Background
Life expectancy at birth is one of the most commonly used measures of the health status of a population and the performance of the healthcare system [
1]. It represents the average age at death if everyone experienced the prevailing death rates throughout their lifetime. Another important dimension is the uncertainty around that expectation (i.e. the variation in ages at death) which is also known as lifespan inequality [
2]. Lifespan inequality has become relevant for policy makers with the growing interest in economic and health inequalities, [
3,
4] in particular because: [
1] it is a marker of heterogeneity in age at death at the macro level, and [
2] it is a marker of uncertainty in the timing of death at the micro level [
5‐
7]. Typically, early deaths are more common in underprivileged groups, simultaneously reducing life expectancy and increasing lifespan inequality [
8‐
11]. Both indicators may have implications for individuals’ decisions over their life course. For instance, when to invest in education or when to retire are decisions based on life expectancy but also on the uncertainty surrounding the eventual time of death [
10].
Life expectancy is lower in Denmark than in Norway and Sweden for females and males. From 1975 to 1995, while their Scandinavian counterparts showed continuous improvement, life expectancy stagnated among Danish women, while Danish men experienced only slow progress. For both sexes, life expectancy improved after 1995, but remains lower than in Sweden and Norway [
12]. Differences between Denmark and Sweden in life expectancy have been thoroughly documented [
13,
14]. Among females, the stagnation in life expectancy resulted mainly from the increased mortality of those born from 1919 to 1939, cohorts with high levels of smoking and alcohol consumption compared to their Swedish contemporaries [
13,
14]. Similarly, smoking-related mortality was considerably higher in Danish compared to Swedish males because of the widespread use of snus instead of cigarettes in Sweden [
15]. While these factors are known contributors to life expectancy differences, [
16] their effects on lifespan inequality differences are unknown. Previous evidence has shown mixed results for the effects of smoking on lifespan inequality: little to no effect on the Finnish population, [
17] while it increased lifespan inequality in some European countries [
18].
The Danish case, juxtaposed with Sweden, is interesting given the shared history, culture and similarities in their healthcare systems [
19]. It is unknown how the different age and cause-of-death mortality trends in the two countries would extend to lifespan inequality patterns.
Because life expectancy and lifespan inequality tend to be negatively correlated [
5,
7] we hypothesize that 1) during the last decades, Denmark experienced higher lifespan inequality relative to Sweden in females and males; 2) the 1975–1995 stagnation in life expectancy of Danish women was accompanied by stagnation in lifespan inequality; 3) the slow increase in life expectancy for males in the same period was accompanied by slow reduction of lifespan inequality. Because it is well-documented that smoking in the interwar Danish female cohorts was a major cause of the 1975–1995 stagnation in Danish female life expectancy, [
14] we hypothesize that 4) any 1975–1995 stagnation in lifespan inequality can be attributed to smoking-related deaths in these cohorts.
Hence, we analyze data since 1960 for Denmark and Sweden to make a cause-by-age analysis of changes in life expectancy and lifespan inequality for both sexes.
Discussion
In this study, we found that the same causes and age groups that held back Danish life expectancy in 1975–1995, especially for females, also held back lifespan equality in the same period. Although lifespan inequality has declined and life expectancy has increased since the late 1990s, Denmark still lags its Scandinavian counterparts, despite similarities in social and healthcare systems. The comparison with Sweden suggests that Denmark can now reduce inequality in lifespans and increase life expectancy through the same policy targets: cancer and infant mortality. This suggests an important social development, but also a clear policy target.
Reducing lifespan inequality cannot be the only policy goal, since this would neglect the interests of those who have already lived to higher ages: The effect of mortality reduction on lifespan inequality is large and negative at age zero, decreases with increasing age, and reverses at a unique threshold age, so that mortality reductions above this threshold age increase lifespan inequality [
5,
35]. Therefore, the causes that extend average lifespan and the causes that reduce lifespan inequality are not necessarily the same [
36]. Smoking-related mortality is a clear example of this. In Denmark, life expectancy stagnated over the 1975–1995 period because mortality reduction from most causes of death was offset by mortality increase from smoking-related causes. These increases in smoking-related mortality had a smaller net impact on lifespan inequality compared to life expectancy over the same period, since smoking-related mortality occurred both below and above the threshold age. By the latest period 1995–2014, however, reduction in smoking-related mortality was comparatively more important for decreases in lifespan inequality (19.4%) than increases in life expectancy (11.2%). In general, the impact of smoking on lifespan inequality is dependent on both the age of smokers compared to non-smokers (the maturity of the smoking epidemic), as well as the actual impact of smoking on mortality at different ages [
17]. Similar to what was found in a comparison of G7 countries, [
36] reductions in injuries and child mortality were relatively more important for lifespan inequality decrease than for life expectancy increase.
