Outbreak diversity in epidemic waves propagating through distinct geographical scales

TL;DR

This study models COVID-19 spread across Brazil's geographical scales using a stochastic metapopulation approach, revealing diverse epidemic dynamics and emphasizing region-specific mitigation strategies validated by real data.

Contribution

It introduces a data-driven stochastic epidemic model capturing multi-scale spatial spread and regional response variability, applicable beyond Brazil.

Findings

Epidemic curves show broad distribution across regions and scales.

Strong correlation between outbreak delay and distance in some states.

Regional responses to mitigation vary significantly.

Abstract

A central feature of an emerging infectious disease in a pandemic scenario is the spread through geographical scales and the impacts on different locations according to the adopted mitigation protocols. We investigated a stochastic epidemic model with the metapopulation approach in which patches represent municipalities. Contagion follows a stochastic compartmental model for municipalities; the latter, in turn, interact with each other through recurrent mobility. As a case of study, we consider the epidemic of COVID-19 in Brazil performing data-driven simulations. Properties of the simulated epidemic curves have very broad distributions across different geographical locations and scales, from states, passing through intermediate and immediate regions down to municipality levels. Correlations between delay of the epidemic outbreak and distance from the respective capital cities were predicted to be strong in several states and weak in others, signaling influences of multiple epidemic foci propagating towards the inland cities. Responses of different regions to a same mitigation protocol can vary enormously implying that the policies of combating the epidemics must be engineered according to the region' specificity but integrated with the overall situation. Real series of reported cases confirm the qualitative scenarios predicted in simulations. Even though we restricted our study to Brazil, the prospects and model can be extended to other geographical organizations with heterogeneous demographic distributions.