By Paul Nicolaus
February 9, 2021 | Because coronavirus can spread before people become aware of their symptoms—or when they experience no symptoms whatsoever—some are turning to wearables as an early detection tool. By picking up on changes in certain metrics, smartwatches may be able to help curb transmission by providing indications of infection and nudging individuals to self-isolate or get tested sooner.
Researchers at California-based Evidation Health along with collaborators at the Bill and Melinda Gates Foundation, the University of California Santa Barbara, and the University of Washington are one group exploring this potential. By analyzing commercial wearable activity tracker data as well as self-reported symptoms, they compared people with flu and COVID-19 to learn more.
Their findings appeared Dec. 12 in the journal Patterns (DOI: 10.1016/j.patter.2020.100188) and revealed that physical activity (like step count) and physiological signs (like resting heart rate) change around the time of symptom onset.
Using wearables as high frequency/low-sensitivity tests may help shorten the detection timeframe and enable people to isolate themselves a day or two earlier. This is important, according to senior author Luca Foschini, Evidation Health’s co-founder and chief data scientist, considering infectivity is highest around the time symptoms first appear.
“Since the COVID-19 virus is so novel, we are constantly figuring out ways technologies can help stop the spread and aid in early detection,” noted Sheneen Lalani, a New York City-based internal medicine doctor working the frontlines with COVID patients (who was not involved in this study).
Wearable activity trackers could be used to help detect general indicators of COVID, such as elevated heart rate, elevated respiratory rate, fever, or decreased oxygenation, she told Diagnostics World, and the recently published research by Foschini and colleagues “does a great job pointing out how the symptoms of COVID last longer than flu.”
One notable challenge faced by efforts looking to use tracking devices, however, is that the illness manifests itself differently in different individuals. Not everyone experiences these or other common symptoms, Lalani added, and it could be difficult to distinguish symptoms like elevated temperature or heart rate from other viruses or general activity.
Impact of Illness on Step Count and Heart Rate
Evidation Health researchers and colleagues analyzed roughly 7,000 patients who self-reported flu or COVID-19 diagnosis. The individuals were broken down into three primary cohorts, Foschini explained. One group consisted of people who contracted the flu before the pandemic began. The other two were made up of people who got the flu during the pandemic and people who got COVID during the pandemic.
To arrive at their findings, they compared a cohort of 230 self-reported COVID-19 cases to 426 non-COVID-19 flu cases that took place during the same time frame and 6,270 cases of flu that took place during the 2019-2020 flu season—before the COVID-19 pandemic.
For a subset of participants, “we have Fitbit data as well,” Foschini explained, “specifically heart rate, step count, and sleep.” The three main cohorts were filtered to include participants in each group with wearable sensor data and a low fraction of missing data, which amounted to 41 COVID-19 patients, 85 non-COVID-19 flu patients, and 1,226 pre-COVID-19 flu patients.
In short, their findings revealed that flu and COVID are both similar and different.
The differences were revealed mainly through the symptom self-reporting. While there is an overlap between flu and COVID symptoms, the data showed that people with COVID tend to have a much higher prevalence of shortness of breath, chest pressure, and difficulty breathing. And even for symptoms that are common between the two, like fatigue, it was apparent that COVID symptoms lasted longer and peaked later.
COVID-19 illnesses lasted a median of 12 days, compared with 9 days for non-COVID-19 flu and 7 days for pre-pandemic flu illnesses. Symptoms peaked 2 to 3 days following illness onset in both of the flu groups (except for shortness of breath for the non-COVID-19 flu cohort). COVID-19 symptoms, on the other hand, peaked 3 to 7 days following illness onset, with most of the symptoms peaking 4 to 5 days after onset.
The similarities were revealed through the lens of wearable device data as the research group noticed a spike in resting heart rate as well as changes in daily step counts that were “of similar magnitudes for both the flu and COVID-19 cohorts.” This is a blurry lens, however, considering a smartwatch lives on the wrist and cannot see a person’s overall holistic health, Foschini told Diagnostics World. Even so, the wearable data did offer several insights.
The percentage of those with COVID-19 who had a highly elevated resting heart rate was higher near the onset of illness (between days −2 and 2), when infectivity is at its height and isolation interventions could have the greatest impact, compared to days −10 to −5 when spread is not as likely.
Furthermore, the percentage of the COVID-19 group with highly elevated resting heart rate around the time of illness onset was higher than that of the non-COVID-19 cohort but did not differ from the pre-pandemic cohort, suggesting that an elevated resting heart rate alone may not be a specific marker of infection.
The research also revealed the impact of illness on daily step count, finding that the decline in daily steps was more extensive and prolonged for those with COVID-19 than those with non-COVID-19 flu and pre-pandemic flu. This may be explained by more strict self-imposed quarantine measures taken after a COVID-19 diagnosis, the researchers acknowledged, but the reduction could also reflect the more prolonged illness durations in the COVID-19 patients.
For those with COVID, the number of steps did not return to normal even after three or four weeks in some cases, which hinted at the existence of “long COVID” cases, where symptoms last for many weeks or even months.
