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

Area of Science:

• Clinical Informatics
• Addiction Medicine
• Health Services Research

Background:

• Predictive models as clinical decision support systems face challenges in adoption.
• Understanding factors influencing the uptake of these tools is crucial for effective implementation.

Purpose of the Study:

• To identify factors impacting the adoption of a predictive model for medication for opioid use disorder (MOUD) selection (methadone vs. buprenorphine) in inpatient settings.
• To explore clinician perspectives on utilizing decision support tools in MOUD treatment.

Main Methods:

• Qualitative study employing semi-structured interviews with 13 clinicians prescribing MOUD.
• Socio-technical model applied for combined deductive and inductive content analysis.

Main Results:

• Non-specialist clinicians defer to specialists and trust tools endorsed by experts and institutions.
• Clinicians expect models to incorporate clinical reasoning beyond electronic health record data, including social determinants of health.
• Trust in predictive models is contingent on perceived alignment with clinical judgment and consideration of multifaceted patient needs.

Conclusions:

• Predictive models for MOUD must be designed to build appropriate trust among users.
• Design considerations should encompass the tool's purpose, process, limitations, and performance to ensure effective clinical integration.