We investigated the death toll of the Covid-19 pandemic in Sweden during 2020 along the dimensions of age, sex, and care status. We used care status-specific life expectancy in combination with confirmed Covid-19 deaths and excess deaths to estimate YLL in 2020 and compared it to other causes of death in 2019 and 2020. The comparison allowed us to put into context the amount of expected remaining life years that have been lost during the pandemic.

Our analysis revealed four key findings. First, remaining life expectancy varies greatly across care states. Second, the total amount of YLL from Covid-19 is reduced by 30% when care-specific life expectancy is considered, which translates to a reduction of around 2 years in the YLL per death. Third, more than 50% of all Covid-19 deaths had an average remaining lifetime of less than 4 years. Fourth, our results suggest that Covid-19 deaths did not replace other causes of death in 2020 but came on top of expected deaths. YLL from Covid-19 rank among the highest in comparison with other causes of death, also when considering care status.

Previous studies have documented a close link between existing comorbidities and an elevated risk for a severe progression and death from Covid-19 [1,2,3]. Our analysis of care status confirms these findings and additionally shows that a detailed understanding of the Covid-19 death toll requires information on health status, care need or similar variables that enables estimation of the variation in remaining life expectancy among those affected. Although the magnitude remains high, YLL are considerably reduced when care status is considered.

In Sweden as well as in other countries Covid-19 mortality has been particularly high among individuals in care homes [3, 22, 23], because they are frail, but also because they have been exposed to the virus to a greater extent than old people living at home. Protecting individuals within geriatric care has thus been a top priority in the public health response to the pandemic. The total number of YLL, especially among women receiving home care or living in care homes, however, show that protection efforts in Sweden have not been successful in the first year of the pandemic. Instead, the comparison to other causes of death shows that the Covid-19 pandemic was an unprecedented health threat to individuals receiving care.

The high number of YLL among men with no care is a worrisome finding. However, these numbers must be interpreted in the context of the gender differences in institutionalized care. Previous studies documented that living together with a potential caregiver considerably reduces the risk of admission to a care home [34, 35]. This likely also applies to the timing and health status at the onset of home care. Given that in the studied cohorts, married or partnered men are on average older than their wives, and women more often live alone, it is more likely that women act as care givers for their male spouses [36, 37]. This would result in an overestimation of the YLL among respective men without care since their remaining life expectancy at death would be somewhat lower than those measured for the no care group as a whole.

The observation that Covid-19 deaths closely resemble the distribution of excess deaths over age speaks towards that Covid-19 did not replace other causes of deaths but rather added deaths on top of the expected “normal” mortality. This is also confirmed when comparing the relative share of different causes on all deaths in 2019 and 2020 when excess deaths have been removed (see supplemental materials). The similar age distribution also shows that the pandemic particularly increased the mortality levels of frail individuals that already have a high death risk. The higher number of excess deaths compared to Covid-19 deaths may also suggest that cause-of-death-based death counts underestimate the true death toll of the pandemic, a problem that has been pointed out previously for many countries worldwide [38].

In most cases, Covid-19 deaths cut individual lifespans in the last years of life, as for instance, indicated by the finding that more than 50% of all deaths had an expected remaining lifetime of less than 4 years. This has implications for future mortality trends in Sweden. First, the concentration of deaths near the end of life suggests that potential mortality displacement may only have a short-term effect. We can thus assume that any adverse consequences following for instance from reduced cancer screening, or postponed care that may affect the total population will likely result in additional excess mortality. Second, increased mortality during 2020 will likely not result in a deviation of future mortality trends in the long run, even if the overall number of YLL caused by the pandemic remains high. Moreover, considering the variation in remaining life spans between affected groups, and the fact that many deaths had a remaining life expectancy well below the average number of YLL per death, this measure does not serve as a good summary indicator and should be used with caution in the case of estimating YLL during the Covid-19 pandemic for total populations.

Our paper has the strength of nationwide high quality administrative register data. This allowed us to use both cause of death statistics from medically signed death certificates, as well as excess mortality, an indicator for the adjusted net effect on mortality. We had access to monthly information about care status and could therefore model transitions in very short intervals, which is an advantage given the dynamics of transitions between care states and death. However, even with this approach, not all short-term transitions have been captured. Such transitions are mostly transitions into a state with higher care demand and into death soon after. Particularly remaining life expectancy for care home residence may thus be slightly overestimated. Moreover, we only had care status for the population aged 70 years and older. For ages below 70, we thus used total life expectancy as expected remaining lifetime, which may also lead to some overestimation of the remaining lifetime for some subgroups. There is likely also variation in death risk within different care status, which we did not consider in our analysis. With care-specific life expectancy, we thus capture variation between the different subgroups but not the variation within subgroups. This could result in an overestimation of YLL as individuals that died during 2020 may had a higher mortality risk compared to their subgroup peers. We also did not address the variation in the exposure to the virus within and across subgroups, which likely explains part of the differences in deaths and YLL between the different care groups. Individual health status and mortality risk would have captured even more variation in remaining life expectancy, yet also increased the complexity of the analysis. Overall, considering the large variation we observe in remaining life expectancy across the three care groups, we believe that our stratification of deaths and our estimates provide a more realistic picture of the YLL compared to previous estimates that stratify along age and sex only.