- Split View
-
Views
-
Cite
Cite
F Janssen, AE Kunst for The Netherlands Epidemiology and Demography Compression of Morbidity research group†, Cohort patterns in mortality trends among the elderly in seven European countries, 1950–99, International Journal of Epidemiology, Volume 34, Issue 5, October 2005, Pages 1149–1159, https://doi.org/10.1093/ije/dyi123
- Share Icon Share
Abstract
Background Secular trends in old-age mortality are of crucial importance to population ageing. For the understanding and prediction of these trends, it is important to determine whether birth cohort effects, i.e. long-lasting effects of exposures earlier in life, are important in determining mortality trends up to old age. This study aimed to identify and describe cohort patterns in trends in mortality among the elderly (>60 years of age) in seven European countries.
Methods A standard age-period-cohort analysis was applied to all-cause and cause-specific mortality data by 5-year age groups and sex, for Denmark, England and Wales, Finland, France, The Netherlands, Norway, and Sweden, in the period 1950–99.
Results Cohort patterns were identified in all countries, for both the sexes and virtually all causes of death. They strongly influenced the trends in all-cause mortality among Danish, Dutch, and Norwegian men, and the trends in mortality from infectious diseases, lung cancer (men only), prostate cancer, breast cancer, and chronic obstructive pulmonary disease (COPD). All-cause mortality decline stagnated among Danish, Dutch, and Norwegian male birth cohorts born between 1890 and 1915, among French men born after 1920, and among women from all countries born after 1920. Where all-cause mortality decline stagnated, cohort patterns in mortality from lung cancer, COPD, and to a lesser extent ischaemic heart diseases, were unfavourable as well. For infectious diseases, stomach cancer, and cerebrovascular diseases, mortality increased among cohorts born before 1890, and decreased strongly thereafter.
Conclusions Cohort effects related to factors such as living conditions in childhood and smoking in adulthood were important in determining the recent trends in mortality among the elderly in seven European countries.
With the increase in life expectancy there has been a shift in the trend from dying at younger ages to dying at older ages. Consequently, the mortality trends among the elderly gained more and more importance in determining the mortality patterns and the extent of ageing of national populations,1 with its far-reaching consequences to the individual, society, and health care policies. Therefore, it is crucial to carefully study old-age mortality trends and their determinants.
Trends in mortality among the elderly in Europe have been characterized by an overall decline since the 1950s.2–5 However, a recent study on old-age mortality trends from 1950 to 1999 in seven European countries showed that the pace of decline varied strongly by country and over time.6 In Denmark, The Netherlands, and for Norwegian men, mortality declines even stagnated during the 1980s and 1990s.6
In order to identify the determinants of national trends in mortality, it is essential to distinguish between period and cohort patterns. Whereas period patterns indicate immediate effects of conditions occurring in late life, cohort patterns may reflect long-lasting effects of determinants located earlier in the life course, e.g. in infancy, childhood, or adulthood. This distinction is of crucial importance in making projections of future developments in mortality. If cohort effects prove to be important, cohort-wise projections may be used to explore the possible future mortality trends.
Interest in cohort effects dates back to the influential works by Kermack et al. in 19347 and by Case and MacMahon et al. in the 1950s.8,9 Since the 1980s, there has been a renewed interest in cohort effects. In epidemiology, this interest was stimulated by results from the new studies on the effects of living conditions in early life on health in later life.10,11 At the same time, a number of demographic studies reported on the debilitating effects of war and famine during early ages on mortality in adulthood.12–14 Despite this revival of interest, it remains unknown whether cohort effects are important in determining the recent trends in mortality of national populations, partly because the identification of cohort effects is complicated by the generally linear nature of the mortality decrease in Europe since the turn of the century.15 Moreover, it is as yet not known whether cohort effects also extend into old age.
Our hypothesis is that cohort effects are important factors in determining old-age mortality trends, and that these effects are especially visible in mortality from causes of death that are known to be related to smoking or to early life circumstances. In addition, we expect that the importance and patterns of cohort effects might differ between countries, due to, for example, differences in the development of the smoking epidemic among adults.16
In this paper, we aimed to identify and describe cohort patterns in all-cause and cause-specific mortality trends among the elderly populations of seven European countries. Using data on old-age mortality trends during the second half of the twentieth century, we assessed for each country individually (i) whether cohort patterns are strong enough to determine trends in all-cause and cause-specific mortality, and (ii) the timing of these cohort patterns.
In our study, we included both all-cause mortality and cause-specific mortality for seven European countries. Previous studies on cohort patterns in causes of death did not always compare the results for different causes of death or different countries. Furthermore, the comparison of the results between different studies is complicated by the application of different techniques to identify the importance of cohort effects. Our approach enabled us to observe not only commonalities but also dissimilarities in the importance and patterns of cohort effects.
Methods
Data
Data on total mortality, the underlying cause of death and population at risk, by year of death (1950–99), sex, and 5-year age groups were obtained for Denmark, England and Wales, Finland, France, The Netherlands, Norway, and Sweden, from national statistical offices and related institutes. For Denmark, Finland, and Norway, data were available from 1951, and for Sweden from 1952. For France data until 1997 were available and for Denmark until 1998. Data on total mortality and population for the highest age groups (80–100+, by single year of age) were obtained from the Kannisto–Thatcher Database on Old Age Mortality (http://www.demogr.mpg.de/databases/ktdb).3 See Janssen et al., 2004 for a more detailed description of the included data.6
In our analysis, we included data for those aged ≥60 years in order to address the question whether the influence of cohort effects extends into old age. The study was not restricted to the very old in order to include sufficient age groups to distinguish between period and cohort effects.
With regard to the causes of death, we selected infectious diseases (ICD-9 001–139), cancer of lung (ICD-9 162), cancer of stomach (ICD-9 151), cancer of prostate (ICD-9 185), cancer of breast (ICD-9 174–175), ischaemic heart disease (ICD-9 410–414), cerebrovascular diseases (ICD-9 430–434, 436–438), and chronic obstructive pulmonary diseases (ICD-9 490–494, 496). These are all causes of death occurring frequently among the elderly. Together they accounted for on average 46% of all deaths among those aged ≥60 years in the late 1990s (see Table 1).
