Communications of the ACM
Wearable Data Predicted COVID Infections
During the pandemic, wearable fitness trackers helped to provide actionable early warning signals that helped quell wider outbreaks.
Credit: Syneos Health
As the COVID pandemic raged around the world in 2020 and 2021, methods such as early testing of suspected cases and social isolation were the best weapons healthcare professionals could muster against the disease prior to the mass introduction of vaccines. For some of the most vigilant of those professionals, wearable fitness trackers helped to provide actionable early warning signals that helped quell wider outbreaks.
Such was the case at Florida State University (FSU) in Tallahassee. The university's athletes, who compete at the highest level of U.S. collegiate sports, are equipped with WHOOP wristbands that regularly monitor their biomarkers, such as heartbeat, temperature, and respiration rate. At the height of the pandemic, FSU athletic department medical staff suspected they might be able to tell early on if an athlete had become infected with the COVID virus. Elisa Angeles, associate director of athletics research at FSU's Institute of Sports Sciences and Medicine (ISSM), noticed subtle changes in some athletes' WHOOP data and played a hunch they might be in the pre-symptomatic phase of the disease.
Angeles and colleagues in the ISSM recently published a study in Sports Health: A Multidisciplinary Approach demonstrating that hunch was indeed accurate; of 113 female student-athletes in soccer, golf, softball, indoor volleyball, beach volleyball, and tennis who were monitored for early signals of the disease, 33 tested positive. The athletes were equipped with WHOOP bands from August 2020 to May 2021. Angeles told Communications the monitoring, the first time FSU staff had tried to tie WHOOP data to a clinical condition, was a vital element in trying to get back to some sort of routine.
WHOOP, Angeles said, had published observations from its own data that COVID could be identified before symptoms persist. "And at that point in time we were looking at return-to-play protocols," Angeles said; "return to life in general, actually, talking about how we would safely return to campus."
The FSU researchers found two distinct time periods in which an underlying biosignal indicated disease was setting in: respiratory rate increased three days prior to a positive COVID test, while resting heart rate and heart rate variability increased one day before a subject tested positive. One of the study's co-authors, David Ormsbee, noted at least one instance in which Angeles acted quickly on the data and may have headed off a wider outbreak.
"One day, after reviewing recovery data and noticing a clear outlier and suspected potential COVID infection, Angeles immediately consulted with sports medicine staff to urgently send the student-athlete for testing and quarantine," Ormsbee said in an FSU news release. "At that point in time, that student-athlete just arrived at a team function and watch party for the Olympics. Sure enough, she tested positive for COVID. However, because of the swift identification and action, no additional team members were contact-traced or quarantined, and no additional team or staff members contracted COVID via that event."
Angeles said the FSU staff has expanded interpreting athletes' wearables data beyond COVID to try to get a jump on other conditions such as influenza, and to try to maximize overall vitality.
Nor is FSU alone in discovering the subtle variations in biosignals that hint at much larger ramifications. From forehead-worn electroencephalograph headbands that detect early signs of Alzheimer's disease to the atrial fibrillation and Parkinson's disease symptoms detection features of the Apple Watch (which have already received regulatory approval as medical devices), the convergence of consumer wearable data and clinical results is accelerating at a rapid pace. The ecosystem surrounding the technical advances, however, is still struggling to take shape.
Black boxes, heterogeneity are obstacles.
Dr. Carsten Skarke, a translational medicine researcher at the University of Pennsylvania, worked with colleagues on a study that appeared in Nature Communications to find that no fewer than 73 diseases or conditions could be predicted to some degree by subtle differences in the amplitude of body temperature changes as measured at the wrist over a specified period of time.
Using a week's worth of such data from 92,000 participants in the United Kingdom's U.K. Biobank database, Skarke and his colleagues found the 73 conditions were significantly associated with decreased temperature rhythm, meaning that participants with a smaller day-night difference in their wrist temperature readings were showing increased rates for the future onset of these diseases. Among the largest associations, nonalcoholic fatty liver disease (NAFLD) emerged with a 91% increased risk, followed by Type 2 diabetes with 69%, renal failure with 25%, hypertension with 23%t, and pneumonia with 22%.
