Emergency Department Admitting Service Triage Using Retrieval-Augmented Language Models

Poster Number: P35
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Large Language Models (LLMs), Information Retrieval, Workflow, Informatics Implementation, Internal Medicine or Medical Subspecialty
Primary Track: Applications
Programmatic Theme: Clinical Informatics

We evaluated the utility of large language models (LLMs) for triaging emergency department admissions an academic medical center. Using retrieval-augmented generation (RAG) with existing policy documents, we tested Anthropic's Claude 2 and OpenAI's GPT-4. Our findings show that GPT-4, combined with a departmental spreadsheet, most accurately determined correct admitting services, suggesting a viable method to reduce clerical burden without additional resource investment.

Speaker(s):
Dong-han Yao, MD
Stanford

Author(s):
Dong-han Yao, MD - Stanford; Debadutta Dash, MD - Stanford; Adi Badhwar, MS - Stanford; Wendy Li, MD - Stanford; Nicholas Hall, MD - Stanford; Yoseph Semma, MD - Stanford;