[Paleopsych] Journal of Clinical Epidemiology: Trends in old-age mortality in seven European countries, 1950-1999

[Premise Checker]

Trends in old-age mortality in seven European countries, 1950-1999 Journal of 
Clinical Epidemiology Volume 57, Issue 2 , February 2004, Pages 203-216

F. Janssen [Corresponding Author Contact Information] , [E-mail The 
Corresponding Author] , a, J. P. Mackenbacha, A. E. Kunsta and for NEDCOMa, a, 
a, a, a, b, b, a, b, 1

a Department of Public Health, Erasmus MC, University Medical Center Rotterdam, 
P.O. Box 1738, 3000 DR, Rotterdam, The Netherlands
b Population Research Centre, Faculty of Spatial Sciences, University of 
Groningen, P.O. Box 800, 9700 AV, Groningen, The Netherlands

Accepted 24 July 2003.  Available online 20 May 2004.


Abstract

Objective

Different from the general observed decline in old-age mortality, for The 
Netherlands and Norway there have been reports of stagnation in the decline 
since the 1980s. We detect periods of stagnation in recent old-age mortality 
trends, and explore for which causes of death the recent stagnation is most 
apparent.

Study design and setting

We applied Poisson regression analysis to total and cause-specific mortality 
data by age (80+), period (1950-1999), and sex for seven European low-mortality 
countries.

Results

We found large heterogeneity in the pace of decline in the countries under 
investigation, with periods of stagnation being widespread. In the 1980s and 
1990s, stagnation was observed in Denmark, The Netherlands, and Norway (males). 
Continued mortality decline was observed especially in France. Although smoking 
has had a marked influence on the trends in old-age mortality, the role of 
smoking in the recent stagnation seems only modest and restricted to Norway. 
Mortality from cardiovascular diseases showed important crossnational 
variations in the pace of decline. Mortality from diseases specifically related 
to old age increased recently in all countries, except France.

Conclusion

Old-age mortality seems highly plastic and susceptible to many factors, with 
both favorable and unfavorable effects on trends over time.

Author Keywords: Mortality; Trends; Causes of death; Elderly; Smoking; Europe

Article Outline

1. Background
2. Data and methods
2.1. Data
2.2. Statistical analysis
2.3. Concordance
3. Results
4. Discussion
4.1. Evaluation of data and methods
4.2. Explanations of the trends observed
4.3. Implications
Acknowledgements
Appendix I. The concordance table used for bridging five revisions of the 
International Classification of Diseases (ICD)
References


1. Background

The general tendency in the trends in old-age mortality (80+) in low-mortality 
countries since the 1950s has been a declining one [1, 2, 3 and 4 ]. This 
tendency has contributed to the current increase in the number, the proportion, 
and the mean age of elderly people in these populations.

Because these developments will have huge implications for policy, society, and 
the demand of health care services, there has been much interest in possible 
future trends in life expectancy, especially in the debate on the limit to life 
expectancy [5, 6, 7, 8, 9, 10, 11 and 12 ]. On the one hand, proponents of "the 
limited-lifespan paradigm" state that biologic and practical constraints on 
reducing old-age mortality set an upper limit to life expectancy of 
approximately 85 years [5, 9 and 10 ]. On the other hand, "the 
mortality-reduction paradigm" is adhered to by researchers who argue that life 
expectancy will continue to increase—at least for some time to come—due to 
substantial reductions in mortality rates at all ages, including the oldest old 
[6, 8 and 11 ]. They support their view by referring to mortality development 
in subpopulations with extreme good health, to foreseen biomedical progress, 
and to historic observations that old-age mortality has been decreasing 
continuously.

Recently, however, research on old-age mortality has shown some exceptions to 
this continuous decline in old-age mortality. For The Netherlands and Norway, 
there have been reports of stagnation or even increases in old-age mortality 
since the 1980s [2, 13 and 14 ]. This raises the question of whether the 
mortality decline among elderly in low-mortality countries has been as 
consistent as previously reported, and whether these developments coincide with 
the idea that life expectancy continues to increase in the future.

However, before making inferences about future trends in old-age mortality, it 
is important to investigate the determinants of trends in old-age mortality. 
Although there have been studies on old-age mortality in the past that 
acknowledge the heterogeneity of the trends between countries [2, 15 and 16 ], 
they were mostly descriptive in nature. As a result, little is known about the 
reasons behind the crossnational differences in the pace of mortality decline 
among the elderly. Knowing about the determinants of trends in old-age 
mortality could provide important clues on future mortality among the elderly 
population, especially in the short and medium term.

