Abstract

Deterministic epidemic models that allow for replenishment of susceptibles display damped oscillatory behaviour. However, dynamics of epidemics are influenced by stochastic effects, particularly when the disease prevalence is low. At the beginning of an epidemic, due to low prevalence levels, stochastic die out is possible and is well studied in the literature. Once an epidemic takes off, extinction is highly unlikely, but the probability of extinction increases again with the decline of the wave. Extinction taking place during this period, that is, during the trough between the first and the potential second wave is known as epidemic fade-out. We consider a set of epidemics that evolve independently. Some of the epidemics may display fade-out while others do not. While fade-out is necessarily a stochastic phenomenon, the probability of this event taking place depends on the parameters associated with the epidemic. Therefore, we investigate whether time-series data of multiple outbreaks can be used to identify key drivers of epidemic fade-out across sub-populations in which the epidemics take place.

In this talk, using synthetic data, I will illustrate how a Bayesian hierarchical modelling approach can 1) identify when the sub-population specific model parameters supporting each epidemic have significant variability and 2) estimate the probability of epidemic fade-out for each sub-population. I will also demonstrate that a hierarchical analysis improves the estimates compared to when the epidemics are considered independently. Our estimation framework is applicable to other similar biological data.