We estimated sex specific all-cause excess mortality in the seasons 2016/17 to 2019/20 (week 27 to week 26 the following year), to investigate differences in sex specific excess mortality during the first half year of the COVID-19 pandemic and in previous periods with excess mortality. For this purpose, we applied the EuroMOMO model on data from European countries reporting all-cause death counts to the statistical office of the European Union, Eurostat. Expected all-cause mortality and excess all-cause mortality were estimated separately for each sex[i].
Data sources
Data on all-cause death counts, from a number of European countries, are freely downloadable from Eurostat[ii], by NUTS-code[iii], sex, 5-years age groups and ISOweek.
Population size on January 1st for each of the countries are also freely available from Eurostat17, by NUTS-code, sex and age.
Data used in this study were downloaded November 4, 2020.
Data preparation and criteria for inclusion
Criteria for inclusion
In order to apply the EuroMOMO model to estimate the expected baseline number of deaths in one winter season (week 27 to week 26 the following year), historical data for the preceding five seasons are required. Therefore, for each season we limited the analyses to countries having data covering at least the previous five seasons before and the full actual season. Further, all analyses were performed on country level (NUTS level 0), and aggregated to the following age groups: 0-14, 15-44, 45-64, 65-74, 75-84, 85+ years and all ages. Analyses were limited to countries with data available on age and sex.
EuroMOMO input data
The EuroMOMO model requires input data to be one record for each death, even though it is aggregated by ISOweek. Therefore, downloaded weekly death counts from Eurostat were randomly distributed over the week to fit the EuroMOMO model. This could be done without loss of information.
Weekly population data
Population size by ISOweek were linearly interpolated from the January 1st population data, by country (NUTS level 0), sex and age group.
Analyses
Expected and excess mortality
We used the EuroMOMO model R package[iv] to estimate the expected (baseline) weekly number of deaths for each country, by sex and age group, for each of the seasons 2016/17 to 2019/20. Number of countries with sufficient retrospective data dropped drastically for seasons before 2016/17 (supplementary 1), why earlier seasons were not included in the analyses. The country baselines and excess number of deaths were pooled, stratified by country, to account for heterogeneity between counties, to provide European estimates of expected (baseline) number of deaths and excess number of deaths for each of the seasons16.
European weekly estimates of excess number of deaths i.e. difference between observed and expected numbers of deaths, as well as their variances were extracted from the pooled data. Furthermore, excess mortality incidences by 100,000 person years by sex and age group were calculated, using population data.
Differences in age-distribution between sexes may make the comparison between sexes on all ages misleading. Therefore, a weekly age standardized total over age groups according to the overall age group distribution of both sex was calculated.
Sex-differences in excess mortality
Differences in weekly female and male excess mortality incidences were expressed as the difference between female and male excess mortality incidences (female minus male). This measure eliminates the underlying difference in mortality between the sexes
Sex-specific excess mortality versus overall excess mortality
A possible association between sex-difference in excess mortality and overall excess mortality was explored by linear regression, using the respective excess mortalities for the seasons 2016/17 to and including the 2019/20 COVID-19 season.