At a Glance:
- AI-based screening flagged patients at risk for opioid use disorder (OUD) and helped reduce 30-day hospital readmissions.
- The tool may offer a cost-effective way to improve access to addiction care in clinical settings.
A novel artificial intelligence (AI) screening tool has shown significant potential in identifying hospitalized patients at risk for opioid use disorder (OUD) and reducing their likelihood of readmission, according to a new NIH-supported study led by Dr. Majid Afshar at the University of Wisconsin–Madison.
OUD continues to be a critical public health crisis in the United States. Hospitalized patients with OUD face elevated risks of overdose, repeated hospital visits, and complications. While hospital-based addiction treatment improves outcomes, inconsistent screening practices often result in missed opportunities for intervention. Many patients are discharged before being seen by addiction specialists, heightening their risk of overdose after leaving the hospital.
To address these gaps, researchers developed an AI screening system that analyzes electronic health records (EHRs) in real time to detect patterns suggestive of OUD. The system can alert clinicians and recommend consultations with addiction specialists or other relevant interventions.
In a study published April 3, 2025, in Nature Medicine, the team evaluated the AI screener’s performance during a clinical trial involving more than 51,000 adult patients admitted to the University of Wisconsin Hospital. The research compared a baseline phase—where providers conducted risk assessments without AI support—to a second phase where the AI tool was actively used.
During the AI implementation phase, more than 17,000 patients were screened, leading to 727 addiction medicine consultations. While AI and provider-only assessments resulted in similar rates of specialist consultations, patients screened by AI were 47% less likely to be readmitted to the hospital within 30 days after discharge.
The cost savings were notable: each avoided readmission saved an estimated $6,800 in health care expenses, highlighting the potential financial and clinical value of deploying AI in hospital workflows.
“AI holds promise in medical settings, but many models have remained stuck in the development phase without integration into real-world environments,” said Dr. Afshar. “Our study is one of the first to demonstrate that AI screening can be embedded effectively into addiction medicine and hospital systems, showing both practical feasibility and clinical impact.”
The findings underscore how AI-driven tools can augment clinician decision-making, streamline patient care, and potentially improve long-term outcomes for individuals struggling with substance use disorders.