. | Relative share (%) . | . | . | . | . | . | . | . | . | . | . | . | . | . | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Males . | . | . | . | . | . | . | Females . | . | . | . | . | . | . | |||||||||||||
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||||||||
All causes (N) | 97 951 | 11 34 834 | 91 963 | 6 42 737 | 2 82 455 | 95 371 | 2 05 139 | 1 08 686 | 13 34 906 | 1 12 878 | 6 94 906 | 3 07 813 | 1 01 918 | 2 17 462 | |||||||||||||
All causes (rate × 1000) | 54.75 | 49.20 | 46.38 | 43.26 | 47.42 | 51.15 | 48.16 | 45.95 | 43.54 | 37.96 | 33.91 | 38.36 | 41.20 | 39.60 | |||||||||||||
Cancer of lung | 7.0 | 7.5 | 6.7 | 7.3 | 10.1 | 5.0 | 3.8 | 4.0 | 3.6 | 1.5 | 1.3 | 2.2 | 2.1 | 2.0 | |||||||||||||
COPD | 6.6 | 6.6 | 4.2 | 4.1 | 7.3 | 4.6 | 3.2 | 5.4 | 4.2 | 1.4 | 2.9 | 3.6 | 3.1 | 2.3 | |||||||||||||
Ischaemic heart disease | 21.8 | 26.5 | 31.3 | 10.6 | 16.9 | 24.2 | 28.0 | 18.7 | 20.4 | 27.4 | 9.0 | 13.1 | 18.9 | 22.4 | |||||||||||||
Infectious diseases | 0.5 | 0.5 | 0.7 | 1.5 | 0.9 | 0.8 | 0.8 | 0.6 | 0.5 | 0.8 | 1.6 | 1.0 | 1.1 | 1.0 | |||||||||||||
Stomach cancer | 0.9 | 1.6 | 1.5 | 1.4 | 1.6 | 1.6 | 1.2 | 0.6 | 0.9 | 1.1 | 0.9 | 1.0 | 1.0 | 0.8 | |||||||||||||
Prostate cancer/breast cancer | 4.0 | 3.7 | 3.9 | 4.2 | 4.1 | 5.7 | 5.7 | 3.6 | 3.2 | 2.3 | 3.5 | 4.0 | 2.7 | 2.5 | |||||||||||||
Cerebrovascular diseases | 8.3 | 8.8 | 9.5 | 7.5 | 7.8 | 10.3 | 9.9 | 10.8 | 13.2 | 13.7 | 10.5 | 11.6 | 14.0 | 13.3 | |||||||||||||
Total share selected causes | 49.2 | 55.1 | 57.8 | 36.6 | 48.8 | 52.2 | 52.7 | 43.8 | 46.0 | 48.2 | 29.6 | 36.4 | 43.0 | 44.4 |
. | Relative share (%) . | . | . | . | . | . | . | . | . | . | . | . | . | . | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Males . | . | . | . | . | . | . | Females . | . | . | . | . | . | . | |||||||||||||
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||||||||
All causes (N) | 97 951 | 11 34 834 | 91 963 | 6 42 737 | 2 82 455 | 95 371 | 2 05 139 | 1 08 686 | 13 34 906 | 1 12 878 | 6 94 906 | 3 07 813 | 1 01 918 | 2 17 462 | |||||||||||||
All causes (rate × 1000) | 54.75 | 49.20 | 46.38 | 43.26 | 47.42 | 51.15 | 48.16 | 45.95 | 43.54 | 37.96 | 33.91 | 38.36 | 41.20 | 39.60 | |||||||||||||
Cancer of lung | 7.0 | 7.5 | 6.7 | 7.3 | 10.1 | 5.0 | 3.8 | 4.0 | 3.6 | 1.5 | 1.3 | 2.2 | 2.1 | 2.0 | |||||||||||||
COPD | 6.6 | 6.6 | 4.2 | 4.1 | 7.3 | 4.6 | 3.2 | 5.4 | 4.2 | 1.4 | 2.9 | 3.6 | 3.1 | 2.3 | |||||||||||||
Ischaemic heart disease | 21.8 | 26.5 | 31.3 | 10.6 | 16.9 | 24.2 | 28.0 | 18.7 | 20.4 | 27.4 | 9.0 | 13.1 | 18.9 | 22.4 | |||||||||||||
Infectious diseases | 0.5 | 0.5 | 0.7 | 1.5 | 0.9 | 0.8 | 0.8 | 0.6 | 0.5 | 0.8 | 1.6 | 1.0 | 1.1 | 1.0 | |||||||||||||
Stomach cancer | 0.9 | 1.6 | 1.5 | 1.4 | 1.6 | 1.6 | 1.2 | 0.6 | 0.9 | 1.1 | 0.9 | 1.0 | 1.0 | 0.8 | |||||||||||||
Prostate cancer/breast cancer | 4.0 | 3.7 | 3.9 | 4.2 | 4.1 | 5.7 | 5.7 | 3.6 | 3.2 | 2.3 | 3.5 | 4.0 | 2.7 | 2.5 | |||||||||||||
Cerebrovascular diseases | 8.3 | 8.8 | 9.5 | 7.5 | 7.8 | 10.3 | 9.9 | 10.8 | 13.2 | 13.7 | 10.5 | 11.6 | 14.0 | 13.3 | |||||||||||||
Total share selected causes | 49.2 | 55.1 | 57.8 | 36.6 | 48.8 | 52.2 | 52.7 | 43.8 | 46.0 | 48.2 | 29.6 | 36.4 | 43.0 | 44.4 |
Or last year available.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
COPD = chronic obstructive pulmonary diseases.
. | Relative share (%) . | . | . | . | . | . | . | . | . | . | . | . | . | . | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Males . | . | . | . | . | . | . | Females . | . | . | . | . | . | . | |||||||||||||
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||||||||
All causes (N) | 97 951 | 11 34 834 | 91 963 | 6 42 737 | 2 82 455 | 95 371 | 2 05 139 | 1 08 686 | 13 34 906 | 1 12 878 | 6 94 906 | 3 07 813 | 1 01 918 | 2 17 462 | |||||||||||||
All causes (rate × 1000) | 54.75 | 49.20 | 46.38 | 43.26 | 47.42 | 51.15 | 48.16 | 45.95 | 43.54 | 37.96 | 33.91 | 38.36 | 41.20 | 39.60 | |||||||||||||
Cancer of lung | 7.0 | 7.5 | 6.7 | 7.3 | 10.1 | 5.0 | 3.8 | 4.0 | 3.6 | 1.5 | 1.3 | 2.2 | 2.1 | 2.0 | |||||||||||||
COPD | 6.6 | 6.6 | 4.2 | 4.1 | 7.3 | 4.6 | 3.2 | 5.4 | 4.2 | 1.4 | 2.9 | 3.6 | 3.1 | 2.3 | |||||||||||||
Ischaemic heart disease | 21.8 | 26.5 | 31.3 | 10.6 | 16.9 | 24.2 | 28.0 | 18.7 | 20.4 | 27.4 | 9.0 | 13.1 | 18.9 | 22.4 | |||||||||||||
Infectious diseases | 0.5 | 0.5 | 0.7 | 1.5 | 0.9 | 0.8 | 0.8 | 0.6 | 0.5 | 0.8 | 1.6 | 1.0 | 1.1 | 1.0 | |||||||||||||
Stomach cancer | 0.9 | 1.6 | 1.5 | 1.4 | 1.6 | 1.6 | 1.2 | 0.6 | 0.9 | 1.1 | 0.9 | 1.0 | 1.0 | 0.8 | |||||||||||||
Prostate cancer/breast cancer | 4.0 | 3.7 | 3.9 | 4.2 | 4.1 | 5.7 | 5.7 | 3.6 | 3.2 | 2.3 | 3.5 | 4.0 | 2.7 | 2.5 | |||||||||||||
Cerebrovascular diseases | 8.3 | 8.8 | 9.5 | 7.5 | 7.8 | 10.3 | 9.9 | 10.8 | 13.2 | 13.7 | 10.5 | 11.6 | 14.0 | 13.3 | |||||||||||||
Total share selected causes | 49.2 | 55.1 | 57.8 | 36.6 | 48.8 | 52.2 | 52.7 | 43.8 | 46.0 | 48.2 | 29.6 | 36.4 | 43.0 | 44.4 |
. | Relative share (%) . | . | . | . | . | . | . | . | . | . | . | . | . | . | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Males . | . | . | . | . | . | . | Females . | . | . | . | . | . | . | |||||||||||||
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||||||||
All causes (N) | 97 951 | 11 34 834 | 91 963 | 6 42 737 | 2 82 455 | 95 371 | 2 05 139 | 1 08 686 | 13 34 906 | 1 12 878 | 6 94 906 | 3 07 813 | 1 01 918 | 2 17 462 | |||||||||||||
All causes (rate × 1000) | 54.75 | 49.20 | 46.38 | 43.26 | 47.42 | 51.15 | 48.16 | 45.95 | 43.54 | 37.96 | 33.91 | 38.36 | 41.20 | 39.60 | |||||||||||||
Cancer of lung | 7.0 | 7.5 | 6.7 | 7.3 | 10.1 | 5.0 | 3.8 | 4.0 | 3.6 | 1.5 | 1.3 | 2.2 | 2.1 | 2.0 | |||||||||||||
COPD | 6.6 | 6.6 | 4.2 | 4.1 | 7.3 | 4.6 | 3.2 | 5.4 | 4.2 | 1.4 | 2.9 | 3.6 | 3.1 | 2.3 | |||||||||||||
Ischaemic heart disease | 21.8 | 26.5 | 31.3 | 10.6 | 16.9 | 24.2 | 28.0 | 18.7 | 20.4 | 27.4 | 9.0 | 13.1 | 18.9 | 22.4 | |||||||||||||
Infectious diseases | 0.5 | 0.5 | 0.7 | 1.5 | 0.9 | 0.8 | 0.8 | 0.6 | 0.5 | 0.8 | 1.6 | 1.0 | 1.1 | 1.0 | |||||||||||||
Stomach cancer | 0.9 | 1.6 | 1.5 | 1.4 | 1.6 | 1.6 | 1.2 | 0.6 | 0.9 | 1.1 | 0.9 | 1.0 | 1.0 | 0.8 | |||||||||||||
Prostate cancer/breast cancer | 4.0 | 3.7 | 3.9 | 4.2 | 4.1 | 5.7 | 5.7 | 3.6 | 3.2 | 2.3 | 3.5 | 4.0 | 2.7 | 2.5 | |||||||||||||
Cerebrovascular diseases | 8.3 | 8.8 | 9.5 | 7.5 | 7.8 | 10.3 | 9.9 | 10.8 | 13.2 | 13.7 | 10.5 | 11.6 | 14.0 | 13.3 | |||||||||||||
Total share selected causes | 49.2 | 55.1 | 57.8 | 36.6 | 48.8 | 52.2 | 52.7 | 43.8 | 46.0 | 48.2 | 29.6 | 36.4 | 43.0 | 44.4 |
Or last year available.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
COPD = chronic obstructive pulmonary diseases.
