Background/Purpose: Among lupus patients, adherence to hydroxychloroquine (HCQ), the backbone of therapy, remains suboptimal. Individual-level factors, including younger age, poverty, and black race, have been associated with HCQ nonadherence. However contextual factors including neighborhood poverty and concentration of healthcare resources have not been examined. We constructed multilevel models to investigate whether zip code, county and state-level characteristics were associated with HCQ nonadherence.

Methods: We identified individuals with SLE enrolled in Medicaid (2000-2010) from 28 U.S. states using a previously defined algorithm. We included new users of HCQ (no use in ≥6 months) with ≥12 months of continuous enrollment with complete drug dispensing data following HCQ initiation. Adherence was measured over this 12-month period using the proportion of days covered (PDC) and defined as ≥80%. We identified individual-level characteristics (demographics, medications, comorbidities) from Medicaid data. We obtained zip code, county and state-level characteristics (percent of the population below the Federal poverty level (FPL), educational attainment and percent black population) from the American Community Survey. Health resource data (per capita hospitals, physicians, pharmacies and health professional shortage areas) were obtained from Area Health Resources Files. We used four-level hierarchical multivariable logistic regression models with Markov Chain Monte Carlo procedures to examine odds (OR [95% credible interval]) of adherence vs. nonadherence.

Results: Among 10,268 HCQ initiators with SLE residing within 4,930 zip codes in 1,414 counties in 28 states, 15% were adherent. After adjusting for individual-level characteristics, we observed lower odds of adherence across zip codes with higher percentages of black residents (highest tertile OR 0.81 [0.68-0.96] vs. lowest) (Table). The association remained after controlling for zip code percent below FPL and educational attainment. Odds of adherence were higher in counties with the greatest number of hospitals vs. the fewest (OR 1.32 [1.08-1.60]), and lower in health professional shortage areas (OR 0.86 [0.75-1.00]). There was no association with county-level per capita number of physicians or pharmacists. There was minimal variation in adherence by geographic area; the greatest was between states (1.4%).

Conclusion: Among Medicaid beneficiaries with lupus, after accounting for individual-level factors, we observed reduced odds of HCQ adherence in areas with higher percentages of black residents, fewer hospitals, and shortages of health professionals. Further studies are needed to assess whether racial residential segregation and poorer availability of high-quality medical care may contribute to racial differences in HCQ nonadherence and in turn, to disparities in lupus outcomes.