Studying the trends in cause-specific mortality can generate evidence on the 
determinants of the trends in all-cause mortality among the elderly, because 
many of the intermediate factors or risk factors that could determine mortality 
trends among the elderly are related to specific causes of death. In doing so, 
we will emphasize the role of smoking. It is commonly known that smoking has a 
strong negative effect on survival. The impact of smoking on mortality trends 
may vary considerably between countries [17 ]. In fact, the stagnation observed 
in some countries could be the result of the increased lifetime exposure to 
smoking for those birth cohorts who reached old age at the end of the 20th 
century [18 and 19].

The objective of this article is to describe mortality trends among the elderly 
in Denmark, England and Wales, Finland, France, The Netherlands, Norway, and 
Sweden, and to contribute to the explanation of these trends. Focus will be on 
the trends from 1980 onwards. The research questions we will address are: (a) 
in which countries did trends in mortality at old age show signs of stagnation 
instead of continued decline, (b) to what extent did mortality increase for 
causes of death related to smoking, and (c) for which other specific causes of 
death did an increase occur?

In this analysis we will extend earlier studies on trends in old-age mortality 
by using detailed mortality data over a long period of time (1950–1999), with a 
distinction by specific causes of death and 5-year age groups up to 100+. 
Furthermore, in our analysis an extensive effort was made to carefully bridge 
the different revisions of the International Classification of Diseases (IDC).

2. Data and methods

2.1. Data

Data were obtained from national statistical offices and related institutes on 
total mortality, cause-specific mortality, and population at risk for Denmark, 
England and Wales, Finland, France, The Netherlands, Norway, and Sweden, for 
the years 1950–1999 (National Institute of Public Health [Denmark], ONS 
[England and Wales] [Twentieth Century Mortality], Statfin [Finland], INDE and 
INSERM [France], Statistics Netherlands, NIDI [The Netherlands], Statistics 
Norway and National Board of Health and Welfare [Sweden]). For Denmark, 
Finland, and Norway data were available only from 1951, for Sweden from 1952. 
Data for France were available until 1997 and for Denmark until 1998.

The selection of countries was restricted to low-mortality countries in North 
Western Europe. The seven countries we included were selected on the basis of 
the quality of the data [1 and 2] and on the availability of cause of death 
data.

For most countries, data on the total number of death by 5-year age groups were 
available, with a maximum age ranging from 85+ to 100+. For The Netherlands and 
France, deaths by single year of age were available to us (no maximum age) (see 
for France, http://www.ined.fr/publications/cdrom_vallin_mesle/continu.htm . 
Tables de mortalité françaises 1806–1997 et projections jusqu'en 2102 by J. 
Vallin and F. Meslé [INED]). To calculate mortality rates we used the 
midyear-population at risk. Population data were available either by single 
year of age (France, The Netherlands, Finland, Sweden, England, and Wales 
1961–1999), or by 5-year age groups (Denmark, Norway, England, and Wales 
1950–1960), with the maximum age ranging from 85+ to no maximum age.

For those aged 80 and over we had additional data on total mortality and 
population data available from the Kannisto-Thatcher Database on Old Age 
Mortality (K-T Database) at the Max Planck Institute for Demographic Research 
(http://www.demogr.mpg.de/databases/ktdb) [2 ]. The data were grouped by both 
year of death and year of birth of the deceased and by single year of age (no 
maximum age). We used these data (1) to redistribute the population numbers and 
deaths in the older age groups over 5-year age groups up to 100+, (2) to 
redistribute the population numbers and total deaths for those aged 80 and over 
from 5-year age groups into data by single year of age, and (3) to check 
whether the population and mortality data from the different data sources 
(Kannisto-Thatcher Database and national data) were consistent.

For the causes of death, we obtained data on their prevalence as the underlying 
cause of death. These data were available by three-digit codes, 5-year age 
groups, sex, and year of death for all countries. The maximum age ranged from 
85+ for England and Wales and The Netherlands (until 1969), and 90+, 95+, and 
100+ for all other countries and periods. The deaths per cause in the older age 
groups were redistributed over 5-year age groups up to 100+ based on the 
distribution of total mortality in these age groups. Table 1 lists the causes 
of death we selected and their relative share in all-cause mortality among 
those aged 80 and over in 1995–1999. Cancers have been designated 
"smoking-related" if their population attributable risk was larger than 0.25 
(according to the American Cancer Study) [20].

