LATEBREAKER
Substance Use
Development of a national index of county-level access to opioid use disorder care and its performance using 2018 Medicaid administrative claims data Jannat Saini* Jannat Saini Helen Powell Ester Villalonga-Olives Nishant Shah Susan dosReis Danya Qato
Introduction: Factors informing access to opioid use disorder (OUD) treatment in the US are multifaceted, yet are often examined singularly as if independent from one another. A composite index of access to OUD care is valuable to quantify OUD resource environment. Medicaid, the largest payer for OUD treatment, is central to addressing the OUD crisis. Therefore, establishing such an index with linkages to Medicaid claims data provides a more holistic view of the OUD population. This study aims to generate a novel county-level OUD Resource Index (OUDRI).
Methods: To ensure OUDRI content validity, we established a conceptual framework with expert feedback to generate index domains representing OUD resource environment. We used publicly-available data (American Community Survey, Census, etc.) to measure characteristics under each domain. We used principal components analysis (PCA) to generate OUDRI. We derived a cohort of beneficiaries with OUD in 2018 from a national sample of Medicaid administrative claims database representing 3 million beneficiaries with OUD (2010-2018). We linked OUDRI to each beneficiary using state-county FIPS codes for a cross-sectional study. Binomial logistic regressions evaluated the relationship between medication for OUD (MOUD) utilization and OUDRI (tertiles; ranked from 1-3, representing lowest to highest potential access to OUD care) to assess construct (convergent) validity.
Results: The OUDRI, constructed from the first PCA component, comprises information from 22 variables representing all theoretical domains. The index was linked to a final sample of 1,237,812 beneficiaries representing 3,003 counties. In multivariate analysis (adjusted for 3 variables excluded from PCA due to low correlation), OUDRI tertiles 1 (33.3% of overall sample) and 2 (33.5%) had lower odds of MOUD utilization (OR: 0.65; 95% CI: [0.65, 0.66] and 0.89; [0.88, 0.90], respectively) compared to those in OUDRI tertile 3 (33.3%).
Conclusions: The OUDRI summarized multiple theoretical domains of OUD resource environment in each county and encompasses characteristics often unmeasured in observational studies of OUD. Likelihood of MOUD utilization was higher in the highest OUDRI tertile compared to the lowest, suggesting construct validity of OUDRI which can be leveraged by other researchers.