"These findings indicate the potential to marry emerging technology with health monitoring in a powerful new way," Skarke said.
The emerging technology of temperature sensors in many consumer trackers, including several Fitbit models and the Apple Watch 8, 9, and Ultra series, would seem to be a natural conduit for tighter integration of at-home measurements and clinical guidance around what Skarke's team found. "In the future, this information may be leveraged with their care teams as a digital biomarker, to understand their risk to develop certain diseases and to navigate treatment or preventative care options," Skarke said.
However, those possibilities come with a big caveat, Skarke told Communications. While he used the open source device by Axivity, which is the U.K. Biobank's endorsed device, translating those results to the polyglot landscape of manufacturers and intellectual property issues are obstacles to widespread clinical adoption of findings such as his.
"There is still hesitancy among clinicians to incorporate these data for various reasons," he said. "You start with the heterogeneity of devices which collect data, so it's unclear how that factors into the clinical interpretation of any data stream. Another big challenge, understandably, is the black boxes around devices where you have a lot of intellectual property. It caters to the business model and to some extent to help the individuals understand more about sleep and activity, and so on, but that is an impediment for these data to be transparent."
Another significant obstacle to quick clinical adoption of digital wearable data, connected to the fidelity and transparency of experimental data, is the traditionally cautious culture around new therapeutics, Skarke said.
"If you want to make anything clinically actionable, the best reference points we have are new molecular entities, like new prescription drugs. The FDA drug development process is the benchmark."
However, the prolonged trial procedures around drug development simply are not compatible with the speed with which new devices are introduced or new features are added to existing ones, he said. So, while the conceptual rigor around vetting the devices might have to resemble that going into drug development, timelines need to be altered significantly.
Pandemic-era investments become manifest.
While the specifics of integrating data from wearables and physiological conditions are still amorphous, the realization by both technology vendors and pioneering scientists of its importance is manifesting itself more often, according to both Skarke and Kit Cangardel, associate principal for health technology at consultancy Blue Matter.
"I have absolutely seen more activity in this space," Cangardel said, "and I feel like this tendency has been happening a lot more than the last six months. That said, a lot of the investment that has been happening in the space that is now paying off was triggered by the pandemic."
The need to invest in more capable devices may have kicked into higher gear during the pandemic, but Cangardel said the industry's "big bang" was probably the U.S. Food and Drug Administration's approval of the Apple Watch's atrial fibrillation detection mechanism in 2018. That clearance, she said, made every serious wearable device vendor realize the stakes had been raised significantly; the recent spate of newly published research such as that from Skarke and Angeles is making it clear the data from these devices can offer vital insight. The next frontier, Cangardel said, is to translate the findings into some sort of workable model that shows improved health and lower costs at a population health level.
Just a matter of time?
In a recent blog series covering the convergence of consumer wearables equipped with biosensors and regulated medical devices, Cangardel and co-author Darya Volgina stated some compelling economics for greater acceptance of everyday personal health data: "It's been estimated that only 12% of U.S. residents are metabolically healthy, indicating a significant need for any solution which can support metabolic health," they wrote. "These consumers are highly motivated to monitor their health, and many will already be using consumer wearables to track wellness activities."
Medical device giant Abbott is on the forefront of this convergence trend. The company is marketing Lingo, a consumer-grade analog of its regulated Freestyle Libre continuous glucose monitor (CGM). The product is in the vanguard of devices for the percentage of the population in the "great middle," those who may have subtle early indicators of disease that are not yet being closely followed. Already, those with chronic diseases are being offered round-the-clock monitoring with sensors such as CGMs, and data from elite athletes such as those at FSU are closely watched by professional medical staffs.
Cangardel said shifts in the consumption of medical service toward remote services and a more equal "partnership" between clinicians and their patients make the eventual inclusion of wearable data, and the potential to heed that data to minimize health risks, a near-certainty.
"Overall, what we're seeing is an overall consumerization of health," she said. "There is less trust in your physician being the holder of the record of your health and much more of the nexus of power is going to you as an individual, and these tools support that."
Gregory Goth is an Oakville, CT-based writer who specializes in science and technology.
No entries found