Related Experiment Video
Updated: Mar 29, 2026
07:31
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
8.3K
Interpretable Machine Learning for Emergency Department Triage: Clinical Insights from 133,198 Patients Using the
MyoungJe Song1, Jongsun Kim1, Eun-Chul Jang2
1Department of Emergency Medicine, International St. Mary's Hospital, Catholic Kwandong University, Incheon 22711, Republic of Korea.
Diagnostics (Basel, Switzerland)
|March 28, 2026
Summary
Explainable AI improves emergency room triage by providing transparent insights for Korea Triage and Acuity Scale (KTAS) prediction. This approach enhances accuracy and supports clinicians in identifying high-risk patients.
Area of Science:
- Artificial Intelligence
- Healthcare Informatics
- Clinical Decision Support
Background:
- Emergency room severity classification (KTAS) is crucial for patient safety but relies on subjective judgment.
- Current machine learning models lack clinical trust due to their opaque 'black box' nature.
Purpose of the Study:
- To develop an explainable machine learning (XAI) framework for transparent and accurate KTAS prediction.
- To overcome the limitations of traditional subjective and black-box AI models in clinical settings.
Main Methods:
- Retrospective analysis of 133,198 emergency room visits (2022-2024).
- Training of Random Forest (RF) and XGBoost models using vital signs and pain scores.
- Application of explainable AI (XAI) techniques for model transparency.
Main Results:
- XGBoost achieved 94.7% accuracy, while RF demonstrated 91.6% accuracy with better interpretability.
- XAI analysis identified pain score, age, and systolic blood pressure as key predictors, aligning with clinical logic.
- The RF model was selected for its balance of predictive power and interpretability.
Conclusions:
- Explainable AI offers transparent insights for KTAS prediction, surpassing black-box model limitations.
- These models can enhance emergency department triage consistency and aid in identifying high-risk patients.
- External validation is necessary before routine clinical implementation.
