The Poisson algorithm: a simple method to simulate stochastic epidemic models with generally distributed residence times

TL;DR

This paper introduces the Poisson algorithm, a simple and efficient method for simulating stochastic epidemic models with arbitrary residence time distributions, overcoming the limitations of traditional Markovian assumptions.

Contribution

The paper presents a novel, efficient simulation algorithm that allows for general residence time distributions in stochastic epidemic models, improving realism over Markovian models.

Findings

The Poisson algorithm effectively simulates non-Markovian epidemic models.

It accommodates any residence time distribution, enhancing model accuracy.

The method is computationally efficient and adaptable.

Abstract

Epidemic models are used to analyze the progression or outcome of an epidemic under different control policies like vaccinations, quarantines, lockdowns, use of face-masks, pharmaceutical interventions, etc. When these models accurately represent real-life situations, they may become an important tool in the decision-making process. Among these models, compartmental models are very popular and assume individuals move along a series of compartments that describe their current health status. Nevertheless, these models are mostly Markovian, that is, the time in each compartment follows an exponential distribution. In epidemic models, exponential sojourn times are most of the times unrealistic, for instance, they imply that the probability that a patient will recover from some disease in the next time unit is independent of the time the patient has been sick. This is an important restriction that prevents these models from being widely accepted and trusted by decision-makers. In spite of the need to incorporate algorithms to tackle the problem, literature on the topic is scarce. Here, we introduce a novel approach to simulate general stochastic epidemic models that accepts any distribution for the sojourn times that is efficient.