Operational AI in Cardiac Arrest Care (2023–2025): Mapping the Gap Between Published Machine Learning Models and Clinically Deployed Resuscitation Systems

This project is a systematic mapping review of artificial intelligence (AI) and machine learning (ML) applications in cardiac arrest and resuscitation published between January 2023 and December 2025. The overarching aim is to quantify and characterize the rapidly growing number of AI/ML models in this field and to contrast them with the very small subset of systems that have actually been deployed, tested, and evaluated in real-world clinical or training workflows. We will search major biomedical and technical databases and screen records in Rayyan according to predefined inclusion and exclusion criteria. We will focus on AI/ML systems related to the cardiac arrest chain of survival (prehospital dispatch, bystander/tele-CPR, EMS and ED care, in-hospital cardiac arrest, post–cardiac arrest care, and simulation/training). For each eligible study, we will extract information on model type, data sources, target task, evaluation design, and outcomes, and then classify models into three broad categories: (1) offline/proof-of-concept models evaluated only on retrospective data, (2) “near-operational” models that yield simple rules or scores but have not yet been prospectively implemented, and (3) operational AI systems that have been integrated into live clinical or educational workflows and assessed prospectively or in pre–/post–implementation studies. The primary outcome of this work is a structured overview of how many AI/ML models in cardiac arrest exist “on paper” versus how many have reached true operational use, highlighting gaps, recurring methodological limitations, and opportunities for future translational research in digital resuscitation care.