In terms of study limitations, the researchers noted that the cohorts examined are not representative of the US population. African Americans, males, and older individuals are all underrepresented, which may limit the generalizability of the findings.
Because only those who sought medical care were included in the analysis, this could have skewed symptom presentation in the direction of more severe issues, like difficulty breathing. Another caveat is that the analysis considered both self-reported symptoms and self-reported diagnostic test confirmation.
In addition, the surveys used to gather symptoms between the earlier pre-pandemic flu group and the COVID-19 and non-COVID flu groups used symptom sets that were not identical. As a result, it wasn’t always possible to make direct comparisons of symptom prevalence.
Asymptomatic Sensitivity of Wearables
The battle against COVID-19 has been hindered by viral infections that present similarly, making detection and monitoring difficult. For wearable-based COVID-19 applications to become a reality, Foschini and colleagues say it will be crucial to consider the flu and other similar respiratory illnesses as a possible source of false positives.
Comparing and contrasting COVID-19 and the flu is significant for screening purposes considering current practices tend to check for more general symptoms. Many screening tests performed at building entrances are temperature-based, for instance, but plenty of people don’t necessarily develop a fever right away. Plus, conditions other than COVID-19 can cause a fever.
Foschini said a spike in resting heart rate seems to be a more sensitive indicator of COVID-19. And for those who use activity trackers, the process could be as simple as asking permission to share that information for screening purposes, similar to taking a temperature reading. More research is needed, however, to better understand the feasibility of such applications.
For the time being, it is still unclear how well wearables can detect COVID-19 in people that have mild or no symptoms. The reason why it’s hard to figure that out, he said, is that it requires a setting in which you have wearable devices on people and ground truth as to whether they have COVID-19 or not. And you need to be able to cross these two datasets, he said.
Because an understanding of the ability of person-generated health data (PGHD) to detect pre-symptomatic and asymptomatic spread cannot be gleaned from data that is based solely on symptoms, additional studies will need to combine PGHD with direct measures of infectivity to shed light on the asymptomatic sensitivity of wearable devices.
This type of research is currently underway with support from the Biomedical Advanced Research and Development Authority (BARDA) and the Bill & Melinda Gates Foundation. The partnership with Evidation Health involves a pilot study that monitors healthcare workers and first responders using PGHD from wearable devices in the hopes of developing an early warning algorithm that could enable real-time interventions.
The study is attempting to enroll 1,000 people who would wear an activity tracking device and take a PCR test once a week. It would make it possible to see the transition from negative to positive (using PCR testing) and learn whether wearable devices can catch that transition, too. The study uses Garmin wearables and has a sub arm with a more sophisticated Empatica device that has more sensors.
The aim is to learn more about people that do get COVID-19 but do not experience symptoms. Will the wearable device still show a jump in resting heart rate? It may be smaller than the jump seen in symptomatic individuals, but perhaps it is still something that could be used as a trigger to find more information or to lead those people to the next test if deployed at scale. Within a couple of months, they hope to have an answer to this question, Foschini said.
Advantages, Drawbacks of Wearables
One of the biggest drawbacks of turning to wearables, he pointed out, pertains to representativeness issues. Roughly 1 in 5 Americans regularly wear a smartwatch or wearable fitness tracker, according to a 2019 Pew Research Center survey, but their usage varies when socioeconomic factors are taken into account.
“I think to find a solution around wearables, you need to have a very specific use case in which, for example, you’re following a cohort of students and trying to get them back to school or a cohort of athletes and trying to have them safely exercise,” he said. “It has to be a controlled environment where you can make sure that wearable devices have been given to everyone, and there are no inequities that are coming from the fact that not everyone has one.”
Wearables do have one key advantage: they are being worn all the time. “Basically, you are running a repeated test on yourself every day,” he said, without the need for reminders to take your temperature or spit in a tube. Wearables are passive and unobtrusive, he added, and can potentially play into people’s behavior if appropriately modulated by a light intervention.
An example would be an app that alerts users when something unusual has been detected on their wearable device. It could suggest that they refrain from going out for the next several days and trigger tests sent to their homes to determine whether they are infected with COVID.
“We know that COVID gets worse slowly,” he said, which can be problematic when people don’t act quickly. In the United States, for instance, most tests require a prescription, so there is a need to first interact with a physician. Then, following a test, there is a wait time involved as the sample ships, and it generally takes two to three days to receive results.
The entire process, from symptom onset to final awareness, can take nearly a week, he explained. So if it is possible to tell someone, look, I could be wrong, but be more careful and get tested, it is probably possible to prevent potential contacts that would otherwise create opportunities for the virus to spread.
Foschini highlighted the need for any wearable-based solutions to emphasize privacy. It’s an issue that has been top of mind during the pandemic with the emergence of contact tracing apps, and he believes it will be even more top of mind if there are tech-based solutions monitoring people in real-time and knowing something about their health.
“I think it’s really, really important to have the solution being built with privacy first,” he added, so that people understand what data is being collected and shared as well as how that data is being used.
Paul Nicolaus is a freelance writer specializing in science, nature, and health. Learn more at www.nicolauswriting.com.