Assessing the impact of COVID-19 interventions on human mobility and SARS-CoV-2 transmission dynamics in the United States

Project Summary The rapid spread of SARS-CoV-2 led countries across the globe to implement strong social distancing and lockdown measures to reduce transmission. The first peak of newly reported COVID-19 cases and deaths in the United States occurred in April 2020, but the number of positive tests began increasing again in June as many states started easing their initial shelter-in-place orders despite ongoing transmission. In the absence of widespread deployment of an effective vaccine or another pharmaceutical intervention, state and local governments will have to rely on a range of non-pharmaceutical interventions (NPIs) to limit further outbreaks over the next 12-24 months. To provide policy makers with actionable information regarding the efficacy of different NPIs under a range of realistic epidemiological contexts, we will examine the impact of different NPIs on both mobility patterns and disease transmission using a geographically realistic, agent-based model. First, we will assemble a comprehensive database of local, county, and state policies related to COVID-19 from public websites and social media and categorize these policies by intervention type. We will also obtain epidemiological data from several different publicly available databases and use county-level case, testing, hospitalization, and mortality data to assess the impact of different county and state policies and NPIs in real- time. We will assess the link between NPIs and SARS-CoV-2 dynamics using cell phone-derived mobility data from a combination of publicly available sources and data sharing agreements with several data providers. First, we will use statistical models to assess the impact of different categories of county and state COVID-19 policies and NPIs on epidemiologically-relevant human mobility and activity patterns, including activity data at different places-of-interest subject to particular COVID-19 related restrictions. These mobility metrics will then be used to inform changes in local contact patterns in our agent-based transmission model. This transmission model will also incorporate detailed information on the demographics, socioeconomic factors, co-morbidities, and occupations that have been shown to be important for SARS-CoV-2 epidemiology. Local populations will be linked using regional connectivity metrics derived from cell phone data. Incorporation of these details will allow us to estimate the impact of different policies on transmission dynamics in a range of settings while accounting for local conditions as well as regional dynamics. Model estimates will be iteratively updated on a weekly basis over the course of the project to provide short-term forecasts of infections, hospitalizations, and deaths based on the current mix of NPIs across the country. These forecasts will be used to validate our NPI-impact estimates by comparing forecasts to future observations.