Related Experiment Video
Updated: Jul 28, 2026
09:52
Setting Up a Stroke Team Algorithm and Conducting Simulation-based Training in the Emergency Department - A Practical Guide
Published on: January 15, 2017
17.1K
Enhancing trauma triage in low-resource settings using machine learning: a performance comparison with the Kampala
Mike Nsubuga1,2,3, Timothy Mwanje Kintu4,5, Helen Please6,7
1The Infectious Diseases Institute, Makerere University, P. O. Box 22418, Kampala, Uganda. nsubugamike021@gmail.com.
BMC Emergency Medicine
|January 24, 2025
Summary
Machine learning models significantly outperform the Kampala Trauma Score (KTS) for predicting trauma triage decisions. These advanced models offer a promising approach to improve trauma care in low- and middle-income countries (LMICs).
Area of Science:
- Medical informatics and artificial intelligence in global health.
- Trauma care and emergency medicine research.
Background:
- Traumatic injuries are a major global cause of death and disability, particularly in low- and middle-income countries (LMICs).
- The Kampala Trauma Score (KTS) is widely used for trauma triage in LMICs, but its predictive accuracy is debated.
Purpose of the Study:
- To evaluate the effectiveness of machine learning (ML) models in predicting trauma triage decisions.
- To compare the performance of ML models against the established KTS.
Main Methods:
- Trained and evaluated four ML models (Logistic Regression, Random Forest, Gradient Boosting, Support Vector Machine) using data from 4,109 trauma patients in Uganda.
- Assessed model performance using accuracy, precision, recall, F1-score, and AUC-ROC.
- Developed a benchmark multinomial logistic regression model based on the KTS.
Main Results:
- All ML models significantly outperformed the KTS (AUC-ROC 0.91 vs. 0.62, p < 0.01).
- The Random Forest model achieved the highest accuracy (0.69), compared to the KTS model (0.54).
- Sex, hours to hospital, and age were the most significant predictors identified by the ML models.
Conclusions:
- Machine learning models demonstrate superior predictive capabilities for trauma triage compared to the KTS, using limited patient data.
- Integrating ML into triage decision-making presents a significant opportunity to advance trauma care in LMICs.
- Further validation is crucial for the reliable integration of ML models into clinical practice for improved resource allocation and patient outcomes.
