The Apple Watch’s heart rate sensors come in handy for knowing how hard your blood is pumping at the gym. But a new, if preliminary, study suggests that the smartwatch also has the potential to spot a much more serious medical condition: an irregular heart rate, known as atrial fibrillation.
The study’s researchers first trained an algorithm to recognize instances of atrial fibrillation in heart rate measurements submitted by people all over the world. The algorithm then accurately detected when a small group of people was experiencing atrial fibrillation in real time, based on data flowing from the Apple Watch on their wrists.
These results are being presented Thursday at the Heart Rhythm Society’s annual conference in Chicago. They have not been published in a scientific journal and need to be validated in larger groups of patients, so don’t expect your Apple Watch to replace a heart check-up any time soon.
Still, cardiology experts say that if the concept is proven to work, the Apple Watch could be a useful tool in helping identify, track, and treat patients with a medical condition that affects an estimated 2.7 million Americans. Atrial fibrillation increases risk of blood clots, stroke, and heart failure — but because it sometimes doesn’t result in symptoms, it can go undetected and untreated, according to the American Heart Association. Catching it early, alerting a doctor, and treating it with blood-thinning medications could save lives.
“This is an important study which gives hope to the notion that someday, it may be possible on a widespread basis for patients or individuals to detect atrial fibrillation with smartwatch technology,” said Hugh Calkins, director of the Cardiac Arrhythmia Service at Johns Hopkins University, who was not involved with the study. He also stressed that it is “no more than an early proof of concept.”
“We were pretty surprised that a device you could go into Best Buy and purchase was capable of this level of accuracy.”
The study, led by researchers at UC San Francisco and the heart rate-analysis startup Cardiogram, illustrates how wearables could help make scientific research and health care more personalized, precise, and effective. Millions of devices sold by companies like Apple, Fitbit, Garmin, and Jawbone are capturing unprecedented quantities of biometric data, from steps to sleep to heart rate, that researchers have never had access to before.
There are already a couple wireless, FDA-cleared devices that atrial fibrillation patients can use to track their heart rate, but because they aren’t meant to be worn all the time, they inevitably miss some data. The Zio Patch sticks to your chest for two weeks, which makes it most useful for monitoring patients right after they’re discharged from the hospital. The AliveCor, a set of electrodes that straps onto the back of your smartphone, produces a heart-rate readout when you press on it.
But people wear Apple Watches all the time. One analyst estimated that Apple sold 6 million units in the last quarter of 2016 alone — nearly 80% of the total smartwatch market. That popularity, along with its high-quality heart-rate sensor, makes it an attractive tool for researchers like Greg Marcus, an atrial fibrillation expert at UCSF and senior author of the study.
Instead of requiring people to buy new gadgets, “the idea here is that we can leverage what people are buying on their own and using anyway,” Marcus told BuzzFeed News.
Marcus is leading an ongoing research project, called the Health e-Heart Study, which aims to study heart disease and health in people scattered throughout the globe. For this study, his team drew from a pool of about 6,400 Apple Watch owners, including 166 people with atrial fibrillation and AliveCor devices. Together, they produced nearly 140 million heart rate measurements and 6,340 AliveCor recordings.
Cardiogram, a startup that’s raised $2 million from Andreessen Horowitz and other investors, collected those data points through its iOS app. Then it used them to train an algorithm to distinguish atrial fibrillation patterns from normal heart rhythms.
To see if it worked, Marcus’ group waited for atrial fibrillation patients to come to UCSF for cardioversions, the procedure for restoring a normal heart rhythm. They gave the patients Apple Watches to wear before, during, and after, and also ran electrocardiograms for a definitive record of their heartbeats. When the algorithm was later applied to the collected heart rate data, it turned out to flag atrial fibrillation episodes with 97% accuracy.
“We were pretty surprised that a device you could go into Best Buy and purchase was capable of this level of accuracy,” said Brandon Ballinger, cofounder of Cardiogram.
It’s an early example of how machine learning can potentially help diagnose people and spot health problems before humans do. But it may be a while before physicians feel totally comfortable relying on an algorithm.
“The downside is there’s a bit of a black-box nature to it,” Marcus said. “By its nature, it’s figuring out the best way to do it and we as investigators may not have as much transparency into the exact algorithm it’s using. That’s going to take some getting used to.”
Other hurdles mean algorithms and wearables are a long way from becoming a mainstay in medical care. When the Apple Watch was being tested on patients in the study, it had to be in workout mode in order to continually capture data. People had to keep still, since the heart rate sensors are potentially less accurate when the wearer is moving around. The algorithm was also tested on a small group of about 50. “This is just a promissory note because they only have a limited number of people they’ve analyzed so far,” said Eric Topol, a cardiologist and geneticist at the Scripps Research Institute, who was not involved with the study.
And while many people have Apple Watches, not all of them are at risk for atrial fibrillation, since the condition is more common in people over age 60. As Calkins put it, “Is your grandmother going to be able to wear this smartwatch to figure out if she has atrial fibrillation or not?”
Still, Topol says that he can see a future where “the Apple Watch and other wearables will get to a point where people will get an alert on their phone or through their devices that says it looks like you may have atrial fibrillation.” He added, “That’s where we’re headed.”