| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2004;78:18-24
© 2004 The Society of Thoracic Surgeons
a Excellus BlueCross BlueShield, Rochester, New York, USA
b Department of Community and Preventive Medicine, University of Rochester School of Medicine and Dentistry, Rochester, New York, USA
Accepted for publication January 28, 2004.
* Address reprint requests to Dr Mukamel, University of California, Irvine, Health Policy Research, 100 Theory, Suite 110, Irvine CA 92697-5800, USA
e-mail: dmukamel{at}uci.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
METHODS: New York State (NYS) hospital discharge data for 1996 and 1997 for patients undergoing CABG surgery were combined with risk-adjusted mortality rates for cardiac surgeons calculated by the NYS Department of Health. Statistical analysis was performed to determine the relationship between patients' race and the quality of the surgeon performing the CABG, as measured by the surgeon's risk-adjusted mortality rate, after controlling for patient characteristics such as comorbidities and socioeconomic status; the hospital where the surgery was performed; and the number of surgeries the surgeon performed over a 3-year period.
RESULTS: African Americans and Asian/Pacific Islanders are treated by surgeons with higher risk-adjusted mortality rates compared with whites. This association does not appear to be a result of inadequate risk adjustment. It is explained to some degree by the hospital to which these patients are admitted, and to a lesser degree by (1) the education and income level in the patient's zipcode of residence and (2) being treated by a low-volume surgeon. After controlling for these factors, race continues to be associated with treatment by a surgeon with a higher risk-adjusted mortality rate.
CONCLUSIONS: Efforts to achieve the "Healthy People 2010" goals of eliminating health disparities should address not only access to care, but also access to high-quality care.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
Racial disparities in outcomes and health status have also been widely documented [3]. African Americans in particular experience worse health outcomes. For example, minorities in New York are more likely to be readmitted for complications after CABG [4].
Even when access to care is similar, care practices can vary with the race of the patient, often compromising quality of care. Virning and associates [5] found that white Medicare+Choice enrollees were more likely than blacks to receive the recommended beta blockers after an acute myocardial infarction. Furthermore, racial minorities are more likely to be treated by providers (hospitals and physicians) with worse performance records. African Americans undergoing CABG surgery in NYS were operated on by surgeons with higher risk-adjusted mortality rates compared with whites [6]. Thus, minorities are disadvantaged even when they gain access to care.
The objective of this study is to examine several factors that may explain the racial differences observed in prior studies. We first examine the possibility that the previously reported racial differences in access to surgeons with low risk-adjusted mortality rates are an artifact resulting from inadequately accounting for race in the risk-adjusted mortality rate. As the data we present suggest that this is not the case, we then proceed to investigate other potential explanations—specifically, whether the observed racial differences in access to high-quality surgeons are due to (1) socioeconomic differences among patients of different racial groups, (2) racial differences in referrals to high-quality hospitals, or (3) racial differences in referrals to low-volume surgeons.
|
Material and methods |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
These data were augmented with the surgeons' risk-adjusted mortality rates calculated by NYS [7, 8]. The NYS Department of Health calculates the risk-adjusted mortality rate for each surgeon based on the outcomes experienced by all patients treated by that surgeon. The risk-adjusted mortality rate reflects the average performance of the surgeon relative to his or her peers after taking into account differences in disease severity and comorbidities (eg, ejection fraction) [9]; it is interpreted as a measure of the quality of the surgeon. These risk-adjusted mortality rate data are considered to be among the most reliable and valid measures of the quality of cardiac surgeons [10] for several reasons. First, unlike many other risk-adjusted outcome measures, the NYS measures are based on clinical risk factors tested specifically for prediction of CABG mortality. Second, the risk adjustment methods are recalculated for each report, so they take into account the impact of surgical innovations that may affect the impact of risk factors. Third, the risk adjustment models have been validated on an external data set [11]. The NYS cardiac report is also one of the oldest public reporting systems, and studies found that patients and managed care organizations view this information as meaningful and use it when choosing cardiac surgeons [12, 13]. Socioeconomic information was obtained from the 2000 census.
