An AI-powered clinical support tool will help prevent relapse in patients receiving buprenorphine treatment for opioid use disorder — a condition that affects hundreds of thousands of people in America every year.
The PROTECT tool, devised by researchers at the University of Florida and the University of Pittsburgh with a $3.6 million grant from the National Institutes of Health and the National Institute on Drug Abuse, uses machine learning algorithms to identify buprenorphine patients who are at high risk of relapsing and provides recommendations for next steps.
“Relapse risk isn’t a single lab value; it’s a pattern hidden across medical charts and in the social context. Our goal is to surface that pattern in time for busy primary care teams to act, so prevention becomes proactive rather than reactive,” said Mahmudul Hasan, Ph.D., an assistant professor in the UF College of Pharmacy Department of Pharmaceutical Outcomes and Policy with a joint appointment in the UF Warrington College of Business Information Systems and Operations Management Department. “PROTECT can flag early, subtle patterns and show an individual’s top risk factors, so teams know who is at risk, why and what to do next.”
Preventing overdoses, saving lives
Opioid use disorder is part of a larger public health crisis, as more than 81,000 Americans died from opioid overdose in 2023. Buprenorphine — one of the most commonly prescribed medications for the disorder — reduces the risk of opioid misuse and overdose, yet up to 60% of patients relapse while taking the drug. This statistic underscores the importance of providing patients at high risk of relapse with a personalized treatment plan, before overdose-related complications may occur.
“Buprenorphine is a lifesaving treatment, yet it works best when we can predict risk early and wrap the right supports around the medication at the right moment,” Hasan said. “Risk of relapse is dynamic, shifting with missed doses, new prescriptions, emergency department visits and job or housing changes. Patients benefit from layered support like timely counseling, dose or formulation adjustments and help with social needs.”
To calculate a patient’s risk of relapse, PROTECT — which stands for Prediction of Relapse on OUD Treatment using Machine Learning-Driven Evidence-based Clinical Decision Support Tool — uses machine-learning algorithms that capture data sets incorporating a patient’s prescriptions, medical office visits, toxicology reports and treatment times, as well as social factors and life changes. Hasan said PROTECT updates as data populates, giving clinicians an up-to-date view of a patient’s risk score.
A proactive and accessible approach
When PROTECT flags a patient susceptible to relapse, their clinician will review the calibrated risk score and its key factors before overseeing a tailored set of actions. Hasan said treatment suggestions might include adjustments to dose or formulation; an altered follow-up schedule; referrals to other facilities for pain, recovery or behavioral health; or resources for housing or transportation. Hasan noted that, because PROTECT augments clinical judgment, a clinician can override all alerts.
“Let’s say a patient misses two medication refills and has a recent emergency room visit, a new benzodiazepine prescription and documented housing instability. PROTECT quietly flags the patient as ‘high risk’ before the visit and proposes a brief, evidence‑based care report,” Hasan said. “The care provider accepts the recommendation, and the care team initiates outreach the same day.”
Hasan’s team has been developing PROTECT since 2023 and the tool should be ready for clinical use by 2030. The goal of PROTECT is simple: to make long-term recovery from opioid use disorder more attainable for all patients. When the tool is ready for clinical use, it will make relapse something to be anticipated and prevented, rather than a surprise that ends in overdose.
“I hope the tool creates a different experience for patients, one where risk is met with timely, compassionate action. A proactive call after a missed medication refill, a quick dose adjustment or a warm handoff to a counselor can interrupt a downward slide before it becomes a crisis,” Hasan said. “By showing clinicians why a patient’s risk is rising and making the right care easy to deliver, we can reduce overdoses and hospitalizations, strengthen trust and bring more stability to families and communities. Our aim is for these benefits to reach those most affected by the crisis, not just those easiest to reach.”