The notification popped up without warning just as I was syncing the Whoop tracker I was testing: The tracker had a new metric, respiratory rate. Truthfully, I almost ignored it. But the notification used the phrase covid-19, and my eyes widened.
At 7 am, I was bleary-eyed, under-caffeinated, and still adjusting to my new reality of sheltering in place. But, from what I could tell, this niche tracker was telling me that there was a possible correlation between my sleeping respiratory rate and the novel coronavirus dominating headlines. Was this real or marketing bullshit?
The Whoop may have been the first wearable tech company to alert me to the relationship between metrics on a wearable and covid-19, but it certainly wasn’t the last. It seemed wearable companies big and small all had the same idea: That their devices might be useful in the fight against covid-19. It looked promising – a logical progression of how wearables in the past few years have increasingly blurred the boundary between wellness tech and medical devices. Apple Watches and a few other smartwatches can now take electrocardiograms – a test that can measure the electrical activity of your heartbeat – straight from the wrist. But wearables have mostly focused on things like sleep, reproductive health, and heart disease. Detecting infectious diseases is newer territory, and arguably not something these devices were really designed for.
For every heartwarming story of an Apple Watch or Fitbit saving someone’s life, there’s another lurking about health tech peddling false promises and shady marketing passed off as science. With the stakes of covid-19 so high, how much of this is a genuine desire by wearables companies to lend their expertise during an unprecedented crisis? How much is a PR play meant to drum up goodwill at a time when consumers are more careful with their purse strings? And crucially, is a future where your smartwatch warns you before you get sick even possible?
It might sound like science fiction, but there’s reason to believe wearables could be useful in detecting infections. Whether researchers can figure it out in time to make a difference against covid-19 is another story.
My Zoom chat with Michael Snyder was the first interview I’ve ever done where someone other than myself was wearing multiple smartwatches – three watches and an Oura smart ring to be exact. There’s not much reason to unless you’re a tech reviewer or researching potential applications of what these devices can do. Snyder, the director of genomics and personalised medicine at Stanford University School of Medicine, is very much the latter.
Snyder heads one of two coronavirus-related studies Fitbit is involved in, the Wearables Data Study, which looks to study whether wearables can predict covid-19. (The other is the DETECT study by Scripps Research Translational Institute, which aims to improve detection and containment of outbreaks.) He told me there is actually clinical evidence that wearables might be capable of detecting infectious diseases early based on a study Stanford published in 2017 that found these devices could be useful in identifying when you get sick by catching physiological abnormalities.
Michael Snyder, director of genomics and personalised medicine at Stanford, straps on a lot of wearables in his hunt to see if they can help detect infectious diseases early. (Image: Steve Fisch/Stanford)
The exact metrics that researchers are studying vary, depending on what the sensors a participant’s smartwatch or fitness tracker can track. Generally speaking, while wearables makers are providing access to hardware, a population for researchers to study (their user base), and data for those who opt-in, medical researchers are the ones trying to find patterns in the data. There are differences depending on the study, but the researchers I spoke with are examining a wide range of metrics that include heart rate, heart rate variability, sleep, blood oxygen saturation levels, respiratory rate, skin temperature, and even general activity to find a link between covid-19 symptoms and the data tracked by wearable devices. If that seems like a lot of metrics, it’s because there’s a lot we don’t know about the novel coronavirus, and researchers are looking for anything that might stand out.
One significant metric is heart rate. Stanford’s 2017 study and a more recent 2020 study from Scripps Research both note a correlation between elevated resting heart rate and infections. But where the Scripps study focused on whether data from wearables could help detect outbreaks in real-time, the earlier Stanford study posits that it might be possible to detect infections before external symptoms appear.
Snyder was actually a participant in the Stanford study, which involved tracking metrics, including heart rate and SpO2 levels, across a variety of devices. At one point during the course of the study, Snyder visited his brother in rural Massachusetts. Two weeks later, he was flying to Norway through Frankfurt, and on the last leg of the flight, he noticed the blood oxygen levels reported on his fitness tracker seemed abnormally low and his heart rate abnormally high. Snyder later developed a low-grade fever and suspected that Lyme disease might be the culprit – a suspicion that was later confirmed via an antibody test. Because Snyder had given blood before visiting his brother, this was a clear cut case – those samples had tested negative for Lyme.
The experience prompted his team to pore over two years’ worth of Snyder’s data. They discovered that in addition to the Lyme episode, he’d been sick three other times – and one of those times, Snyder had actually been asymptomatic.