Analyses
To evaluate the possibility that the observed racial disparities are an artifact, we examined the evidence about race as a risk factor in predicting inpatient CABG mortality. We also investigated the distribution of minorities across surgeons to determine whether they tend to congregate among a few surgeons and therefore have a large impact on these surgeons' risk-adjusted mortality rates.
We then estimated multivariate regression models to determine which factors are associated with the surgeon's risk-adjusted mortality rate. Note that we did not develop a risk adjustment model for CABG mortality. Rather, we used the risk-adjusted mortality rate calculated by NYS as a measure of quality, and then tested which patient characteristics hypothesized to influence choice of surgeon were associated with being treated by surgeons with higher or lower risk-adjusted mortality rates.
The base model included sex, age, payer, region, admission type (scheduled or emergency), hospital transfer, rural residence, discharge year, distance relative to the closest hospital that performs CABG, comorbidities if present on admission to the hospital, as well as race (white, African American, and Asian/Pacific Islander). We then added to the base model socioeconomic status, measured as income and education levels in the patient's zipcode (model 2), hospital indicator variables (model 3), and surgeon volume (model 4). Inferences in all models were based on Huber-White standard errors to account for the potential correlations among observations for patients treated by the same surgeon. To correct for heteroscedasticity, observations were weighted by the square root of the sample used to calculate the risk-adjusted mortality rate for each surgeon. We compared the coefficient for the race variable in the base model and in each successive model. A decline in this coefficient suggests that the association between race and surgeon quality is not entirely due to race, but is at least partially explained by the variables added to the model. For example, a decline in the race coefficient when comparing the base model to the model with added socioeconomic variables (which are correlated with race) would suggest that socioeconomic status explains at least some of the observed racial disparities in access to high-quality surgeons.
More details about the data and the statistical methods used in these analyses are found in the Appendix.
|
Results |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
|
African Americans and Asian/Pacific Islanders were also more likely than whites to have diabetes, and African Americans were more likely than whites to have renal failure. Chronic obstructive pulmonary disease was more prevalent among whites. Whites were also more likely to have had a prior CABG.
Are observed racial disparities due to inadequate risk adjustment?
One possible explanation of the prior findings [6] that African Americans are more likely to be treated by low quality surgeons is that this association is only a reflection of inadequate risk adjustment for race. Indeed, the NYS risk-adjustment methodology does not include race as a risk factor [7–9]. Race was, however, included in the initial analyses. These analyses showed that after controlling for clinical risk factors such as ejection fraction, race was not a significant predictor of in-hospital mortality. The final risk adjustment methodology, which is the one used to calculate the risk-adjusted mortality rate used in this study, therefore excluded race.
Most other studies of CABG mortality, performed on different databases and populations, also failed to identify race as a significant predictor [11, 15–17]. The two studies that did find a significant relationship between CABG mortality and race [18, 19] were both based on the same data. As noted in one of the studies[19], when race was introduced into the model other clinical variables dropped out (eg, congestive heart failure was dropped from the 1995 model when sex and race were added). Furthermore, the goodness-of-fit statistics did not change when race was added into a model composed of clinical risk factors only. This finding suggests that the association between race and CABG mortality is tenuous, present in only some samples, and is not robust to choice of clinical risk factors.
This was recognized by the Cooperative CABG Database Project [15], which included representatives of seven large cardiovascular data bases (including New York's) and other experts. One of their objectives was to identify variables useful for monitoring the quality of care for CABG patients. By consensus, they classified race as a variable that has not clearly been shown to relate directly to short-term CABG mortality.
Further evidence of the external validity of the NYS risk adjustment model, which excludes race, was provided by a comparison of this model with several other CABG risk adjustment models using the same data from the Cooperative Cardiovascular Project [11]. The four models (including New York's) had similar discriminating ability. The predicted mortality rates from the NYS model were similar to the actual rates for all five risk groups, which were identified a priori.