To reconstruct mortality trends from these specific causes of death over the period from 1950 to 1999, we had to bridge five different revisions of the International Classification of Diseases (ICD-6 to ICD-10) for all countries except France for which coherent series of causes of death were already available.17 For this purpose, we constructed a general concordance table based on three-digit codes of the ICD, and the four digit code 422.1 for ischaemic heart diseases under ICD-6/7 (see Janssen et al., 2004).18 Remaining mortality discontinuities—caused by the use of three-digit instead of four-digit codes and by incidental changes in coding rules—were identified and adjusted for in our analysis. These discontinuities were identified by (i) visual analysis of cause-specific mortality trends, (ii) country-specific background information on these mortality discontinuities, and (iii) a regression-based method that quantified the size of the discontinuities (see Janssen et al., 2004).18 Mortality discontinuities that were regarded as owing to changes in coding rules were controlled for in our regression method by means of cause- and sex-specific transition coefficients. These transition coefficients are the parameter estimates of variables associated with a coding change (e.g. ICD-8toICD-9) and were obtained through sex-specific age-period regression models applied to cause-specific mortality among those aged ≥60 years.6,18 Lung cancer and stomach cancer were not affected by coding changes. Infectious diseases showed a mortality discontinuity associated with the revision from ICD-6/7 to ICD-8 in five countries. For the remaining causes of death, adjustments had to be made in one or two countries, mainly for incidental coding changes, for example the generally applied coding change in England and Wales between 1984 and 1992 (see Janssen et al., 2004).18
Statistical analysis
For statistical analysis, we aggregated the data into 5-year age groups, 5-year periods and consequently 10-year overlapping cohort groups. To these data we applied log-linear age-period-cohort (APC) models through the GENMOD procedure of the SAS 8.0 package.
In APC analyses, it is impossible to identify the role of age, period, and cohort separately, owing to the interdependency between these variables. We dealt with this identification problem19–21 by using the standard Clayton & Schifflers procedure in which mortality is decomposed in a common linear trend (drift), a non-linear period effect, and a non-linear cohort effect.20,21 See Box 1 for the different models that we fitted accordingly.
Model parameters | Statistical notation |
Age (A) | E[lnYa] = αa |
Age + drift (AD) | E[lnYad] = αa + δ |
Age + period (AP) | E[lnYap] = αa + δ + βp |
Age + period + cohort (APC) | E[lnYapc] = αa + βp + γ c |
Model parameters | Statistical notation |
Age (A) | E[lnYa] = αa |
Age + drift (AD) | E[lnYad] = αa + δ |
Age + period (AP) | E[lnYap] = αa + δ + βp |
Age + period + cohort (APC) | E[lnYapc] = αa + βp + γ c |
Model parameters | Statistical notation |
Age (A) | E[lnYa] = αa |
Age + drift (AD) | E[lnYad] = αa + δ |
Age + period (AP) | E[lnYap] = αa + δ + βp |
Age + period + cohort (APC) | E[lnYapc] = αa + βp + γ c |
Model parameters | Statistical notation |
Age (A) | E[lnYa] = αa |
Age + drift (AD) | E[lnYad] = αa + δ |
Age + period (AP) | E[lnYap] = αa + δ + βp |
Age + period + cohort (APC) | E[lnYapc] = αa + βp + γ c |
E[lnY] is the expected value of the natural log of the mortality rate, with the number of deaths being Poisson distributed. α, δ, β, and γ are the age, drift, period, and cohort effect, respectively. The variables for age (a) and cohort (c) had one baseline class. For period (p) a second baseline class was chosen c = A–a+p. The variables were indexed a = 1, 2, …, 7, 8; p = 1, 2, …, 9, 10; and c = 1, …, 13.
We measured the contribution of drift, the non-linear period effects, and the non-linear cohort effects in the mortality trends by assessing the reduction in scaled deviances (a measure of unexplained variance) when comparing the subsequent models (i) AD with A, (ii) AP with AD, and (iii) APC with AP, respectively. We expressed these reductions as percentages of the reduction in scaled deviance observed between the model with age only and the full APC model. These percentage reductions were used to measure the contribution of the different effects to systematic mortality trends, and to make comparisons of this contribution between countries and causes of death. The statistical significance of the difference in scaled deviances between the subsequent models was assessed by a log-likelihood ratio test, resulting in one-sided P-values.
In order to describe the timing of the identified non-linear cohort effects, we calculated cohort-specific mortality levels derived from the parameter estimates of the cohort variables out of the full age-period-cohort model. We thus controlled for age and period, and by choosing two baseline period groups (1950–54 and 1995–99) and only one baseline cohort group (1920–29), we included the drift parameter in the measurement of cohort differences in mortality. In our description of the cohort patterns we focused on the (cohort) deviations from linearity in the observed patterns. The cohort-specific mortality risks were calculated relative to the unweighted average of the mortality levels of all cohorts together. As not all birth cohorts have complete death counts for those aged ≥60 years (e.g. for those born in 1865 we have observations only for those aged ≥85 years, and those born in 1935 were aged 60–64 years at the end of our observation period) we were cautious when describing the mortality levels of the extreme cohorts.
Results
In all-cause mortality, drift (the linear component of both period and cohort effects) contributed to a large extent to the systematic trends, especially among elderly women (Figure 1). However, for men in Denmark, The Netherlands, and Norway, non-linear cohort effects and non-linear period effects contributed substantially to old-age mortality trends.
For all selected causes of death, the contribution of the non-linear cohort effect to the systematic trends was statistically significant in each country and for both sexes, except for mortality from stomach cancer among Norwegian women (Table 2). Non-linear cohort effects contributed substantially to the mortality trends in the majority of countries for infectious diseases, lung cancer (men only), prostate cancer, cancer of breast, and chronic obstructive pulmonary diseases (COPD). For Finland and France, the non-linear cohort effects also contributed considerably to the mortality trends for ischaemic heart diseases (IHD).