Table 1. List of selected causes of death and their relative share in all-cause 
mortality in the period 1995–1999,a males and females, aged 80 and over
[Full Size Table]

DK = Denmark; E&W = England and Wales; FIN = Finland; F = France; NL = The 
Netherlands; NO = Norway; S = Sweden.

COPD = chronic obstructive pulmonary diseases.

2.2. Statistical analysis

We analyzed the mortality data by means of a (log-linear) Poisson regression 
model with linear splines. The dependent variable was the number of deaths, 
with the person-years at risk as offset variable. As independent variables, we 
used age (single year of age for total mortality and 5-year age groups for 
cause-specific mortality) and year of death. Spline functions divide the 
overall trend into a number of separate, adjacent segments [21 ]. In our 
analysis, we used five segments each covering a period of 10 years (1950–1959, 
1960–1969, 1970–1979, 1980–1989, 1990–1999). The analysis using splines thus 
yielded estimates of annual changes in mortality within each 10-year period, 
thereby taking into account the overall trend. Comparison of these five 
decade-specific rates of change enabled us to detect and quantify changes in 
the secular trend in mortality, such as a stagnation of the decrease in 
mortality for a specific country. "Stagnation" of the mortality decline was 
defined as either a leveling off of the mortality decline leading to small 
declines or a reversal into increasing mortality. All our analyses are 
conducted using SAS package version 8.

In addition, in Fig. 1 for total mortality, directly standardized observed and 
fitted mortality rates were calculated, using the total population of England 
and Wales in 1999 as the reference.


[Enlarge Image] (34K)
[Enlarge Image] (31K)

Fig. 1.. Trends in standardized observed and fitted all-cause mortality rates 
in seven countries, 1950–1999, aged 80 and over. (A) Males. (B) Females.

The use of 5-year age groups in the cause of death analyses was due to the 
restriction that these data were not available to us by single year of age. To 
evaluate to what extent a possible change over time in the distribution of 
deaths within a 5-year age group could affect our results, we compared the 
results for total mortality using data by 5 years of age with the results for 
total mortality using data by single year of age. Because this comparison 
generated virtually the same results, we expected that, although age patterns 
can differ for the specific causes of death, the bias when using the data by 
5-year age groups will be minimal.

2.3. Concordance

When analyzing trends in causes of death for a longer period, numerous 
revisions of the World Health Organization (WHO) International Classification 
of Diseases (ICD) have to be taken into account, because they can lead to 
biases in the trends. In our analysis we had to bridge four or five different 
ICD revisions per country. For this purpose we constructed a general 
concordance table in which the different three-digit codes for a specific cause 
of death in successive ICD revisions were linked (see Janssen et al. [14] for 
more information; see also Appendix 1 ). To accurately bridge the revision from 
ICD6/7 to ICD8 for ischemic heart disease we obtained the numbers of death for 
one additional four-digit code (422.1) under ICD6/7. These numbers were not 
available for Finland until 1963, and Sweden until 1961. We estimated them on 
the basis of the ratio of the number of deaths from ischemic heart diseases 
with and without 422.1 calculated for the first year in which 422.1 was coded.

On the basis of this general concordance table, data on the required 
three-digit codes were obtained from the different countries. For Finland, 
however, not all causes of death were available by three-digit code and part of 
the data we had to request by short list consisting of slightly different 
groups of causes of death. In depth analysis of the Finnish mortality trends 
suggested that this approximation has not led to irregularities.

Cause-specific trends based on the concordance table can, however, still 
contain irregularities due to (1) remaining problems with ICD revisions, 
because at the level of three-digit codes the continuity of the medical content 
of some causes of death could not always be optimized; (2) incidental changes 
in coding rules, for example, changes within ICD revisions in the reporting of 
causes of death by physicians or in coding rules applied at the statistical 
offices; and (3) incidental outliers, that is, causes of death with a single 
year of exceptional mortality levels. In addition to the use of the concordance 
table, it was therefore necessary to trace these irregularities and to control 
for them when assessing long-term trends in mortality.

