Agentic Artificial Intelligence for Hospital Readmission Review: A Single-Center Blinded Evaluation and Exploratory Qualitative Analysis

The study evaluated agentic artificial intelligence for reviewing unplanned hospital readmissions within 30 days of discharge. The AI system analyzed clinical data and completed the same standardized review form used by physicians. In a comparison of 20 cases, AI classified 45 percent of readmissions as preventable, compared to 47.5 percent by physicians. Review quality was similar between AI and physicians (4.35 versus 4.20 on a 1-5 scale). AI analysis cost $0.23 per case, while physicians spent an average of 15 minutes per case (estimated $42.43 per case). The AI system showed no hallucinations and identified recurring issues in post-discharge follow-up plans and medication safety. Results suggest AI can be a useful tool, but preventability judgments remain variable and require human oversight.