A Stochastic Model for the Early Stages of Highly Contagious Epidemics by using a State-Dependent Point Process

TL;DR

This paper introduces a state-dependent self-exciting point process model to better capture the overdispersion and rapid growth in infections during early epidemic stages, addressing limitations of traditional stochastic models.

Contribution

It proposes a novel stochastic modeling approach using state-dependent point processes that incorporate current infection and recovery states, enabling more accurate early epidemic dynamics simulation.

Findings

The model captures overdispersion in infection counts.

Exact moment expressions are derived for the process.

Simulation algorithms for the proposed model are provided.

Abstract

The recent COVID-19 pandemic has shown that when the reproduction number is high and there are no proper measurements in place, the number of infected people can increase dramatically in a short time, producing a phenomenon that many stochastic SIR-like models cannot describe: overdispersion of the number of infected people (i.e., the variance of the number of infected people during any interval is very high compared to the average). To address this issue, in this paper we explore the possibility of modeling the total number of infections as a state dependent self-exciting point process. In this way, infections are not independent among themselves, but any infection will increase the likelihood of a new infection while also the number of currently infected and recovered individuals are included into determining the likelihood of new infections, Since long term simulation is extremely computationally intensive, exact expressions for the moments of the processes determining the number of infected and recovered individuals are computed, while also simulation algorithms for these state-dependent processes are provided.