racial and ethnic disparities. As a result of New York state policies to limit the number of hospitals authorized to perform certain cardiac procedures (Chassin 2002; Hannan et al. 1998), CABG surgery, angioplasty, and pediatric cardiac surgery were done at only 22, 29, and 19 hospitals, respectively, in 1995-1996, compared to most other procedures examined that were done at more than 90 hospitals. Although the modest volume-related disparities in angioplasty in 1995-1996 largely disappeared by 2001-2002, large disparities in both CABG and pediatric cardiac surgery can be seen in both pairs of years. Even so, the magnitude of disparities was smaller compared to other cardiovascular procedures.

A different policy strategy affected AIDS care and may help explain why the typical minority disadvantages were not found for AIDS patients. New York subsidized the formation of large numbers of "designated AIDS centers," strategically placed in hospitals located primarily in neighborhoods with a high prevalence of AIDS (New York State Department of Health 2005). Our findings show that, among patients hospitalized for AIDS, minority patients were actually more likely than whites to use high-volume hospitals, even though the number of high-volume hospitals for AIDS was only slightly greater than the number of high-volume facilities for prostate problems or AMIs, where disparities were substantial.

Limitations

Our findings should be interpreted in light of several limitations. First, data on patient race and ethnicity are as reported by hospitals, and it is apparent from the missing data and from the use of the "other" race category that their procedures for collecting and recording this information vary. It has been shown previously that there is measurement error, especially for nonblack and nonwhite categories (Blustein 1994). However, this should not bias differential rates across high- versus low-volume hospitals. Second, thresholds used to define high-volume hospitals are inherently arbitrary and may mask more subtle variations either above or below the threshold. Because hospitals that fall slightly below the high-volume threshold are treated as having a similar volume as hospitals well below the threshold, our findings may overstate or understate the significance of the volume-disparity association. This factor should be kept in mind in comparing disparities over time, since relatively minor shifts in volume could alter the number of hospitals identified as "high volume."

Third, because our study is limited to hospitals in the New York City metropolitan area, hospitalizations of area residents outside this geographic area are excluded. However, 99.6% of the hospital use in New York state by residents of the New York City area was in hospitals in our catchment area. Fourth, because of low prevalence, we were forced to exclude from our analysis some procedures for which there were large disparities in the raw data (e.g., esophageal cancer surgery, cerebral aneurysm repair). Because lower frequency procedures are often associated with large volume-related mortality differentials (Birkmeyer et al. 2002), our findings may understate the extent of clinically meaningful disparities. However, precisely because these are uncommon procedures, disparities in their use will have relatively small consequences for the overall health of minority populations.

Other limitations highlight the need for additional research. Clearly, the New York metropolitan area is not representative of the country as a whole. In locales where patients must travel longer distances to get to hospitals, the dynamics of hospital use and the determinants of racial and ethnic utilization patterns may be different from those identified in this study. Also, we did not examine disparities in physician volume in this study, but the literature shows that physician volume has an independent effect on quality (Birkmeyer et al 2003).

Conclusions

Our findings show that in a large, racially diverse metropolitan area, where high-volume hospitals are relatively accessible geographically to the entire population, substantial racial and ethnic differences exist in the use of high-volume hospitals for services for which high volume is associated with better outcomes. Disparities between whites and blacks are large in magnitude, consistent across different services, and persistent over time. Hispanics and Asians exhibit similar patterns of disparities, although the magnitude and consistency is not as strong compared to blacks. These disparities cannot be attributed to differences in socioeconomic status, insurance coverage, or neighborhood of residence.

As noted earlier, our findings raise important questions about the processes by which racial and ethnic groups gain access to medical care. While researchers probe such questions, it is important to consider whether racial and ethnic disparities in the use of high-volume hospitals may be amenable to policy intervention, at least in large metropolitan areas. In particular, policies that increase the number of high-volume hospitals among providers of a given service seem especially promising. In New York, the designated AIDS care center program has directed resources to hospitals located in areas where cases are concentrated, facilitating high use in multiple facilities. Similarly, New York's certificate-of-need program that limited the number of institutions that perform a procedure as common as angioplasty resulted in a majority of hospitals exceeding the volume threshold. When the majority of patients use high-volume hospitals, there are fewer racial and ethnic disparities.

References

Adams, E. K., R. Houchens, G. E. Wright, and J. Robbins. 1991. Predicting Hospital Choice for Rural Medicare Beneficiaries: The Role of Severity of Illness. Health Services Research 26(12):583-612.

Bach, P. B., H. H. Pham, D. Schrag, R. C. Tate, and J. L. Hargraves. 2004. Primary Care Physicians Who Treat Blacks and Whites. New England Journal of Medicine 351 (6):575-584.

