A surrogate Bayesian framework for a SARS-CoV-2 data driven stochastic model

Abstract

Abstract Dynamic compartmentalized data (DCD) and compartmentalized differential equations (CDEs) are key instruments for modeling transmission of pathogens such as the SARS-CoV-2 virus. We describe an effi-cient nowcasting algorithm for modeling transmission of SARS-CoV-2 with uncertainty quantification for the COVID-19 impact. A key concern for transmission of SARS-CoV-2 is under-reporting of cases, and this is addressed in our data-driven model by providing an estimate for the detection rate. Our novel top-down model is based on CDEs with stochastic constitutive parameters obtained from the DCD using Bayesian inference. We demonstrate the robustness of our algorithm for simulation studies using synthetic DCD, and nowcasting COVID-19 using real DCD from several regions across five continents.