Dataset_of_Study_compare_model_of_prediction_triage

Emergency departments (EDs) play a crucial role in urgent care for patients with varying acuity levels. Accurate Emergency Severity Index (ESI) triage is vital for optimizing patient flow and resource allocation. However, the increasing volume and complexity of ED visits challenge manual triage, potentially causing inconsistencies and delays. This study compares the performance of logistic regression, gradient boosting, neural network, and random forest models in predicting ESI triage levels for non-traumatic patients at Lampang Hospital's ED. A retrospective observational study was conducted using data from January 1, 2026, to April 30, 2027. After data cleaning, 45,246 complete records were analyzed. The dataset was divided into training and testing sets using stratified cross-validation (k=10). Model performance was evaluated using precision-recall graphs, accuracy, precision, recall, F1 score, confusion matrix, and feature importance. The gradient boosting model demonstrated the highest performance (accuracy, precision, recall, F1 score of 0.81). The neural network model also performed well (metrics at 0.78), while logistic regression had the lowest performance (accuracy 0.64, F1 score 0.60). Significant predictors included pain scale, sex, and mean arterial pressure. Gradient boosting and neural network models outperform logistic regression in predicting ESI triage levels, potentially enhancing triage accuracy and efficiency in ED.