Balsa, A. I., T. G. McGuire, and L. S. Meredith. 2005. Testing for Statistical Discrimination in Health Care. Health Services Research 40(1): 227-252.

Barnato, A. E., F. L. Lucas, D. Staiger, D. E. Wennberg, and A. Chandra. 2005. Hospital-level Racial Disparities in Acute Myocardial Infarction Treatment and Outcomes. Medical Care 43(4):308-319.

Basu, J. 2005. Severity of Illness, Race, and Choice of Local Versus Distant Hospitals Among the Elderly. Journal of Health Care for the Poor and Underserved 16(2):391-405.

Birkmeyer, J. D., A. E. Siewers, E. V. Finlayson, T. A. Stukel, F. L. Lucas, I. Batista, H. G. Welch, and D. E. Wennberg. 2002. Hospital Volume and Surgical Mortality in the United States. New England Journal of Medicine 346(15):1128-1137.

Birkmeyer, J. D., T. A. Stukel, A. E. Siewers, P. P. Goodney, D. E. Wennberg, and F. L. Lucas. 2003. Surgeon Volume and Operative Mortality in the United States. New England Journal of Medicine 349(22):2117-2127.

Birkmeyer, J. D., and R. A. Dudley. 2004. Evidence-Based Hospital Referral Factsheet. Washington, D.C.: The Leapfrog Group for Patient Safety.

Blustein, J. 1994. The Reliability of Racial Classifications in Hospital Discharge Abstract Data. American Journal of Public Health 84(6):1018-1021.

Boscarino, J., and R. Adams. 2004. Public Perceptions of Quality of Care and Provider Profiling in New York: Implications for Improving Quality of Care and Public Health. Journal of Public Health Management and Practice 10(3):241-250.

Bradley, E. H., J. Herrin, Y. Wang, R. L. McNamara, T. R. Webster, D. J. Magid, M. Blaney, E. D. Peterson, J. G. Canto, C. V. Pollack Jr., and H. M. Krumholz. 2004. Racial and Ethnic Differences in Time to Acute Reperfusion Therapy for Patients Hospitalized with Myocardial Infarction. Journal of the American Medical Association 292(13): 1563-1572.

Burns, L. R., and D. R. Wholey. 1992. The Impact of Physician Characteristics in Conditional Choice Models for Hospital Care. Journal of Health Economics 11(1):43-62.

Chassin, M. R. 2002. Achieving and Sustaining Improved Quality: Lessons from New York State and Cardiac Surgery. Health Affairs 21(4):40-51.

Chernew, M., D. Scanlon, and R. Hayward. 1998. Insurance Type and Choice of Hospital for Coronary Artery Graft Surgery. Health Services Research 33(3 P.1):447-466.

Comorbidity Software. 2003. Rockville, Md.: Agency for Healthcare Research and Quality.

Cook, T. 2002. Advanced Statistics: Up with Odds Ratios! A Case for Odds Ratios When Outcomes Are Common. Academic Emergency Medicine 9(12):1430-1434.

Dudley, R. A., K. L. Johansen, R. Brand, D. J. Rennie, and A. Milstein. 2000. Selective Referral to High-volume Hospitals: Estimating Potentially Avoidable Deaths. Journal of the American Medical Association 283(9): 1159-1166.

Educational Commission for Foreign Medical Graduates: Annual Report. 1992. Philadelphia, Pa.: Educational Commission for Foreign Medical Graduates.

Elixhauser, A., C. Steiner, D. R. Harris, and R. M. Coffey. 1998. Comorbidity Measures for Use with Administrative Data. Medical Care 36(1):8-27.

Epstein, A. M. 2004. Health Care in America Still too Separate, Not Yet Equal. New England Journal of Medicine 351(6):603-605.

Gandjour, A., A. Bannenberg, and K. W. Lauterbach. 2003. Threshold Volumes Associated with Higher Survival in Health Care: A Systematic Review. Medical Care 41(10): 1129-1141.

Gray, B. H., R. Scheinmann, P. Rosenfeld, and R. Finkelstein. 2006. Aging without Medicare: Evidence from New York City. Inquiry 43(3):211-221.

Green, A. R., D. R. Carney, D. J. Pallin, L. H. Ngo, K. L. Raymond, L. I. Iezzoni, and M. R. Banaji. 2007. Implicit Bias Among Physicians and its Prediction of Thrombolysis Decisions for Black and White Patients. Journal of General Internal Medicine 22:1231-1238.

