Abstract Timely stroke diagnosis is essential for delivering life-saving treatments, yet current prehospital stroke assessment tools often lead to missed or delayed diagnoses. Emergency Medical Services (EMS) increasingly collect detailed data during transport, offering an opportunity to develop AI-based tools to support early stroke detection. In this retrospective study, we evaluated the availability and reliability of prehospital data compared to Emergency Department (ED) records. Using this data, we tested the potential of machine learning to support EMS stroke triage. Our cohort included 4,754 patients across 8,796 ambulance encounters from 2015 to 2020 with stroke rate of 2.2% (68% severe strokes). Vital signs, recorded in over 88%, were generally higher than their ED counterparts, with increasing transit time significantly associated with decreasing ED values. We trained and evaluated random forest, XGBoost, and sequential neural networks models for detecting strokes and severe strokes. The random forest model achieved the best performance for stroke detection (ROC-AUC 0.827, PR-AUC 0.230), while XGBoost performed best for identifying severe strokes (ROC-AUC 0.871, PR-AUC 0.237). Models were calibrated to improve reliability, and feature importance was assessed using SHAP to enhance interpretability. These findings highlight the promise of AI-based tools in improving prehospital stroke triage with real-time EMS data.
Saban et al. (Mon,) studied this question.