Statistical Analysis

We calculated descriptive statistics for the past-year OUD treatment need sample using survey weights to derive nationally representative estimates, as well as the proportion within each subgroup reporting past-year MOUD, non-MOUD services, or no treatment. Then we fit multinomial regression models to identify the association of predisposing, enabling, and need variables with MOUD, as compared with both non-MOUD services and no drug treatment. We also examined past-year contacts with the health care and criminal legal systems to describe rates of MOUD among people encountering each system. In sensitivity analyses, the model included detailed categories of public insurance and an additional indicator of past-year criminal legal system involvement to examine associations independent of criminal legal contact. All analyses were conducted using svy command estimations in Stata, version 15MP (StataCorp LLC), with standard errors accounting for complex survey data design using Taylor linearization, and a 2-sided P value significance threshold of less than .05.