Wearables and COVID-19
Remote monitoring with wearables
As the world grapples with the COVID-19 pandemic, employers and sports organizations are contemplating return-to-work/return-to-play scenarios. One of the concepts being actively discussed is the role of consumer wearable devices. Does this make sense? Yes, in theory wearables could be used as an early warning for possible COVID-19. While these devices cannot diagnose COVID-19, they could raise a red flag early in the course of illness. This would alert an individual to stay at home and if possible, obtain a COVID-19 test, before exposing others in the workplace. This could be particularly helpful since the most common physical screening procedure at this time - temperature checks - may capture less than half of individuals with COVID-19.
So what measures and in what combination are most effective for detecting likely COVID-19? That is exactly what several research studies are working on right now. Results of these studies will take some time to emerge. They are further challenged by how little we truly know about the disease, how rapidly information is evolving about it, and the largely non-specific symptoms of the disease. Nonetheless, we can take current knowledge about the disease, pair it with some common medical sense and understanding of wearable technology, and begin to formulate a remote monitoring strategy.
What to look for in a device
The first step is to identify a suitable device. It should provide the right types of data and should be sufficiently easy to use and comfortable. Ideally, it also provides data to a master dashboard, so all users within the organization can be monitored remotely on a daily basis.
A multi-modal approach. The likely best bet is to merge several different physiological measurements rather than relying on a single measurement. Many symptoms have been associated with COVID-19, and each by themselves could be caused by myriad diseases. In theory, examining several relevant measures should improve detection accuracy. Pairing wearable data with self-reported symptoms (see CDC Symptom List) will further improve the monitoring procedure.
Research indicates that increased core temperature (fever) is an important sign of COVID-19. However, fever is not a sufficient standalone predictor and may not appear until later in the course of disease - not as helpful for early detection. Respiratory distress is another common feature. Decreased blood oxygenation (Sp02) has been found in infected patients that do not appear severely ill otherwise, which could aid early detection. In lieu of Sp02, an increased breathing rate could also signal respiratory challenges. Finally, increased resting heart rate and decreased heart rate variability often serve as an early sign of impending illness in general and make sense to consider here.
Examine trends. In general, wearables are best for looking at how an individual’s measurements change over time rather than providing absolute values. Few wearables have proven to have the accuracy of medical-grade measures (i.e., thermometers for temperature). However, many have been shown to be consistent enough from reading-to-reading that trends can emerge over time. Significant deviations from normal (tbd what “significant” is for COVID-19, but 1 standard deviation seems reasonable for a starting point) may signal an underlying issue. Additionally, “normal” for many of these measures, such as resting heart rate and heart rate variability, depend on the individual. In these cases, absolute values are largely meaningless on their own anyway, and it is only with examining individual trends over time that these metrics provide valuable insight.
Opt for resting measures. Taking the proposed measurements while the individual is at rest or even sleeping will improve consistency and remove confounding factors (e.g., exercise, environmental temperature). In the case of illness, these measures also have more meaning in a resting context. If a device does not automatically detect these measurements during rest, it will be important to establish procedures to maximize consistency of when and how these data are recorded.
Vet the data. There is no certification standard for non-medical grade consumer wearables so it is critical to carefully vet data from any wearable device. Peer-reviewed published validation studies are very helpful for understanding the strengths and limitations of the device but may take some expert assistance to properly evaluate. Some wearable devices have sought medical device certification from the FDA, which lends some trust in the device in the absence of validation studies. White papers and other company-published “validation” should be reviewed skeptically. While these can be informative, they typically lack sufficient detail to fully judge the findings and have not had the scientific scrutiny provided by the peer-review process. If possible, speaking directly with the company can be highly valuable for evaluating their openness and transparency. In my experience, the companies that are the most open about their data and algorithms – and willing to admit where the device struggles - will provide the best data in the long run. Finally, it’s important to get a sample device and run in-house sanity checks of the data (and determine how user-friendly it actually is!) before committing to a wearable.
A starting list for wearable devices
With these considerations in mind, I’ve compiled a list of wearable devices that could support remote COVID-19 monitoring. This does not represent an endorsement of any specific devices. I have not had hands on experience with the majority of these devices, so this is my scientific scoping of devices based on information available online. Of note, a few of these devices (Fitbit, Oura, Cosinuss) have announced research studies to develop algorithms for COVID-19 detection. These devices may start providing automatic illness flagging in the coming months, once these studies are completed.
The list is prioritized for devices that fit the following:
Provides metrics in at least 2 of 3 target categories: temperature (core temp, skin temp), respiratory (oxygen saturation, breathing rate), heart rate (heart rate, heart rate variability)
Has a reasonable form factor for adoption by a healthy, mobile population
Appears to allow remote monitoring through dashboards/API access
Appears to be available for purchase now
A couple other notes
Published validation does not equal a valid device. The list includes links to peer-reviewed validation papers on one or more relevant metrics from the devices. These were found via Google Scholar searches for “DEVICE METRIC validation,” from the references of relevant review papers, or from the company’s website. However, the presence of papers does not mean that the device is valid. Interpretations of validation results is nuanced - always read the papers and reach out to experts to weigh in on the findings if possible.
Skin temp does not equal core temp but could still be useful. I distinguish Core Temperature (CT) and Skin Temperature (ST) in the chart because these readings are not equal. CT is used to derive indicators of fever and is typically measured at the mouth, ear, rectum, and forehead. Values do vary among these locations but are generally interpretable in a medical sense. Devices measuring ST on the torso (armpit, chest) may serve as good approximations of CT once a correction factor is incorporated (which many devices do automatically). However, these still may be best for trends rather than absolute interpretation. ST at the finger, wrist, or other peripheral areas are best for looking at day to day trends rather than absolute values. The values likely will not match CT; are more affected by the temperature of the environment, clothing, etc.; and may have different phasing than CT.
Final Thoughts
Monitoring health from wearables is not an easy thing. Monitoring a novel illness based on limited and rapidly changing data takes that challenge to a whole new level. Hopefully this provides some information to get you started thinking about this task while continuing to monitor new advances in COVID-19 detection as they emerge. Once you identify a device, you will need to deploy and train users on how to work with the device and upload their data (usually through a mobile app), and you will need someone with some expertise in medical physiology to dedicate significant time to monitoring the data and flagging possible cases (or developing an automated algorithm). This is not a simple task but with the right team it is doable. Wishing you good luck and good health in these challenging times.