In the 1975–1995 period, non-smoking cancers also contributed (albeit to a small extent) to reductions in life expectancy and increases in lifespan inequality. The conservative definition of smoking-related cancers in this paper is one explanation for this phenomenon. Competing risks is another: people who previously died of other causes could die of cancer, and these increased cancer rates would show up as holding back life expectancy. In this respect, we note that non-smoking related cancer was on the rise also in Sweden, so it is likely not a phenomenon specific to Denmark. Specifically for Danish females, other risk-taking behavior may have led to increased cancer rates in general [
13,
14].
Causes of death that drive within-country changes in lifespan inequality are not necessarily the same as the causes of death that drive contemporary gaps between countries [
37]. However, the comparison with Sweden suggests that Denmark can simultaneous increase life expectancy and decrease lifespan inequality by targeting two main causes of death: cancer and infant mortality. Reducing lifespan inequality towards Sweden by these conditions would lead to an increase of 0.7 and 0.8 years in life expectancy for females and males in Denmark, respectively. To put this in perspective, in 2014 the infant mortality rate in Denmark is twice as high as in Sweden, which is one of the lowest among developed countries [
12]. Although mortality at very young ages may be affected by different registration practices in high income countries (e.g. non-viable live births registered as stillbirths), [
38] the Nordic countries do not show evidence of such patterns [
39]. Moreover, even after controlling for gestational age Sweden showed lower infant mortality rates than Denmark [
40]. Thus, infant mortality is the largest single contributor to the gap with Sweden in terms of lifespan inequality. Preventive policies focusing on prenatal risk factors and improving maternal health before and during pregnancy, [
41] as well as efforts to reduce the risk of sudden infant death syndrome [
42] could help to reduce infant mortality towards Swedish levels.
Targeting cancer is another clear public health intervention to reduce lifespan inequality and increase life expectancy in Denmark, confirming the priority given to this objective for the last two decades through the National Cancer Plans [
43]. Our results show that improvements in cancer mortality have had an effect on both health indicators over the last 20 years. However, Denmark had the highest mortality rates from all neoplasms in the European region, and the female population exhibited the highest lung cancer mortality rates [
24]. This is in line with our comparison with Sweden and with previous evidence highlighting the role of smoking behaviors on life expectancy trends [
14].
For Sweden, the decomposition results suggest that young-age external mortality can be further reduced. According to the WHO, males in Denmark have lower age-standardized external mortality rates (39 per 100,000) than Sweden and Norway (50.6 and 52 respectively) in 2014 [
44]. Our results further show that these differences are concentrated between ages 15 and 40. Moreover, since the late 1990s, Swedish and Norwegian males have experienced higher suicide rates between ages 15 and 24 [
45].
The mere observation that Sweden is doing better than Denmark for most causes of death does not mean that Denmark could easily do better. However, it does provide a starting point for public health intervention. For instance, previous evidence suggests that focusing on vulnerable and less socially advantaged subgroups may reduce suicide rates among the young [
45,
46].
For other countries that lag a comparable country, similar decompositions can be made. This may not result in a clear and consistent message: causes of death that hold back life expectancy may not be the same as the causes of death that hold back equality. Yet if it does, as in the case of Denmark when compared to Sweden, the benefits are substantial, because the policy goals can be so clearly stated. We therefore suggest that this method could be a valuable tool for epidemiologists and policy makers alike.
As any cause of death analysis, our study has the limitations that: 1) causes of death are treated as mutually exclusive, while they may not be (e.g., poor sight due to diabetes may lead to an accident); 2) medical doctors and even coroners have imperfect knowledge about causes of death; and 3) trends in awareness of certain diseases and changing insights in disease processes affect classification. Yet through using otherwise high-quality data and broad categories of causes of death, we believe we have achieved a useful, workable grouping of causes of death. In addition, we performed a sensitivity analysis to assure consistency of grouping across ICD versions and did not find significant variation when ICD revisions changed (Additional file
2: Figure S2). In addition, although the correlation between lifespan indicators suggest that our results would not differ had we used a different indicator, relative inequality indicators (e.g. coefficient of variation) differ in properties from indicators that measure absolute lifespan inequality (e.g. standard deviation). To alleviate any concern we replicated our results using the standard deviation (Additional file
4: Figure S3, Additional file
5: Figure S4, Additional file
6: Figure S5, Additional file
7: Figure S6) and did not find major differences.
Lifespan inequality is an important dimension of population health. By looking at this dimension we could disclose how lifespans differ within Denmark and Sweden. Moreover, our decomposition by age and cause of death allowed us to identify conditions and ages that contribute the most to lifespan inequality changes, and we were able to translate them into potential gains in life expectancy if efforts were concentrated in these ages and causes of death.