During an infection, Snyder explained, your immune system produces something called a C-reactive protein. A serious infection, such as Lyme disease, would result in extremely high levels of this particular protein—something that Snyder noticed in subsequent blood tests, even though at the time he hadn’t actually felt sick. The team then found every time his C-reactive protein levels rose, Snyder’s heart rate and skin temperature were elevated before symptoms appeared. Those results held true for three other people involved in the study. Each time they found elevated heart rates before people fell ill.
“We’re talking tens of millions of people, all with these smartwatches that could be health protectors for infectious diseases like covid-19.”
News reports, social media, and government press conferences have all stressed that asymptomatic carriers can still pass on covid-19 without even knowing. Infected people can also spread the virus in the days before the coughs and fevers manifest. That said, being “asymptomatic” might be a misnomer if invisible metrics like heart rate, respiratory rate, or even skin temperature could possibly reveal infection. If a wearables company could build an algorithm that would alert every person with a tracker or smartwatch that they were potentially sick before external symptoms appeared, the benefits would be numerous. People would know when to self-isolate. Frontline workers would get a heads-up, and health care professionals could potentially have a more accurate picture of infection rates. You could triage care for vulnerable populations more effectively. Most importantly, you might be able to drastically cut the infection rate.
That’s the end goal.
Right now, Stanford’s Wearables Data Study is looking for participants – specifically, people who’ve had a confirmed or suspected case of covid-19, have been exposed to someone who had or may have had covid-19, or those who are at a higher risk of exposure, like essential workers. Once enough people have opted-in via Stanford’s site and their data’s been collected, the second phase involves building a personal dashboard that can tell people when they’re getting sick. And while the original Stanford study’s algorithm was developed using a Basis watch and a few other discontinued devices, this new study aims to be device-agnostic. Fitbits, Apple Watches, and Oura Rings are just some of the wearables included.
“We’re getting a tonne of people enrolling who have a smartwatch and have been ill,” Snyder says. “There’s lots of smartwatch wearers out there. There’s 30 million active users from Fitbit, millions from Apple Watch. We’re talking tens of millions of people, all with these smartwatches that could be health protectors for infectious diseases like covid-19.”
It almost sounds too good to be true, and truthfully, many obstacles stand in the way. Snyder told me they are working at “full blast” around the clock at Stanford, and he believes phase one of the study will be done in a matter of weeks. Still, in America, where many major wearables companies are based, said companies will have to win clearance from the US Food and Drug Administration before rolling out covid-19 predictor features, which is an entirely different process. Take the Withings Move ECG, a watch that offers ECG capability, just like the Apple Watch, which was announced at CES 2019. Despite applying for FDA clearance and receiving a CE Mark, it’s still not available in the US. (It is, however, available in the UK.)
Plus, there’s the simple fact that medicine doesn’t move as quickly as technology – for good reason. It’s true that the first human trial for a covid-19 vaccine trial is already underway, but we’re still several months to years away from a viable, mass-produced vaccine. Researchers may have found a potential relationship between biometric data and covid-19, but that doesn’t mean by this time next year we’ll be checking our wrists to see if we’re infected. While the FDA does have a pilot programme to fast-track software-based medical features, it’s not clear whether covid-19 related software would be included as part of this programme.
“It’ll have to roll out as a research study [first] because these are not medically approved devices,” Snyder concedes. “As soon as you start getting into the medical side it gets very, very regulated.”
You’ve probably never heard of Ava, a reproductive health tracker, unless you or your partner has had trouble conceiving. But in Lichtenstein, the government has already outfitted roughly 2,000 citizens – even those not trying to have a baby – with an Ava bracelet to see if early covid-19 detection is possible.
Among wearables companies, Ava stands out as one of the few that proactively engages in clinical research and publishes studies in peer-reviewed medical journals. The bracelet is a licensed medical product, receiving FDA approval for the US market as a Class One medical device back in 2016. As the global pandemic worsened, Ava founder Lea von Bidder was keen to put the word out that the company was seeking research partners to see if their device and clinical expertise could be put to good use.
Ava’s commitment to scientific research makes it seem like the ideal candidate for rapidly developing a wearable solution to covid-19. The company has experience obtaining regulatory approval, employs a clinical team in addition to a data team, and already had the necessary permits in place to hit the ground running. And, like Stanford, Ava had previously looked into infection – in its case during pregnancy – and already has some data. The result is the COVI-GAPP study, which will try to see if the five parameters measured by Ava – skin temperature, resting pulse rate, perfusion, breathing rate, and heart rate variability – can be used to create an algorithm that will identify covid-19 at an early stage, “even when no typical disease symptoms are present.”
