Large Language Models Improve Trauma Triage Accuracy

Large language models enhance prehospital trauma triage, potentially improving patient outcomes.

• LLMs achieved 83.5% triage accuracy, outperforming human clinicians’ 78.9% accuracy (p<0.01).
• Under-triage with LLMs improved slightly to 4.8%, compared to 5.1% with human triage (p=0.73).
• Essential transcripts reduced communication length by 80.8% while maintaining accuracy.

Integrating LLMs in trauma assessment can lead to better decisions, but further research is necessary for validation.