Methods: We drew on an original, longitudinal dataset for a cohort of people on probation in San Francisco, California (N = 772). We tested the relationship between weekly behavioral health service use and the outcome of recidivism (defined as rearrest) with a series of Cox proportional hazard models. We treated service use as a time-varying predictor and adjusted for competing exposures and potential confounders measured at baseline, including demographic, criminal, social and behavioral health-related factors. To assess effect variation, we interacted treatment with disorder type (mental, substance use, or comorbid) and service type (mental or substance use). To strengthen confidence that results reflect the true relationship between treatment and recidivism, and not the mode of analysis, we verified results with two additional analytic methods, a Marginal Structural Models weighting approach and a propensity score matching approach.
Results: Weekly service use was associated with a nearly 33% reduction in the risk of recidivism (HR = 0.67, p = .005), even after adjusting for demographic, criminal, social and behavioral health-related covariates. Both disorder and treatment type interacted with behavioral health services to affect recidivism, with the greatest effects for people on probation with mental disorder diagnoses and mental health-focused services. These results were stable across analytic approach.
Conclusion and Implications: We conclude that behavioral health service attendance overall has a meaningful and statistically significant recidivism reducing effect. However, this reduction is borne primarily by people with mental disorder diagnoses and mental health-focused treatment. These findings suggest efforts to reduce recidivism should include increasing access to community-based behavioral health services, especially for justice-involved persons with mental disorder diagnoses. These findings also suggest a need to identify and test interventions for reducing recidivism among people with substance use disorders and a need to identify factors that reduce the effectiveness of existing substance use interventions for justice-involved persons.