2,000 citizens in Lichtenstein are wearing the Ava tracker in a large study to see if it can detect covid-19 early. (Photo: Victoria Song/Gizmodo)
In terms of goals, this COVI-GAPP study and Stanford’s Wearable Data Study aren’t much different. Both will collect data, look for patterns, and, hopefully build an algorithm that can catch covid-19 before symptoms ever appear. There’s a notable difference, however, in terms of expectations.
On the phone, von Bidder sounds determined, but also grounded when discussing what the study hopes to accomplish. For starters, while Snyder is hoping to develop a workable algorithm – or at least a prototype of one – to help the crisis within the next few months, von Bidder is aiming to take on a potential second wave of covid-19 infections. In fact, the COVI-GAPP study’s first “tangible results” aren’t expected until fall 2020. After that, the study will move onto a second phase, which would include the entire population of Lichtenstein – and this is assuming that the researchers find anything to begin with.
“We don’t even know yet how these parameters interact with covid-19,” von Bidder told me. “Let’s start right there. This whole thing might never work. I think there’s enough reason to believe it might work, but then we still don’t know if it works fast enough. My assumption is that we will get there, otherwise, I wouldn’t invest all this time and money in trying to figure it out, but it’s not that simple.”
“You could tell people that they have covid when they don’t and tell other people that they don’t have it but they will.”
Even if researchers find a relationship between the metrics Ava can track and covid-19, there’s still a question of sensitivity and specificity. It’s not terribly helpful if a wearable can detect that you’ve got a fever when you already know you have one. There’s also the issue of how to notify someone of a potentially traumatic diagnosis.
“Do we tell you, ‘You definitely have covid-19,’ or do we tell you, ‘Hey, you might have covid-19, go get tested,’” von Bidder said. “That’s a highly medical application, because you’re going to really interfere, and you need to be very sure about what you’re doing. You could tell people that they have covid when they don’t and tell other people that they don’t have it but they will.”
Snyder was also frank about potential limitations. Skin temperature, while one logical metric to determine whether a person might have covid-19, isn’t something that all trackers can measure. Plus, many people infected with the novel coronavirus may never have a fever. The accuracy of skin temperature readings also depends on how tightly or loosely a person wears their device. Lastly, it’s not yet clear whether any algorithm could distinguish between types of viral infections – as in, could the device tell the difference between the flu, the common cold, and covid-19?
Then there’s the question of whether wearables will be able to even detect infection fast enough.
The thing about data is that it can be very noisy.
“My one concern is that even among symptomatic individuals, the highest degree of viral shedding and transmission is actually in the pre-symptomatic period,” Thomas Tsai, assistant professor in the Department of Health Policy and Management at Harvard T.H. Chan School of Public Health, told Gizmodo over the phone. “So you know, the wearables are unlikely to be able to presage the period when they’re most infectious.”
Tsai went on to note that it is a good thing that this sort of research is happening. The more that we can learn about covid-19, the better public health officials can tackle the pandemic effectively. The thing about data is that it can be very noisy. Your heart rate can be elevated because you’re watching a scary movie, and maybe your sleeping respiratory rate isn’t ideal because your cat screams every night at 4 am.
Going back to the 2017 Stanford study, Tsai is a bit sceptical about the relationship between C-reactive proteins and heart rate. A quick refresher: the study identified high C-reactive protein levels with elevated heart rate as a sign of illness. However, commercially available wearables may not even have the appropriate sensors to consistently and accurately measure C-reactive protein levels. Even if that correlation is confirmed in the current ongoing study, it’s not guaranteed to have an impact on how doctors treat covid-19. “C-reactive protein is a very nonspecific marker for inflammation,” Tsai explains. “My worry is when we look backward and try to find patterns between heart rate and some markers that we see, retrospectively, patterns in the noise. They may not actually change the clinical course.”
But even if the signal in the noise is found, the red tape hasn’t gone anywhere. The COVI-GAPP study is a bit of an outlier. Not only is Ava experienced at navigating regulatory hurdles, but Lichtenstein’s government was also quick, proactive, and ready to reach into its pockets to fund this type of medical wearables research. Lichtenstein also has a smaller, contained population.
