English  Français

Robustness and bias of excess death estimates under varying model specifications

Various procedures are in use to calculate excess deaths during the ongoing COVID-19 pandemic. Using weekly death counts from 20 European countries, we evaluate the robustness of excess death estimates to the choice of model for expected deaths and perform a cross-validation analysis to assess the error and bias in each model's predicted death counts. We find that the different models produce very similar patterns of weekly excess deaths but disagree substantially on the level of excess. The country ranking of percent excess death in 2020 is sensitive to the exclusion of an exposure variable but otherwise stable across models that do account for population structure. The five-year average death rate model consistently produces the lowest excess death estimates, whereas high excess deaths are produced by the popular five-year average death count and Euromomo models. Cross-validation revealed these estimates to be biased under a causal interpretation of "expected deaths had COVID-19 not happened."

Keywords: COVID-19, Methodology, Mortality

See extended abstract.

  Presented in Session 118. Estimating and Modelling Mortality