Particle.news
Download on the App Store

AI Reads Medical Records and EKGs to Flag Sudden Cardiac Arrest Risk

The study signals a possible low-cost screening path using AI-read EKGs pending external checks before clinical use.

Overview

  • Researchers reported in JACC: Advances in May 2026 that models trained on about 1.7 million patients used electronic health records and 12‑lead EKGs to predict out‑of‑hospital cardiac arrest risk.
  • In a real‑world group of 39,911 people who had EKGs in 2021 and were tracked for two years, the combined model later identified 153 of 228 people who then had a cardiac arrest.
  • The tools concentrated risk from roughly 1 in 1,000 people to about 1 in 100, giving clinicians a smaller group to evaluate and monitor more closely.
  • The system also highlighted risks beyond classic heart disease, including electrolyte problems, substance use, and medication interactions that clinicians can sometimes correct.
  • An EKG‑only model performed only modestly worse than the full version, but the authors say broader validation across other health systems and clear follow‑up steps are needed before routine use.