Gregory, P. M., E. S. Malka, J. B. Kostis, A. C. Wilson, J. K. Arora, and G. G. Rhodes. 2000. Impact of Geographic Proximity to Cardiac Revascularization Services on Service Utilization. Medical Care 38(1):45-57.

Groeneveld, P. W., S. W. Laufer, and A. M. Garber. 2005. Technology Diffusion, Hospital Variation, and Racial Disparities among Elderly Medicare Beneficiaries. Medical Care 43(4):320-329.

Halm, E. A., C. Lee, and M. R. Chassin. 2002. Is Volume Related to Outcome in Health Care? A Systematic Review and Methodologic Critique of the Literature. Annals of Internal Medicine 137(6):511-520.

Hannan, E. L., M. Racz, R. E. Kavey, J. M. Quaegebeur, and R. Williams. 1998. Pediatric Cardiac Surgery: The Effect of Hospital and Surgeon Volume on In-hospital Mortality. Pediatrics 101(6):963-969.

Hargraves, J. L., P. J. Cunningham, and R. G. Hughes. 2001. Racial and Ethnic Differences in Access to Medical Care in Managed Care Plans. Health Services Research 36(5):853-868.

Hasnain-Wynia, R., D. W. Baker, D. Nerenz, J. Feinglass, A. C. Beal, M. B. Landrum, R. Behal, and J. S. Weissman. 2007. Disparities in Health Care Are Driven by Where Minority Patients Seek Care: Examination of the Hospital Quality Alliance Measures. Archives of Internal Medicine 167:1233-1239.

Institute of Medicine. 2002. Unequal Treatment: Confronting Racial and Ethnic Disparities in Health Care. Washington, D.C.: National Academies Press.

Jha, A. K., E. J. Orav, J. Zheng, and A. M. Epstein. 2008. The Characteristics and Performance of Hospitals that Care for Elderly Hispanic Americans. Health Affairs 27(2): 528-537.

Kahn, K. L., M. L. Pearson, E. R. Harrison, K. A. Desmond, W. H. Rogers, L. V. Rubenstein, R. H. Brook, and E. B. Keeler. 1994. Health Care for Black and Poor Hospitalized Medicare Patients. Journal of the American Medical Association 271(15):1169-1174.

Kaiser Family Foundation/Agency for Healthcare Research and Quality. 2000. National Survey on Americans as Health Care Consumers: An Update on the Role of Quality Information. Menlo Park, Calif.: Kaiser Family Foundation.

Leape, L. L., L. H. Hilborne, R. Bell, C. Kamberg, and R. H. Brook. 1999. Underuse of Cardiac Procedures: Do Women, Ethnic Minorities, and the Uninsured Fail to Receive Needed Revascularization? Annals of Internal Medicine 130(3):183-192.

Lillie-Blanton, M., R. M. Martinez, and A. Salganicoff. 2001. Site of Medical Care: Do Racial and Ethnic Differences Persist? Yale Journal of Health Policy, Law, and Ethics 1:15-32.

Little, R. A., and D. B. Rubin. 1987. Statistical Analysis with Missing Data. New York: John Wiley.

Liu, J. H., D. S. Zingmond, M. L. McGory, N. F. SooHoo, S. L. Ettner, R. H. Brook, and C. Y. Ko. 2006. Disparities in the Utilization of High-Volume Hospitals for Complex Surgery. Journal of the American Medical Association 296(16):1973-1980.

Luft, H. S. 2003. From Observing the Relationship Between Volume and Outcome to Making Policy Recommendations. Medical Care 41(10): 1118-1122.

McGuirk, M. A., and F. W. Porell. 1984. Spatial Patterns of Hospital Utilization: The Impact of Distance and Time. Inquiry 21(1):84-95.

McNutt, L. A., C. Wu, X. Xue, and J. P. Hafner. 2003. Estimating the Relative Risk in Cohort Studies and Clinical Trials of Common Outcomes. American Journal of Epidemiology 157(10):940-943.

Mukamel, D. B., A. S. Murthy, and D. L. Weimer. 2000. Racial Differences in Access to High-Quality Cardiac Surgeons. American Journal of Public Health 90(11):1774-1777.

National Survey on Consumer Experiences With and Attitudes Toward Health Plans. 2001. Menlo Park, Calif.: Kaiser Family Foundation/ Harvard School of Public Health.

New York State Department of Health. 2005. Health Care Services. Designated AIDS Centers. http://www.health.state.ny.us/diseaseslaids/ about/hlthcare.htm**dacs. Accessed November 15, 2006.

Rask, K. J., M. V. Williams, R. M. Parker, and S. E. McNagny. 1994. Obstacles Predicting Lack of a Regular Provider and Delays in Seeking Care for Patients at an Urban Public Hospital. Journal of the American Medical Association 271(24):1931-1933.