We identified outliers and incidental changes in coding rules on the basis of 
visual analysis of cause-specific trends for males and females combined for 
those aged 60 and over, and using country-specific background information on 
the irregularities observed. To evaluate the existence of remaining problems 
with ICD revisions, we identified possible mortality jumps due to the ICD 
revisions, using cause-specific regression models (with splines) applied to 
data for males and females combined, aged 60 and over. To these regression 
models we added transition variables indicating the ICD revisions, that is, 
ICD6/7to8, ICD8to9 or ICD9to10. In this way these regression models generated 
parameter estimates for the transition variables.

A transition variable was included in our final regression model if: (1) the 
parameter estimate corresponding to this variable was statistically 
significant, (2) the significant effect could not be attributed to nonlinear 
trends or to a single outlier, for instance, an influenza epidemic nearby; and 
(3) the observed effect could be traced back to a ICD transition problem, for 
example, due to a four-digit code not included in the concordance table or an 
effect on one cause of death mirrored by an opposite effect on a complementary 
cause of death.

For France, no check on the concordance was needed, because J. Vallin and F. 
Meslé reconstructed coherent series of data for causes of death, the results of 
which are available from a Web site 
(http://matisse.ined.fr/%7Etania/causfra/data/) [22 and 23]. From this Web site 
the cause of death data we distinguished could be extracted using ICD9 codes.

3. Results

Old-age mortality showed an overall decline and a convergence in the mortality 
level between countries over time (Fig. 1 ). However, there was a large 
heterogeneity in the pace of decline, with periods of stagnation (defined as 
either a leveling off of the mortality decline leading to small declines or a 
reversal into increasing mortality) being widespread. In the 1950s, and to a 
lesser extent in the 1960s, mortality in the Nordic countries declined slightly 
or even increased. From the 1980s onwards, small declines or even increases 
were observed in Denmark, The Netherlands, and among Norwegian males. In 
contrast, mortality decline continued in England and Wales (females), and 
especially in France, resulting in the lowest mortality level in the 1990s. 
Striking is the huge mortality decline in Finland in the 1970s, which was 
followed by a much more modest decline in the 1980s. The mortality rates were 
higher among males compared to females, with the mortality level of males in 
the 1990s being about equal to the mortality level of females in the 1950s.

To identify the periods of stagnation more precisely, we quantified the pace of 
all-cause mortality decline within each of the 5 decades by means of annual 
changes (%) (Table 2 ). If the confidence intervals for successive decades do 
not overlap the changes in the pace of mortality decline are statistically 
significant. Stagnation was more widespread among males compared to females. In 
Denmark, mortality decline leveled off from the 1970s onwards for males, and 
from the 1980s onwards for females. Mortality decline among Dutch males turned 
into increase in the 1980s, followed by a small decline in the 1990s. Among 
Dutch females mortality decline levelled off in the 1980s and 1990s. Among 
Norwegian males the small (nonsignificant) increase in the 1980s changed into a 
modest decline in the 1990s.

Table 2. Annual trends in total mortality by country and decade (1950-1999), 
males and females, aged 80 and over
[Full Size Table]

Bold indicates a significant increase.

Underlined indicates annual changes (%)??0.60.

The annual trends for the "smoking-related diseases," that is, smoking-related 
cancers and chronic obstructive pulmonary disease (COPD), showed an overall 
pattern of long-term increases (Table 3 ). Due to this overall increase these 
diseases had an increased share in total mortality, and thus their trends had 
increasingly more effect on the trends observed for all-cause mortality. Since 
the 1980s, mortality from smoking-related diseases among men decreased in 
Finland and France, whereas in Denmark, The Netherlands, and Norway in the 
1980s mortality still increased rapidly. However, only Norway showed a clear 
persistence of the increase among males in the 1990s. For females, recent 
increases are more frequent, and are most pronounced in Norway, Denmark, and 
The Netherlands. Overall, lung cancer reveals the strongest increases, but COPD 
showed a more unfavorable trend in the 1990s.

Table 3. Annual trends in mortality from smoking-related diseases by country 
and decade (1950–1999), males and females, aged 80 and over
[Full Size Table]

Bold indicates an increase.

COPD = chronic obstructive pulmonary disease.

