As coronavirus cases surged worldwide in March, Fitbit observed a global decline in physical activity among its 30 million activity tracker and smartwatch users. As governments implemented shelter-in-place orders, the company recorded a nearly 20 percent decline in weekly step counts in New York and San Francisco. The pandemic may be hampering regular fitness routines, but now Fitbit is exploring whether its wearable devices could help researchers predict if users have a coronavirus infection before they even notice any symptoms.
“We think wearables have a really strong value proposition to add, both to detect, as well as help to track and contain infectious diseases like COVID-19,” says Amy McDonough, senior vice president and general manager, Fitbit Health Solutions.
Fitbit this week launched a consortium with the Stanford University School of Medicine and the Scripps Translational Research Institute, where teams of researchers have already started enrolling wearable device users in two studies looking to train algorithms to find early warning signs of COVID-19 or other infections hiding within their activity data.
Michael Snyder, a professor of genetics and director of the Center for Genomics and Personalized Medicine at Stanford, says a typical wearable device “will measure 250,000 data points per day,” offering a treasure trove of health data for researchers, including heart rate, sleep patterns and skin temperature.
Snyder is a wearables evangelist and uses multiple devices daily, including three different smartwatches on his wrists, a ring with sensors on one finger, a continuous glucose monitor, an environmental exposure monitor and a radiation monitor. All of the devices feed data into his iPhone, which he also uses to measure his steps.
It was in a 2017 study published in PLOS Biology that Snyder made the connection between an increase in his resting heart rate and skin temperature before he started feeling symptoms from what would later turn out to be Lyme disease. “That’s the key,” he says, “We think your smartwatch tells you when you’re getting ill before you know it.”
This is particularly relevant to COVID-19, since there has been widespread asymptomatic transmission of the virus. Fitbit is donating 1,000 devices each to Stanford and Scripps, which will be given to frontline workers at higher risk of testing positive, such as healthcare providers and their family members, grocery clerks and pharmacy staff.
This isn’t the first time that Fitbit, whose pending $2.1 billion acquisition by Google is currently under regulatory review, has used its devices to support healthcare research. Fitbit has collaborated with various research institutions for more than a decade and there are more than 900 published studies using Fitbit data and devices, McDonough says.
Both the Stanford study and the Scripps study, dubbed “DETECT,” involve crowdsourcing participants (in addition to the frontline workers), who use different wearable devices, including Fitbit, Apple, Garmin and others. They also hope people will give permission to access their electronic medical records, which would help validate the data. The first step is training algorithms to establish ranges for participants’ baselines readings, so they can then recognize when there is a significant change from normal levels.
Snyder and his team at Stanford are looking to recruit as many people as possible, who can pinpoint when they’ve been sick over the last few months. They will retrospectively analyze their data and use it to train the algorithms to identify the signs associated with infection, whether it ends up being COVID-19, the flu, or a cold. Snyder doesn’t yet know if different infectious diseases will display different warning signatures within the wearables data. The second part of the study, which still needs Institutional Review Board approval, would send a signal back to the user if the algorithm notices certain sustained signals, such as elevated heart rate.
Doctors have been measuring heart rates for centuries, but wearables have changed the game when it comes to frequency and accessibility outside of a medical setting, says Dr. Steven Steinhubl, a cardiologist and director of digital medicine at the Scripps Research Translation Institute. “It’s like a new vital sign,” he says. “Even though pulse is an old vital sign, that day-to-day variation in an individual’s pulse rate is brand new, and there’s a lot we have to learn about it.”
Steinhubl and the Scripps team are also collecting participants’ real-time data and hoping to match it against health records. After establishing a baseline for heart rate and other signs, the algorithm will send the users a message if it notices any significant changes.
“Noticing when somebody is mildly sick, or maybe doesn’t even feel symptomatic, that would really be the goal,” says Steinhubl. This is especially true in light of COVID-19, which can be asymptomatic in a large part of the population, but still contagious. “One of the benefits of this algorithm is we want to be able to give people a personalized warning that says, ‘Hey, it’s a good idea for you to stay home today.’”
Steinhubl was the co-author of a study published in The Lancet earlier this year, which found increases in resting heart rate to be an indication of flu-like illness based on anonymized data from more than 47,000 Fitbit users.
The quality of both of these new studies will ultimately depend on how many people they are able to enroll. Fitbit is helping spread the word by encouraging users to participate on a COVID-19 information tab within the company’s app, alongside other resources, such as the latest guidelines from the World Health Organization.
Steinhubl says they have enrolled around 6,000 people, but are looking to get several hundred thousand, if not closer to a million. And the study will continue regardless of what ends up happening with COVID-19. “Viral illnesses are not going to go away and there are going to be future pandemics. They’re going to be future bad flu seasons,” he says. “And we think this could be a valuable tool.”