There is considerable evidence suggesting that race is not an independent predictor of CABG inpatient mortality after controlling for relevant clinical factors. However, if despite the evidence cited above, excluding race from the risk adjustment introduces a bias that penalizes surgeons who are treating minorities by underestimating their expected mortality rate, the finding by Mukamel and associates [6] that African Americans access lower quality surgeons (as measured by higher risk-adjusted mortality rates) would be erroneous only if minority patients tend to congregate among a few surgeons and thus dominate their practice.
To investigate this possibility, we examined the distribution of African-American patients across physicians. In 1996, the mean percentage of African Americans in surgeons' practices in NYS (after excluding surgeons with fewer than 6 CABG patients per year) was 7.8% (standard deviation = 10.3%). To evaluate the sensitivity of our conclusions to the percentage of blacks in a surgeon's practice, we repeated the analyses reported below excluding the top 5% of surgeons with the most African-American patients (with more than 18% African-American patients). The results were similar before and after these surgeons were excluded.
Thus, because the evidence suggests that race is not likely to be a predictor of CABG mortality at the individual level and because racial minorities do not tend to dominate surgeons' practices, it is unlikely that the observation that minorities are treated by surgeons with higher risk-adjusted mortality rates is due to inadequate risk-adjustment.
Factors associated with surgeons' risk-adjusted mortality rates: results of multivariate analyses
Association between race and surgeons' risk-adjusted mortality rate
Table 2
presents the associations between race and surgeons' risk-adjusted mortality rates from the four regression models: the base model and the models adding socioeconomic, hospital, and surgeon volume variables. In these analyses, surgeons' risk-adjusted mortality rates (the dependent variable) serve as a measure of quality. Patients treated by surgeons with lower risk-adjusted mortality rates are assumed to receive better care than those treated by a surgeon with a higher risk-adjusted mortality rate. Note that the associations with race are estimates of the race effect when controlling for a large number of individual characteristics (listed in the analysis section), which may also influence which patient is treated by which surgeon.
|
Does socioeconomic status explain the race effect?
The relationship between race and surgeons' risk-adjusted mortality rates changes only minimally after controlling for differences in education and income (see Table 2, model 2). The coefficient for African Americans (that is, the difference in surgeons' risk-adjusted mortality rates for African Americans versus whites) declines from 0.311 to 0.296, and for Asian/Pacific Islanders, from 0.383 to 0.378. This suggests that the race effect cannot be explained by differences in socioeconomic characteristics among racial groups.
A more refined analysis, stratifying by income, reveals that racial disparities do not exist for all income groups. In Table 3 we report the race coefficients for a model in which race was interacted with income. The reference category is high-income whites. We find no differences among whites in different income categories or among whites and high-income African Americans. However, African Americans of medium and especially low income are more likely to be treated by surgeons with higher risk-adjusted mortality rates. Among Asian/Pacific Islanders, the largest disparity is observed for the medium income group, although the estimates may be less precise because of the smaller sample size.
|
Does surgeon volume explain the race effect?
As shown in Table 1, minorities tend to be treated by surgeons who perform fewer CABG surgeries than whites (p < 0.001 compared with whites). As low volume has been shown to be associated with worse CABG outcomes [20], this referral pattern may explain the higher risk-adjusted mortality rate for surgeons treating African Americans and Asian/Pacific Islanders. Table 2, model 4, reports the race coefficient in a model that controls for surgeon volume as well. We find that surgeon volume has only a minimal effect; the coefficient for African Americans declines from 0.191 to 0.174 and for Asian/Pacific Islanders, from 0.194 to 0.166. This suggests that after controlling for hospital, surgeon volume cannot explain the remaining disparity in access to high-quality surgeons.
In summary, after controlling for all of these factors—patient characteristics including socioeconomic status, the selection of hospital, and the surgeon volume—the association between race and surgeon's risk-adjusted mortality rate is smaller, but it is still significantly positive. Both African-American and Asian/Pacific Islander patients are treated by surgeons whose risk-adjusted mortality rates are about 0.17 percentage points higher on average (p < 0.01). Race is associated with the surgeon's risk-adjusted mortality rate independently of all of these other factors.
