Minding the Black-White Gap in Cardiovascular Medication Adherence among Seniors: Potential Drivers across the Distribution of Adherence Behavior

Background: Extant evidence demonstrates a persistently large and clinically consequential black-white disparity in cardiovascular medication adherence among the elderly. Despite a large literature, the underlying factors driving this disparity remain poorly understood. A systematic comparison of how social, behavioral, and health-system determinants of adherence might be operating differently between blacks and whites is needed.

Objective: To estimate the contributions of racial differentials in the determinants of adherence to the observed black-white adherence gap across the distribution of adherence behavior.

Methods: We analyzed pooled longitudinal data (panels 11-14, years 2006-10) of the Medical Expenditure Panel Survey on 3,288 respondents, nationally representing 19 million Medicare recipients 65 years or older who were white or black, and used ≥1 class of maintenance cardiovascular medications. Class-specific adherence was measured as the proportion of days covered (PDC) over an average of 342 days. The main outcome variable was an overall PDC, averaged over respondent-specific medication classes. Using a simple recentered influence function-based methodology (Firpo, Fortin, Lemieux 2009), we estimated race-stratified unconditional quantile regressions for the 10th to 80th deciles of the continuous PDC distribution, as a function of respondents’ baseline sociodemographic, health and functional status, beliefs, characteristics of prescription drug coverage, access to primary care, and experience with providers. We then applied linear Oaxaca-Blinder decomposition to estimate each covariate’s contribution to the estimated racial difference at each PDC decile.

Results: The black-white disparity in adherence was largest below the median (PDC~77%), ranging from -6.40 percentage points at the 10^th percentile to -9.00 at the 40^th percentile. Among seniors with adherence below the median, the lower rates of having employer-sponsored (majorly retiree) drug coverage among blacks relative to whites (10% vs 23%, respectively) accounted for 25% of the adherence disparity. This is despite a much stronger positive association between this coverage with adherence among blacks than whites. More frequent physician visits were associated with better adherence among whites but worse adherence among blacks, accounting for 35% of the disparity, although self-reported satisfaction and experience with providers were comparable by race. Among seniors with very poor adherence (10^thpercentile, PDC~30%), although the prevalence of (asymptomatic) hyperlipidemia is lower among blacks, it was associated with a much more negative effect on adherence than among whites. Had this effect been similar to whites', adherence among blacks would have been 50% better than whites (disparity reversal), ceteris paribus. The roles of racial differences in dual-eligibility and copay levels were small and less clear. Racial differences in beliefs regarding risk and care-seeking, and depressive symptoms contributed negligibly to the adherence disparity.

Conclusion: We found racial disparities to be largest among subpopulations with poor adherence. Findings suggest immediate roles for drug coverage generosity and a quality patient-provider relationship in improving adherence among black seniors. However, it is important not to lose sight of unequal social forces as fundamental culprits in shaping, for example, blacks’ access to employment earlier in life with subsequently generous retiree coverage, access to competent, compassionate providers, and having compelling literacy that statins are important to take regardless of symptomatology.