Mortality from cardiovascular diseases declined, at least after the 1960s 
(Table 4 ). During the last 2 decades, declines were largest in France and 
England and Wales and smallest in Norway and The Netherlands. Trends in 
mortality from ischemic heart disease (IHD) were least favorable with increases 
in mortality in the 1950s and 1960s. Mortality from IHD also increased in the 
1980s in Norway (males) and Finland. Mortality from cerebrovascular diseases 
increased in the 1950s, followed by a sustained decrease in mortality. Striking 
were the enormous decreases in mortality from stroke in France since the 1980s. 
For both IHD and stroke, the mortality decrease in France sets in later 
compared to the other countries. Mortality from "other cardiovascular diseases" 
decreased, especially in Finland and England and Wales, whereas in Denmark and 
Norway mortality increased since the 1980s.

Table 4. Annual trends in mortality from cardiovascular diseases by country and 
decade (1950–1999), males and females, aged 80 and over
[Full Size Table]

Bold indicates an increase.

For the "other causes of death," that is, total mortality minus 
"smoking-related diseases" and all cardiovascular diseases, mortality since the 
1980s increased in all countries, except France (Table 5 ). The increases are 
most pronounced in Denmark and The Netherlands, in the 1990s. Especially 
mortality from diseases specifically related to old-age (infectious diseases, 
pneumonia, dementia, and ill-defined conditions) increased substantially. 
Differences between the countries, however, exist in which of these causes 
showed the largest mortality increases. In France, the increase in diseases 
specifically related to old age has been offset by the continuation of the 
strong decline for "other diseases." Caution, however, should be exercised when 
interpreting the annual changes for causes of death with a low mortality level 
in 1980 and 1990, especially infectious diseases, diabetes mellitus, dementia 
(males), and all ill-defined causes (in England and Wales, Finland, and 
Sweden). Annual changes for such causes tend to be large, even though the 
change is small in absolute terms.

Table 5. Annual trends in mortality from "other causes of death" by country and 
decade (1980–1999), males and females, aged 80 and over
[Full Size Table]

Bold indicates an increase.

4. Discussion

This article provides important new findings on the trends in old-age mortality 
(80+) in seven low-mortality countries. First, although old-age mortality tends 
to decline and to converge, there is large heterogeneity in the pace of decline 
in the countries under investigation, with periods of stagnation being 
widespread. Since the 1980s, stagnation was observed in Denmark, The 
Netherlands, and Norway, whereas England and Wales and especially France showed 
a continuation of a strong mortality decline. Second, the long-term mortality 
increase for smoking-related cancers and COPD turned into declines since the 
1980s among males. Only for Norway the mortality increase among males clearly 
persisted. Third, mortality from cardiovascular diseases showed clear 
crossnational variations in the general decline since the 1970s. Fourth, 
mortality from diseases specifically related to old age (infectious diseases, 
pneumonia, dementia, and ill-defined conditions) increased recently in all 
countries except France.

Previous studies on trends in old-age mortality also report some heterogeneity 
in the speed of the mortality decline, but found only rare periods of 
stagnation [2 and 15]. The overall conclusion reached in most of these studies 
therefore has been a continued decline in old-age mortality [2, 4, 15, 24 and 
25 ]. The present study, however, shows that periods of stagnation are more 
frequent than previously observed. This conclusion could be reached by, on the 
one hand, extending the observation period to the 1990s, and on the other hand, 
by looking in more detail at the trends within specific decades in stead of 
broader periods. Our results thus show a more differentiated picture than 
suggested by previous analyses on old-age mortality.

The heterogeneity between countries and periods as observed in this analysis 
and in previous analyses suggests that the findings obtained in this study may 
not be considered to be representative of other countries nor of earlier 
periods.

4.1. 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 [1 and 2]. 
Reported survivorship counts are highly accurate [2 and 26 ]. Comparison of our 
mortality data to the mortality data from the Kannisto-Thatcher Database—in 
which the data was checked for age-heaping and were subjected to a number of 
checks for plausibility—showed only small discrepancies, that had no 
appreciable effects on our results.

Analyses of long-term trends in causes of death have to deal with several 
transitions between ICD revisions. Considerable effort was made to bridge these 
ICD revisions by carefully constructing a concordance table and by controlling 
for the remaining biases due to these revisions in our regression model. 
Further checks showed minimal sensitivity. Thus, even though some residual 
effects of problems with ICD revisions could not be excluded, we expect that 
these problems do not affect the results to any substantial extent.

