Outbreak-size distributions under fluctuating rates

TL;DR

This paper analyzes how stochastic fluctuations in infection and recovery rates affect outbreak size distributions in the SIR model, providing analytical solutions and insights into noise effects on epidemic variability.

Contribution

It introduces a model with Ornstein-Uhlenbeck processes for rates, deriving analytical outbreak-size distributions for different noise regimes, and compares demographic and reaction-rate noise impacts.

Findings

Outbreak-size distribution can be highly skewed with large fluctuations.

Slowly varying reaction-rate noise diminishes demographic noise effects.

Crossover to white-noise regime occurs near the recovery time scale.

Abstract

We study the effect of noisy infection (contact) and recovery rates on the distribution of outbreak sizes in the stochastic SIR model. The rates are modeled as Ornstein-Uhlenbeck processes with finite correlation time and variance, which we illustrate using outbreak data from the RSV 2019-2020 season in the US. In the limit of large populations, we find analytical solutions for the outbreak-size distribution in the long-correlated (adiabatic) and short-correlated (white) noise regimes, and demonstrate that the distribution can be highly skewed with significant probabilities for large fluctuations away from mean-field theory. Furthermore, we assess the relative contribution of demographic and reaction-rate noise on the outbreak-size variance, and show that demographic noise becomes irrelevant in the presence of slowly varying reaction-rate noise but persists for large system sizes if the noise is fast. Finally, we show that the crossover to the white-noise regime typically occurs for correlation times that are on the same order as the characteristic recovery time in the model.