Rothenberg, B. M., T. Pearson, J. Zwanziger, and D. Mukamel. 2004. Explaining Disparities in Access to High-Quality Cardiac Surgeons. Annals of Thoracic Surgery 78(1):18-24.

Schnittker, J., and K. Liang. 2006. The Promise and Limits of Racial/Ethnic Concordance in Physician-Patient Interaction. Journal of Health Politics, Policy and Law 31(4):811838.

Schulman, K. A., J. A. Berlin, W. Harless, J. F. Kerner, S. Sistrunk, B. J. Gersh, R. Dube, C. K. Taleghani, J. E. Burke, S. Williams, J. M. Eisenberg, and J. J. Escarce. 1999. The Effect of Race and Sex on Physicians' Recommendations for Cardiac Catheterization. New England Journal of Medicine 340(8):618-626.

Skinner, J., A. Chandra, D. Staiger, J. Lee, and M. McClellan. 2005. Mortality after Acute Myocardial Infarction in Hospitals that Disproportionately Treat Black Patients. Circulation 112: 2634-2641.

The SAS System. Cary, N.C.: The SAS Institute.

Wan-Tzu, C. T., F. W. Porell, and E. K. Adams. 2004. Hospital Choice of Rural Medicare Beneficiaries: Patient, Hospital Attributes, and the Patient-Physician Relationship. Health Services Research 39(6 Pt 1): 1903 1922.

Notes

The authors acknowledge the advice and assistance of Peter Bach, John Billings, Diane Blum, Jan Blustein, Benjamin Chu, Spencer Foreman, Ethan Halm, Karen Heller, Robert Knauf Arthur Levin, Dana Mukamel, Steve Schoenbaum, John Shaw, Marjorie Shulman, James Tallon, Gerald Thompson, and Lynn Wozniak.

(1) Hospitals in all seven counties belong to the Greater New York Hospital Association.

(2) The use of imputed racial and ethnic data in this manner will not bias our estimates, even if there is a relationship between missing data on race/ethnicity and the probability of a hospital exceeding the volume threshold for a given service. (We did not observe any obvious relationship of this sort.) Imputation adds to the imprecision of measuring race and ethnicity (which must always be understood as markers that incorporate considerable subjective interpretation) and may increase the standard errors on the regression estimates, but should not bias those estimates.

(3) Patients who received surgical services were not counted twice since they received these services only at the hospital to which they were transferred. Some patients with medical diagnoses (AIDS, AMI) could have been in data from more than one hospital.

(4) Reporting probabilities does introduce some complications. The probabilities reported in the text are calculated from the regression coefficients and the average values of the explanatory variables for each minority group. Because the models are estimated for the combined sample, they are appropriate for the full-sample means and not necessarily for the subpopulations represented for each minority group. We assessed the potential biases this might introduce for several procedures and they were small, making the results reported in the text reasonable approximations of the true probability differences among subgroups in our sample.

(5) We infer racial and ethnic differences from whites based on the statistical significance of the racial and ethnic coefficients in the logit models. Although we report in the text results for identifiable ethnic and racial groups, those who were identified as "other race" were also significantly less likely than whites to use high-volume facilities. These "other race" disparities were reasonably consistent across the 17 volume-sensitive procedures studied here and of equal magnitude as those for other minorities, with the exception of some cancer surgeries.

(6) These hospitals typically coded many patients as "other" on race and did not complete the "Hispanic" ethnicity question.

(7) Our statements about statistical significance are inferred from the statistical significance of the race/ethnicity coefficients in the logit model.

(8) We also calculated results with an adjustment for the nonlinear nature of the estimating model and found the results were virtually unchanged. The adjustment accounted for less than .1% of the change when rates were adjusted.

(9) Our estimation of mortality effects was done quite simply by multiplying the estimated mortality differences for each procedure, based on the research reported by Halm, Lee, and Chassin (2002), by the probability that patients from different racial and ethnic backgrounds used a high-volume facility, and the prevalence of the procedures.