More difficult to tackle were changes within an ICD revision, either in coding 
practices at the Statistical Offices or in the reporting of causes of death by 
physicians. Although all countries under investigation use the coding rules of 
the WHO, the national statistical offices might change, over time, their 
interpretation of these international coding rules. Because changes in coding 
practices most probably led to quite abrupt and sometimes temporal changes, the 
changes could be traced and controlled for. Changes in the reporting of causes 
of death on the death certificate by physicians, however, result in more 
gradual shifts that could not be controlled for. The causes of death expected 
to suffer substantial effects are diabetes mellitus, dementia, and "all 
ill-defined causes." The huge increases for dementia that were observed 
especially in the 1980s could partly be explained by a growing propensity of 
physicians to report dementia as an underlying cause of death. The recent 
increases in "all ill-defined causes" (especially in England and Wales, 
Denmark, Sweden, and The Netherlands) could be the result of less detailed and 
less accurate diagnosis and reporting by physicians. An increase in the 
doctor's tendency to list only one cause on the death certificate could perhaps 
explain the huge increases observed for pneumonia in The Netherlands and 
Denmark in the 1990s, because diseases that were formerly reported mainly as 
secondary causes of death could increasingly be listed as primary causes of 
death. The more stable trends for diabetes mellitus seem to be less affected by 
this coding problem.

Thus, especially the recent increases found for dementia, "all ill-defined 
causes" and pneumonia should be viewed with caution. Moreover, one should be 
aware of the opposite effects these increases in mortality might have on other 
diseases. For example, an increasing tendency to report these diseases as the 
underlying cause of death may result in an overestimation of the recent 
decreases observed for cardiovascular diseases. However, this cannot account 
for the recent declines observed for cardiovascular diseases, because in 
absolute terms, the decline in cardiovascular diseases was much larger than the 
increases for diseases prone to changes in reporting causes of death (data not 
shown). Diseases that have a more straightforward diagnosis, like 
smoking-related cancers and likely COPD as well, are probably relatively 
resistant to changes in coding practices.

4.2. Explanations of the trends observed

Our findings suggest both favorable trends in old-age mortality (e.g., the 
continued decline in England and Wales, and especially France, and the huge 
decline in Finland in the 1970s), but also unfavorable trends (e.g., stagnation 
since the 1980s in Denmark, The Netherlands, and to a lesser extent Norway). In 
this section, we will discuss possible explanations of these trends.

Although many smokers die already before the age of 80, our results indicate 
that smoking is a possible determinant of mortality trends even among the 
oldest old. Smoking-related cancers and COPD contributed to the recent 
stagnation, especially in Norway, for which mortality from these causes of 
death increased recently. The absence of clear increases in Denmark and The 
Netherlands in the 1990s suggests that the contribution of smoking-related 
diseases to the recent stagnation as observed in The Netherlands and Denmark 
was more modest. However, by looking merely at the trends in these 
smoking-related diseases, other diseases, of which smoking is a risk factor as 
well, would be neglected. The recent declines observed for cardiovascular 
diseases—that are not correlated with the trends for lung cancer—suggest that 
smoking did not contribute much to the recent stagnation as observed in The 
Netherlands and Denmark. On the contrary, the significant decline in smoking 
prevalence among elderly men in recent periods [27 ], suggests a favorable 
effect of smoking on recent trends in mortality from cardiovascular diseases 
(and other causes of death), for which current smoking levels are more 
important than life-time smoking exposures [28 ]. Thus, although smoking has 
had a marked influence on the trends in old-age mortality, the role of smoking 
in the stagnation since the 1980s seems only modest and restricted to Norway.

Another explanation for the recent stagnation in Denmark, The Netherlands, and 
Norway could be that further improvement in old-age mortality in these 
countries is no longer possible, due to low levels of mortality already 
attained and a limit to life expectancy being approached. However, we observed 
further improvements in old-age mortality in both England and Wales and France, 
who had reached the same low level of mortality as Denmark, The Netherlands, 
and Norway around the 1980s. Life expectancy at age 80 in 1980–1990 was 
slightly higher in France (7.52) compared to Denmark (7.43) and Norway (7.49), 
suggesting that the limit to life expectancy is not yet being approached in the 
latter countries [29].

