|
Article Excerpt
In contrast to an extensive literature documenting worse health outcomes among blacks than whites throughout the health care system (Smedley, Stith, and Nelson 2002), a number of studies of veterans hospitalized in the Veterans Affairs (VA) health care system have found either that blacks have better outcomes than whites (Jha et al. 2001; Deswal et al. 2004) or that there are no race/ethnicity based disparities in patient outcomes (Horner et al. 2002; Petersen et al. 2002; Freeman et al. 2003; Goldstein et al. 2003; Selim et al. 2004).
There are several potential reasons why clinical care and outcomes may be relatively more favorable toward blacks within VA than in the rest of the U.S. health care system. White and black veterans who use VA are much more homogeneous in terms of socioeconomic status than whites and blacks outside the VA system. As socioeconomic status has been shown to have a large impact on health (Lynch and Kaplan 2001; Adler and Newman 2002; Marmot 2002) and to account for some (but not all) racial disparities in health (Geronimus et al. 1996; Carlisle, Leake, and Shapiro 1997; Shen, Wan, and Perlin 2001; Goldman and Smith 2002), this homogeneity could explain why racial disparities might be smaller within the VA. In addition, because blacks in the United States on average have lower incomes and are less likely to have health insurance than whites (Smedley, Stith, and Nelson 2002), the public funding of care in the VA should act to reduce racial disparities. Finally, the military has played an important role within American society in bringing about desegregation (Moskos and Butler 1997), and the better integration of minorities within the military may be reflected in more equal treatment of racial minorities within the VA.
However, while some studies have found equal outcomes between blacks and whites within VA, the observation of better outcomes among blacks remains unexplained (Hermos et al. 2001; Mark 2001). There are at least five potential explanations for better outcomes among blacks that require further examination.
First, recent work has found that differential sorting into quality hospitals may explain racial differences in treatment and outcome (Barnato et al. 2005). It is possible that such sorting could account for the better outcomes among blacks if blacks are more likely to be admitted to VA hospitals with better outcomes. Hospital effects were not considered in previous studies of racial differences in the VA.
Second, previous studies (Jha et al. 2001) have used the VA Beneficiary Identification Record Locator System Death File (BIRLS) and in-hospital data to measure deaths. Death rate ascertainment with BIRLS has been considered to have an accuracy rate of about 95 percent (Fisher, Weber, and Goldberg 1995; Cowper et al. 2002), but it is unclear whether inaccuracy in the BIRLS data varies by race. Thus, confirming that previous findings were not due to incomplete ascertainment of death among blacks would be important.
Third, if blacks have more limited access to outpatient care because of either geographic location or fewer resources to use outside facilities (Pappas et al. 1997; Basu and Clancy 2001), they may be more likely than whites to be hospitalized for conditions that could have been treated in outpatient settings. Therefore, among conditions for which hospital admission is discretionary, higher admission rates among blacks could result in lower observed mortality rates because the average hospitalized black patient would be less severely ill. Examination of conditions for which hospital admission is not discretionary would likely address this potential bias (Miller et al. 1994). Fourth, the studies that found better outcomes for black veterans were based on hospital discharges that occurred before the significant reforms undertaken within the VA health care system in the mid- 1990s (Jha et al. 2003). As quality improvement may target the worst performing groups within a system (Sequist et al. 2006), it is plausible that such reforms may have led to a reduction in the mortality difference between blacks and whites. As a consequence of the changes in the VA health care system, the racial differences in mortality observed in the mid-1990s may no longer exist. Finally, differential selection into the VA by race may be a mechanism for lower VA mortality among blacks. Given that selection into VA may be based on financial barriers to non-VA care and that Medicare reduces these financial barriers for those over 65, if selection matters we would expect racial differences in mortality within VA to be different in the over age 65 and under age 65 populations.
We undertook an analysis to determine if racial disparities in 30-day mortality exist for veterans hospitalized within VA from FY1996 to FY2002 for three conditions for which hospital admission is nondiscretionary (acute myocardial infarction [AMI], hip fracture, and stroke), and for three conditions for which hospital admission is discretionary (congestive heart failure [CHF], gastrointestinal bleeding [GI bleed], and pneumonia). We estimated racial differences in mortality for each of these six conditions and extended previous work by examining whether (1) between-site variation or (2) more complete ascertainment of death accounted for the observed differences in race-specific mortality; (3) racial differences in outcomes were similar for conditions for which admission is or is not discretionary; (4) racial differences persisted over time; and (5) effects were similar in the subgroups under and over 65 years of age.
METHODS
This study was approved by the Institutional Review Boards of the Philadelphia VA Medical Center and the University of Pennsylvania.
Data
Our primary data source was hospital discharge data from the VA Patient Treatment File (PTF) for FY1996-FY2002. The PTF contains information about primary and secondary diagnoses, age, gender, discharge disposition, transfer status, length of stay, patient's zip codes, and race and means test eligibility for every hospital discharge within the Veterans Health Administration. The initial analysis included all deaths identified in the PTF and BIRLS within 30 days of hospital admission. Date of death was verified with the National Death Index (NDI) for all veterans with evidence of death in either the PTF or BIRLS and all veterans with unknown vital status 30 days after each admission (i.e., no active follow-up within the VA system). The 2000 Census was the source for socioeconomic characteristics, which were linked to the PTF...
|