Bradford H. Gray, Ph.D., is a senior fellow at the Urban Institute. Mark Schlesinger, Ph.D., is a professor of public health at the Yale University School of Medicine. Shannon Mitchell Siegfried, Ph.D., is an independent scholar. Emily Horowitz, Ph.D., is an assistant professor at St. Francis College. This research was supported by grants to the New York Academy of Medicine by the Agency for Healthcare Research and Quality and the Commonwealth Fund Address correspondence to Dr. Gray at the Urban Institute, 2100 M St., N. W., Washington DC20037. Email: bgray@urban.org
Table 1. Number of hospitals meeting high-volume threshold and number
of procedures

                                                    Study years

                                               1995-1996

Procedure/conditions          High-volume    High-volume       Total
                               threshold      hospitals      procedures

Cancer surgery
  Breast cancer                 151 (a)           10           10,323
  Colorectal cancer             115 (a)            7            7,484
  Gastric cancer                 63 (a)            1            1,106
  Lung cancer                    19 (a)           25            2,598
  Pancreatic cancer              11 (b)            5              293
Cardiovascular
  AMI                           238 (a)           38           36,941
  CABG                          450 (b)           13           17,834
  Angioplasty                   400 (b)           14           19,991
  Abdominal aortic               50 (a)            7            1,871
    aneurysms
  Carotid endarterectomy         50 (a)           18            5,191
  Pediatric cardiac surgery     300 (a)            1            1,187
Orthopedic
  Hip fracture repair           136 (c)           10           11,404
  Hip replacement               100 (a)           19           12,144
  Knee replacement              200 (a)            4            7,902
Prostate surgery
  Open prostatectomy             98 (c)            4            3,624
  TURP                           61 (c)           51           12,758
AIDS                            100 (a)           60           65,279

                                    Study years

                                2001-2002

Procedure/conditions          High-volume       Total
                               hospitals      procedures

Cancer surgery
  Breast cancer                     6            7,921
  Colorectal cancer                 9            7,434
  Gastric cancer                    1            1,038
  Lung cancer                      21            2,725
  Pancreatic cancer                 9              375
Cardiovascular
  AMI                              39           37,830
  CABG                              9           16,226
  Angioplasty                      18           37,599
  Abdominal aortic                  3            1,082
    aneurysms
  Carotid endarterectomy           18            4,690
  Pediatric cardiac surgery         1            1,036
Orthopedic
  Hip fracture repair               6            9,784
  Hip replacement                  29           14,086
  Knee replacement                  7           11,616
Prostate surgery
  Open prostatectomy                6            3,862
  TURP                             36            8,547
AIDS                               51           36,813

(a) Based on median threshold reported in Halm, Lee, and Chassin
(2002).

(b) The Leapfrog Group recommended volume standard.

(c) Based on mean threshold (only two studies available) reported in
Halm, Lee, and Chassin (2002).

Table 2. Descriptive statistics of study population, New York
metropolitan area, 2001-2002

                                            Racial and ethnic status

                                              White,         Black,
                                           non-Hispanic   non-Hispanic

Sociodemographics
  Mean age                                     69.7          57.3 *
  Percent male                                 48.2          42.8 *
  Mean income of census tract (000$)           58.9          36.9 *
  Percent college educated in census           39.2          23.3 *
    tract
Insurance characteristics
  Percent covered by commercial                17.6          15.1 *
    insurance
  Percent covered by Medicare                  52.1          29.7 *
  Percent covered by Medicaid                   7.2          30.4 *
  Percent enrolled in managed care plan        21.6          19.6 *
Treatment process
  Mean number of comorbidities                   .99          1.01 *
  Percent scheduled admission                  28.6          21.1 *
  Percent transfer from other hospital          6.4           5.3 *
  Prevalence of procedure in census             1.0           1.9 *
    tract (per 10,000 residents)
Geographic accessibility
  Mean miles to nearest hospital                4.5           2.3 *
  Mean miles to nearest high-volume             7.0           5.1 *

                                           Racial and ethnic status

                                           Hispanic    Asian

Sociodemographics
  Mean age                                  56.5 *     62.9 *
  Percent male                              46.7 *     51.2 *
  Mean income of census tract (000$)        35.6 *     45.5 *
  Percent college educated in census        24.7 *     34.7 *
    tract
Insurance characteristics
  Percent covered by commercial             14.2 *     16.6 *
    insurance
  Percent covered by Medicare               30.7 *     30.8 *
  Percent covered by Medicaid               35.7 *     22.3 *
  Percent enrolled in managed care plan     15.1 *     22.6 *
Treatment process
  Mean number of comorbidities                .89 *      .88 *
  Percent scheduled admission               23.9 *     29.1
  Percent transfer from other hospital       5.8 *     7.4 *
  Prevalence of procedure in census          2.2 *     1.2 *
    tract (per 10,000 residents)
Geographic accessibility
  Mean miles to nearest hospital             2.5 *     3.8 *
  Mean miles to nearest high-volume          4.1 *     4.6 *
    hospital

* Difference from white non-Hispanic statistically significant at
p < .05.

