Background: Delayed Code Stroke activation contributes to worse outcomes in acute stroke. Emergency Department (ED) triage notes contain free-text clinical information that could enable automated, real-time pathway activation. We evaluated the diagnostic accuracy of a multi-pass large language model (LLM) pipeline for identifying patients meeting Code Stroke criteria from ED triage notes.
Diagnostic Accuracy of a locally deployed Large Language Model Algorithm for Automated Code Stroke Pathway Identification in Emergency Department Triage Notes
calendar_today
August 5, 2026
person
Valente, M. J., Sharobeam, A., Vuong, J., Chan, W. K.
domain
medrxiv