By contrast, the US is a big, sprawling nation that was slow and ill-prepared for preventing the spread of covid-19. Back in February, the FDA did issue an Emergency Use Authorisation (EUA) for medical devices in light of the pandemic. Whether the EUA has been effective or an easy process to navigate remains a mystery. Yet it has led to folks getting creative in finding new ways to use technology to help with shortages of life-saving equipment, such as ventilators and personal protective equipment.
Fitbit is also working with researchers to find ways to detect covid-19 early. (Photo: Victoria Song/Gizmodo)
It could also theoretically be used by wearable companies to roll out detection algorithms. I asked the FDA to clarify whether it would consider fast-tracking approvals for an algorithm to early detect covid-19 using software for existing wearables. In response, the FDA sent back its digital health policy for covid-19. According to the policy, most apps and software systems for “public health surveillance and communication” don’t require clearance or approval, as they’re not medical devices.
The policy also states the FDA “does not intend to enforce requirements under the Food, Drug and Cosmetic Act at this time for certain lower risk” software – but it’s unclear whether early detection algorithms fall under this category. They most likely do not. For higher-risk software – any sort of diagnostic feature, for example – the policy refers back to the EUA process. That means even with expedited processes in place, vetting this sort of experimental tech will likely take longer than anyone would like. That said, clear FDA guidance is absolutely necessary from a safety standpoint.
To the average consumer, the lines between educational features and diagnostic devices might sometimes blur. Take Whoop, the sleep tracker I was testing when I saw a covid-19 notification. The FDA’s digital health policy explains why Whoop can quickly roll out a respiratory rate metric: While the Whoop app itself explains the potential relationship between your breathing rate and covid-19, the company provides the metric for passive monitoring or journalling by the user. Whoop isn’t going to tell you if you have covid-19, or what to do if your respiratory rate drastically changes overnight.
Emily Capodilupo, Whoop’s vice president of data science and research, said on a recent podcast that the tracker “is not a medical device” and that the goal isn’t to “diagnose any disease or condition, especially not covid-19 or the flu.”
The idea is to make users aware of their baselines, with the knowledge that a sudden, drastic change could signal...something. That said, Whoop is also participating in a study with Cleveland Clinic in the US and Central Queensland University in Australia to see if there is a potential link between respiratory rate and covid-19. It bears repeating: No wearable devices on the market can diagnose covid-19.
But the research part isn’t exactly the problem. It’s what comes after.
And there’s no indication that Whoop (or Fitbit, Oura, Apple, or other big-name wearable makers) are advertising the ability to do so. But that hasn’t stopped people from panic-buying anything that might give them a better chance of preventing (or surviving) illness. According to the American Lung Association, people are also snapping up pulse oximeters – at the expense of hospitals and people who actually need them – because they might early detect or quantify shortness of breath, a well-known symptom of covid-19.
“Fixating on pulse oximeter readings may provide a false sense of security,” Dr Albert Rizzo, chief medical officer at the American Lung Association, said in a statement. “In some cases, they do catch lung problems before you struggle with shortness of breath. However, it is also possible for the device to show healthy oxygen saturation levels even when experiencing trouble breathing, which may lead some individuals to delay seeking urgently needed care.”
It’s not absurd to worry that consumers might look at all these studies and develop unrealistic expectations that might cause more harm than intended, or leave healthcare professionals cleaning up after Big Tech’s mess.
“I wouldn’t tell people to buy an Ava bracelet for this [purpose] tomorrow,” Ava’s von Bidder says. “We clearly tell people to not do it.”
In the event an accurate, thoughtfully implemented, government-cleared, device-agnostic algorithm is built and rolled out to fitness trackers and smartwatches the world over, privacy is a big asterisk. After all, tech companies train their algorithms with lots and lots of valuable health data.
I spoke with Whoop, Fitbit, Stanford, and Ava about how they’re approaching data privacy with regard to studying covid-19. The answers across the board boiled down as follows: All ethical standards are being adhered to, all data is aggregated, none of it is being shared with advertisers, and every study is completely opt-in. Whoop’s Capodilupo told me that Whoop only began looking into covid-19 once users who tested positive reached out and volunteered their data. That said, Snyder mentioned that one big draw of collaborating with Fitbit is the company’s expertise in managing lots of data in real-time. When I asked Fitbit about that, the company reiterated its commitment to privacy and said that no data collected is personally identifiable. And, again, everyone has to opt-in when they enrol in Stanford’s study.
