A multi-type branching process model for epidemics with application to COVID-19.

Abstract

In this paper we model an infectious disease epidemic using Multi-type Branching Process where the number of offsprings of different types follow non-identical Poisson distributions whose parameters may vary over time. We allow for variation in parameters due to the behavior of citizens, government interventions in the form of lockdown, testing and contact tracing and the infectiousness of the variant of the virus in circulation at a time-point in a location. The model can be used to estimate several unknown quantities of interest in an epidemic such as the number of undetected cases and number of people quarantined following contact tracing. The model is fitted to the publicly available COVID-19 caseload data of India, South Korea, UK and US and is seen to provide good fit. It also provides good short-term forecast of the caseload for these countries. This model can be useful for health policy planners in assessing the impact of various intervention strategies such as testing, contact tracing, quarantine etc.