A new U.S. study finds that combining AI-enhanced electrocardiography (ECG) with electronic health records (EHR) can predict out-of-hospital cardiac arrest risk, potentially enabling earlier intervention. The case-control study developed multimodal AI models using ECG waveforms and EHR data, demonstrating superior performance over standalone ECG or EHR-based tools, with an area under the receiver operating characteristic curve of 0.83.
Researchers tested the AI-driven approach in a real-world healthcare setting over a two-year period. The combined model identified about two-thirds of individuals who later experienced cardiac arrest. Among high-risk patients identified by the AI-enhanced ECG, the 2-year cumulative incidence of out-of-hospital cardiac arrest reached 2.4%, compared to 0.5% for low-risk individuals.
These findings suggest AI-enhanced ECG screening could be scalable for population-level risk assessment, especially for those without previously diagnosed structural heart disease or severely reduced cardiac function. The research highlights the growing role of artificial intelligence in cardiology, using routinely collected health data to improve prevention strategies and reduce sudden cardiac death.