Early in the pandemic, I predicted that Fitbit would put its ubiquitous fitness trackers to work to predict whether users might be getting Covid-19. It turns out I was right.
On May 27, Fitbit began enrolling users in a study that aimed to correlate Fitbit sensor data with Covid-19 infections. The company recruited 30,529 users for the study, which ran through September 2020. Fitbit published its results in the journal npj Digital Medicine in November.
Fitbit found that it could predict up to 43% of Covid-19 cases the day after patients experienced symptoms, using sensor data alone. They could even predict nearly a quarter of cases one day before users experienced obvious physical symptoms. The company also found that the severity of symptoms and sensor data correlated with the need for hospitalization, allowing Fitbit to predict which patients would need medical support to make it through the illness.
Fitbit found that it could predict up to 43% of Covid-19 cases the day after patients experienced symptoms, using sensor data alone.
But here’s the best part about the study: It showed that Fitbit could predict Covid-19 infections using data that is already available to users of the company’s Fitbit Premium service. If you have a modern Fitbit and subscribe to Premium, you don’t need some fancy, proprietary algorithm to spot a potential infection. You can open your Fitbit app right now, eyeball your data, and determine whether you might be getting sick. The technique presumably works with illnesses other than Covid-19, too — a quick look at your Fitbit data could potentially show you if you’re getting a cold, flu, or another disease.
Before we get started, let’s make one thing clear: If you have any symptoms of Covid-19 or suspect you could be infected, you should stop reading this, self-quarantine, get tested, and call your doctor. You should never rely on a consumer wearable alone to diagnose a disease. But as Fitbit showed in their study, data from the company’s devices might reveal Covid-19 infections, which are otherwise asymptomatic. If you learn to understand your Fitbit data, you could potentially spot an asymptomatic infection that would otherwise go undetected. And if you have a confirmed infection, the data might help determine how severe your disease course will be. Even if you shouldn’t use Fitbit data to diagnose the disease, there’s still plenty of value in analyzing and understanding it.
To check your own data, open the Fitbit app, pull up the health metrics tab, and switch to the month view. This view shows trends in several health-related metrics for the last 30 days. In their study, Fitbit found that at least three health metrics were useful in predicting Covid-19 infections: respiration rate (called breathing rate in the app), heart rate (called resting heart rate), and heart rate variability.
Start by looking at your breathing rate. Fitbit’s study revealed that breathing rate tends to increase when people come down with a disease like Covid-19 (other studies have shown that the metric correlates with fevers, a common Covid-19 symptom). You should see a chart of your own rate for the last 30 days in the app. Take a look at the chart, as well as the grayed-out area above and below the graph of your breathing rate. This grayed-out area represents your personal range.
To calculate your personal range, Fitbit says that it looks at your last 30 days of breathing rate data and calculates a mean and standard deviation. According to the app, the baseline is set at three standard deviations above and below your own personal average. Think back to high school statistics. Most physiological data follows a normal distribution, and 99.7% of data fall within three standard deviations of the mean. Under normal circumstances, your Breathing Rate should only go outside your Personal Range 0.3% of the time. If it goes outside this range more than 0.3% of the time, that’s an indication that something abnormal is probably going on.
Look back at your own breathing rate data. Did your breathing rate go above your personal range in the last 30 days? If it did, you’ll see a blue dot on that day. Make a note of any days where you went above your range.
Next, let’s add another data point. Take a look at your heart rate variability (HRV). Fitbit found in their study that HRV tended to decrease when people got Covid-19, and other studies cited by Fitbit show that a lowered HRV may predict an increased risk of dying from the disease. Again, look at your own data, making note of any days where your HRV went below your normal range.
Finally, take a look at resting heart rate (RHR). Fitbit’s study showed that RHR tends to increase during a Covid-19 infection, noting that “A 1 °C rise in body temperature can increase heart rate by 8.5 beats per minute (bpm) on average.” In other words, an elevated heart rate might predict fever or other symptoms of an illness. Take a look at your own data and note days that your RHR went above your personal range.
Now, go back and take a look at the three data points together. Were there any days that your breathing rate and RHR went above your personal range, and your HRV went below it? Those are days when you might have been ill — either with Covid-19 or potentially with something else. If you’ve seen a worrisome trend in the three data points over the last several days, you should get tested right now, even if you have no obvious symptoms.
