We examined mortality rates among Medicare enrollees and found that for all six common medical conditions, 30-day mortality rates were lower for black Americans than white Americans. However, after hospitalization, the survival advantage initially seen among blacks dissipated, and by the end of 2 years, blacks had higher mortality than whites for five of the six conditions. When we compared this with outcomes within the VA, results were similar at 30 days after admission but long-term mortality did not worsen to the same degree as in the Medicare population for blacks relative to whites. The increases in mortality rates for blacks relative to whites at longer time intervals after hospitalization—and the slower rate at which relative mortality for blacks increased in the integrated VA health care system—suggest that while racial differences in hospital care, if any, do not lead to worse outcomes for blacks within the first 30 days after admission, factors related to care outside the hospital, such as type of social support, quality of ambulatory care, or access to prescription drugs are likely to play an important role in longer-term outcomes.
It is not obvious a priori that 30-day mortality rates would be lower for blacks given the lower life expectancy for blacks (National Center for Health Statistics 2005
), prior findings that blacks receive a lower intensity of recommended therapies (Meehan et al. 1995
), and some data suggesting that blacks are sicker at admission than whites (Buckle et al. 1992
; Ebell et al. 1995
; Johnson et al. 1995
; Williams et al. 1995
; Sonel et al. 2005
). Nevertheless, the few non-VA studies that directly examine racial differences in 30-day hospital mortality—mostly for heart failure and AMI in nationwide samples—have also found similar results (Rathore et al. 2003
; Barnato et al. 2005
). Studies of hospitalizations in the VA also show lower 30-day mortality rates for blacks across a range of conditions ( Jha et al. 2001
; Deswal et al. 2004
; Volpp et al. 2007
) and the lower 30-day mortality rates for blacks older than age 65 in the VA were found to be comparable to non-VA hospitalizations in Pennsylvania and California (Polsky et al. 2007
). It is possible that the lower 30-day mortality rates for elderly blacks are due to lower rates of procedures which confer short-term risk (Meehan et al. 1995
; Barnato et al., 2005
), differences in the average severity of whites and blacks who access hospital care in ways that are not observable with administrative data, or a “survivorship bias,” in which blacks who survive to age 65 are hardier than whites because the sickest blacks have already died by this point in time. However, a “survivorship bias” would be less compatible with the observed relative worsening in mortality for blacks with increasing time from hospital admission. While we cannot differentiate between these speculative explanations, to our knowledge this is the first study to demonstrate the consistency of the pattern of racial differences in hospital mortality in the elderly population throughout the United States across common conditions.
Several potential factors could explain our findings. First, unmeasured confounders may affect the findings of lower initial mortality rates for blacks. However, adjustments for unbalanced comorbid conditions, socioeconomic status, and hospital selection by race lowered 30-day mortality for blacks relative to whites to a greater degree than observed in the unadjusted analyses and did not modify the relative increases in the mortality rate for blacks from 30 days to 2 years. This suggests that these measured factors do not explain the lower 30-day mortality for blacks, nor can they explain the relative increases in mortality for blacks as time from admission increases. Second, linear and nonlinear adjustment models produced results similar to each other and to unadjusted mortality, suggesting that model choice or specification does not explain the findings.
Although we cannot explain why blacks have lower mortality initially and higher mortality as the time from admission increases, the increasing rates of mortality for blacks may be generated by the greater vulnerability of blacks to environmental factors (more dangerous neighborhoods, impact of poverty, differences in health behaviors), by a fragmented health care system ill-equipped to address these factors, or poor access to high-quality outpatient care (Sherkat et al. 2005
; Trivedi et al. 2006
). This idea is supported by the comparison of racial differences in mortality over time in the VA and in the non-VA hospital system. In this comparison we see that the VA's integrated health care system, which includes prescription drug coverage with small or nonexistent copayments, may make a difference in the long-term relative mortality risk of blacks for conditions for which coordination of care and access to prescription drugs are important. An alternative explanation is that among the VA population, difference in socioeconomic status, lifestyle, and psychosocial factors between blacks and whites may be less pronounced than in non-VA settings. It is not possible with the available data to determine the relative importance of these factors in explaining the observed differences in mortality over time. Differences in the observed patterns for congestive heart failure compared with other conditions may relate to the fact that it is a chronic condition, which necessitates more frequent hospital readmission than the other conditions. At each hospital admission, patients presumably get a careful workup and revision of their treatment plan as needed. The factors that lead to worsened outcomes for blacks as the amount of time from hospital admission increases thus could be “overridden” by the frequent admissions for patients at high risk for mortality. Our data provide some empirical support for this hypothesis; after adding back in the readmissions excluded from our analytic files, 32.6 percent of patients with a principal diagnosis of heart failure had more than one heart failure admission (38.3 percent for blacks, 31.9 percent for whites) while for other conditions the proportion of patients with more than one admission ranged from 6.1 percent for hip fracture to 10.6 percent for stroke, 14.4 percent for GI Bleed, 14.7 percent for AMI, and 21.4 percent for pneumonia.
The primary limitation of our study is the fact that it is descriptive and does not attempt to identify the mechanism that might explain the pattern of racial differences observed. Our findings do not imply that hospitals discriminate against whites or that discrimination against blacks does not exist. However, describing this robust pattern brings the literature closer to identifying the sources of racial disparities in population-based outcomes, which will aid in the design of effective interventions aimed at ameliorating racial disparities. A second important limitation is the use of administrative data to study health outcomes. These administrative data have only limited measures of health status and socioeconomic status. However, we attempted to overcome these limitations by using a well-validated approach to risk adjustment that has a high degree of ability to discriminate between patients who die and patients who do not (Southern, Quan, and Ghali 2004
) and by building on previous studies of racial disparities in outcomes that use administrative data (Jha et al. 2001
; Smedley, Stith, and Nelson 2003; Volpp et al. 2007
Among Medicare enrollees we found black Americans have lower 30-day mortality than white Americans. However, by 2 years after admission, blacks had higher mortality than whites for five of the six study conditions. This relative deterioration of outcomes after discharge for blacks was not as great among patients hospitalized within the VA, which suggests the possibility that an integrated health care delivery system like the VA may attenuate racial disparities in health. Further research should attempt to differentiate between competing explanations for lower 30-day mortality for blacks and the observed patterns of higher mortality for blacks with increasing time from hospital admission. Medical system efforts to reduce racial disparities may need to increasingly focus on post-hospital care and environmental factors rather than hospital-targeted initiatives.