University of Wisconsin Hospital

Screening for opioid use disorder in hospitals remains inconsistent, with hospitalized patients frequently leaving before seeing an addiction specialist — a factor linked to a tenfold increase in overdose rates. In overwhelmed hospital settings, resource-intensive procedures like addiction screening are easily overlooked, limiting access to treatment.

Researchers at the University of Wisconsin School of Medicine deployed an AI screening tool embedded into the hospital's EHR workflow. The tool analyzed clinical notes, medical history, and other real-time documentation to identify patterns associated with opioid use disorder, then issued alerts to providers when opening a patient's chart, recommending addiction medicine consultation and withdrawal monitoring.

The AI screening group saw approximately 8% 30-day readmission rates compared to 14% in the provider-led group — a 47% reduction in odds of readmission. The tool generated an estimated $108,800 in healthcare savings over the 8-month deployment period, at a net cost of $6,801 per readmission avoided, even after accounting for AI software maintenance costs.

• AI-assisted screening matched provider-initiated consultation quality while offering a more scalable, automated approach that doesn't depend on overstretched staff remembering to screen.
• Embedding AI alerts directly into the EHR chart-opening workflow drove real-world adoption without requiring separate tools or workflows.
• Alert fatigue and cross-system validation remain challenges; broader deployment across different health systems is needed before generalizing results.