If you have a Fitbit Sense, Ionic, or Versa, you can add a few more metrics. Skin temperature and oxygen saturation weren’t officially analyzed in the Fitbit study, but Fitbit notes that many Covid-19 positive study participants had fevers, and the company says in the Fitbit app that its skin temperature metric may indicate the “onset of fever.” Other studies have shown that oxygen saturation correlates with Covid-19 illness as well.
If you have a compatible Fitbit, look for days when your oxygen saturation went below your baseline and days when your skin temperature went above it. Note, though, that Fitbit calculates your personal range for skin temperature by looking at two standard deviations instead of three. That means there’s likely to be more days when your skin temperature goes outside your personal range, even if you weren’t actually sick. Take this metric with a grain of salt.
Finally, consider looking at your overall activity levels, as measured by your step count. An independent study by Stanford University showed that step count could be predictive of Covid-19 infection since sick people tend to rest more and move around less. If you see a dramatic, unexplained drop on a particular day, that could be a worrying sign.
I looked at my own data for the last 30 days, and thankfully I saw no concerning trends. I was within my personal range for all the metrics Fitbit studied. My skin temperature was above the normal range on one day, but because Fitbit calculates the range differently for this metric, I wasn’t worried. To double-check my results, I wrote down all my data points, put them into a Google spreadsheet, and calculated the mean and standard deviation myself. I confirmed that Fitbit was indeed calculating my personal range correctly.
If you have a modern Fitbit and subscribe to Premium, you don’t need some fancy, proprietary algorithm to spot a potential infection.
Going forward, I’m planning to monitor my metrics every few days — or more often if I start to feel ill for any reason. If I see them trending in a concerning direction, I can notify my doctor and get tested, even if I feel okay. If you have a Fitbit, you might consider doing the same.
While you can analyze your Fitbit data yourself, there are also ways to get help. Stanford University is currently enrolling wearable users (including those who wear Fitbits, the Apple Watch, or Garmin devices) in a Covid-19 study of their own. Participating requires signing consent forms, meeting eligibility requirements, providing personal information, and giving Stanford access to your data. If you don’t mind doing that, Stanford will let you download an app that automatically analyzes your data every one to two days and raises an alert if you’re outside your baseline values. Coviddeep, an app from a Princeton University spin-off company, performs a similar function. Its creators claim that it is 90% effective at detecting infections.
There are some challenges to predicting Covid-19 with Fitbit data, though. Fitbits can’t record accurate data if you’re moving around, so the company resorts to measuring your metrics when you’re in a deep sleep. That means you have to wear your Fitbit to bed to get health metrics data, and it’s delayed by one day. If you want to take instantaneous readings (or monitor your data without paying for Fitbit Premium), consider a dedicated HRV app such as Welltory (which has also studied Covid-19), as well as investing in a pulse oximeter, which measures your heart rate and oxygen level.
It’s also important to note that Google recently acquired Fitbit. In part to placate the European Union, Google agreed not to use Fitbit data for ad targeting. But it could likely use your data — and any Covid-19 insights it may contain — for other internal purposes. If you find the idea of Google knowing your Covid-19 status concerning, consider choosing a different wearable.
If you’re already a Fitbit devotee, though, you’ll likely appreciate the ability to analyze your data for signs of illness. Especially with Google in its corner, expect Fitbit to roll out more tools that allow you to monitor for specific health conditions. The FDA already cleared Fitbit to use the Fitbit Sense watch to monitor for AFib, a potentially deadly heart condition. A formal Fitbit Covid-19 detector — or more likely, an app that monitors for better-studied conditions like sleep apnea — might be just around the corner.
More broadly, Fitbit’s health tracking capabilities are a positive step towards personalized, data-driven medicine. Everyone’s heart rate and breathing patterns are different. That makes it hard to diagnose conditions like Covid-19 using these data points alone. If you breathe more frequently than the average person, an algorithm that compares your breathing rate to an average value might flag you as sick when you’re actually fine. Because Fitbit knows your personal baseline — rather than an average for the whole population — it can spot concerning personal deviations while reducing false positives.
I found that learning to monitor my own Fitbit data felt empowering, too. Eyeballing my own data made me feel more connected to my health. It feels like a positive way to keep tabs on your body and a tool for noticing changes that might otherwise remain invisible. So if you have a Fitbit, fire up the app, take a look at your own data, and see if you learn anything helpful or interesting about your own health.