Frontline Communities and Sars-Cov-2 - Multi-Population Modeling With an Assessment of Disparity by Race/Ethnicity Using Ensemble Data Assimilation

Abstract

The COVID-19 pandemic has imposed many strenuous effects on the global economy, community, and medical infrastructure. Since the out- break, researchers and policymakers have scrambled to develop ways to identify how COVID-19 will affect specific sub-populations so that good public health decisions can be made. To this end, we adapt the work of Evensen et al [1] which introduces a SEIR model that incorporates an age-stratified contact matrix, a time dependent effective reproduction number R, and uses ensemble data assimilation to estimate model parameters. The adaptation is an extension of Evensen’s modeling framework, in which we model sub-populations with varying risks of contracting SARS-CoV-2 (the virus that causes COVID-19) in a particular state, each with a characteristic age-stratified contact matrix. In this work, we will focus on 9 U.S. states as well as the District of Columbia. We estimate the effective reproductive number as a function of time for our different sub-populations and then divide them into two groups: frontline communities (FLCs) and the complement (NFLCs). Our model will account for mixing both within populations (intra-population mixing) and between populations (inter-population mixing). Our data is conditioned on the daily numbers of accumulated deaths for each sub-population. We aim to test and demonstrate methodologies that can be used to assess critical metrics of the pandemic’s evolution which are difficult to directly measure. The output may ultimately be of use to measure the success or failures of the pandemic response and provide experts and policymakers a tool to create better plans for a future outbreak or pandemic. We consider the results of this work to be a reanalysis of pandemic evolution across differently affected sub-populations which may also be used to improve modeling and forecasts.