The Race to Predict and Prevent Sudden Cardiac Arrest

Can artificial intelligence and consumer wearable devices help identify who is at risk of sudden cardiac arrest years before it happens-or in the hours leading up to it?

Supported by more than $5 million in new funding from the National Institutes of Health (NIH) and the American Heart Association (AHA), Cedars-Sinai Health Sciences University investigators Sumeet Chugh, MD, and Kyndaron Reinier, PhD, MPH, are building on two decades of collaborative research to improve long-term risk prediction and spot warning signs shortly before sudden cardiac arrest strikes.

"We're cautiously optimistic that incorporating AI tools and wearables data will improve prevention of cardiac arrest and outcomes for this lethal condition," said Chugh, who directs the Center for Cardiac Arrest Prevention and the Artificial Intelligence in Medicine Research Center at Cedars-Sinai.

Out-of-hospital sudden cardiac arrest affects about 1,000 people in the U.S. each day. Unlike a heart attack-a circulation problem caused by a blocked artery-cardiac arrest is an electrical failure in which the heart suddenly develops a chaotic rhythm and stops pumping blood. Without immediate CPR and defibrillation, death follows within minutes.

"Because many cases occur without warning, and cardiac arrest is fatal in about 90% of cases, improving risk identification and prevention is critical," said Reinier, associate director of Epidemiology in the Center for Cardiac Arrest Prevention.

To mark Sudden Cardiac Arrest Awareness Month in October, the Cedars-Sinai Newsroom spoke with Chugh and Reinier about the latest advances in cardiac arrest prediction and prevention.

Why is sudden cardiac arrest so difficult to predict?

Reinier: For up to half of patients, cardiac arrest is the first sign of underlying heart disease. Historically, clinicians relied on left ventricular ejection fraction, or LVEF-a measure of heart pumping function-to gauge risk. But LVEF has become less predictive as treatments for reduced heart function have improved. Also, many people who suffer cardiac arrest never had severely reduced LVEF. So, better risk predictors are essential.

Chugh: Implantable defibrillators-which act like personal paramedics, delivering a shock within seconds when they detect a dangerous heart rhythm-were once effective at protecting people with weak heart function. In the past, implanting 100 defibrillators based on LVEF saved 10 to 15 lives a year. Today, that number is closer to three to five. LVEF simply isn't catching enough high-risk patients anymore.

What has earlier research revealed about prediction and prevention?

Chugh: Over two decades of community-based research in Oregon and California allowed us to study cardiac arrest in approximately 2 million people. With help from local paramedics, police, firefighters and the coroner, we identified nearly every cardiac arrest that occurred. We combined the data collected by paramedics with the individuals' electronic health record to develop a unique learning health system. This resource helped Dr. Reinier and the Center for Cardiac Arrest Prevention team develop a ventricular fibrillation risk score, called VFRisk, that outperformed LVEF in identifying at-risk people in those populations.

Paramedics also collected blood samples, which revealed genetic and protein patterns unique to cardiac arrest patients.

Reinier: VFRisk uses 13 variables, including electrocardiogram [ECG] measures, heart structure imaging from echocardiograms, and clinical history, such as cardiovascular risk factors and other medical conditions. We also evaluated warning symptoms-like chest pain, shortness of breath and palpitations-alongside blood proteins. We've proved that combining dynamic risk (how a person's heart health changes over time) and static risk (a single snapshot in time) can provide a clearer, earlier warning.

How does your new research build on these findings?

Chugh: The new grants allow us to test new tools in a real-world health system setting. For our NIH-funded study, we're using de-identified electronic health records from the O.S.C.A.R. [Observational Study of Cardiac Arrest Risk] cohort-about 400,000 Los Angeles County residents treated at Cedars-Sinai-to evaluate how well VFRisk predicts events over time and determine whether adding dynamic risk markers improves risk prediction.

How can AI help?

Chugh: Conventional tools can analyze numbers and text well, but it is much more difficult to integrate digital ECG signals, moving echocardiogram images, biological samples and health records. AI can unify complex information from disparate sources and identify subtle patterns that would be difficult to detect by examining only a few individual markers. We will compare AI-based methods with conventional approaches to determine whether they improve prediction.

Where do wearable devices fit into the picture?

Reinier: The AHA-funded study uses wearables to track near-term risk. Over three years, we will monitor 600 high-risk O.S.C.A.R. patients identified by their VFRisk and provide them with a fitness tracker that logs heart rate, heart-rate variability, oxygen saturation and activity levels. By establishing personal baselines, we can analyze rolling 48-hour windows for subtle deviations before an event occurs. Paired with clinical history and AI analysis, wearable data could signal when an event is imminent.

What is the bottom line for the future of cardiac arrest prevention?

Reinier: By broadening risk markers, we hope to pinpoint high-risk individuals earlier while reassuring low-risk patients. Eventually, if wearable data and symptoms indicate rising short-term risk, we could provide clearer guidance on when to seek urgent care.

In the immediate term, if you witness a cardiac arrest, fast action is essential: Call 911, start CPR and use an automated external defibrillator [AED] if available.

Cedars-Sinai Health Sciences University is advancing groundbreaking research and educating future leaders in medicine, biomedical sciences and allied health sciences. Learn moreabout the university.