. | Share of scaled devianceaexplained by adding the non-linear cohort effect to the age-period model . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 34.8 | 4.6 | 4.8 | 1.8 | 44.9 | 20.3 | 5.2 | |||||||
Cancer of lung | 12.2 | 44.1 | 41.2 | 6.4 | 21.0 | 7.1 | 18.5 | |||||||
COPD | 7.3 | 31.2 | 20.2 | 73.5 | 40.4 | 17.3 | 15.5 | |||||||
Ischaemic heart disease | 1.9 | 8.5 | 28.7 | 36.6 | 8.5 | 7.4 | 1.0 | |||||||
Infectious diseases | 9.0 | 13.4 | 10.8 | 43.4 | 19.5 | 23.6 | 35.0 | |||||||
Cancer of stomach | 1.1 | 6.7 | 4.0 | 5.5 | 0.4 | 2.0 | 1.3 | |||||||
Cancer of prostate | 4.3 | 11.4 | 24.8 | 28.3 | 15.6 | 18.3 | 10.9 | |||||||
Cerebrovascular diseases | 0.9 | 4.1 | 1.9 | 7.2 | 1.6 | 4.1 | 4.2 | |||||||
Women | ||||||||||||||
All-cause mortality | 7.8 | 1.1 | 3.3 | 5.8 | 2.9 | 3.1 | 3.8 | |||||||
Cancer of lung | 3.6 | 3.3 | 8.9 | 7.5 | 9.7 | 2.3 | 9.0 | |||||||
COPD | 13.9 | 22.6 | 43.0 | 64.7 | 6.8 | 16.8 | 20.1 | |||||||
Ischaemic heart disease | 5.1 | 3.3 | 43.5 | 59.5 | 1.3 | 6.1 | 1.4 | |||||||
Infectious diseases | 6.8 | 18.0 | 17.1 | 62.0 | 14.0 | 19.7 | 40.9 | |||||||
Cancer of stomach | 1.1 | 1.7 | 3.1 | 6.5 | 0.9 | (0.4) | 1.0 | |||||||
Cancer of breast | 46.5 | 11.0 | 11.8 | 5.4 | 33.1 | 31.3 | 16.7 | |||||||
Cerebrovascular diseases | 5.2 | 8.3 | 6.0 | 18.1 | 6.0 | 10.3 | 14.2 |
. | Share of scaled devianceaexplained by adding the non-linear cohort effect to the age-period model . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 34.8 | 4.6 | 4.8 | 1.8 | 44.9 | 20.3 | 5.2 | |||||||
Cancer of lung | 12.2 | 44.1 | 41.2 | 6.4 | 21.0 | 7.1 | 18.5 | |||||||
COPD | 7.3 | 31.2 | 20.2 | 73.5 | 40.4 | 17.3 | 15.5 | |||||||
Ischaemic heart disease | 1.9 | 8.5 | 28.7 | 36.6 | 8.5 | 7.4 | 1.0 | |||||||
Infectious diseases | 9.0 | 13.4 | 10.8 | 43.4 | 19.5 | 23.6 | 35.0 | |||||||
Cancer of stomach | 1.1 | 6.7 | 4.0 | 5.5 | 0.4 | 2.0 | 1.3 | |||||||
Cancer of prostate | 4.3 | 11.4 | 24.8 | 28.3 | 15.6 | 18.3 | 10.9 | |||||||
Cerebrovascular diseases | 0.9 | 4.1 | 1.9 | 7.2 | 1.6 | 4.1 | 4.2 | |||||||
Women | ||||||||||||||
All-cause mortality | 7.8 | 1.1 | 3.3 | 5.8 | 2.9 | 3.1 | 3.8 | |||||||
Cancer of lung | 3.6 | 3.3 | 8.9 | 7.5 | 9.7 | 2.3 | 9.0 | |||||||
COPD | 13.9 | 22.6 | 43.0 | 64.7 | 6.8 | 16.8 | 20.1 | |||||||
Ischaemic heart disease | 5.1 | 3.3 | 43.5 | 59.5 | 1.3 | 6.1 | 1.4 | |||||||
Infectious diseases | 6.8 | 18.0 | 17.1 | 62.0 | 14.0 | 19.7 | 40.9 | |||||||
Cancer of stomach | 1.1 | 1.7 | 3.1 | 6.5 | 0.9 | (0.4) | 1.0 | |||||||
Cancer of breast | 46.5 | 11.0 | 11.8 | 5.4 | 33.1 | 31.3 | 16.7 | |||||||
Cerebrovascular diseases | 5.2 | 8.3 | 6.0 | 18.1 | 6.0 | 10.3 | 14.2 |
The scaled deviance that is explained by the full age-period-cohort model as compared with the model including only age.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
( ) No statistically significant reduction in the scaled deviance (P < 0.05 one sided).
COPD = chronic obstructive pulmonary diseases.
. | Share of scaled devianceaexplained by adding the non-linear cohort effect to the age-period model . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 34.8 | 4.6 | 4.8 | 1.8 | 44.9 | 20.3 | 5.2 | |||||||
Cancer of lung | 12.2 | 44.1 | 41.2 | 6.4 | 21.0 | 7.1 | 18.5 | |||||||
COPD | 7.3 | 31.2 | 20.2 | 73.5 | 40.4 | 17.3 | 15.5 | |||||||
Ischaemic heart disease | 1.9 | 8.5 | 28.7 | 36.6 | 8.5 | 7.4 | 1.0 | |||||||
Infectious diseases | 9.0 | 13.4 | 10.8 | 43.4 | 19.5 | 23.6 | 35.0 | |||||||
Cancer of stomach | 1.1 | 6.7 | 4.0 | 5.5 | 0.4 | 2.0 | 1.3 | |||||||
Cancer of prostate | 4.3 | 11.4 | 24.8 | 28.3 | 15.6 | 18.3 | 10.9 | |||||||
Cerebrovascular diseases | 0.9 | 4.1 | 1.9 | 7.2 | 1.6 | 4.1 | 4.2 | |||||||
Women | ||||||||||||||
All-cause mortality | 7.8 | 1.1 | 3.3 | 5.8 | 2.9 | 3.1 | 3.8 | |||||||
Cancer of lung | 3.6 | 3.3 | 8.9 | 7.5 | 9.7 | 2.3 | 9.0 | |||||||
COPD | 13.9 | 22.6 | 43.0 | 64.7 | 6.8 | 16.8 | 20.1 | |||||||
Ischaemic heart disease | 5.1 | 3.3 | 43.5 | 59.5 | 1.3 | 6.1 | 1.4 | |||||||
Infectious diseases | 6.8 | 18.0 | 17.1 | 62.0 | 14.0 | 19.7 | 40.9 | |||||||
Cancer of stomach | 1.1 | 1.7 | 3.1 | 6.5 | 0.9 | (0.4) | 1.0 | |||||||
Cancer of breast | 46.5 | 11.0 | 11.8 | 5.4 | 33.1 | 31.3 | 16.7 | |||||||
Cerebrovascular diseases | 5.2 | 8.3 | 6.0 | 18.1 | 6.0 | 10.3 | 14.2 |
. | Share of scaled devianceaexplained by adding the non-linear cohort effect to the age-period model . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 34.8 | 4.6 | 4.8 | 1.8 | 44.9 | 20.3 | 5.2 | |||||||
Cancer of lung | 12.2 | 44.1 | 41.2 | 6.4 | 21.0 | 7.1 | 18.5 | |||||||
COPD | 7.3 | 31.2 | 20.2 | 73.5 | 40.4 | 17.3 | 15.5 | |||||||
Ischaemic heart disease | 1.9 | 8.5 | 28.7 | 36.6 | 8.5 | 7.4 | 1.0 | |||||||
Infectious diseases | 9.0 | 13.4 | 10.8 | 43.4 | 19.5 | 23.6 | 35.0 | |||||||
Cancer of stomach | 1.1 | 6.7 | 4.0 | 5.5 | 0.4 | 2.0 | 1.3 | |||||||
Cancer of prostate | 4.3 | 11.4 | 24.8 | 28.3 | 15.6 | 18.3 | 10.9 | |||||||
Cerebrovascular diseases | 0.9 | 4.1 | 1.9 | 7.2 | 1.6 | 4.1 | 4.2 | |||||||
Women | ||||||||||||||
All-cause mortality | 7.8 | 1.1 | 3.3 | 5.8 | 2.9 | 3.1 | 3.8 | |||||||
Cancer of lung | 3.6 | 3.3 | 8.9 | 7.5 | 9.7 | 2.3 | 9.0 | |||||||
COPD | 13.9 | 22.6 | 43.0 | 64.7 | 6.8 | 16.8 | 20.1 | |||||||
Ischaemic heart disease | 5.1 | 3.3 | 43.5 | 59.5 | 1.3 | 6.1 | 1.4 | |||||||
Infectious diseases | 6.8 | 18.0 | 17.1 | 62.0 | 14.0 | 19.7 | 40.9 | |||||||
Cancer of stomach | 1.1 | 1.7 | 3.1 | 6.5 | 0.9 | (0.4) | 1.0 | |||||||
Cancer of breast | 46.5 | 11.0 | 11.8 | 5.4 | 33.1 | 31.3 | 16.7 | |||||||
Cerebrovascular diseases | 5.2 | 8.3 | 6.0 | 18.1 | 6.0 | 10.3 | 14.2 |
The scaled deviance that is explained by the full age-period-cohort model as compared with the model including only age.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
( ) No statistically significant reduction in the scaled deviance (P < 0.05 one sided).
