A research team at the University of Wisconsin School of Medicine and Public Health has demonstrated the real-world potential of artificial intelligence in addressing the opioid crisis, according to a press release. In a clinical trial funded by the National Institutes of Health and published in Nature Medicine, the team embedded an AI-driven screening tool into hospital workflows to detect patients at risk for opioid use disorder. The tool successfully identified patients in need of addiction care and prompted timely referrals to addiction specialists.
Compared to traditional provider-initiated screenings, the AI tool was equally effective at initiating consultations and recommending withdrawal monitoring. However, the AI approach significantly reduced hospital readmissions. Patients identified through AI and treated by addiction specialists had 47% lower odds of returning to the hospital within 30 days of discharge. This translated to nearly $109,000 in health care savings over the eight-month study period.
The trial involved over 51,000 hospitalizations between 2021 and 2023, with 727 addiction consultations completed. When the AI screener was deployed, 1.51% of patients received a referral, compared to 1.35% during the provider-only phase. Readmission rates also dropped, with 8% of AI-identified patients returning to the hospital, versus 14% in the traditional group. These findings remained consistent even when accounting for demographic and health-related variables.
Lead researcher Dr. Majid Afshar called the trial a breakthrough in translating AI from concept to clinical application, citing its pragmatic integration with electronic health records. Still, challenges such as provider alert fatigue and the need for broader validation persist. Future efforts will focus on refining the tool and expanding its use in diverse hospital systems.
