Estimating community-level prevalence of opioid use disorder: Extrapolating from Medicaid claims data and other publicly available data sources in Ohio, USA

Background and aims: Addressing the opioid crisis requires opioid use disorder (OUD) prevalence estimates at the community level for targeted interventions. This study presents a new method that utilizes Medicaid claims data and publicly available data to estimate OUD prevalence at the United States (US) county level and compares it with other existing estimates. Design, setting and participants: This study utilized data on OUD diagnoses among Medicaid beneficiaries in combination with national survey data, Census data and published literature to estimate the prevalence of OUD (including undiagnosed OUD) in each of the US state of Ohio's 88 counties for the years 2019 to 2021 among the population aged 12 years and older. Prevalence estimates were adjusted for misclassification in claims data and for variation in healthcare utilization among individuals with OUD. Measurements: Counts and proportions of the population aged 12 and older with OUD at the US state and county level. Findings: OUD prevalence for Ohioans aged 12 years and older was relatively stable at 3.6% [95% uncertainty interval (UI) = 3.5%–3.8%] in 2019 and 3.7% (95% UI = 3.5%–3.9%) in 2021. County-level prevalence estimates ranged from 0.7% to 14.2% in 2021. Southern counties generally had higher OUD prevalence than northern counties. The prevalence estimates were strongly correlated (Pearson's r = 0.88) with prevalence estimates for 19 Ohio counties from a previous study. Compared with the previous estimates, estimates from the current study tended to be lower for most communities but remained within the 95% credible intervals of previous estimates. Conclusions: This approach for estimating opioid use disorder prevalence within United States communities by using Medicaid claims data and publicly available data is a robust alternative to methods relying on individual-level data or multiple linked datasets.