WVU School of Pharmacy develops AI tool targeting 50% reduction in 30-day patient readmissions through medication reconciliation

Medication reconciliation errors account for 85% of discharge-related errors, with at least 1.5 million people harmed by medication errors annually at an estimated cost of $3.5 billion. Pharmacists at large facilities like J.W. Ruby Memorial Hospital must manually review records for ~200 discharge patients per day, spending 30–50 minutes per patient pulling data from multiple clinicians across fragmented electronic systems.

The WVU HealBig Research Lab is developing an AI tool that uses deep learning for natural language processing to parse physician, nurse, and pharmacist notes from electronic health records. The tool builds a comprehensive patient profile and generates a risk score for 30-, 60-, and 90-day readmission, triggering an alert system for pharmacists when a patient is flagged as high risk so the care team can intervene before discharge.

The tool is currently in prototype development, funded by a $100,000 two-year grant from the American College of Clinical Pharmacy. Pilot integration into a hospital EHR system is planned as the next phase. Referenced benchmarks indicate that pharmacist-led transition-of-care programs have achieved 50% reductions in 30-day readmission rates, which this AI tool aims to replicate at scale.

• Medication reconciliation is the highest-risk step in the discharge process, accounting for 85% of errors — automating it with NLP-driven profiling directly targets this failure point.
• Combining readmission risk scoring with a real-time alert system allows pharmacists to escalate borderline cases back to the care team before the patient leaves the hospital.
• Academic research labs can serve as the development engine for clinical AI tools, with a clear path from prototype to EHR-integrated pilot at a partnered health system.