Dynamic Treatment Regimes for Personalized Opioid Use Disorder Care
Dynamic Treatment Regimes for Personalized Opioid Use Disorder Care
This study aims to develop and evaluate dynamic treatment regimes (DTRs) to improve personalized care for individuals with opioid use disorder (OUD). Using machine learning methods and longitudinal data from a national behavioral health provider, the investigators will identify optimal treatment sequences that minimize the risk of overdose and improve recovery outcomes. A pilot hybrid factorial SMART trial will be conducted to assess the feasibility and acceptability of implementing these personalized treatment decision rules in real-world clinical settings.
This project supports the development and testing of dynamic treatment regimes (DTRs) for individuals with opioid use disorder (OUD), leveraging clinical and behavioral data from Discovery Behavioral Health and linked administrative records. In the first phase, multiple machine learning approaches (e.g., Q-learning, causal forests) will be used to estimate and validate DTRs that recommend tailored sequences of care based on patient characteristics and treatment response. The DTRs will be evaluated based on their ability to reduce fatal and non-fatal opioid overdose risk and hospital-based service use.
In the second phase, the investigators will conduct a pilot hybrid factorial Sequential Multiple Assignment Randomized Trial (SMART) to test the feasibility and acceptability of implementing the highest-performing DTR in a clinical setting. This pilot study will assess recruitment and retention, adherence to assigned treatment paths, and the practicality of integrating the DTR into routine care. The findings will inform a future full-scale trial aimed at improving personalized care for OUD and outcomes.
Inclusion Criteria:
Exclusion Criteria:
jbgibbons@bwh.harvard.edu7734011266