AI study maps new uses across cardiac arrest care

A scoping review of 114 studies finds artificial intelligence is expanding across cardiac arrest care, from early warning and emergency response to resuscitation support and recovery planning. The authors say the tools could improve time-critical decisions, but broader clinical testing is still needed before widespread adoption.

Why it matters: - Cardiac arrest is a race against time, and AI could help clinicians spot risk earlier, guide resuscitation, and improve care after survival. - The review suggests AI is moving beyond prediction and into multiple points in the cardiac arrest pathway. - Better tools could matter most in out-of-hospital cardiac arrest, where survival remains especially limited.

What happened: - Researchers from Sun Yat-sen University and Sun Yat-sen Memorial Hospital published a scoping review in the 2026 issue of World Journal of Emergency Medicine. - The review examined AI applications across in-hospital and out-of-hospital cardiac arrest care. - The study covered prediction, resuscitation support, prognosis, large language models, emergency call handling, wearable detection, rhythm identification, education and extracorporeal cardiopulmonary resuscitation candidate identification. - The paper is available as the original study.

The details: - The review followed PRISMA guidelines and searched PubMed, Embase, the Cochrane Library and Web of Science from database inception to June 10, 2025. - After screening 2,108 records, the authors included 114 studies and assessed 92 AI models. - Most studies focused on pre-arrest prediction, especially in-hospital cardiac arrest. - A multilayer perceptron model reached the highest reported area under the receiver operating characteristic curve for in-hospital prediction at 0.998. - For out-of-hospital cardiac arrest prediction, extreme gradient boosting and random forest models reached a reported area under the receiver operating characteristic curve of 0.950. - For CPR-related decision support, convolutional neural network models reached a best reported area under the receiver operating characteristic curve of 0.990. - Prognostic models were widely studied after out-of-hospital cardiac arrest, with one multilayer perceptron model reaching 0.976. - The review also identified newer work on GPT-style large language models, emergency call recognition, wearable-based detection and AI-assisted education. - The authors noted the review was supported by a grant from the National Natural Science Foundation of China, grant 82372207.

Between the lines: - The review shows AI research is spreading across the full chain of survival, not just one stage of care. - High model scores are promising, but many studies remain retrospective and may not translate cleanly into frontline emergency care. - Data imbalance, limited external validation, infrastructure gaps, privacy concerns and algorithmic bias remain major barriers. - The strongest near-term value may come from systems that support clinicians rather than replace decisions.

What's next: - The authors want multicenter clinical testing, not just stronger algorithms. - Future work should focus on explainable models, prospective trials and equitable deployment in both high-resource and resource-limited settings. - Hospitals and emergency systems may next test AI for risk detection, dispatch support, AED localization, CPR feedback, rehabilitation planning and clinical trial design.

The bottom line: - AI is starting to touch every stage of cardiac arrest care, but real-world validation will decide whether those tools improve survival and recovery.