Comment
Among patients undergoing CABG surgery in NYS, being African American or Asian/Pacific Islander is significantly associated with being treated by surgeons of poorer quality, when quality is measured by risk-adjusted mortality rates. This association is due in some part to the fact that these patients tend to be admitted to different hospitals, hospitals that on average have higher risk-adjusted mortality rates. It is also associated, although to a much lesser degree, with differences in income and education among whites, African Americans and Asian/Pacific Islanders, as well as the higher likelihood that these minorities are referred to lower volume surgeons.
Two potential data limitations should be considered. First, the information on race was obtained from the SPARCS data and is based on the impression of the hospital admitting clerk. However, Blustein [21] has demonstrated that race recorded in the SPARCS data set for the same individual on two different admissions was concordant 93.7% of the time. This suggests that the race information used in this study is reliable. Furthermore, if the reasons for the disparities we observe are due to health system factors as opposed to patient choice, then race based on the perception of health providers is likely to be the more relevant variable to use.
Another limitation is that income and education data were available only at the race/zipcode. This introduces a measurement error that tends to bias regression coefficients for these variables toward zero. However, prior studies have shown that zipcode level socioeconomic data do not compromise these types of analyses. Fiscella and Franks [22] found no difference in their analyses after adjusting for educational attainment levels using zipcode level census data, block-level census data, and self-reported data. Gornick and associates [23] examined the impact of income on mortality and utilization. They found that census data and self-reported data yielded the same results, although the impact was usually larger using the self-reported data. Based on these prior studies, education and income may explain more of the racial disparities we observe than our data allow us to detect.
Despite these limitations, our study confirms the existence of racial disparities in access to high-quality cardiac surgeons and offers insights into potential mechanisms that may contribute to it. The importance of the hospital in explaining these disparities suggests that general referral patterns may be important determinants of where racial minorities receive treatment. Such referral patterns are likely to be dependent on many factors and not only the quality of cardiac surgery services. Additional factors may include geographic proximity of patients to hospitals and the referral networks of the health care system within which patients receive all their care, not just cardiac care. To the degree that such broader referral patterns contribute to these disparities, policy interventions to ameliorate the disparities may need to address the factors underlying these broader referral patterns.
The fact that racial disparities persist even when the estimated models control for the selection of the hospital indicates that even among patients at the same hospital, whites are treated by surgeons with better outcomes. Furthermore, the disparities we observe cannot be attributed to the fact that the better surgeons may be unavailable to conduct emergency surgery because their time is filled with elective surgeries [24]. Thus, other factors, not identified in this study, may contribute to the disparity we observe. Identifying such factors is an important first step in designing effective strategies to equate access to high-quality surgeons.
The racial disparity in access to high-quality surgeons probably contributes to poorer health outcomes and health status among patients with coronary artery disease. If indeed the risk-adjusted mortality rate is a measure of the performance of the cardiac surgeon, it is probable that patients treated by surgeons with higher risk-adjusted mortality rates would have other complications and other poorer outcomes, not only higher mortality. In a recent study, Hannan and associates [4] found that African-American CABG patients were more likely to be readmitted to the hospital within 30 days than patients of other races. Thus, achieving the national goal of eliminating racial and ethnic disparities in health, as presented in Healthy People 2010 [25], requires addressing not only the disparities in access to care but also the disparities in access to high-quality providers.
Furthermore, improving access to high-quality providers might help lower the disparity in utilization as well. There is some evidence indicating that racial minorities are more likely to refuse recommendations for revascularization than whites [26, 27]. These studies suggest that some of the disparities in procedure rates can therefore be traced to patient preferences and expectations. If minorities tend to be treated by lower quality providers, they would have lower expectations of favorable outcomes compared with nonminority patients. Such expectations may lead to increased reluctance to undergo surgery. Thus, addressing disparities in access to high-quality providers may have an added benefit of decreasing the disparity in procedure rates.
