Enlisting smartwatches, fitness trackers in fight against COVID-19
Smartwatch and fitness tracker data may serve as a useful complement to symptom screening in identifying patients positive for the novel coronavirus disease (COVID-19), according to a recent study.
“Our results show that individual changes in physiological measures captured by most smartwatches and activity trackers are able to significantly improve the distinction between symptomatic individuals with and without a diagnosis of COVID-19 beyond symptoms alone,” the researchers said. “Although encouraging, these results are based on a relatively small sample of participants.”
In a cohort of 30,529 participants with available sensor data, 3,811 had symptoms indicative of a COVID-19 infection. Ultimately, 54 patients were diagnosed with COVID-19, while 279 symptomatic participants underwent testing but were negative for the infection. Groups were statistically comparable in terms of baseline demographics and type of device used. [Nat Med 2020;doi:10.1038/s41591-020-1123-x]
The research team developed a smartphone app that connected to the participants’ personal trackers, collecting data such as resting heart rate (RHR), sleep duration, and activity level. Changes in these metrics from baseline (–27 to –7 days before symptom onset) to the symptomatic period (0–7 days after onset) were collected and compared between diagnosis groups.
The mean change in sleep was significantly greater in those who tested positive for the infection, growing by a mean of 57±92 minutes. In comparison, sleep lengthened by only 4±68 minutes in COVID-19-negative comparators (p<0.01).
Similarly, changes in activity patterns differed significantly according to COVID-19 status. Those who were diagnosed with the infection saw a drop of –3,533±4,418 steps from the baseline to the symptomatic period, as opposed to only –208±3,086 steps in negative controls (p<0.01).
In contrast, no significant between-group difference was seen in the change in RHR (1.51±4.83 vs 0.61±3.68 bpm; p=0.330. However, 30.3 percent of COVD-19 patients showed a spike in RHR of up to two standard deviations above the baseline average.
To examine the potential of each factor as a predictor of COVID-19 infection, the researchers transformed the device measurements into corresponding uniform statistical metrics (SleepMetric, ActivityMetric, and RHRMetric). A composite of these three metrics was called the SensorMetric.
An overall variable, called SymptomMetric, was also adapted to capture the symptoms and common risk factors of COVID-19 patients, including coughs, fatigue, abnormalities in taste and smell, age, and sex.
Both SleepMetric and ActivityMetric were good predictors of COVID-19 with areas under the curve (AUCs) of 0.68 and 0.69, respectively. RHR was less so and had a value of 0.52. SensorMetric further improved the discriminatory value of the wearable devices, returning an AUC of 0.72.
However, combining SensorMetric with SymptomMetric yielded the best results, providing an AUC of 0.80 (95 percent confidence interval, 0.73–0.86).
“These results suggest that sensor data can incrementally improve symptom-only-based models to differentiate between COVID-19-positive and COVID-19-negative symptomatic individuals, with the potential to enhance our ability to identify a cluster before more spread occurs,” the researchers said.
“Such a passive monitoring strategy may be complementary to virus testing, which is generally a one-off, or infrequent, sampling assay,” they added.