Abnormal Heartbeat? There's an App for That -- And it Could Be 97 Percent Accurate

Save ArticleSave Article

Failed to save article

Please try again

This article is more than 4 years old.
With nearly 6 million people in the U.S. predicted to develop AF by 2050, the app could prevent large numbers of strokes and even death. (iStock)

A mobile app was able to detect irregular heart rhythms with 97 percent accuracy, a finding by UC San Francisco researchers that could lead to life-saving preventative screenings.

One in four adults over the age of 40 is at risk for a heart condition known as atrial fibrillation (AF). It's one of the leading causes of stroke and often goes undetected until it's too late.

The study, which relied on an app specifically designed for the Apple Watch, is reportedly the first to use a smartwatch to detect abnormal heart rhythms.

“By identifying candidates for appropriate anti-coagulation treatment, we might ultimately leverage common wearable devices to reduce ... complications, even death," said Gregory Marcus, director of clinical research in the UCSF Division of Cardiology and lead author of the study.

AF occurs when electrical impulses in the the upper chambers of the heart become erratic, causing the atrium's walls to quiver as blood passes through, potentially triggering blood clots.


Researchers say the findings, published in JAMA Cardiology, could lead to the development of more effective screenings. And with nearly 6 million people in the U.S. estimated to develop AF by 2050, the app could prevent large numbers of strokes and death.

With a 97 percent accuracy rate, the app performed better than a competing FDA-approved product sold as Kardia Band.

The UCSF-tested app, designed by Cardiogram, relies on deep neural networks, a type of sophisticated machine-learning algorithm that provides a continuous stream of data. More than 139 million measurements of heart rates and step counts were collected from 9,750 participants.  

Researchers cited certain limitations to their study, including a number of participants who failed to link their Cardiogram accounts. In addition, researchers focused on those with known AF and did not look into the app's ability to identify undiagnosed conditions.