Comparison of racial and ethnic disparities in COVID-19 mortality between Veterans Health Administration and U.S. populations

Abstract: In the USA, racial and ethnic disparities in COVID-19 outcomes for minoritized groups compared with non-Hispanic White (hereafter “White”) individuals have in part been attributed to structural racism and social determinants of health (SDOH) resulting in inadequate healthcare access. There is evidence that minoritized versus White disparities in certain outcomes, including mortality, are attenuated in the Veterans Health Administration (VA) compared to the general US population. This may reflect reduced access barriers and VA’s efforts to address SDOH2. This study aimed to determine if the magnitude of racial and ethnic disparities in COVID-19 mortality differed between VA and general US populations. We performed a cross-sectional analysis of COVID-19-positive cohorts from VA and general US populations, using individual-level data from 3/1/2020 to 5/7/2021. Data sources included VA Office of Health Equity consortium national database of VA users evaluated for respiratory illness or COVID-19 exposure and the Center for Disease Control (CDC) public use database on cases and mortality among patients with confirmed COVID-19, reported by age, sex, and race and ethnicity. We included patients with non-missing mortality data (52.1%). This work received a determination of non-research from the VA Greater Los Angeles Healthcare System Institutional Review Board. Our primary predictors were race, ethnicity, VA/US data source, and sex. We limited our analysis to non-Hispanic Black, Hispanic, and White adults (aged 20-and-older), due to limited representation of other groups. Our outcome was COVID-19-related mortality. We tested whether the magnitudes of racial and ethnic disparities in COVID-19 mortality differed between VA/US data sources by sex using sex-stratified logistic regression models with a race and ethnicity-by-population interaction term. We report sex-stratified findings to account for substantial VA/US between-group differences in sex compositions. We controlled for age and obtained bootstrapped standard errors. We calculated sex-stratified Black-White and Hispanic-White differences in COVID-19 adjusted mortality proportions (AMP), followed by VA/US differences between these estimates. We considered coefficients with two-sided p < 0.05 as significant. There were 225,230 patients in the VA population and 10,977,877 in the general US population (Table 1). In VA, Black/White mortality disparities were present for men (Black AMP 1.7 percentage points (pp) higher than for White, 95% CI 1.4, 2.0), but not for women (Fig. 1). General US Black/White mortality disparities were present for both men (AMP + 3.1 pp, 95% CI 3.1, 3.2) and women (AMP + 2.0 pp, 95% CI 1.9, 2.0). Hispanic/White mortality disparities were present for men in the VA (AMP + 1.7 pp, 95% CI 1.2, 2.1) and US (AMP + 1.5 pp, 95% CI 1.5, 1.6), but not for women in either population. US Black/White disparities exceeded those in VA for both men (AMP difference + 1.5 pp, 95% CI 1.2, 1.8) and women (AMP difference + 2.0 pp, 95% CI 1.7, 2.3) (Fig. 1). VA/US Hispanic/White disparities comparisons did not differ for men or women. We found that COVID-19 mortality disparities were greater in the US compared with VA for Black men, but similar for Hispanic men. Disparities were present for Black women in the US, but not in VA, and were not present for Hispanic women in either setting. Study limitations include potential for selection bias in VA use; inability to adjust for unavailable covariates (e.g., comorbidities, geography); large amount of missing CDC mortality and race and ethnicity data, which may bias findings (however excess mortality among minoritized groups and men and AMPs are consistent with other studies); and insufficient sample size to include other racial groups also experiencing COVID-19 disparities.