Detecting Human Bias in Emergency Triage Using LLMs: Literature Review, Preliminary Study, and Experimental Plan

Marta Avalos (1, 2, 3) , Dalia Cohen (2, 3) , Dylan Russon (2, 3) , Melissa Davids (3, 2) , Océane Dorémus (2, 3) , Gabrielle Chenais (2, 3) , Eric Tellier (2, 3) , Cédric Gil-Jardiné (2, 3, 4) , Emmanuel Lagarde (2, 3)

1 SISTM - Statistics In System biology and Translational Medicine
2 BPH - Bordeaux population health
3 UB - Université de Bordeaux
4 CHU Bordeaux

The surge in AI-based research for emergency healthcare poses challenges such as data protection compliance and the risk of exacerbating health inequalities. Human biases in demographic data used to train AI systems may indeed be replicated. Yet, AI also offers achance for a paradigm shift, acting as a tool to counteract human biases. Our study focuses on emergency triage, rapidly categorizing patients by severity upon arrival. Objectives include conducting a literature review to identify potential human biases in triage and presenting a preliminary study. This involves a qualitative survey to complement the review on factors influencing triage scores. Moreover, we analyze triage data descriptively and pilot AI-driven triage using an LLM with data from the local hospital. Finally, assembling these pieces, we outline an experimental plan to assess AI’s effectiveness in detecting human biases in triage data.

Mots clés: Emergency Department; Triage; Human Bias; Large Language Models; Natural Language Processing

Domaines: Santé publique et épidémiologie; Intelligence artificielle [cs.AI]; Statistiques [stat]

Marta Avalos, Dalia Cohen, Dylan Russon, Melissa Davids, Océane Dorémus, et al.. Detecting Human Bias in Emergency Triage Using LLMs: Literature Review, Preliminary Study, and Experimental Plan. FLAIRS 2024 - 37th International Florida Artificial Intelligence Research Society Conference, The Florida Artificial Intelligence Research Society, May 2024, Miramar Beach, United States. pp.6. ⟨hal-04575557⟩