A machine learning approach to unravel client and program-specific effects in opioid treatment retention

Abstract

This study examines the impact of workforce diversity, particularly the presence of Black/African American staff, on client retention in opioid use disorder (OUD) treatment, recognizing the historically low retention rates among Black and Hispanic populations in such programs. Using a novel machine learning technique called “causal forest,” we explored the heterogeneous treatment effects of staff diversity on client retention, aiming to identify strategies that enhance client retention and improve treatment outcomes. Analyzing data from four waves of the National Drug Abuse Treatment System Survey spanning the years 2000, 2005, 2014, and 2017 (n = 627), we focus on the relationship between workforce diversity and retention. The findings revealed diversity-related variations in retention across 61 out of 627 OUD treatment programs (<10%), with potential beneficial effects attenuated by other program characteristics. These characteristics include programs that are more likely to be private-for-profit, have lower percentages of Black and Latino clients, lower staff-to-client ratios, higher proportions of staff with graduate degrees, and lower percentages of unemployed clients. Our results suggest that workforce diversity alone is insufficient for improving retention. Programs with characteristics linked to greater retention are better positioned to leverage a diverse workforce to enhance retention, offering important implications for policy and program design to better support Black clients with OUDs.