Can AI Predict the Future of Emergency Patients? A Review of Prognostic Models from Triage to Discharge

With the increasing volume of data in emergency settings and the need for rapid and accurate decision-making, artificial intelligence has emerged as a promising tool for prognostic prediction in emergency patients. This narrative review examines the application of AI in predicting outcomes such as mortality, need for intensive care, hospital admission, and cardiac arrest. Machine learning and deep learning models—including random forests, gradient boosting, and convolutional neural networks—using structured data, clinical text, physiological signals, and even short video clips have demonstrated performance equal to or better than traditional triage and early warning systems. However, significant challenges remain, including limited external validation, insufficient calibration reporting, potential biases across demographic subgroups, and a lack of prospective evaluations and randomized trials. These issues hinder the translation of such models into reliable clinical tools. The review emphasizes the need for outcome standardization, improved transparency, fairness assessments, and the design of human–AI interactive systems suited to the high-pressure emergency care environment.