A caveat to using wearable sensor data for COVID-19 detection: The role of behavioral change after receipt of test results

This article has been Reviewed by the following groups

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Abstract

Recent studies indicate that wearable sensors can capture subtle within-person changes caused by SARS-CoV-2 infection and play a role in detecting COVID-19 infections. However, in addition to direct effects of infection, wearable sensor data may capture changes in behavior after the receipt of COVID test results. At present, it remains unclear to what extent the observed discriminative performance of the wearable sensor data is affected by behavioral changes upon receipt of the test results.

Methods

We conducted a retrospective study of wearable sensor data in a sample of medical interns who had symptoms and received COVID-19 test results from March to December 2020, and calculated wearable sensor metrics incorporating changes in step, sleep, and resting heart rate for interns who tested positive (cases, n = 22) and negative (controls, n = 83) after symptom onset. All these interns had wearable sensor data available for > 50% of the days in pre- and post-symptom onset periods. We assessed discriminative accuracy of the metrics via area under the curve (AUC) and tested the impact of behavior changes after receiving test results by comparing AUCs of three models: all data, pre-test-result-only data, and post-test-result-only data.

Results

Wearable sensor metrics differentiated between symptomatic COVID-19 positive and negative individuals with good accuracy (AUC = 0.75). However, the discriminative capacity of the model with pre-test-result-only data substantially decreased (AUC from 0.75 to 0.63; change = -0.12, p = 0.013). The model with post-test-result-only data did not produce similar reductions in discriminative capacity.

Conclusions

Changes in wearable sensor data, especially physical activity and sleep, are robust indicators of COVID-19 infection, though they may be reflective of a person’s behavior change after receiving a positive test result as opposed to a physiological signature of the virus. Thus, wearable sensor data could facilitate the monitoring of COVID-19 prevalence, but not yet replace SARS-CoV-2 testing.

Article activity feed

  1. SciScore for 10.1101/2021.04.17.21255513: (What is this?)

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

    Table 1: Rigor

    Ethicsnot detected.
    Sex as a biological variablenot detected.
    RandomizationCreate B=1000 data sets, each by randomly removing the same amount of data for each person as in Step 2; based on each of B random reduced data sets, calculate an AUC and the difference from the AUC in Step 1, resulting in B=1000 values of change in AUC; Step 4.
    Blindingnot detected.
    Power Analysisnot detected.

    Table 2: Resources

    No key resources detected.


    Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).


    Results from LimitationRecognizer: An explicit section about the limitations of the techniques employed in this study was not found. We encourage authors to address study limitations.

    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.