But the research part isn’t exactly the problem. It’s what comes after. Who owns any potential algorithm that gets made? How do you convince wearables companies to implement tech they don’t own on their devices (especially if it’s not clear how it’ll be monetised)? Who gets to see how often you get sick, and who owns that data? Does medical privacy legislation protect that information if it’s not being directly shared with healthcare providers, but instead sent to users as a notification? If the government expedites approvals for wearable diagnostic features during the pandemic, does that set a new precedent once this is all over? While Stanford is working on a device-agnostic algorithm, it’s less clear whether a potential algorithm out of the COVI-GAPP study will work universally.
Whoop claims it began looking into the relationship between data it gathers and covid-19 after users reached out. (Photo: Victoria Song/Gizmodo)
Wearable makers are collaborating with researchers and letting them take the lead now. Yet after the research is done those makers could create their own proprietary covid-19 detection algorithms. Ideally, like Apple and Google’s forthcoming contact tracing tech, any covid-19 detection algorithm for wearables would work on any available device for quick, widespread adoption, not just a specific brand. In the event of proprietary algorithms, we should all be a little wary.
Fitbit and Withings are two wearables companies that have already released trends of how their users have been impacted by covid-19. All of the data is aggregated, but you get a glimpse at just how much these companies know about us. Withings, for instance, published a blog that includes a breakdown by country of how much weight people have gained (or haven’t) since lockdown policies began, what types of workouts people are doing, and how much sleep people are getting. Fitbit published several blogs, complete with graphs, maps, and charts about physical activity, sleep patterns, and whether people around the world are complying with lockdown orders based on step count.
Fitbit uses its aggregated data to produce flashy graphs. (Image: Fitbit)
This data, while valuable, isn’t quite on the same level as the data in your medical records. But early infection predictions? That’s more of a grey area, and it’s unclear whether people will trade their health data for what could be a life-saving diagnosis.
“The concern over privacy is warranted. It’s one of the leading reasons why people will not get a fitness tracker or a smartwatch,” Ramon Llamas, a research director at IDC who covers wearables technology, told Gizmodo. “But if we add covid-19 to the mix and we come up with a workable solution, some people may change their tune. It should be on these vendors to get it right.”
Llamas said all the research points to a simple fact: Convenience trumps everything. While there are potential downsides to most new technology, a killer app or function makes the privacy trade-offs easier to bear – especially ones that could save your life. For example, the very first Apple Watch was characterised as an unnecessary luxury, but millions have since warmed up to the idea of owning one – or gifting one – based on the premise it could improve your health. If a smartwatch could reliably tell you when you’re sick – or about to be – it might make giving up a degree of privacy more palatable for some people. It’s also possible that early detection capabilities might not be convenient enough to quell privacy concerns. Instead of sparking excitement or relief, covid-19 diagnostic tools might instead conjure dystopian images of Minority Report and Black Mirror. It’s impossible to judge either way until the technology is fully developed and ready for mass market.
“Some people might think about saying, alright, what do I have to give up in order to have this little bit of convenience and peace of mind?” Llamas said. “It’s a matter yet to be sorted out, but if anything it’s going to be enough to give people pause.”
So, is it possible that smartwatches could help mitigate the spread of covid-19? The answer is unsatisfying. It’s not all marketing bullshit – there is credible evidence that wearables might be able to help us better understand, and in the best-case scenario, early detect infectious diseases. But there are also a lot of unanswered questions and unknown variables that make it hard to say with certainty that wearables will be helpful right now. Even if a breakthrough is found, it’s not evident that’ll change how public health officials approach this pandemic, Tsai said. What’s needed is more testing – which is what health officials have been repeating ad nauseam since this all began.
The impulse to at least try and solve the problems caused by covid-19 with wearable technology is understandable. But it’s also important to remember that whatever algorithms, software, or devices eventually hit the market, they won’t magically restore the world back to the way it was. After all, there’s no putting the novel coronavirus back in a box. Scientists say the pandemic will likely stick around until a vaccine is developed or widespread natural immunity is here. There’s a lot of potential for a wearables-based solution – and a lot of regulatory hurdles that exist to protect the public from any cut corners. If one day soon your wrist buzzes and your smartwatch asks if you’d like to participate in a covid-19 study, it might be worth considering volunteering your data for science. Maybe doing so will help make a future pandemic less of a nightmare.
Featured image: Jim Cooke (Gizmodo)