COPD = chronic obstructive pulmonary diseases.
Turning to the timing of the identified cohort effects (including drift), all-cause mortality generally declined among subsequent cohorts (Figure 2). For men, the decline stagnated among Danish, Dutch, and Norwegian birth cohorts born between 1890 and 1915, and among French generations born after 1920. For women, the decline in all-cause mortality accelerated among cohorts born after 1880, but stagnated among the cohorts from 1920 onwards. This stagnation was most pronounced in Denmark and did not occur in England and Wales or Finland.
The increase in mortality from lung cancer among men accelerated among cohorts born from 1880 onwards. This increase levelled off again among those born after 1895, except for Norway and France, and resulted in a decrease for The Netherlands, England and Wales, and Finland later on. For women, the increase in lung cancer mortality accelerated from 1915 onwards in The Netherlands, Denmark, Norway, and Sweden, while this mortality increase levelled off among women in the other countries.
For men in The Netherlands, Norway, Sweden, and Denmark, the cohort patterns for COPD resembled that for lung cancer, i.e. initial modest increases accelerated among the birth cohorts from 1880 onwards, and decelerated later on. The onset of mortality decline was later as compared with lung cancer. For men in England and Wales, France, and Finland, the increase in COPD mortality was initially strong, but reversed into mortality decline from birth cohorts 1895 onwards. For women, the cohort trends in COPD mortality were heterogeneous and erratic until birth cohort 1900. In cohorts from 1905 onwards, mortality generally increased, especially in Denmark.
For France, Finland, and Norway, IHD mortality increased for both men and women up to cohorts 1900–15, and declined thereafter. In Denmark, mortality increased until birth cohort 1880, but declined thereafter, especially among women. IHD mortality in Sweden, England and Wales, and The Netherlands declined among all birth cohorts, although this decline stagnated among male birth cohorts from 1890 onwards, especially among Dutch men.
For infectious diseases, stomach cancer, and cerebrovascular diseases, mortality generally declined after an initial increase among cohorts born in the nineteenth century (Figure 3). This initial mortality increase was most pronounced for infectious diseases, for which the increase persisted until cohorts born around 1890 (men) and 1885 (women). Stomach cancer increased only among the earliest birth cohorts of Finland, France, England and Wales (men), and Sweden (men). For cerebrovascular diseases, mortality increased among cohorts born before 1880 (men) and before 1885 (women). This initial increase was stronger for women.
Mortality from prostate cancer and breast cancer generally increased among subsequent cohorts, with only small deviations from the linear trend. For prostate cancer, mortality increases were strongest among the earlier birth cohorts, especially in Sweden.
Discussion
Cohort patterns were identified in all countries, for both the sexes and virtually all causes of death. Non-linear trends in mortality among subsequent birth cohorts contributed substantially to the trends in all-cause mortality among Danish, Dutch, and Norwegian men, and especially to the trends in mortality from infectious diseases, lung cancer (men only), prostate cancer, breast cancer, and COPD in most countries. We found that the secular decline in all-cause mortality stagnated among Danish, Dutch and Norwegian male birth cohorts born between 1890 and 1915, among French men born after 1920, and among women from all countries born after 1920. Where the decline in all-cause mortality stagnated, cohort-specific trends in mortality from lung cancer, COPD, and IHD were generally unfavourable as well. For infectious diseases, stomach cancer, and cerebrovascular diseases, mortality increased among cohorts born before 1890, and decreased rapidly among later generations.
The substantial (non-linear) cohort effects that we found in all-cause mortality trends among elderly Danish, Dutch, and Norwegian men were not identified in the few previous studies on cohort effects in all-cause mortality trends among the elderly.3,22 Kannisto suggested that cohort patterns do not have important effects on trends in all-cause mortality among the elderly in developed countries.3 However, his time series analyses did not include Denmark and The Netherlands, and his shorter observed period (1950–84) did not fully represent the birth cohorts that experienced a stagnation of mortality decline.6,23 Jacobsen et al. did not identify non-linear cohort effects in mortality trends among Danish men aged 40–84 years during 1960–99.22 However, their analysis was restricted to birth cohorts born after 1895, and therefore missed the important non-linearities that would become visible when extending the observation to cohorts born before 1895. The non-linearities that we observed for women born from 1910 onwards were found in their study as well.
Previous cohort studies on mortality from specific causes of death observed the same general tendencies as were observed in our study.24–37 However, in some cases, the exact timing and form of the observed cohort trends differed, probably because of the use of different techniques, shorter study periods, or different age groups. By applying the same technique to an identical selection of eight different causes of death among men and women in seven European countries, we were able to observe not only commonalities but also dissimilarities between these different countries.
Evaluation of data and methods
The mortality and population data used in this study stem from countries considered to have good or excellent population and vital registries,2,3 and highly accurate reported survivorship counts.3,38 Total mortality and population data for those aged ≥80 years were obtained from the Kannisto–Thatcher Database in which the data were checked for age-heaping and were subjected to a number of checks for plausibility.3
Particularly at older ages, the validity of the underlying cause of death may be questioned owing to the presence of more than one chronic disease contributing to death.39,40 Because changes over time in the quality of reporting underlying causes of death are likely to be gradual rather than abrupt, this may have resulted in only a modest overestimation of period effects. The non-linear cohort patterns in cause-specific mortality trends are likely to be unaffected.
We made considerable effort to control for the abrupt coding changes between and within revisions of the International Classification of Diseases (ICD), which otherwise could have biased the analyses of long-term trends in causes of death, and could overestimate period effects in APC analyses. We expect that any residual bias of coding problems is removed from our analyses by including period as a control variable in all our analyses of cohort patterns of mortality.
The identification problem inherent to APC analyses was dealt with by dividing mortality into a linear component (drift) and non-linear components. This enabled us to identify non-linear cohort effects and to quantify their importance. A focus on non-linear cohort effects leads to underestimation of the importance of all cohort effects to an unknown degree, because linear cohort effects are, together with linear period effects, included in the drift term. There can be substantial linear cohort effects when there is a large effect of drift, i.e. when mortality trends are largely linear. When we observed small non-linear cohort effects, they occurred in general together with large drift effects. The only exception concerned trends in IHD in Danish, Dutch, and Norwegian men. Only in these cases, it is certain that mortality trends are not strongly determined by cohort effects (see Appendix 1). Our method thus yields a conservative but informative estimate of the contribution of cohort effects.
