AnRt- based model for predicting multiple epidemic waves in a heterogeneous population

Abstract

AbstractRelaxing the homogeneous mixing assumption in a population is often necessary to improve fits of epidemic models to observed infection counts. Establishing a link between observed infections and the underlying network of contacts is paramount to understanding how the network structure affects the speed of spread of a pathogen. In this paper we argue that introducing a flexible structure for the effective reproductive number (Rt) over the course of an epidemic allows for a more realistic description of the network of social contacts. This, in turn, produces better retrospective fits, as well as more accurate prospective predictions of observed epidemic curves. We extend this framework to fit multi-wave epidemics, and to accommodate public health restrictions on mobility. We demonstrate the performance of this model by doing a prediction study over two years of the SARS-CoV2 pandemic.