Abstract
Background Adherence to hydroxychloroquine (HCQ), the standard-of-care treatment among patients with systemic lupus erythematous (SLE) is suboptimal and interventions to date to improve this have been largely unsuccessful. Individual-level factors, including younger age and non-white race/ethnicity, have been implicated. While contextual influences such as area-level poverty are known to contribute to health behaviors and to SLE-related damage, no studies to date investigate the role of these factors on HCQ adherence. We therefore aimed to use multilevel models to understand the effects of sociodemographic variables (area-level racial composition, poverty, educational attainment) and health resource concentration (numbers of physicians, pharmacists, hospitals) on HCQ adherence while also accounting for potential individual-level factors.
Methods We identified SLE patients with new use of HCQ (no use in ≥6 months) in Medicaid (2000–2010) from 28 U.S. states. We required 12 months of continuous enrollment after HCQ initiation with complete drug dispensing data and measured adherence using the proportion of days covered (PDC) during this period. We identified individual-level variables from Medicaid, zip code, county and state-level sociodemographic variables from the American Community Survey, and health resource data from Area Health Resources Files. We used 4-level hierarchical multivariable logistic regression models to examine odds (OR (95% Credible Interval)) of adherence (PDC ≥80%) vs nonadherence.
Results Among 10,268 HCQ initiators with SLE, 15% were adherent (PDC ≥80%). After adjusting for individual-level demographic and SLE-related characteristics, we observed lower odds of adherence in zip codes with higher percentages of black individuals (highest tertile OR 0.81 (0.69–0.96) vs lowest) (table 1). This association persisted after controlling for zip code educational attainment, percent below federal poverty level (FPL), urbanicity and healthcare resources. We did not find statistically significant associations with zip code-level percent Hispanic, percent White, education or percent below FPL. Odds of adherence were higher in counties with the highest concentration of hospitals vs the fewest (OR 1.30 (1.07–1.58)); no statistically significant associations were found with other health resource characteristics.
Conclusions Among Medicaid beneficiaries with SLE, we observed significant effects of racial composition and hospital concentration on HCQ adherence. Further studies with smaller geographic units and data on spatial relationships are needed to investigate the potential role of residential racial segregation on adherence. Interventions that acknowledge and address contextual factors should be considered to reduce high rates of nonadherence in vulnerable populations.