In conclusion, this paper suggests that the policy agenda addressing disparities in health and in access to health care should be expanded to address issues related to the quality of the providers treating minorities. While some policy interventions may be successful in addressing both lack of access to services and access to lower quality providers, there is a need to study the referral process to identify factors that influence primarily the choice of provider and that may present a barrier to access to high-quality providers.
|
Acknowledgments |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
|
Appendix |
|---|
The final sample size was 27,969 patients
Variables
For each CABG patient, we determined whether the median education level in the patient's zipcode of residence was high school or college graduate, using 2000 census data. We also determined the median household income for residents of the same age and race as the patient in his or her zipcode of residence.
We included regional indicators based on patient migration patterns for CABG surgery [6]. These indicators correspond to the areas surrounding the major urban centers in the state.
These data were augmented with the surgeons' RAMRs obtained from the NYS Cardiac Surgery Reports for 1994 to 1996 and 1995 to 1997 [7, 8]. The RAMR for 1994 to 1996 was used for the 1996 discharges and the RAMR for 1995 to 1997 was used for the 1997 discharges. Data on volume in the regressions also came from the NYS Cardiac Surgery Reports and not from the number of surgeries by each surgeon in the final sample used for this analysis.
Statistical estimation
Because surgeons treating more CABG patients have RAMRs with smaller standard errors, all regressions were estimated as weighted ordinary least squares, with the weights proportional to the reciprocal of the square root of the surgeon's surgery volume. To account for correlations between observations for patients treated by the same surgeon, inferences were based on Huber-White robust standard errors, allowing for clustering of patients treated by the same surgeon.
Because the dependent variable is correlated with the volume of surgeries each surgeon performs, adding volume to the model (model 4) may introduce endogeneity. We therefore ran a separate model (results not shown) with surgeon volume as the dependent variable. The results showed a clear association between being African American and going to a low-volume surgeon, after controlling for the same patient characteristics as in the other models.
The computer program, SAS, was used to create the data set, and STATA was used to run the regressions.
Calculation of clinical impact of estimated race effect
We estimate the impact of the race effect in the base model of about 0.35 RAMR differential between surgeons treating white and minority patients to be an additional 170 minority deaths nationwide. This assumes that 8% of the 607,000 CABG surgeries performed each year (1997 to 1999 Medicare NYS data and AHRQ data) are performed on minority patients.
|
References |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsAcknowledgmentsReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. Neuburger, A. Hutchings, N. Black, and J. H. van der Meulen Socioeconomic differences in patient-reported outcomes after a hip or knee replacement in the English National Health Service J Public Health, March 1, 2013; 35(1): 115 - 124. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. T. Kolstad and M. E. Chernew Quality and Consumer Decision Making in the Market for Health Insurance and Health Care Services Med Care Res Rev, February 1, 2009; 66(1_suppl): 28S - 52S. [Abstract] [PDF]
|
|
|
|
 |
|
 D. J. Gaskin, C. S. Spencer, P. Richard, G. F. Anderson, N. R. Powe, and T. A. LaVeist Do Hospitals Provide Lower-Quality Care To Minorities Than To Whites? Health Aff., March 1, 2008; 27(2): 518 - 527. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Menachemi, A. Chukmaitov, L. S. Brown, C. Saunders, and R. G. Brooks Quality of Care Differs by Patient Characteristics: Outcome Disparities After Ambulatory Surgical Procedures American Journal of Medical Quality, December 1, 2007; 22(6): 395 - 401. [Abstract] [PDF]
|
|
|
|
 |
|
 J. Skinner, A. Chandra, D. Staiger, J. Lee, and M. McClellan Mortality After Acute Myocardial Infarction in Hospitals That Disproportionately Treat Black Patients Circulation, October 25, 2005; 112(17): 2634 - 2641. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. A. Mensah, A. H. Mokdad, E. S. Ford, K. J. Greenlund, and J. B. Croft State of Disparities in Cardiovascular Health in the United States Circulation, March 15, 2005; 111(10): 1233 - 1241. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. R. Nerenz Health Care Organizations' Use Of Race/Ethnicity Data To Address Quality Disparities Health Aff., March 1, 2005; 24(2): 409 - 416. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
