Clustering-based methodology for comparing multi-characteristic epidemiological dynamics with application to COVID-19 epidemiology in Europe.

Abstract

This study utilized a clustering-based approach to investigate whether countries with similar COVID-19 dynamics also share similar public health and selected sociodemographic factors. The pairwise distances between 42 European countries for six characteristics were calculated, including COVID-19 incidence, mortality, vaccination, SARS-CoV-2 genetic diversity, cross-country mobility and sociodemographic data. Hierarchical clustering trees were constructed, and the strengths of association between the pairs of trees were quantified using cophenetic correlation and Baker's Gamma correlation measures. The analysis revealed distinct patterns of agreement between clusterings. Vaccination clusterings showed moderate agreement with incidence but no strong agreement with mortality. Mortality-based clustering only agreed with population health clustering. Incidence-based clustering aligned with population health, genetic diversity and selected sociodemographic parameters. Genetic diversity clusterings agreed with mobility and related sociodemographic characteristics. The utility of the cluster-based methods for the time-series is illustrated, and these findings provide insights into the underlying mechanisms driving epidemiological disparities across localities and subpopulations.