Andjela Rodic1*, Sofija Markovic1, Igor Salom2, and Marko Djordjevic1
1Faculty of Biology, University of Belgrade, Studentski trg 16, 11000 Belgrade, Serbia
2Institute of Physics, National Institute of the Republic of Serbia, Pregrevica 118, University of Belgrade, 11000 Belgrade, Serbia
andjela.rodic [at] bio.bg.ac.rs
Abstract
The Global Health Security Index (GHSI) is designed to assess the preparedness of countries to deal with infectious disease outbreaks. However, the COVID-19 pandemic has revealed a paradoxical relationship between the GHSI and the COVID-19 mortality, with higher GHSI scores being associated with higher death rates. We aimed to explain this puzzle. To rely on an accurate and robust measure of COVID-19 severity across countries, we used our model-derived measure instead of the standard Case Fatality Rate. We employed a range of statistical learning techniques, including non-parametric machine learning methods, to identify the factors that influence COVID-19 severity in 85 countries. Also, we searched for the predictors of the largely unexplored excess mortality counts. Our results suggest that the association of higher preparedness, measured by the GHSI, with higher COVID-19 mortality may be an artifact of oversimplified statistical analyses used in published studies. In addition, it could be a consequence of misclassified COVID‑19 deaths, combined with the higher median age of the population and earlier epidemics onset in countries with high GHSI scores.
Keywords: bioinformatics, modeling epidemics, machine learning, COVID-19 severity, excess deaths