Table 3. Prevalence of use of high-volume hospitals in the New York
metropolitan area, 2001-2002

                                  Racial and ethnic status

                                Unadjusted probabilities of use
                                              (%)

                          White,         Black,
Procedure/conditions   non-Hispanic   non-Hispanic   Hispanic    Asian

Cancer surgery
  Breast cancer            35.4          16.9 *       26.2 *    20.1 *
  Colorectal cancer        37.5          16.4 *       21.6 *    21.0 *
  Gastric cancer           17.8          10.5 *       13.2 *     2.7 *
  Lung cancer              84.4          51.2 *       66.0 *    66.7 *
  Pancreatic cancer        71.4          41.9 *       45.8 *    42.1 *
Cardiovascular
  AMI                      81.4          56.0 *       68.6 *    81.4
  CABG                     71.9          52.4 *       60.0 *    56.3 *
  Angioplasty              99.2          89.9 *       99.7      97.6
  Abdominal aortic
    aneurysms              25.8          10.5 *       31.5 *    23.8 *
  Carotid
    endarterectomy         77.3          45.1 *       62.4 *    77.6
  Pediatric cardiac
    surgery                43.2          25.0 *       26.1 *     4.4 *
Orthopedic
  Hip fracture repair      18.9           8.3 *       12.0 *    14.9 *
  Hip replacement          78.7          65.5 *       69.5 *    67.6 *
  Knee replacement         45.1          38.8 *       52.2 *    53.2 *
Prostate
  Open prostatectomy       42.0          23.7 *       26.1 *    25.4 *
  TURP                     76.5          62.1 *       62.0 *    65.7 *
AIDS                       84.7          91.8 *       92.3 *    83.8

                                 Racial and ethnic status

                         Regression-adjusted probabilities of use
                                          (%) (a)

                          White,         Black,
Procedure/conditions   non-Hispanic   non-Hispanic   Hispanic    Asian

Cancer surgery
  Breast cancer            31.8          16.6 *       24.0 *    16.5 *
  Colorectal cancer        32.1          15.2 *       17.6 *    16.4 *
  Gastric cancer            6.0           6.3          4.7        .6 *
  Lung cancer              85.8          60.5 *       71.7 *    77.3 *
  Pancreatic cancer        76.4          56.7 *       66.9      37.9 *
Cardiovascular
  AMI                      86.7          64.2 *       73.7 *    82.8 *
  CABG                     72.7          64.9 *       67.0 *    63.1 *
  Angioplasty              99.6          98.0 *       99.9 *    99.3 *
  Abdominal aortic
    aneurysms              23.2           8.7 *       22.6      17.3
  Carotid
    endarterectomy         80.1          46.6 *       56.3 *    78.8
  Pediatric cardiac
    surgery                43.5          24.7 *       23.9 *     5.1
Orthopedic
  Hip fracture repair       5.7           3.1 *        3.8 *     4.1 *
  Hip replacement          82.5          71.0 *       73.5 *    74.1 *
  Knee replacement         45.9          37.0 *       44.2      52.9 *
Prostate
  Open prostatectomy       39.0          22.1 *       24.1 *    24.6 *
  TURP                     78.7          63.5 *       60.9 *    64.0 *
AIDS                       90.8          93.2 *       92.4 *    87.2 *

(a) Regression models control for age, gender, insurance source
(commercial, Medicare, Medicaid) and type (HMO and FFS), average
household income and educational attainment of census tract, admission
type (scheduled, transfer), number of comorbidities, prevalence of
procedure-condition in that census tract, distance from home to
nearest hospital and nearest high-volume hospital.

* Difference from white non-Hispanic statistically significant at
p < .05.

Table 4. Prevalence of use of high-volume hospitals in the New York
metropolitan area, 1995-1996

                                 Racial and ethnic status

                              Unadjusted probabilities of use
                                             (%)

                       White, non-   Black, non-
Procedure/conditions    Hispanic      Hispanic     Hispanic   Asian

Cancer surgery
  Breast cancer           48.9         20.3 *      18.1 *     41.3 *
  Colorectal cancer       28.3          8.8 *       8.3 *     20.4 *
  Gastric cancer          14.0          2.5 *       6.5 *     10.4 *
  Lung cancer             75.2         39.9 *      60.9 *     59.6 *
  Pancreatic cancer       50.2         15.6 *       9.5 *     15.3 *
Cardiovascular
  AMI                     74.9         53.8 *      75.4       62.1 *
  CABG                    90.3         73.9 *      61.6 *     79.5 *
  Angioplasty             93.7         77.6 *      82.8 *     80.1 *
  Abdominal aortic
    aneurysms             43.4         15.6 *       4.3 *     42.9
  Carotid
    endarterectomy        73.2         45.2 *      72.1       58.8 *
  Pediatric cardiac
    surgery               34.3         10.9 *       1.2 *     11.8 *
Orthopedic
  Hip repair              24.8          9.8 *      44.0 *     20.4 *
  Hip replacement         58.7         49.7 *      63.0       39.8 *
  Knee replacement        37.9         33.7 *      40.8       30.1 *
Prostate
  Open prostatectomy      28.1         10.3 *      11.9 *     25.4
  TURP                    86.9         72.8 *      81.5 *     88.1
AIDS                      92.3         95.1 *      96.8 *     96.5 *

