Clinician Perspectives on a Predictive Model for Recommending Opioid Use Disorder Treatment

Background: Predictive models that have been made available as clinical decision support systems have not always been used. Objectives: This qualitative study aimed to identify factors that might impact the uptake of a predictive model recommending either methadone or buprenorphine as medication for opioid use disorder (MOUD) in the inpatient setting. Methods: We conducted semi-structured interviews with clinicians who prescribe MOUD and performed a combined deductive and inductive content analysis using a socio-technical model. Results: Thirteen clinicians were interviewed. Non-specialists trusted their specialist peers to lead MOUD decisions and claimed they would trust a tool endorsed by experts and the institution. Clinicians expected the model to follow clinical reasoning, which involves considering factors that are not well-captured by the electronic health record (e.g., housing status, access to care, facility preferences). Conclusion: Predictive models for MOUD should be designed to foster appropriate trust given the tool's purpose, process, limitation, and performance.