Alliance for Pandemic Preparedness
August 27, 2020
Population-Scale Longitudinal Mapping of COVID-19 Symptoms, Behaviour and Testing
Category: Article Summary
Topic: Modeling and Prediction
Keywords (Tags): transmission
- How We Feel (http://www.howwefeel.org), a web and mobile application, collected 3.6 million longitudinal self-reported survey responses on health, behavior related to the COVID-19 pandemic, and demographics from over 500,000 users across all 50 US states and several US territories (83% female, 76% white) during April 2 – May 12, 2020.
- Among 3,829 users who used the app within ±2 weeks of their reported PCR test results for SARS-CoV-2, Allen et al. evaluated a set of prediction models to distinguish positive and negative results. The best predictive model included common COVID-19 symptoms plus features observed in the survey (AUC=0.92), followed by a model with the four most predictive features (AUC=0.87), and a symptoms-only model (AUC=0.76).
Allen et al. (Aug 26, 2020). Population-Scale Longitudinal Mapping of COVID-19 Symptoms, Behaviour and Testing. Nature Human Behaviour. https://doi.org/10.1038/s41562-020-00944-2