Years of life lost during the Covid-19 pandemic in Sweden considering variation in life expectancy by level of geriatric care

It is clear that a large share of Covid-19 deaths have occurred among individuals with preexisting health problems [1‐3]. This has led to speculations that deceased individuals would have died soon even in the absence of the pandemic. The controversy fueled debates about the effectiveness and costs of non-pharmaceutical interventions, such as lockdowns, considering also their potential unintended adverse consequences, including mental health problems [4‐8]. An essential component of this discussion is the question to what extent Covid-19 and the further consequences of the pandemic have shortened lifespans, or in other words, how many life years have been lost because of the direct and indirect effects the Covid-19 pandemic.

Almost all countries experienced excess mortality during the pandemic [9, 10]. The increase in mortality has resulted in declining life expectancy across most countries where information is available [11, 12]. Both excess mortality as well as life expectancy have been argued to be among the most objective indicators for the consequences of the pandemic on population-level mortality [12‐14]. Both indicators provide a net estimate of the mortality change, including additional deaths, such as those from Covid-19 itself but also other adverse health consequences of the pandemic, and averted deaths during the pandemic, such as potential lives saved due to the declining mobility [15].

In Sweden, life expectancy (LE) in 2020 has been estimated to be reduced by 0.6 years for females and 0.9 years for males compared to 2019 [12]. Considering that LE has increased continuously for decades, such a reduction is substantial.

An alternative perspective on the impact of the pandemic on mortality are years of life lost (YLL) [16‐18]. YLL indicate the average remaining number of expected life years (remaining life expectancy) for an individual at the time of death. Life expectancies are usually drawn from life tables, and thus, reflect population-level estimates. Both the decline in life expectancy and the amount of YLL are the highest in Sweden compared to other Scandinavian countries. However, Sweden is no outlier if the comparison is extended towards other European countries, the US, or Brazil [12, 19, 20]. For Sweden, Kolk et al. [21] estimate 36,304 lost life years for women (7.99 years per death), and 48,174 lost life years for men (9.14 years per death) in 2020, while Pifarré I Arolas et al. [20] estimated 63,169 lost life years for Sweden with an average of 8.83 YLL for females and 10.28 for males using data until the end of November 2020. YLL estimates in both studies are based age- and sex-specific life expectancy combined with Covid-19 deaths.

Most studies that quantify the mortality impact of Covid-19 do not account for health status, and implicitly assume that individuals dying from Covid-19 had the same remaining life expectancy as the total population. However, in many countries including Sweden, the majority of deaths with Covid-19 took place among individuals receiving formal care, and thus, within a subpopulation, which has on average a poorer health status and lower remaining life expectancy compared to the total population [3, 22, 23]. Yet only a small number of studies have incorporated health information in the YLL estimation. For instance, Hanlon et al. [24] reported a decrease of YLL from 12 years to 9.4 years per death and 14 years to 11.6 years for women and men in Italy, while Ferenci [17] found a decrease from 10.5 to 9.2 years per death across Hungarian women and men. Both studies used the number and type of long-term health conditions as measure for health status. For Scotland, Burton et al. [25] found a reduction in total YLL from 7.05 to 4.88 years for women, and 7.39 years to 5.39 years for men, when distinguishing between care home residents and non-care-home residents.

In Sweden, municipalities are legally obliged to provide publicly funded geriatric care to older individuals in need. These services are allocated according to individuals’ needs. Most commonly individuals are offered home care, and only when a person’s needs can no longer be met in their own home do they move into a care home. This makes care status an important stratification criterion when analyzing life years lost in the context of Covid-19.

In this study, we investigate the death toll caused by the Covid-19 pandemic in Sweden along the dimensions of age, sex, and care status. We use life expectancy estimates derived from an incidence-based multistate model stratified by age, sex, and care status to analyze the questions of how remaining life expectancy differs between individuals with and without formal care, and how many life years have been lost due to Covid-19 when taking this aspect into account. To put the YLL into context, we additionally compare the YLL from Covid-19 to the YLL by other causes of death.

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 [1‐3]. 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.

The large differences in remaining life expectancy between individuals with and without formal care, and the fact that most Covid-19 deaths in 2020 in Sweden occurred in care homes or among individuals receiving home care illustrates the need for subgroup specific mortality rates when estimating the burden of Covid-19 on overall mortality. Nevertheless, despite the shorter life expectancies of many individuals that died from Covid-19, the data are in line with the view that Covid-19 did not replace other causes of death but were additional deaths that came on top of expected mortality, and thus, reduced lifespans in all groups. Even after taking care status into account, YLL to Covid-19 in 2020 remain comparable to YLL from ischemic heart disease.