Pitt PI: Wei-Hsuan Jenny Lo-Ciganic, PhD, MS, MSPharm
Funding Source: NIH/NIDA
March 2026 - February 2028
This project proposes to reduce buprenorphine care discontinuity—and the associated risks of relapse and mortality— among patients with opioid use disorder (OUD) by developing a machine learning-based clinical decision support (CDS) tool for primary care providers. More than half of patients prescribed buprenorphine, an evidence-based OUD treatment that reduces overdose and death, discontinue it within six months of initiation. Primary care providers (PCPs), who prescribe the majority of buprenorphine, often lack systematic tools to efficiently identify patients at highest risk of stopping treatment. The BUP-CARE team will use electronic health record (EHR) data from OneFlorida+, a PCORnet Clinical Data Research Network, to develop and validate machine-learning algorithms that predict buprenorphine discontinuation, incorporating social determinants of health and advanced deep learning techniques such as recurrent neural networks. The team will then apply a user-centered design process to prototype a CDS e-tool that integrates these algorithms directly into the EHR workflow, alerting PCPs when their patients are at high risk of discontinuation so that tailored interventions, such as long-acting injectable buprenorphine or peer recovery support, can be offered proactively. This work builds on Dr. Lo-Ciganic's prior machine learning research.