AI-Enhanced Risk Stratification for Out-of-Hospital Cardiac Arrest in General Populations

Article Impact Level: HIGH
Data Quality: STRONG
Summary of  JACC: Advances https://doi.org/10.1016/j.jacadv.2026.102787
Dr. Surbhi Sharma et al.

Points

Researchers developed AI models using data from 1.7 million patients to identify individuals at high risk for sudden cardiac arrest, a condition with a staggering 90 percent mortality rate.
The integrated EHR-EKG model successfully predicted 153 out of 228 cardiac arrest cases in a real-world cohort, improving risk detection from one in 1,000 to one in 100 patients.
The study found that AI-enhanced 12-lead EKG analysis alone provides a low-cost and highly effective method for stratifying cardiac risk in communities that may lack extensive electronic health records.
Beyond traditional cardiovascular disease markers, the AI identified significant modifiable risk factors including substance use, electrolyte imbalances, and dangerous medication interactions that often precede a sudden cardiac emergency.
Clinical application of these models may allow physicians to review high-risk patient histories and implement preventative surveillance, though further research is required to standardize specific medical interventions.

Summary

This study evaluated the efficacy of artificial intelligence (AI) models in predicting out-of-hospital cardiac arrest (OHCA) within the general population, a condition characterized by a 90% mortality rate. Utilizing a dataset of 1.7 million patients from a major U.S. healthcare system, researchers developed three distinct models: an EKG-only model, an EHR-only model weighing 156 clinical features, and a combined EHR-EKG integrated model. The models were developed using a training cohort of 993 OHCA cases and 5,479 age- and sex-matched controls, then validated against a testing cohort of 463 cases and 2,979 controls to ensure accurate risk discrimination.

In a prospective real-world cohort of 39,911 individuals who received an EKG in 2021, the combined EHR–EKG model demonstrated high predictive accuracy, correctly identifying 153 out of 228 individuals who subsequently experienced an OHCA over a two-year follow-up period. The implementation of these AI algorithms enriched risk stratification significantly, narrowing the identification of at-risk patients from 1 in 1,000 to 1 in 100. Notably, the AI-enhanced 12-lead EKG analysis alone maintained strong predictive utility, suggesting a cost-effective pathway for global risk screening in diverse clinical settings.

The investigation further identified non-traditional modifiable risk factors, including electrolyte imbalances, substance use, and adverse medication interactions, which could serve as immediate clinical red flags. While the models provide a feasible framework for identifying high-risk individuals before an emergency occurs, the authors note that the study was limited to a single healthcare system and warrants further investigation into standardized clinical interventions. These findings suggest that AI-enabled screening could transition cardiac arrest from a largely unpredictable event to a manageable clinical risk through targeted surveillance and medication review.

Link to the article: https://www.sciencedirect.com/science/article/pii/S2772963X26002085

References

Sharma , S., & A. Brody , J. (n.d.). Artificial intelligence-enhanced electrocardiography and health records to predict cardiac arrest. JACC: Advances. Retrieved https://doi.org/10.1016/j.jacadv.2026.102787