Andrey Popov
smart watch
Using data from smartwatches, a new algorithm reads heart rate as a proxy for bodily stress, potentially alerting wearers they’re falling ill before they have symptoms.
Researchers led by Michael Snyder, PhD, professor and chair of genetics, have enrolled thousands of participants in a study that employs the algorithm to look for extended periods during which heart rate is higher than normal — a telltale sign that something may be amiss.
But figuring out what may be wrong takes a little sleuthing. During the study, many stressors triggered an alert. Some folks received them while traveling; some while running a marathon; others after over-indulging at the bar.
But the most exciting finding, Snyder said, was that the algorithm was able detect 80% of confirmed COVID-19 cases before or when participants were symptomatic.
“The idea is for people to eventually use this information to decide whether they need to get a COVID-19 test or self-isolate,” Snyder said. “We’re not there yet — we still need to test this in clinical trials — but that’s the ultimate goal.”
The algorithm can’t differentiate between someone who’s knocked back a few too many and someone who’s ill with a virus. Although it pinged users who had COVID-19, more refining is needed before people can depend on their smartwatches to warn them of an impending infection with SARS-CoV-2 or other viruses.
Stress detection
During the study, which ran for about eight months in 2020 and 2021, 2,155 participants donned a smartwatch, which tracked “stress events” via heart rate. When notified of a stress event, through an alert paired with an app on their phone, participants recorded what they were doing. To trigger an alert, their heart rate needed to be elevated for more than a few hours, so a quick jog around the block or a sudden loud noise didn’t set it off.
“What’s great about this is people can contextualize their alerts,” Snyder said. “If you’re traveling via airline and you receive an alert, you know that air travel is likely the culprit.”
If, however, you’re sitting on the couch with a cup of chamomile tea and you receive an alert, that may be a sign that something else — an infection, perhaps — is brewing. Snyder hopes wearers will be able to discern when an alert means they should consider getting tested.
Of 84 people who were diagnosed with COVID-19 during the study, the algorithm flagged 67. Most alerts fell into other categories, such as travel, eating a large meal, menstruation, mental stress, intoxication or non-COVID-19 infections. The algorithm also flagged a period of stress after many participants received a COVID-19 vaccine, reflecting the uptick in immune response prompted by the shot.
Refining the algorithm
As Snyder and the team recruit more participants into the study, they’re planning to hone the specificity of the alerts by adding data — including step count, sleep patterns and body temperature — in the hope that data patterns can correspond to and flag distinct stress events. In addition, the researchers plan to run a clinical trial to determine if the alerts can reliably detect a COVID-19 infection and be used to guide medical choices.
Source: The Stanford University School of Medicine consistently ranks among the nation’s top medical schools, integrating research, medical education, patient care and community service. For more news about the school, please visit http://med.stanford.edu/school.html. The medical school is part of Stanford Medicine, which includes Stanford Health Care and Stanford Children’s Health. For information about all three, please visit http://med.stanford.edu.
A paper detailing the study was published online in Nature Medicine on Nov. 29. Snyder, the Stanford W. Ascherman, MD, FACS, Professor of Genetics, and Amir Bahmani, PhD, lecturer and director of Stanford’s Deep Data Research Computing Center, are co-senior authors. Arash Alavi, PhD, research and development lead in Stanford’s Deep Data Research Computing Center; research scientist Meng Wang, PhD; and postdoctoral scholars Gireesh Bogu, PhD, Ekanath Srihari Rangan, PhD, and Andrew Brooks, PhD, share lead authorship. The alert system was built using MyPHD, a scalable, secure platform for health data.
Other Stanford co-authors of the study are software engineers Rajat Bhasin, Shrinivas Panchamukhi and Qiwen Wang; clinical research coordinators Emily Higgs and Alessandra Celli; postdoctoral scholars Tejaswini Mishra, PhD, and Ahmed Metwally, PhD; user experience designer Kexin Cha; undergraduate researchers Erika Hunting and Peter Knowles; software developer Amir Alavi, PhD; research scientists Alexander Honkala and Diego Celis; research intern Tagore Aditya; Benjamin Rolnik, director of the Stanford Healthcare Innovation Lab; research associate Orit Dagan-Rosenfeld; clinical research coordinator Arshdeep Chauhan; research analyst Jessi Li; assistant clinical research coordinator Caroline Bejikian; bioinformatics researcher Vandhana Krishnan; and web developer Lettie McGuire.
A researcher from Case Western University also contributed to the work.
This study was funded by the National Institutes of Health (grants 1R01NR02010501, 1S10OD023452-01 and UL1 TR001085), Amazon Web Services Diagnostic Development Initiative, Google for Education and the Schmidt Futures program.