Modelagem hierárquica bayesiana de contatos: uma aplicação em modelos epidemiológicos compartimentais

Abstract

Hierarchical Linear Models, also called Multilevel Regression Models or Mixed-Effects Models, is a modeling method for nested data sets that present a hierarchical structure, being used for studies that pursuit to investigate the effects of variables at the individual level and at group levels, as well as for longitudinal studies, which rely on the presence of repeated measures. This type of adjustment explores the relationship between individuals and the environment to be studied, and understands that for this reason, all possible associations must be analyzed. The study in question aims to propose a multilevel Bayesian model to estimate contact rates among residents of Aglomerado da Serra by age groups and social circles, based on the studies of the POLYMOD project (Mossong et al., 2008) and the article by Prem et al. (2017). The estimated rates will be projected for regions of the city of Belo Horizonte in order to apply them in a SIR (Susceptible-Infected-Removed) model, as part of the studies to mitigate the impacts of COVID-19. Caused by the new coronavirus, SARS-CoV-2, the transmission of the virus occurs from one infected person to another and with millions of cases and deaths around the world, seeking to understand the patterns of contact networks considering the variations that may occur due to age groups and places of interaction, is of paramount t importance, as they can lead to differences in the effect of social distancing measures.