Explanation of observed cohort patterns
Where the decline in all-cause mortality stagnated among subsequent cohorts, cohort-specific trends in mortality from lung cancer, COPD, and to a lesser extent IHD, were generally unfavourable as well. This parallel development indicates that changed smoking behaviour among adults, as reflected by the cohort patterns in lung cancer mortality,41 largely influenced the trends in COPD and IHD and also influenced the temporal stagnation of male all-cause mortality in The Netherlands, Denmark, and Norway among birth cohorts 1895–1920, and the stagnation of all-cause mortality decline for women born from 1920 onwards. Available data on smoking prevalence strengthen this conclusion. Dutch men born between 1897 and 1917 had a higher life-time exposure to smoking as compared with other generations of men.42 Among Danish adult men in 1954, 86% were present or former smokers, with a higher percentage for those aged 35–54 years (cohorts 1900–19) as compared with older men (cohorts before 1900).43 For Norwegian men, the proportion of ever smokers increased for the birth cohorts 1890–1914, after which it stayed at the same level up to cohort 1930–34 (83%).44 Among men in the other countries, lung cancer also increased among the earlier birth cohorts, as observed in previous studies.37 However, this increase seemed not strong enough in absolute terms to substantially influence the trends for all-cause mortality. For women, the large increase in lung cancer mortality for the cohorts born from 1920 onwards also parallels increasing smoking rates among subsequent generations.44,45 The low contribution of non-linear cohort effects to trends in lung cancer mortality among older women might be owing to the recent timing and the generalized nature (among all birth cohorts) of the spread of smoking among women.
The onset of the mortality decline in infectious diseases, cerebrovascular diseases, and stomach cancer for cohorts born at the end of the 19th century, which occurred in all countries, seems to indicate a contribution of the improvements in living conditions in early life. Earlier, a poor environment during infancy and childhood, and especially a high load of respiratory or gastrointestinal infections, has been proposed as a determinant of mortality from stomach cancer, tuberculosis, and cerebrovascular diseases at older ages.46 Mortality from infectious diseases at older ages is related to the disease load experienced during the first year of life, especially exposure to airborne infectious diseases.47 Infant mortality—often used as a proxy of living conditions in infancy48 and of the disease load during the birth year47—was still high in the late 19th century in most European countries, and began to fall at the turn of the century.49 The fact that infant mortality started to decline for about the same cohorts as adult mortality from the abovementioned causes of death suggests that improvements in living conditions have left an imprint on the mortality experience of cohorts up to old age. It is important to note, however, that this possible impact is visible only in mortality trends from some specific causes of death, while it is not reflected in all-cause mortality trends.
The strong non-linear cohort patterns in IHD mortality trends in France, Finland, and to a lesser extent Norway may reflect another cohort-specific determinant of old-age mortality trends. As these cohort patterns (strong increases among the birth cohorts born before 1900 or 1910, followed by substantial decreases afterwards) were observed for both men and women, changed smoking behaviour among adults could not be the determining factor in these countries, given the fact that men and women differed in their smoking history.16 Late effects of circumstances in early life have been suggested in relation to IHD,48,50,51 but the effects have been debated and seem to be less pronounced than for stroke and COPD.46,52 Moreover, during the late 19th and early 20th century, trends in infant mortality and Gross Domestic Product (an indicator of general socioeconomic development53) were not markedly different in France, Finland, and Norway as compared with the other countries (data not shown). Another possibility is that the initial rise in IHD mortality risk in France, Finland, and Norway is related to national patterns of behavioural change that are similar for both men and women, such as diet, alcohol use, or physical inactivity.54,55 This hypothesis, which was generated by observing variations between countries in cohort patterns for IHD, needs further testing by in-depth analyses within specific countries using the data on trends in these possible determinants.
Implications
Cohort effects appear to be important in determining secular trends in all-cause and cause-specific mortality among the elderly in seven European countries. Our results indicate that factors occurring in adulthood (predominantly due to smoking) and in infancy or childhood (such as poor living conditions) have left an imprint on the mortality experience of birth cohorts up to old age. Therefore, in order to understand the recent trends in old-age mortality and to make projections of possible future developments, one should take into account the determinants that are located earlier in the life course of subsequent cohorts.
. | Share of scaled devianceaexplained by adding drift to the model including only age . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 39.3 | 86.8 | 87.3 | 93.2 | 26.9 | 23.1 | 80.8 | |||||||
Cancer of lung | 63.1 | 4.3 | 0.8 | 78.6 | 42.7 | 85.4 | 47.4 | |||||||
COPD | 75.4 | 58.0 | 10.6 | 0.8 | 50.5 | 81.1 | 64.2 | |||||||
Ischaemic heart disease | 4.3 | 18.4 | 3.2 | 8.1 | 2.1 | 0.4 | 18.7 | |||||||
Infectious diseases | 65.8 | 75.7 | 86.3 | 52.3 | 36.7 | 50.1 | 50.6 | |||||||
Cancer of stomach | 98.2 | 88.4 | 95.8 | 92.9 | 98.7 | 97.6 | 97.7 | |||||||
Cancer of prostate | 75.7 | 68.9 | 69.9 | 45.8 | 81.3 | 79.9 | 85.7 | |||||||
Cerebrovascular diseases | 95.2 | 90.0 | 91.3 | 82.3 | 94.4 | 80.4 | 92.0 | |||||||
Women | ||||||||||||||
All-cause mortality | 81.9 | 98.5 | 93.1 | 93.8 | 91.7 | 93.3 | 95.4 | |||||||
Cancer of lung | 95.9 | 92.3 | 83.4 | 90.2 | 85.7 | 94.5 | 90.1 | |||||||
COPD | 83.2 | 24.2 | (0.1) | 0.5 | 15.3 | 42.9 | 67.1 | |||||||
Ischaemic heart disease | 44.0 | 56.6 | 4.8 | 5.7 | 55.1 | 22.1 | 69.0 | |||||||
Infectious diseases | 49.6 | 37.1 | 75.1 | 16.2 | 33.4 | 26.1 | 20.4 | |||||||
Cancer of stomach | 98.7 | 96.7 | 96.5 | 92.3 | 98.9 | 99.4 | 98.7 | |||||||
Cancer of breast | 27.5 | 50.4 | 68.7 | 88.6 | 45.4 | 50.2 | 25.3 | |||||||
Cerebrovascular diseases | 88.2 | 87.7 | 90.4 | 74.7 | 92.2 | 83.7 | 81.7 |
. | Share of scaled devianceaexplained by adding drift to the model including only age . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 39.3 | 86.8 | 87.3 | 93.2 | 26.9 | 23.1 | 80.8 | |||||||
Cancer of lung | 63.1 | 4.3 | 0.8 | 78.6 | 42.7 | 85.4 | 47.4 | |||||||
COPD | 75.4 | 58.0 | 10.6 | 0.8 | 50.5 | 81.1 | 64.2 | |||||||
Ischaemic heart disease | 4.3 | 18.4 | 3.2 | 8.1 | 2.1 | 0.4 | 18.7 | |||||||
Infectious diseases | 65.8 | 75.7 | 86.3 | 52.3 | 36.7 | 50.1 | 50.6 | |||||||
Cancer of stomach | 98.2 | 88.4 | 95.8 | 92.9 | 98.7 | 97.6 | 97.7 | |||||||
Cancer of prostate | 75.7 | 68.9 | 69.9 | 45.8 | 81.3 | 79.9 | 85.7 | |||||||
Cerebrovascular diseases | 95.2 | 90.0 | 91.3 | 82.3 | 94.4 | 80.4 | 92.0 | |||||||
Women | ||||||||||||||
All-cause mortality | 81.9 | 98.5 | 93.1 | 93.8 | 91.7 | 93.3 | 95.4 | |||||||
Cancer of lung | 95.9 | 92.3 | 83.4 | 90.2 | 85.7 | 94.5 | 90.1 | |||||||
COPD | 83.2 | 24.2 | (0.1) | 0.5 | 15.3 | 42.9 | 67.1 | |||||||
Ischaemic heart disease | 44.0 | 56.6 | 4.8 | 5.7 | 55.1 | 22.1 | 69.0 | |||||||
Infectious diseases | 49.6 | 37.1 | 75.1 | 16.2 | 33.4 | 26.1 | 20.4 | |||||||
Cancer of stomach | 98.7 | 96.7 | 96.5 | 92.3 | 98.9 | 99.4 | 98.7 | |||||||
Cancer of breast | 27.5 | 50.4 | 68.7 | 88.6 | 45.4 | 50.2 | 25.3 | |||||||
Cerebrovascular diseases | 88.2 | 87.7 | 90.4 | 74.7 | 92.2 | 83.7 | 81.7 |
The scaled deviance that is explained by the full age-period-cohort model as compared with the model including only age.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
( ) No statistically significant reduction in the scaled deviance (P < 0.05 one sided).
