The effects of local homogeneity assumptions in metapopulation models of infectious disease

• Published in: Royal Society open science 2022-07, Vol.9 (7), p.211919
• Main Authors: Zachreson, Cameron, Chang, Sheryl, Harding, Nathan, Prokopenko, Mikhail
• Format: Article
• Language: English
• Subjects: agent-based models; compartmental models; Computer Science and Artificial Intelligence; epidemic models; heterogeneity
• ISSN: 2054-5703; 2054-5703
• DOI: 10.1098/rsos.211919

Computational models of infectious disease can be broadly categorized into two types: individual-based (agent-based) or compartmental models. While there are clear conceptual distinctions between these methodologies, a fair comparison of the approaches is difficult to achieve. Here, we carry out such a comparison by building a set of compartmental metapopulation models from an agent-based representation of a real population. By adjusting the compartmental model to approximately match the dynamics of the agent-based model, we identify two key qualitative properties of the individual-based dynamics which are lost upon aggregation into metapopulations. These are (i) the local depletion of susceptibility to infection and (ii) decoupling of different regional groups due to correlation between commuting behaviours and contact rates. The first of these effects is a general consequence of aggregating small, closely connected groups (i.e. families) into larger homogeneous metapopulations. The second can be interpreted as a consequence of aggregating two distinct types of individuals: school children, who travel short distances but have many potentially infectious contacts, and adults, who travel further but tend to have fewer contacts capable of transmitting infection. Our results could be generalized to other types of correlations between the characteristics of individuals and the behaviours that distinguish them.