Pre-symptomatic detection of COVID-19 from smartwatch data

  1. Tejaswini Mishra
  2. Meng Wang
  3. Ahmed A. Metwally
  4. Gireesh K. Bogu
  5. Andrew W. Brooks
  6. Amir Bahmani
  7. Arash Alavi
  8. Alessandra Celli
  9. Emily Higgs
  10. Orit Dagan-Rosenfeld
  11. Bethany Fay
  12. Susan Kirkpatrick
  13. Ryan Kellogg
  14. Michelle Gibson
  15. Tao Wang
  16. Erika M. Hunting
  17. Petra Mamic
  18. Ariel B. Ganz
  19. Benjamin Rolnik
  20. Xiao Li
  21. Michael P. Snyder

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

No abstract available

Article activity feed

  1. ScreenIT

    SciScore for 10.1101/2020.07.06.20147512: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    NIH rigor criteria are not applicable to paper type.

    Table 2: Resources

    Software and Algorithms
    SentencesResources
    Recruitment was done through social media, word of mouth, COVID-19 registries, presentations, as well as from Stanford HealthCare.
    Stanford HealthCare
    suggested: None
    MyPHD App for Wearables Data Collection: After participants enrolled on REDCap, we provided them with MyPHD, a smartphone app developed by our study team to collect their wearables data in a de-identified and encrypted manner.
    REDCap
    suggested: (REDCap, RRID:SCR_003445)
    Symptoms/Other Metadata Processing: Participant metadata and symptom surveys were downloaded and processed using a custom R and Python script.
    Python
    suggested: (IPython, RRID:SCR_001658)
    Visualization methods: We used ggplot2, Matplotlib, and matlab for plotting the most of the figures 30,31
    ggplot2
    suggested: (ggplot2, RRID:SCR_014601)
    Matplotlib
    suggested: (MatPlotLib, RRID:SCR_008624)
    matlab
    suggested: (MATLAB, RRID:SCR_001622)

    Results from OddPub: Thank you for sharing your data.

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    Another limitation we observed is that some individuals do not wear their devices (or let their charge expire) when symptomatic, which may affect monitoring patterns. Patterns of non-use were observed both with Fitbit and for different devices and we expect that devices which require daily charging will also have more missing data. Nonetheless, devices whose charge lasts for several days should be powerful for early detection prior to loss of device function. It is currently unclear as to whether our approach can distinguish infections from SARS-COV2 from those caused by other illnesses. A challenge is that COVID-19 has heterogeneous physiologic presentation between individuals 21,22, as observed in our study. Regardless, any illness onset information is valuable, especially during a pandemic and can be followed up with appropriate testing. It is also likely that other types of physiological measurements (e.g. heart rate variability, respiration rate, skin temperature, blood oxygen saturation, electrocardiogram) obtainable from wearable devices will be valuable for both distinguishing illnesses from different infectious agents and could be used to increase diagnostic sensitivity and perhaps even predict illness severity and symptoms 23–26. Data from reported respiratory rates and blood oxygen is expected to be particularly useful in COVID-19 prediction 27. At the time of this writing, such data was not available to us; however, such data, along with increased participant size...

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

    Results from JetFighter: We did not find any issues relating to colormaps.

    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.

  2. Version published to 10.1038/s41551-020-00640-6
  3. NCRC

    Our take

    Physiological data derived from wearable technology in 24 US COVID-19 patients revealed latent physiological disturbance patterns—including heightened heart rate, decreased physical activity, and increased sleep duration—prior to symptom onset. A wearable technology detection algorithm correctly identified physiological abnormalities associated with pre-symptomatic COVID-19 infection in two-thirds of COVID-19 patients prior to symptom onset.

    Study design

    retrospective-cohort

    Study population and setting

    Investigators collected survey data and physiological markers, generated from wearable technologies (e.g., Fitbits, Smart Watches), through a smartphone application from a cohort of 5,262 US-based individuals. Investigators integrated participants’ physiological and activity data from wearable technologies with other self-reported metadata (demographics, medical history, daily COVID-19 symptoms, and COVID-19 testing/diagnoses) to: 1) identify physiological changes associated with COVID-19 infection and 2) determine precision with which wearable technologies could detect these physiological changes by, or prior to, symptom onset.