COPD = chronic obstructive pulmonary diseases.
. | Share of scaled devianceaexplained by adding drift to the model including only age . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 39.3 | 86.8 | 87.3 | 93.2 | 26.9 | 23.1 | 80.8 | |||||||
Cancer of lung | 63.1 | 4.3 | 0.8 | 78.6 | 42.7 | 85.4 | 47.4 | |||||||
COPD | 75.4 | 58.0 | 10.6 | 0.8 | 50.5 | 81.1 | 64.2 | |||||||
Ischaemic heart disease | 4.3 | 18.4 | 3.2 | 8.1 | 2.1 | 0.4 | 18.7 | |||||||
Infectious diseases | 65.8 | 75.7 | 86.3 | 52.3 | 36.7 | 50.1 | 50.6 | |||||||
Cancer of stomach | 98.2 | 88.4 | 95.8 | 92.9 | 98.7 | 97.6 | 97.7 | |||||||
Cancer of prostate | 75.7 | 68.9 | 69.9 | 45.8 | 81.3 | 79.9 | 85.7 | |||||||
Cerebrovascular diseases | 95.2 | 90.0 | 91.3 | 82.3 | 94.4 | 80.4 | 92.0 | |||||||
Women | ||||||||||||||
All-cause mortality | 81.9 | 98.5 | 93.1 | 93.8 | 91.7 | 93.3 | 95.4 | |||||||
Cancer of lung | 95.9 | 92.3 | 83.4 | 90.2 | 85.7 | 94.5 | 90.1 | |||||||
COPD | 83.2 | 24.2 | (0.1) | 0.5 | 15.3 | 42.9 | 67.1 | |||||||
Ischaemic heart disease | 44.0 | 56.6 | 4.8 | 5.7 | 55.1 | 22.1 | 69.0 | |||||||
Infectious diseases | 49.6 | 37.1 | 75.1 | 16.2 | 33.4 | 26.1 | 20.4 | |||||||
Cancer of stomach | 98.7 | 96.7 | 96.5 | 92.3 | 98.9 | 99.4 | 98.7 | |||||||
Cancer of breast | 27.5 | 50.4 | 68.7 | 88.6 | 45.4 | 50.2 | 25.3 | |||||||
Cerebrovascular diseases | 88.2 | 87.7 | 90.4 | 74.7 | 92.2 | 83.7 | 81.7 |
. | Share of scaled devianceaexplained by adding drift to the model including only age . | . | . | . | . | . | . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | DK . | E&W . | FIN . | F . | NL . | NO . | S . | |||||||
Men | ||||||||||||||
All-cause mortality | 39.3 | 86.8 | 87.3 | 93.2 | 26.9 | 23.1 | 80.8 | |||||||
Cancer of lung | 63.1 | 4.3 | 0.8 | 78.6 | 42.7 | 85.4 | 47.4 | |||||||
COPD | 75.4 | 58.0 | 10.6 | 0.8 | 50.5 | 81.1 | 64.2 | |||||||
Ischaemic heart disease | 4.3 | 18.4 | 3.2 | 8.1 | 2.1 | 0.4 | 18.7 | |||||||
Infectious diseases | 65.8 | 75.7 | 86.3 | 52.3 | 36.7 | 50.1 | 50.6 | |||||||
Cancer of stomach | 98.2 | 88.4 | 95.8 | 92.9 | 98.7 | 97.6 | 97.7 | |||||||
Cancer of prostate | 75.7 | 68.9 | 69.9 | 45.8 | 81.3 | 79.9 | 85.7 | |||||||
Cerebrovascular diseases | 95.2 | 90.0 | 91.3 | 82.3 | 94.4 | 80.4 | 92.0 | |||||||
Women | ||||||||||||||
All-cause mortality | 81.9 | 98.5 | 93.1 | 93.8 | 91.7 | 93.3 | 95.4 | |||||||
Cancer of lung | 95.9 | 92.3 | 83.4 | 90.2 | 85.7 | 94.5 | 90.1 | |||||||
COPD | 83.2 | 24.2 | (0.1) | 0.5 | 15.3 | 42.9 | 67.1 | |||||||
Ischaemic heart disease | 44.0 | 56.6 | 4.8 | 5.7 | 55.1 | 22.1 | 69.0 | |||||||
Infectious diseases | 49.6 | 37.1 | 75.1 | 16.2 | 33.4 | 26.1 | 20.4 | |||||||
Cancer of stomach | 98.7 | 96.7 | 96.5 | 92.3 | 98.9 | 99.4 | 98.7 | |||||||
Cancer of breast | 27.5 | 50.4 | 68.7 | 88.6 | 45.4 | 50.2 | 25.3 | |||||||
Cerebrovascular diseases | 88.2 | 87.7 | 90.4 | 74.7 | 92.2 | 83.7 | 81.7 |
The scaled deviance that is explained by the full age-period-cohort model as compared with the model including only age.
DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The Netherlands; NO = Norway; S = Sweden.
( ) No statistically significant reduction in the scaled deviance (P < 0.05 one sided).
COPD = chronic obstructive pulmonary diseases.
Cohort patterns were identified in old-age mortality trends in all seven European countries, for both the sexes and virtually all causes of death.
Parallel unfavourable cohort trends were observed for all-cause mortality and mortality from lung cancer, COPD, and to a lesser extent IHD, among Danish, Dutch, and Norwegian men born between 1890 and 1915.
For infectious diseases, stomach cancer, and cerebrovascular diseases, mortality increased among cohorts born before 1890, and decreased strongly thereafter.
Both living conditions in childhood and smoking in adulthood seem to have left an imprint on the mortality experience of birth cohorts up to high ages.
NEDCOM is a cooperation between the Department of Public Health in Rotterdam and the Population Research Centre at the University of Groningen, and includes: F Janssen, AE Kunst, J Barendregt, L Bonneux, C de Laet, W Nusselder, A Peeters, A Al Mamun, F Willekens, and JP Mackenbach.
This article is part of a project that is financed by the sector of Medical Sciences of the Organisation for Scientific Research, The Netherlands (ZonMw) (#904-66-093). We are grateful to Jacques Vallin (INED, France), Martine Bovet (INSERM, France), Hilkka Ahonen (Statfin, Finland), Annika Edberg (National Board of Health and Welfare, Sweden), Örjan Hemström (Sweden), Allan Baker & Glenn Meredith (ONS, England and Wales), Knud Juel (National Institute of Public Health, Denmark), and Jens-Kristian Borgan (Statistics Norway) for providing cause-specific mortality and population data. We gratefully acknowledge James Vaupel and Vladimir Shkolnikov (Max Planck Institute of Demographic Research) for the use of the Kannisto–Thatcher Database on Old Age Mortality. Furthermore, we would like to thank Caspar Looman and Gerard Borsboom for their statistical assistance, and Lex Burdorf, Frans van Poppel, and France Meslé for their useful comments.
