Hybrid AI-Driven Predictive Triage for Emergency Care Using Patient Clustering, Temporal Deep Learning, and IIoT-Based Monitoring

Emergency departments (EDs) face increasing patient volumes and limited clinical resources, making efficient and accurate triage essential for prioritizing high-risk patients and ensuring timely medical intervention. This study proposes a hybrid artificial intelligence–driven predictive triage framework that integrates Gaussian Mixture Model–based patient clustering, Long Short-Term Memory (LSTM) networks for temporal analysis, and XGBoost for risk classification. The framework incorporates Industrial Internet of Things (IIoT)–based monitoring for real-time physiological data acquisition and predictive analytics. The model is evaluated using MIMIC-IV and eICU datasets, achieving an accuracy of 92.4% and an AUC–ROC score of 0.91. The results demonstrate that integrating patient stratification, temporal deep learning, and real-time monitoring significantly improves early detection of patient deterioration and supports clinical decision-making in emergency care settings. This work is intended for submission to a peer-reviewed journal and has not yet undergone formal peer review.