                                 Racial and ethnic status

                         Regression-adjusted probabilities of use
                                        (%) (a)

                       White, non-   Black, non
Procedure/conditions    Hispanic      Hispanic    Hispanic   Asian

Cancer surgery
  Breast cancer           46.7         21.5 *     16.6 *     35.4 *
  Colorectal cancer       24.9          8.5 *      6.5 *     18.0 *
  Gastric cancer           6.6          5.1 *      2.4 *      4.5
  Lung cancer             74.5         57.7 *     72.4       60.7 *
  Pancreatic cancer       44.6         17.5 *     11.9 *     14.6 *
Cardiovascular
  AMI                     81.5         59.4 *     80.5       64.2 *
  CABG                    91.1         80.8 *     74.3 *     83.8 *
  Angioplasty             94.3         86.7 *     88.9 *     85.1 *
  Abdominal aortic
    aneurysms             42.1         16.3 *      3.5 *     58.5
  Carotid
    endarterectomy        74.4         52.4 *     74.9       63.4
  Pediatric cardiac
    surgery               35.9          9.5 *      1.3 *     10.5 *
Orthopedic
  Hip repair              22.4         12.4 *     45.0 *     21.3
  Hip replacement         60.4         47.4 *     67.3 *     42.6 *
  Knee replacement        37.1         32.4 *     34.8       28.2
Prostate
  Open prostatectomy      23.4          8.7 *      9.8 *     23.0
  TURP                    89.2         79.6 *     83.5 *     87.7
AIDS                      95.6         96.4 *     96.8 *     97.7 *

(a) Regression models control for age, gender, insurance source
(commercial, Medicare, Medicaid) and type (HMO and FFS), average
household income and educational attainment of census tract,
admission type (scheduled, transfer), number of comorbidities,
prevalence of procedure/condition in that census tract, distance
from home to  nearest hospital and nearest high-volume hospital.

* Difference from white non-Hispanic statistically significant at
p < .05.

Table 5. Predictors of use of high-volume hospitals for four
procedures, New York metropolitan area, 2001-2002

                                        Coronary artery bypass graft

Explanatory variable                    Coefficient   Standard error

Intercept                                 .3642          .2037
Black                                    -.3608 *        .0633
Hispanic                                 -.2690 *        .0666
Asian                                    -.4393 *        .0954
Other race                               -.7767 *        .0538
Age                                      -.00469 **      .00201
Male                                      .0467          .0406
Commercial insurance                      .1246          .1466
Managed care (all payers)                 .3345 **       .1441
Medicare                                  .4374 **       .1457
Medicaid                                 -.0420          .1474
Transfer from other hospital              .3279 *        .0464
Scheduled admission                      -.0956          .0402
Comorbidity index                         .0562 **       .0156
Nearest hospital (miles)                  .1188 *        .00896
Nearest high-volume hospital (miles)     -.4749 *        .0131
Median income (census tract)              .00796 *       .000941
Median education (census tract)           .0676 *        .0143
Frequency of procedure (census tract)    -.00004 *      9.074E-6

                                           Breast cancer surgery

Explanatory variable                    Coefficient   Standard error

Intercept                               -1.0648 **       .2827
Black                                    -.8317 *        .0830
Hispanic                                 -.3877 *        .0981
Asian                                    -.8542 *        .1538
Other race                               -.7629 *        .1088
Age                                      -.0287 *        .00257
Male                                      .4774          .2693
Commercial insurance                     1.4700 *        .2238
Managed care (all payers)                 .9326 *        .2254
Medicare                                 1.5207 *        .2324
Medicaid                                  .7447 **       .2383
Transfer from other hospital              .0787          .3442
Scheduled admission                       .7964 *        .0876
Comorbidity index                         .0837 **       .0348
Nearest hospital (miles)                  .00336         .0278
Nearest high-volume hospital (miles)     -.1061 **       .0300
Median income (census tract)              .0102 *        .00106
Median education (census tract)           .0588          .0211
Frequency of procedure (census tract)     .00004         .00003

