2025/03/08 by Sasikiran Kandula, Anja Bråthen Kristoffersen, Gunnar Rø +2 · 1 voice
Mathematics · Medicine · Social Sciences · #COVID-19 and healthcare impacts #COVID-19 epidemiological studies #Insurance, Mortality, Demography, Risk Management
paper · doi:10.1016/j.puhe.2025.02.032
openalex publication_date 2025/03/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/23
OBJECTIVES: Excess mortality has been one of the commonly used measures of the population health effects of the COVID-19 pandemic. Denmark, Finland, Norway and Sweden share several health and socioeconomic characteristics but adopted different control measures and experienced varying degrees of case and hospitalization burden during the pandemic. Using mortality trends between 2001 and 2019 and a combination of models, we estimated and compared annual and monthly excess mortality in these countries nationally as well as stratified by age, sex and subnational regions between 2020 and 2023. STUDY DESIGN: Multi-model study. METHODS: Three methods were used to estimate mortality: i) a Bayesian spatial model with a random effect component for spatial dependence among subregions and trend and seasonality terms; ii) a Bayesian GAMM model with terms for annual trend (a thin-plate spline) and within-year seasonality (a cyclic cubic spline); and, iii) a combination of autoregressive and exponential trend smoothing methods. Estimates from these approaches were combined using model averaging. RESULTS: Based on age-standardized mortality rates (per 100,000 population) of the averaged estimates: Finland had the highest cumulative excess mortality of the four countries; older age groups (70+ year) accounted for nearly all excess mortality; men had higher excess rates than women; and capital regions had some of the lowest rates relative to other regions in each country. With a few exceptions, mortality in 2023 returned to pre-pandemic levels. Model verification indicated good calibration and superior skill of the combination model relative to its component models. CONCLUSIONS: We believe our approach better quantifies uncertainty than individual models, and our estimates are comprehensive, spatially, temporally and demographically well-resolved, and can support further association studies.