AI Tool Predicts In-Hospital Cardiac Arrest Minutes in Advance

Clinicians must sift through faint signals from routine monitoring while limiting alarm fatigue that strains limited staff. Earlier, reliable prediction could enable timely intervention and better resource allocation. To help address this challenge, researchers have developed an artificial intelligence model that forecasts impending cardiac events minutes before they occur.

Called the Cardiac Autoregressive Model for ECG Language-Modeling (CAMEL), the system was created by clinicians at Penn Medicine (Philadelphia, PA, USA) working with computer scientists at the University of Pennsylvania’s School of Engineering and Applied Science. It treats electrocardiographic (ECG) data as a time-evolving signal whose patterns can be learned rather than as static snapshots. The work is published as a 2026 preprint on arXiv.

CAMEL converts segments of ECG waveforms into a representation that can be interpreted alongside clinical text, including physician notes and laboratory results. By integrating signal and text, the model reasons about how subtle rhythm variations may foreshadow deterioration. Unlike tools limited to 10‑second strips, it is designed to analyze hours of hospital telemetry, expanding the detection window to enable warnings within 10 to 15 minutes before an event.

According to the investigators, the model has shown promising results when analyzing normal sinus rhythm to surface early indicators of high risk for in‑hospital cardiac arrest. The approach targets dangerous arrhythmias such as ventricular fibrillation and ventricular tachycardia. The aim is to identify risk before overt abnormalities appear on standard monitors.

The team plans background testing that processes real‑time patient data without alerting staff, then compares predictions with outcomes to assess performance against current care and minimize alarm burden. They also see potential applications beyond the hospital through consumer wearable devices. These steps are intended to determine whether proactive forecasts can be delivered safely and reliably at the bedside.

“The last thing I want to do is alert nurses and technicians on the floor at 2 a.m. to intervene based on a false signal. Every time we trigger an alarm, we are diverting a finite resource from another patient who may be in need. In a clinical setting, we have to be certain,” said Rajat Deo, cardiologist at the Perelman School of Medicine.

Related Links
Penn Medicine
University of Pennsylvania’s School of Engineering and Applied Science