Enhancing Emergency Room Triage with Predictive Analytics: A Comparative Study of Random Forest, Gradient Boosting, and Neural Networks

Abstract

This research paper investigates the application of predictive analytics to enhance triage processes in emergency rooms, aiming to improve patient outcomes and optimize resource allocation. The study focuses on three advanced machine learning algorithms: Random Forest, Gradient Boosting, and Neural Networks. By employing a comprehensive dataset comprising patient demographics, clinical variables, and historical triage outcomes, the research constructs predictive models to evaluate and compare the performance of each algorithm. Key metrics include accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). Random Forest demonstrated robust performance with high interpretability, appealing for real-world implementation. Gradient Boosting achieved superior precision and recall, particularly in distinguishing high-risk patients, suggesting its potential to minimize critical oversight. Neural Networks, while computationally intensive, offered exceptional insights into complex non-linear relationships within the dataset, yielding competitive accuracy. The study further examines the scalability, ease of integration into existing systems, and real-time predictive capabilities of these models. Findings underscore the transformative potential of machine learning in medical triage, advocating for a hybrid approach that harnesses the strengths of multiple algorithms. The paper concludes by discussing practical implications, challenges in data privacy and ethics, and directions for future research to address limitations and enhance model robustness in diverse clinical settings.