Dr. Ziad Obermeyer calls sudden cardiac death one of medicine's most tragic puzzles: the cure exists—an implantable defibrillator—but doctors have had no reliable way to know who needs one before it's too late. Now, research from UC Berkeley published in Nature offers something that could change that calculus entirely.
Obermeyer and his team trained an artificial intelligence model on more than 440,000 electrocardiograms paired with death certificate data from Sweden. They fed the system scans from healthy people, at-risk patients, and those who later suffered sudden cardiac death, until the algorithm learned to recognize hidden waveform patterns invisible to the human eye. Over multiple years, they tested it on thousands of patient files from both the United States and Taiwan.
The results outperformed the standard clinical test, which measures how much blood the heart pumps with each beat. That test identifies high-risk patients who face a 4.6% annual rate of sudden cardiac death. The AI system flagged a higher-risk group with a 7% annual rate—a difference representing thousands of patients who would have appeared low-risk by current standards. All of this came from EKG images already widely available at medical centers around the world.
"Medical decisions are really hard, and I think that's why AI is so exciting for me," Obermeyer said. "We can not only make better decisions, but also start to understand what's actually going on with these patients before their hearts stop."
More than 300,000 people die from sudden cardiac arrest in the United States each year. The condition strikes without warning, killing both older adults with known risk factors and young athletes with no prior symptoms. While internal defibrillators can save lives, current screening methods miss most at-risk patients—and two-thirds of implants placed based on those methods never actually fire.
The research could lead doctors to better identify who needs a defibrillator before tragedy strikes. It also opens a new window into understanding the physiological mechanism behind the heart's sudden, fatal misfiring. For Obermeyer, an emergency physician who researches the intersection of machine learning and health policy, the appeal is both practical and deeply human.
"If you knew you were one of the people who was going to drop dead, you would go to a cardiologist and you'd get a defibrillator implanted," he said. "The problem is that doctors can't figure out who needs one before it's too late." This study suggests they may soon have a powerful new tool to change that.