PRP040: Bayesian Analysis of a Medical Legal Partnership's Effect on Mental Health and Utilization: A Randomized Controlled Trial
Winston Liaw, MD, MPH; Robert Suchting, PhD; Thomas Northrup, PhD; Angela Stotts, PhD; Abigail Hernandez, BS
Context: Despite calls to address social determinants of health, family physicians lack resources to deal with social needs, including health-harming legal needs (HHLNs). One promising resource is the medical-legal partnership (MLP), which embeds legal screening and referral in clinics, but more evidence is needed prior to broader implementation. Objective: The objective is to examine whether an MLP reduces stress, depression, and medical overutilization. Study Design: Randomized controlled trial. Setting: Urban, primary care clinic in Houston. Population studied: Low-income individuals, aged 18 or older, who are English or Spanish speaking with positive screens for HHLNs. Intervention: Randomization to an MLP referral or 6-month waitlist. Outcome Measures: The primary outcome was perceived stress (Perceived Stress Scale (PSS); 0-40 range). Secondary outcomes were depression (Center for Epidemiologic Studies Depression Scale (CES-D); 0-60 range with 16 or higher being consistent with a depressive disorder diagnosis), and self-reported emergency department, urgent care, and hospital visits. Assessments occur at 3, 6, and 9 months, and we report 3 month data. Generalized linear modeling was used to fit each outcome at 3-months as a function of MLP group, controlling for baseline. Bayesian statistical inference with weakly informative priors and a 75% posterior probability (PP) threshold was used to identify noteworthy differences. For this interim analysis, groups remained blinded. Results: N=160 were randomized to intervention (80) and control (80) groups. After removing assessments with missing values and non-responders, the follow up rates ranged from 73.8% to 86.3% for Group 1 and 72.5% to 90.0% for Group 2, across the outcomes. Analyses did not find meaningful group differences on the PSS (PP = 50.6%). Being in Group 2 was associated with 44% lower odds of having a CES-D score of 16 or higher (PP = 84.3%) and being hospitalized at 0.5 the rate of Group 1 (PP = 85.5%). The PP for other outcomes (urgent care, and emergency department) did not reach the 75% threshold. Conclusion: Group 2 was associated with lower hospital visits and depression scores. The six and nine month assessments are ongoing and will determine whether these differences persist and new differences emerge.