Harpreet Rai, the CEO of smart ring company Oura, often tells a story about a March 2020 Facebook post. An Oura ring user posted that the device said that his overall health score had dropped below his normal level, which prompted him to get tested for COVID-19 — and the test ended up being positive. The company heard from other users, too.

The anecdotal reports encouraged Oura to partner with research teams to try to figure out how well the ring could predict who might be sick with COVID-19. Their studies were part of a wave of interest over the past year in wearable devices as illness detectors. Now, flush with data, researchers and wearable companies are looking toward their next steps.

Research done over the past year showed that it’s probably possible to flag when someone is sick. But differentiating which illness someone might have will be much harder. Experts think it might eventually be possible, but in the near future, illness detection programs might look more like warning lights: they could tell a user that they might be getting sick, but just not with what.

“It’s just like the warning light for your car — take it into the mechanic, we don’t know what’s wrong, but something looks off,” Rai says. “I think that’s where the industry is heading.”

Illness signatures

Even before the pandemic, researchers were checking wearables’ data to see if they could find telltale signatures that might predict illnesses. One study published in early 2020 found that data from Fitbits could predict state-level trends in flu-like illnesses, for example. Other research found that wearable devices could detect signs of Lyme disease. A research team at Mount Sinai Health System in New York used wearables to predict disease flare-ups in patients with inflammatory bowel diseases (IBD) like Crohn’s.

When COVID-19 hit, many of those research teams adjusted their focus. “We decided to shift some of our emphasis to how we can evaluate and identify COVID-19 infections, using the same techniques and technology,” says Robert Hirten, a gastroenterologist at Mount Sinai who worked on wearables and IBD.

Hirten’s research showed that Apple Watches could detect changes in the heart rate variability of healthcare workers up to seven days before they were diagnosed with COVID-19. Heart rate variability, which tracks the time between heartbeats, is a good proxy for how the nervous system is working, he says. “Often it seems to be very telling of something going on in the body, even before people realize something is happening.”

Other types of data were also useful. A Stanford University study found that heart rate, daily steps, and time asleep as measured by smartwatches changed in a small group of users before they developed symptoms of COVID-19. The first report from the TemPredict study at the University of California, San Francisco found that the Oura ring could detect increases in body temperature before wearers developed COVID-19 symptoms. Through a partnership with New York-based Northwell Health, Fitbit showed that its devices tracked changes in heart rate and breathing rate in the days before someone started feeling sick.

The research is ongoing. Groups at UCSF and the West Virginia University Rockefeller Neuroscience Institute continue to run studies with Oura ring, and Fitbit is still working on research with Northwell Health. Fitbit is also part of projects out of the Scripps Research Translational Institute and the Stanford Medicine Healthcare Innovation Lab. Apple launched a study on respiratory disease prediction and Apple Watch in April.

The big wearable companies have a good reason to pursue this line of research; the studies done so far are promising. “People are really learning better ways to identify and predict conditions,” Hirten says. “This has really taken the field of wearable technologies forward significantly.”

A warning light

That doesn’t mean that smartwatches will have apps that tell wearers when they have COVID-19. There’s a big difference between being able to detect a general change in the body that could be an illness and detecting a specific illness, says Jennifer Radin, an epidemiologist with the digital medicine division at Scripps Research Translational Institute who’s run studies on wearables and COVID-19.

“If your heart rate goes up compared to your normal rate, it can be caused by many other things besides just a viral infection. It could just be that you had too many drinks last night,” she says.

None of the metrics researchers pull from wearables are direct measures of a respiratory illness. “They’re all just markers of if the body is feeling good or not,” Hirten says. The systems are very different from the features on wearable devices that can detect atrial fibrillation, a type of abnormal heart rhythm. In that case, the wearable is directly measuring the marker — heart rate — that changes in a clear, unique way when someone experiences atrial fibrillation.

“Something like COVID-19 is much more complicated. You’re trying to look for alternative markers in the body that track viral illness,” he says. “But it’s a lot more complicated because those markers also track other things going on in the body.”

Fitbit saw the overlap with other illnesses on one of its illness prediction projects this year, which asked participants to take flu tests. There were similar signals in the data when people had the flu as when they had COVID-19, said Conor Heneghan, a director of research at Fitbit. “My instinct is that it’s going to be hard to reliably distinguish between them,” he says.

The task gets even harder as the levels of coronavirus circulating in a community drops. The studies done on wearables and COVID-19 all ran over the past year when the coronavirus was the primary thing making people sick. Flu season hardly existed. There was a pretty good chance that, if someone had a certain signature on their wearable that tracked with respiratory illness, the respiratory illness could be COVID-19. As the disease becomes rare, that gets harder. “It’s not as specific anymore as the prevalence drops,” Hirten says.

But tools to alert people when they might be getting sick are still useful, even if they can’t say what someone is sick with. Heneghan says that’s the likeliest path forward for Fitbit. “It’ll be a general, hey, something has changed in your physiological signs, you might want to consider that you could be getting ill,” he says. “That’s probably ok from our point of view.”

Oura’s Rai says that the Oura ring serves that function already. The ring gives users a readiness score, which incorporates metrics like sleep quality, heart rate, and body temperature. If it detects that someone’s body temperature is elevated, it gives people the option to pause activity goals and enter a rest mode. That’s the product’s warning light, Rai says. He thinks it’ll be the main approach for the next few years.

“These devices cannot diagnose or treat, but what they can do is say that something’s off in your baseline,” he says.

Adding those types of features to more wearable devices will take good communication with users, says Radin. “We don’t want to scare people,” she says. It’d be a problem if people who wore smartwatches thought they had COVID-19 any time their data changed. But the devices could explain to users that there is a range of reasons someone might be seeing a change, including illness, she says. “It’s just a heads up that something is out of your normal range, and it could be something to keep an eye on.”