COVID-19, Climate, & Socioeconomic Status in the U.S.: A Bayesian Analysis of County-Level Case Data

Mentors: Brian Reich & Mohamed Abba
Team: Jordan Bramble, Frederick Donahey, Charlie Frazier, Liam Hanson, & Aaron Marshall

Project Overview Video and Corresponding Slides

Since emerging in 2019, the novel coronavirus SARS-CoV-2 has infected over 33 million individuals in the United States. SARS-CoV-2 spread has varied across the U.S., with counties experiencing differing levels at different times. Heterogeneity in viral spread may be partially explained by spatial and temporal variations in meteorological and demographic variables. Literature suggests that meteorological factors such as temperature and humidity impact COVID-19 case rates, but their influence remains unclear. While the impact of demographic factors such as socioeconomic status, race, and ethnicity is contested in current research, disparate case rates among different demographic groups have been observed. Meanwhile, population density and age structure may influence per capita contact rates and thus SARS-CoV-2 transmission rates. We utilize a Bayesian SIR model with county-dependent parameters to simulate SARS-CoV-2 spread across the U.S.; transmission rates are assumed to evolve temporally according to an AR(1) process. We fit this model to county-level COVID-19 case data and obtain parameter estimates via Markov Chain Monte Carlo sampling. We then explore the above factors’ relative influence on SARS-CoV-2 spread with a Bayesian regression model, in which the estimated means of county-level transmission rates are the response and both meteorological and demographic variables are covariates.