Results from a large U.S. study suggest that information collected from fitness trackers could help identify who has Covid-19 more accurately than tracking symptoms alone, something the researchers hope will help control the spread of the virus.
“We found that if you identify each individual’s normal values when they’re not sick, then you can identify these subtle changes that indicates something is happening to their health,” explained study investigator and senior researcher Jennifer Radin, an epidemiologist based at Scripps Research Translational Institute in California.
In an ideal world, anyone who suspects they may have contracted an infectious virus such as SARS-CoV-2 should be able to get a test to let them know whether they need to isolate themselves from other people or not.
However, unfortunately life is not that simple. Tests for Covid-19 are expensive, in great demand and even in wealthy Western countries they can be hard to access. Finding a way to more accurately predict who has been infected without relying on lab tests is therefore of high importance.
The problem is that using symptoms alone it can be hard to differentiate Covid-19 from other respiratory infections like colds and flu, particularly in individuals who have mild-to-moderate symptoms.
One in 5 Americans now wears some sort of smartwatch or fitness tracker on a regular basis, measuring a range of factors such as heart rate, sleep, and daily activity levels. Radin and colleagues at the Scripps Institute decided to test whether this data could help predict cases of Covid-19.
In March this year they launched a study called ‘Detect’. The team used an app called MyDataHelps to allow people to opt in to securely share their device data.
Between the end of March and the beginning of June this year the team recruited 30,529 participants into the study. Of these, 3811 said they had possible Covid-19 symptoms during this time. Only 333 of these individuals were tested and 54 had a confirmed infection.
When the researchers compared the accuracy of predicting infection based on symptoms alone versus symptoms plus fitness tracker data, the latter combination was more accurate.
“We found that when you add wearable data to the self-reported symptom data that significantly improved our ability to differentiate who in this study had Covid-19 versus who was Covid-19 negative,” said Radin.
Using a statistical test, where a value of 1 = 100% correct, the researchers found that predicting infection using symptoms and fitness tracker data scored 0.80, whereas symptoms alone only scored 0.71.
“We found that individuals who had Covid-19 typically slept a lot more than those who had symptoms, but were Covid-19 negative. Also, these individuals were a lot less active,” noted Radin.
“Resting heart rate was less of a differentiator in this study, but we did find that 30% of the individuals who had Covid-19 had a resting heart rate that went up to two standard deviations above their normal rate. So, it may be something that changes for some individuals, but not everybody.”
The Detect study is continuing to recruit participants across the U.S. with the aim of reaching a total of at least 100,000 participants. It’s open to anyone over the age of 18 who has some form of activity tracker. Consistency is more important than the type of device, so someone who has changed devices frequently would not be eligible, but most types of devices (regardless of whether they measure heart rate or not) can be linked up with the app.
Radin and team hope that the infection predictions will become even more accurate with data from more participants. In addition to encouraging members of the public to participate, they are also encouraging more ‘at risk’ groups such as frontline workers to take part in the study.
“We’re looking at seeing if we can identify wearable changes, both in the pre-symptomatic phase and also for asymptomatic individuals, since those are the really hard people to identify and likely still spread virus to others.”
The researchers are also keen to do comparisons with other viral respiratory illnesses, such as flu, to see if it’s possible to differentiate between different infections using fitness tracker data — something they have not been able to investigate extensively to date.
Other similar studies have been set up in different countries. Most notably the German app Corona-Datenspende, developed by the Robert Koch Institute, has already enrolled 532,254 participants. It is also still recruiting and analyzing data, but has so far had a focus on detecting fever using smartphone and activity tracker data.
As not everyone with Covid-19 actually develops a fever, having a broader detection method that includes multiple factors has merit, says Radin.
“When we have limited testing resources, it’s a way to potentially better screen people than just taking their temperature at the door before they go to work and asking if they feel sick, which isn’t very useful since only 30% of people who go to the hospital with Covid-19 present with fever.”
“What’s exciting here is that we now have a validated digital signal for COVID-19. The next step is to use this to prevent emerging outbreaks from spreading,” says Eric Topol, director and founder of the Scripps Research Translational Institute.
“Roughly 100 million Americans already have a wearable tracker or smartwatch and can help us; all we need is a tiny fraction of them—just 1 percent or 2 percent—to use the app.”