Mortality due to massive events like the COVID-19 pandemic is underestimated because of several reasons, among which the impossibility to track all positive cases and the inadequacy of coding systems are presumably the most relevant. Therefore, the most affordable method to estimate COVID-19-related mortality is excess mortality (EM). Very often, though, EM is calculated on large spatial units that may entail different EM patterns and without stratifying deaths by age or sex, while, especially in the case of epidemics, it is important to identify the areas that suffered a higher death toll or that were spared. We developed the Stata COVID19_EM.ado procedure that estimates EM within municipalities in six subgroups defined by sex and age class using official data provided by ISTAT (Italian National Statistics Bureau) on deaths occurred from 2015 to 2020. Using simple linear regression models, we estimated the mortality trend in each age-and-sex subgroup and obtained the expected deaths of 2020 by extrapolating the linear trend. The results are then displayed using choropleth maps. Subsequently, local autocorrelation maps, which allow to appreciate the presence of local clusters of high or low EM, may be obtained using an R procedure that we developed.•We focused on estimating excess mortality in small-scale spatial units (municipalities) and in population strata defined by age and sex.•This method gives a deeper insight on excess mortality than summary figures at regional or national level, enabling to identify the local areas that suffered the most and the high-risk population subgroups within them.•This type of analysis could be replicated on different time frames, which might correspond to successive epidemic waves, as well as to periods in which containment measures were enforced and for different age classes; moreover, it could be applied in every instance of mortality crisis within a region or a country.
Read full abstract