                                           Total knee replacement

Explanatory variable                    Coefficient   Standard error

Intercept                                2.1716 *        .2748
Black                                    -.3682 *        .0577
Hispanic                                 -.0653          .0683
Asian                                     .2811          .1290
Other race                               -.4076 *        .0818
Age                                       .0171 *        .00221
Male                                     -.0126          .0439
Commercial insurance                    -1.9330 *        .2053
Managed care (all payers)               -2.1028 *        .2048
Medicare                                -1.7870 *        .2039
Medicaid                                -2.0439 *        .2156
Transfer from other hospital            -2.3029 *        .5223
Scheduled admission                      1.0774 *        .1028
Comorbidity index                         .0769 *        .0190
Nearest hospital (miles)                 -.0476 **       .0198
Nearest high-volume hospital (miles)     -.0652 **       .0220
Median income (census tract)             -.00108         .000863
Median education (census tract)           .0871 *        .0152
Frequency of procedure (census tract)     .000045        .0000019

                                                   TURP

Explanatory variable                    Coefficient   Standard error

Intercept                                 .3089          .3163
Black                                    -.7518 *        .0763
Hispanic                                 -.8647 *        .0885
Asian                                    -.7332 *        .1361
Other race                               -.4184 *        .1076
Age                                       .00553         .0031
Male
Commercial insurance                      .8867 *        .2354
Managed care (all payers)                1.0619 **       .2347
Medicare                                  .6612 **       .2325
Medicaid                                  .4759 **       .2424
Transfer from other hospital              .5140 **       .1685
Scheduled admission                       .6860 *        .0552
Comorbidity index                         .000468        .0246
Nearest hospital (miles)                 -.0449          .0248
Nearest high-volume hospital (miles)     -.4007 *        .0317
Median income (census tract)              .00138         .00115
Median education (census tract)          -.0753 **       .0201
Frequency of procedure (census tract)     .000054        .00003

* p < .001.

** p < .05.

Table 6. Potential barriers to identification and selection of
high-volume providers: Kaiser data, 2000

Measures related to identification and selection of providers

                                Perceptions

                   Percent of patients who don't expect
                        much quality variation from:

Respondent group
                     Primary MD    Specialist   Hospital

Black                    17          23 *          10
Latino                   14          14            10
White                    12          12            12

Measures related to identification and selection of providers

                           Perceptions

                   Percent who believe volume
                   doesn't predict quality of:

Respondent group
                     Physicians   Hospitals

Black                    13            9
Latino                   16 *         11
White                    9             9

Measures related to identification and selection of providers

                                                  Selection
                           Information
                                              Percent of patients who
                   Patient didn't see any      prefer familiar over
                   performance measures (%)    higher-rated provider

Respondent group
                         (Past year)          Surgeons   Hospitals

Black                         68                 63 *       71 *
Latino                        77 *               60 *       52 *
White                         66                 45         61

Source: Calculated by the authors using data from the Kaiser Family
Foundation (2000).

* Difference from white statistically significant at the 5%
confidence level.

Table 7. Estimated mortality differentials due to disparities in use
of high-volume hospitals, New York City, 2001-2002

                               Predicted differential mortality
                                rates (a) for minority groups,
                               relative to otherwise comparable
                                      white patients

Procedure                      African American   Hispanic
Less common procedures
  Pediatric cardiac surgery          1.28           1.30
  Pancreatic cancer surgery          1.26           1.13
  Open prostatectomy                 1.08           1.07
Moderately common procedures
  Colorectal cancer surgery          1.05           1.05
  Hip replacement                    1.10           1.08
  TURP                               1.20           1.24

                               Predicted differential mortality
                               rates (a) for minority groups,
                               relative to otherwise comparable
                                        white patients

Procedure                               Asian American
Less common procedures
  Pediatric cardiac surgery                  1.58
  Pancreatic cancer surgery                  1.53
  Open prostatectomy                         1.07
Moderately common procedures
  Colorectal cancer surgery                  1.05
  Hip replacement                            1.07
  TURP                                       1.20

(a) Based on mortality differentials reported in Halm, Lee, and
Chassin (2002) for high- and low-volume hospitals, multiplied
by the prevalence of high-volume hospital use identified in
this study.
COPYRIGHT 2009 Excellus Health Plan, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2009 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Gray, Bradford H.; Schlesinger, Mark; Siegfried, Shannon Mitchell; Horowitz, Emily
Publication:Inquiry
Geographic Code:1USA
Date:Sep 22, 2009
Words:10054
Previous Article:Private long-term care insurance and state tax incentives.
Next Article:Line authority for nurse staffing and costs for acute inpatient care.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |