Early Detection of COVID-19 at Scale Using Wearables

Presymptomatic detection of COVID-19 has been a challenge in containing the ongoing pandemic. Wearable monitoring devices provide an exciting opportunity for early detection for real-time infectious disease monitoring at scale.
Early Detection of COVID-19 at Scale Using Wearables

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

A new model for early disease detection

A hallmark of COVID-19 is transmission in the absence of symptoms: a long presymptomatic period, as well as individuals who are asymptomatic (1). Approaches for early, pre-symptomatic and asymptomatic detection of COVID-19 are expected to have a major impact in the direction of the ongoing pandemic.

The current state of testing technology is inadequate. Traditional testing methods, including qPCR tests, are expensive, slow, typically performed infrequently and often ineffective at early detection (qPCR has over 50% false-negative rate within the first five days of exposure (2)). Thus, new methods are needed.

Wearable monitoring devices have the potential to fill this void. In 2017, our team demonstrated that early detection of infectious disease (Lyme, respiratory viral infection) using wearable devices is possible even days before self-reported symptoms and in asymptomatic cases (3). We wrote an algorithm, Change of Heart, to implement this concept. Our new study shows that measurements from commercially available smartwatches can be used to track and even predict respiratory viral infections (RVI) (4).

Wearable devices are an ideal tool for early RVI detection for a variety of reasons. Firstly, these devices provide continuous, real-time, 24/7 diagnostic data. Secondly, wearable devices are relatively low-cost and have already reached a significant level of widespread adoption. Recent data suggests that 20% of Americans use a wearable device (5) and they are only becoming more popular. Globally, predictions forecast wearables shipment volume to increase to total 637.1 million units by 2024, demonstrating a 5-year compound annual growth rate (CAGR) of 12.4% (6). Transforming existing wearable devices into effective health monitoring tools would provide millions of individuals with immediate access to early COVID-19 detection through something as simple as downloading a mobile application.

Mobilizing the Research Infrastructure

Scaling early detection methods poses many challenges, including cost, speed, and the handling and processing of copious amounts of data. To address these challenges, we created the Personal Health Dashboard, a secure and scalable system for analyzing and visualizing participants’ data. In February, 2020, at an early point in the pandemic, we adapted our disease detection algorithms and Personal Health Dashboard to focus on SARS-CoV-2 wearable detection methods with the ability to implement them at scale.

This project quickly escalated into an interdisciplinary army composed of eight key working groups borne from the Stanford Healthcare Innovation Lab, the Stanford Center for Genomics and Personalized Medicine, and the Snyder Lab at the Stanford University School of Medicine.

  1. Outreach and Partnerships
  2. Front-end and Customer Service 
  3. Regulatory
  4. Scalability 
  5. Security and Privacy 
  6. Mobile Programming (iOS/Android)
  7. Algorithm 
  8. UI/UX

On April 14th, we announced a collaboration with Fitbit and the Scripps Research Institute (7-8), which greatly accelerated our work. Almost overnight, our enrollment increased by over 3,000%. We were also very fortunate to collaborate with Survivor Corps, the largest COVID-19 patient advocacy group in the country. Diana Berrent, COVID-19 survivor and founder of Survivor Corps, is an enthusiastic fitness tracker and immediately understood the purpose and impact of our work.

Early results were very encouraging. In our very first case in April, 2020, our algorithm was effective in detecting COVID-19 9.5 days before symptom onset.

With proof-of-concept in hand, our subsequent study yielded great insight into the body’s response to SARS-CoV-2 infection and our ability to detect it. Altered resting heart rate, sleep, and step count were observed in 80% of COVID-19 infection cases, and we could detect these alterations in over 85% of positive cases prior to, or at, symptom onset. Two separate algorithms were developed:

1) In the RHR Difference (RHR- Diff) method, we detect and identify periods of elevated resting heart rate (RHR) compared to a healthy baseline constructed from a 28-day sliding window.

2) In the anomaly detection method (HROS-AD), we created a new feature known as HROS (Heart Rate Over Steps) by dividing heart rate with step count in hourly intervals. We then compared HROS of each period with all other intervals using Gaussian density estimation to find anomalous periods.

In regards to false positives/negatives, we found that our method relied heavily on having sufficient stable baseline data prior to infection. We encountered difficulty detecting abnormal RHR in individuals where this was lacking. Data for some of these participants were also presumably influenced by other medical conditions and/or medication.

We have proceeded to modify these algorithms for online, real-time detection methods. We note that these detection methods currently have limitations: 1) COVID-19 is difficult to distinguish from other RVIs (although COVID-19 does have a longer pre-symptomatic period than, for example, influenza B infection).  Nonetheless, we believe these alerts are still useful, given that early detection of infectious illnesses could mitigate future contagions. Furthermore, even if the algorithm cannot distinguish the specific identity of the infection, subsequent tests can be administered to identify the illness. 2) Since additional stressors can trigger our algorithm’s detection alarms, the incoming data needs to be contextualized. The use of additional data types (e.g. respiration rate and skin temperature), and new analysis approaches should enable us to distinguish between many different types of stressors and, perhaps, different types of RVIs.

We are all very excited about the promise of this work which has just been published (4). Having adapted our wearable detection system to multiple, independent diseases, we are now in the process of extending these algorithms across a variety of diseases for which early detection is key, including sepsis and Lyme disease, chronic conditions such as heart disease, diabetes/insulin resistance, and even mental health crises. We envision a future in which these wearable devices will be simultaneously protecting us from a variety of diseases.

Ultimately, we see this work as directly interfacing with our healthcare system to provide individuals more personalized and precise care. 

We are currently in the process of launching phase 2 of this early COVID-19 wearables project, where we will be alerting individuals to our algorithm’s detection results in real time. If you are interested in signing up or supporting Phase 2 of our study, please go to www.innovations.stanford.edu/wearables.


  1. He, X., Lau, E.H.Y., Wu, P. et al. Temporal dynamics in viral shedding and transmissibility of COVID-19. Nat Med 26, 672–675 (2020). https://doi.org/10.1038/s41591-020-0869-5
  2. Kucirka LM, Lauer SA, Laeyendecker O, Boon D, Lessler J. Variation in False-Negative Rate of Reverse Transcriptase Polymerase Chain Reaction-Based SARS-CoV-2 Tests by Time Since Exposure. Ann Intern Med. 2020;173(4):262-267. doi:10.7326/M20-1495
  3. Li X, Dunn J, Salins D, Zhou G, Zhou W, et al. (2017) Digital Health: Tracking Physiomes and Activity Using Wearable Biosensors Reveals Useful Health-Related Information. PLOS Biology 15(1): e2001402. https://doi.org/10.1371/journal.pbio.2001402
  4. Mishra, T., et al. Pre-symptomatic detection of COVID-19 from smartwatch data. Nat Biomed Eng (2020). https://doi.org/10.1038/s41551-020-00640-6
  5. Vogels E. About one-in-five Americans use a smartwatch or fitness tracker. Pew Research Center. 9 Jan 2020.  Accessed 5 Nov 2020.  https://www.pewresearch.org/fact-tank/2020/01/09/about-one-in-five-americans-use-a-smart-watch-or-fitness-tracker/
  6. International Data Corporation (IDC). Worldwide Wearables Market Forecast to Maintain Double-Digit Growth in 2020 and Through 2024. Accessed 5 Dec 2020.  https://www.idc.com/getdoc.jsp?containerId=prUS46885820
  7. Armitage H. Stanford Medicine scientists hope to use data from wearable devices to predict illness, including COVID-19. Stanford School of Medicine. 14 April 2020.  Accessed 5 Nov 2020. https://med.stanford.edu/news/all-news/2020/04/wearable-devices-for-predicting-illness-.html
  8. Fitbit Collaborates with Scripps Research and Stanford Medicine to Study the Role of Wearables to Detect, Track and Contain Infectious Diseases like COVID-19. Business Wire. 14 April 2020.  Accessed 5 Nov 2020.  https://www.businesswire.com/news/home/20200414005330/en/

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in