Emergency Department Revisit Quality Review Screening: Exploring Human Decision-Making and Artificial Intelligence Support
A retrospective study found GPT-4 over-flagged emergency department revisit cases while an LLM knowledge-graph screener achieved 83-100% positive predictive value.
In an exploratory retrospective study of 99 emergency department diagnosis pairs from a multihospital health system, clinicians and GPT-4 independently judged whether revisit pairs warranted further assessment. GPT-4 responses correlated poorly with clinicians, flagging 94% of pairs for follow-up, 4.4-13.3 times more than clinicians, though prompt engineering was minimal. An algorithm leveraging an LLM-populated knowledge graph (KGA) achieved 83-100% positive predictive value against at least one clinician rater, suggesting LLM-based screening could broaden revisit quality review without substantially increasing reviewer workload.
- 99 ED revisit diagnosis pairs assessed by clinicians and GPT-4
- GPT-4 flagged 94% of pairs, 4.4-13.3x more than clinicians
- LLM knowledge-graph algorithm achieved 83-100% positive predictive value
- Revisit medical gravity consistently associated with clinician target judgments
- Authors call for validation before LLM-supported screening deployment
From the original article
1 min read at arxiv.orgBackground: Emergency Department (ED) return visits are commonly reviewed for quality assurance, but are often limited (e. g. , to revisits within 48-72 hours) to increase actionable finding yield while minimizing chart review burden.
By Jonathan A. Handler · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.