    Summary of main findings

    Using Fitbit data from enrolled participants (n = 24) with self-reported COVID-19 diagnoses and complete physiological markers (from 14 days prior to symptom onset to at least 7 days after), COVID-19 diagnosis was associated with increased heart rate (median: 7 beats/minute increase) 3-7 days before symptom onset. Decreases in daily steps and increased sleep duration were observed primarily in pre-symptomatic periods but following onset of resting heart rate signals associated with COVID-19 illness. There was high variability between individuals’ physiological markers and the progression/severity of COVID-19 illness. Based on these findings, investigators developed an algorithm detecting abnormal resting heart rates associated with pre-symptomatic COVID-19 infection. The algorithm detected 67% of COVID-19 cases prior to symptom onset in 24 participants supplying 28 days of physiological data ahead of symptom onset.

    Study strengths

    The authors collated large quantities of physiological data, collected in pre- and post-symptomatic periods among participants with COVID-19 infections, to characterize latent physiological markers of early COVID-19 disease. Additionally, investigators drew from various measurement approaches and statistical modeling techniques to appraise the robustness of their findings.

    Limitations

    Despite recruiting a large participant cohort, inferences from this study are drawn only from 24 participants with complete Fitbit records and self-reported COVID-19 diagnoses. As information about participants’ activities or behaviors was limited, some observed physiological changes could be attributed to unmeasured events (i.e., stress, other illness), rather than pre-symptomatic COVID-19 infection. In some cases, the time interval between the detection of physiological aberrations and symptoms onset stretched credulity (e.g, 15 days) given the 5 day median incubation period of COVID-19.The high volume of incomplete physiological records indicates participant data may not be missing at random, given individuals with more severe illness symptoms may have temporarily discontinued use of wearable technology. Lastly, Fitbit technology is not a gold standard for measurement of specific physiological markers and could bias the magnitudes of association reported.

    Value added

    This is the first study to characterize pre-symptomatic physiological changes in individuals with COVID-19 using wearable technology data, and to determine the precision with which wearable technologies, like Fitbits, can detect these physiological changes before symptom onset.

  4. ScreenIT

    SciScore for 10.1101/2020.07.06.20147512: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    NIH rigor criteria are not applicable to paper type.

    Table 2: Resources

    Software and Algorithms
    SentencesResources
    Metadata collection and surveys Study metadata such as demographic information, reports of past illnesses, daily symptom tracking etc. were collected via REDCap.
    REDCap
    suggested: (REDCap, SCR_003445)
    Symptoms/Other Metadata Processing Participant metadata and symptom surveys were downloaded and processed using a custom R and Python script .
    Python
    suggested: (IPython, SCR_001658)
    EllipticEvelope class from Scikit-learn package 15,16,29 to fit a Gaussian distribution of the data , pointing out the anomalies that might be contaminating our dataset because they are extreme points in the general distribution of the dataset.
    Scikit-learn
    suggested: (scikit-learn, SCR_002577)
    Visualization methods We used ggplot2 , Matplotlib , and matlab for plotting the most of the figures 30,31 Data availability:
    ggplot2
    suggested: (ggplot2, SCR_014601)
          <div style="margin-bottom:8px">
        <div><b>Matplotlib</b></div>
        <div>suggested: (MatPlotLib, <a href="https://scicrunch.org/resources/Any/search?q=SCR_008624">SCR_008624</a>)</div>
      </div>
    
      <div style="margin-bottom:8px">
        <div><b>matlab</b></div>
        <div>suggested: (MATLAB, <a href="https://scicrunch.org/resources/Any/search?q=SCR_001622">SCR_001622</a>)</div>
      </div>
    </td></tr><tr><td style="min-width:100px;vertical-align:top;border-bottom:1px solid lightgray">The Stanford Healthcare Innovation Lab gratefully acknowledges the support of Alexandra Duisberg.</td><td style="min-width:100px;border-bottom:1px solid lightgray">
      <div style="margin-bottom:8px">
        <div><b>Stanford Healthcare</b></div>
        <div>suggested: None</div>
      </div>
    </td></tr></table>

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:

    • Another limitation we observed is that some individuals do not wear their devices (or let their charge expire) when symptomatic, which may affect monitoring patterns.

    Results from OddPub: Thank you for sharing your data.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore is not a substitute for expert review. SciScore checks for the presence and correctness of RRIDs (research resource identifiers) in the manuscript, and detects sentences that appear to be missing RRIDs. SciScore also checks to make sure that rigor criteria are addressed by authors. It does this by detecting sentences that discuss criteria such as blinding or power analysis. SciScore does not guarantee that the rigor criteria that it detects are appropriate for the particular study. Instead it assists authors, editors, and reviewers by drawing attention to sections of the manuscript that contain or should contain various rigor criteria and key resources. For details on the results shown here, including references cited, please follow this link.

  5. Version published to 10.1101/2020.07.06.20147512v1 on medRxiv