Detection of COVID-19 using multimodal data from a wearable device: results from the first TemPredict Study.
Level 3 - non-randomized controlled study
Prospective cohort diagnostic algorithm development and validation study
PubMed 35236896 · doi:10.1038/s41598-022-07314-0
What was done
Researchers gathered continuous physiological metrics via a consumer wearable (Oura Ring) and daily symptom questionnaires from 63,153 participants in the TemPredict study. To identify COVID-19 onset, a machine learning classification algorithm was trained on 73 individuals with PCR-confirmed infection and high-quality data selected from 704 self-reported possible cases. The model was also tested on a separate validation group of 10 participants with antibody-confirmed infection.
What was found
The algorithm detected COVID-19 an average of 2.75 days before participants sought diagnostic testing, with a sensitivity of 82%, a specificity of 63%, and an ROC AUC of 0.819 (95% CI [0.809, 0.830]). Adding continuous temperature metrics increased ROC AUC by 4.9% compared to excluding temperature. In the antibody-confirmed validation cohort (n = 10), the model demonstrated an ROC AUC of 0.819 (95% CI [0.809, 0.830]), 90% sensitivity, and 80% specificity. Model accuracy showed substantial variation across age groups and biological sex.
Why it matters
Continuous multimodal physiological tracking from consumer wearables can provide presymptomatic or early warning of viral infection, potentially allowing earlier isolation and reduced transmission before standard diagnostic testing is sought.
Limits
The algorithm was trained on only 73 PCR-confirmed cases and validated on 10 antibody-confirmed cases despite the large overall cohort. Specificity was moderate (63%), which implies high false-positive rates in community deployment. COVID-19 testing dates and symptoms were based on participant self-reports.
Cited by
- supports Wearable-based algorithms tracking parameters like respiratory rate and body temperature can detect COVID-19 infection before symptoms appear.