Scenario Dataset - Can We Rely on Generative AI for Emergency Patient Triage Classification?

Description

This dataset contains clinical triage cases designed to compare expert medical decisions with AI-generated recommendations. Each entry captures the triage system used, expert judgment, and AI output. The Actual and Predicted columns use numerical values representing the expert and AI triage classifications, enabling easier quantitative analysis of performance and agreement across patient groups and acuity levels. The dataset includes the following variables: TRIAGE_CODING: The triage coding system or framework applied in the scenario. EXPERT_SOLUTION: The triage classification provided by a medical expert (ground truth). CHAT_GPT_SOLUTION: The triage classification generated by the AI model. Adult/Pediatric: Indicates whether the case involves an adult or pediatric patient. Actual: Numerical representation of the expert’s triage classification. Predicted: Numerical representation of the AI model’s triage classification. Acuity: Severity level of the case, indicating urgency of care. Potential Applications: Evaluation of AI performance in triage classification Quantitative comparison of expert vs AI decisions Development and benchmarking of clinical decision-support systems Analysis of triage consistency across acuity levels and patient populations

Institutions

• Texas Tech University

  Texas, Lubbock

Categories

Funders

• U.S. National Science Foundation

  Alexandria

  Grant ID: 2319802

Licence