References
Preston SH, Himes C, Eggers M. Demographic conditions responsible for population aging.
Kannisto V, Lauritsen J, Thatcher AR, Vaupel JW. Reductions in mortality at advanced ages: several decades of evidence from 27 countries.
Kannisto V. Development of oldest-old mortality, 1950–1990: evidence from 28 developed countries. Odense, Denmark: Odense University Press, 1994.
Thatcher AR. Trends in numbers and mortality at high ages in England and Wales.
Vaupel JW. The average French baby may live 95 to 100 years. In: Robine J-M, Vaupel J, Jeune B, Allard M (eds). Longevity: to the limits and beyond. Berlin: Springer-Verlag, 1997, pp. 11–27.
Janssen F, Mackenbach JP, Kunst AE. Trends in old-age mortality in seven European countries, 1950–1999.
Kermack WO, McKendrick AG, McKinlay PL. Death-rates in Great Britain and Sweden. Some general regularities and their significance.
Case RA. Cohort analysis of mortality rates as an historical or narrative technique.
Macmahon B, Terry WD. Application of cohort analysis to the study of time trends in neoplastic disease.
Kuh D, Ben-Shlomo Y. A life course approach to chronic disease epidemiology. Oxford: Oxford University Press, 1997.
Ben-Shlomo Y, Kuh D. A life course approach to chronic disease epidemiology: conceptual models, empirical challenges and interdisciplinary perspectives.
Caselli G, Vallin J, Vaupel JW, Yashin A. Age-specific mortality trends in France and Italy since 1900: period and cohort effects.
Horiuchi S. The long-term impact of war on mortality: old-age mortality of the First World War survivors in the Federal Republic of Germany.
Wilmoth J, Vallin J, Caselli G. Differences between observed and expected cohort mortality.
Caselli G. The influence of cohort effects on differentials and trends in mortality. In: Vallin J, D'souza S, Palloni A (eds). Measurement and analysis of mortality. New approaches. Oxford: Clarendon Press, 1990, pp. 229–49.
Lopez AD, Collishaw NE, Piha T. A descriptive model of the cigarette epidemic in developed countries.
Vallin J, Meslé F. Les causes de décès en France depuis 1925 [Causes of death in France since 1925]. Available at: http://matisse.ined.fr/%7Etania/causfra/data (In French).
Janssen F, Kunst AE. ICD coding changes and discontinuities in trends in cause-specific mortality in six European countries, 1950–99.
Hobcraft J, Gilles W. Age, period and cohort analysis in mortality studies. In: Vallin J, Pollard JH, Heligman L (eds). Methodologies for the collection and analysis of mortality data. Liege: Ordina Editions, 1984.
Clayton D, Schifflers E. Models for temporal variation in cancer rates. I: age-period and age-cohort models.
Clayton D, Schifflers E. Models for temporal variation in cancer rates. II: Age-period-cohort models.
Jacobsen R, Keiding N, Lynge E. Long term mortality trends behind low life expectancy of Danish women.
Janssen F, Nusselder WJ, Looman CWN, Mackenbach JP, Kunst AE. Stagnation in mortality decline among elders in The Netherlands.
La Vecchia C, Negri E, Levi F, Decarli A, Boyle P. Cancer mortality in Europe: effects of age, cohort of birth and period of death.
Peltonen M, Asplund K. Age-period-cohort effects on ischaemic heart disease mortality in Sweden from 1969 to 1993, and forecasts up to 2003.
Peltonen M, Asplund K. Age-period-cohort effects on stroke mortality in Sweden 1969–1993 and forecasts up to the year 2003.
Sverre JM. Secular trends in coronary heart disease mortality in Norway, 1966–1986.
Osmond C. Coronary heart disease mortality trends in England and Wales, 1952–1991.
Aragones N, Pollan M, Rodero I, Lopez-Abente G. Gastric cancer in the European Union (1968–1992): mortality trends and cohort effect.
Wolfe CD, Burney PG. Is stroke mortality on the decline in England?
Tarone RE, Chu KC. Implications of birth cohort patterns in interpreting trends in breast cancer rates.
Bornefalk A, Persson I, Bergstrom R. Trends in breast cancer mortality among Swedish women 1953–92: analyses by age, period and birth cohort.
Hermon C, Beral V. Breast cancer mortality rates are levelling off or beginning to decline in many western countries: analysis of time trends, age-cohort and age-period models of breast cancer mortality in 20 countries.
Post PN, Straatman H, Kiemeney LA, Coebergh JW. Increased risk of fatal prostate cancer may explain the rise in mortality in The Netherlands.
van der Gulden JW, Kiemeney LA, Verbeek AL, Straatman H. Mortality trend from prostate cancer in The Netherlands (1950–1989).
Caselli G. Future longevity among the elderly. In: Caselli G, Lopez A (eds). Health and Mortality Among Elderly Populations. Oxford: Clarendon Press,
Osmond C, Gardner MJ, Acheson ED. Analysis of trends in cancer mortality in England and Wales during 1951–80 separating changes associated with period of birth and period of death.
Condran AG, Himes CL, Preston SH. Old-age mortality patterns in low-mortality countries: an evaluation of population and death data at advanced ages, 1950 to the present.
Mackenbach JP, Kunst AE, Lautenbach H, Bijlsma F, Oei YB. Competing causes of death: an analysis using multiple-cause-of-death data from The Netherlands.
Manton KG. Cause specific mortality patterns among the oldest old: multiple cause of death trends 1968 to 1980.
Peto R, Lopez AD, Boreham J, Thun M, Heath C Jr. Mortality from tobacco in developed countries: indirect estimation from national vital statistics.
Gunning-Schepers LJ. The Health Benefits of Prevention. A Simulation Approach. Rotterdam: Erasmus University Rotterdam,
Hamtoft H, Lindhardt M. Tobacco consumption in Denmark. II. The Danish national morbidity survey of
Ronneberg A, Lund KE, Hafstad A. Lifetime smoking habits among Norwegian men and women born between 1890 and 1974.
Levi F, Lucchini F, Negri E, Boyle P, La Vecchia C. Changed trends of cancer mortality in the elderly.
Leon DA, Davey Smith G. Infant mortality, stomach cancer, stroke, and coronary heart disease: ecological analysis.
Bengtsson T, Lindstrom M. Airborne infectious diseases during infancy and mortality in later life in southern Sweden, 1766–1894.
Forsdahl A. Are poor living conditions in childhood and adolescence an important risk factor for arteriosclerotic heart disease?
Corsini CA, Viazzo PP (eds). The decline of infant and child mortality. The European experience: 1750–1990. The Hague: Martinus Nijhoff Publishers,
Barker DJ, Osmond C. Infant mortality, childhood nutrition, and ischaemic heart disease in England and Wales.
Williams DR, Roberts SJ, Davies TW. Deaths from ischaemic heart disease and infant mortality in England and Wales.
Ben-Shlomo Y, Davey Smith G. Deprivation in infancy or in adult life: which is more important for mortality risk?
Smits JP, Horlings E, Luiten van Zanden J. Dutch GNP and its components, 1800–1913. Groningen: Groningen Growth and Development Centre, Groningen University,
Kuulasmaa K, Tunstall-Pedoe H, Dobson A et al. Estimation of contribution of changes in classic risk factors to trends in coronary-event rates across the WHO MONICA Project populations.