An informational approach to uncover the age group interactions in epidemic spreading from macro analysis

TL;DR

This paper explores the use of transfer entropy to analyze age group interactions in epidemic spreading, demonstrating its effectiveness on models and real COVID-19 data to understand behavioral changes over time.

Contribution

It introduces transfer entropy as a tool to uncover contact patterns and behavioral dynamics in epidemic data, providing a macro-level analysis approach.

Findings

Transfer entropy accurately reflects age-mixing patterns in models.

TE reveals behavioral changes during the COVID-19 pandemic.

Coarse-grained representations outperform raw data in analysis.

Abstract

We investigate the use of transfer entropy (TE) as a proxy to detect the contact patterns of the population in epidemic processes. We first apply the measure to a classical age-stratified SIR model and observe that the recovered patterns are consistent with the age-mixing matrix that encodes the interaction of the population. We then apply the TE analysis to real data from the COVID-19 pandemic in Spain and show that it can provide information on how the behavior of individuals changed through time. We also demonstrate how the underlying dynamics of the process allow us to build a coarse-grained representation of the time series that provides more information than raw time series. The macro-level representation is a more effective scale for analysis, which is an interesting result within the context of causal analysis across different scales. These results open the path for more research on the potential use of informational approaches to extract retrospective information on how individuals change and adapt their behavior during a pandemic, which is essential for devising adequate strategies for an efficient control of the spreading.