An Analysis of COVID-19 Recovery Duration: Smartwatch Activity and Self-Report

This study investigates the relationship between smartphone tracking data and self-reported COVID-19 experiences to better understand recovery patterns after COVID-19 infection in patients. We developed COronaVIden, an app that collects activity data through the iOS Health and Google Fit APIs, along with questionnaires on the timeline of infection, symptoms, and WHO-5 well-being scores. Of more than 5,000 participants (with 632 positive cases) who provided data between January 2019 and November 2021, 177 met our inclusion criteria. The average recovery duration based on activity data is 23.04±15.52 days, while self-perceived recovery duration average is 16.51±10.33 days. We developed two predictive models: In model 1 we used physical activity and demographic data, and in model 2 we used physical activity, demographic data, as well as additional post-infection symptoms that were made available to us by the users. We achieved a mean absolute error of 10.66 and 9.42 respectively, showing 29% and 37% improvement over the baseline, which is the median of recovery duration. The Shapley method was used to assess the contributions of features to the duration of recovery. Key findings indicate that being female, older and experiencing fatigue are correlated with longer recovery periods. Higher physical activity before infection was correlated with a shorter recovery time, possibly indicating better baseline health, although this was not significant in our population. Our findings demonstrate that wearables and smartphones can effectively track disease recovery patterns, offering valuable information for public health strategies, although more research is needed to account for potential confounding factors.