Mortality trends were unfavorable for infectious diseases, pneumonia, dementia, 
and ill-defined causes of death. The recent increase of mortality from these 
causes of death—which were not only restricted to Denmark and The Netherlands 
but appeared in all countries except France—could be partly artificial due to 
changes in the coding and reporting of causes of death (as discussed above). 
However, part of the recent increase in these causes, which are specifically 
related to old age, might be real, and perhaps due to increased frailty. 
Increased comorbidity or increased frailty among elderly in itself could easily 
lead to an increased occurrence of symptoms or comorbid conditions as causes of 
death. Increased frailty could be due to decreased mortality selection as a 
result of the declines in all-cause mortality and cardiovascular mortality in 
earlier periods and ages, and indirectly due to improvements in medical care. 
The increasing proportions of elderly people surviving to old ages might be 
expected to be less healthy compared to their more selected predecessors [30], 
resulting in increases in (cardiovascular) morbidity [31 ] and in mortality 
increases from diseases specifically related to old age. Although recent 
studies on disability among the elderly suggest that disability among the 
elderly is declining [32, 33 and 34 ], we still think that increased frailty 
might have played a role in explaining the recent increases in diseases 
specifically related to old age.

Another important finding from our study is the crossnational variations in the 
pace of decline in cardiovascular disease, which is still the most important 
cause of death among the elderly. In the last 2 decades declines for 
cardiovascular disease mortality were highest in England and Wales and France 
and lowest in Norway, The Netherlands, and Denmark (1980s only). Previous 
research on trends in cardiovascular mortality [35, 36 and 37 ] suggested the 
importance of several factors like physical activity, hypertension control, 
diet, smoking, and accessibility of medical care. It might be possible that the 
developments in some of these risk factors were more beneficial in England and 
Wales and France compared to the other countries.

Strikingly favorable were the trends in old-age mortality in France and Finland 
(1970s). France showed the strongest decline for cardiovascular diseases, and, 
contrary to the other countries, did not witness an increase from "other causes 
of death." Within "other causes of death" the declining trend for "other 
diseases" is striking. The latter trend seems the result of the reported 
declines of acute respiratory diseases and digestive diseases among the French 
elderly [38 ]. Changes in alcohol consumption might be one of the factors 
involved. The sharp decline in all-cause mortality in the 1970s in Finland was 
also observed among those aged 60 and over [39], and seems the result of rapid 
economic and social progress in this period [39, 40 and 41 ]. Together with the 
development of a national health and social care system, this has resulted in 
improvements in the living conditions and health care, also for the elderly [41 
]. In addition, a favorable effect of rapid declines in behavioral risk 
factors, like diet, blood pressure, and serum cholesterol levels, in part 
caused by preventive campaigns, can be expected [42 ]. In Finland, old-age 
mortality thus seems to have been highly susceptible to improvements in living 
conditions and in both preventive and secondary health care.

4.3. Implications

Our finding that periods of stagnation were widespread both in the 1950s and 
since the 1980s, challenges the idea that mortality at old age is bound to 
decline steadily in the near future [6, 8 and 11 ]. Old-age mortality seems 
highly plastic and susceptible to many factors, both favorable and unfavorable. 
These factors should be taken into account when making projections of future 
old-age mortality and its implications on social and health care policies.

Acknowledgements

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). 
We are grateful to Jacques Vallin (INED, France), Martine Bovet (INSERM, 
France), Hillka Ahonen (Statfin, Finland), Annika Edberg (National Board of 
Health and Welfare, Sweden), Örjan Hemström (Sweden), Allan Baker and 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, and for giving useful information 
on national coding practices. We gratefully acknowledge James Vaupel and 
Vladimir Scholnikov (Max Planck Institute of Demographic Research) for the use 
of the Kannisto-Thather Database on old-age mortality. We thank Tapani 
Valkonen, Tujia Martelin, and France Meslé for their useful information on 
trends in respectively Finland and France.

References

1. V. Kannisto, J. Lauritsen, A.R. Thatcher and J.W. Vaupel, Reductions in 
mortality at advanced ages: several decades of evidence from 27 countries. 
Popul Dev Rev 20 4 (1994), pp. 793–810. Abstract-GEOBASE | Abstract-EconLit   | 
$Order Document

2. V. Kannisto, Development of oldest-old mortality, 1950–1990: evidence from 
28 developed countries. , Odense University Press, Odense, Denmark (1994).

3. A.R. Thatcher, Trends in numbers and mortality at high ages in England and 
Wales. Popul Stud 46 (1992), pp. 411–426. Abstract-GEOBASE | Abstract-EconLit | 
Abstract-MEDLINE   | $Order Document | Full Text via CrossRef

4. J.W. Vaupel, The average French baby may live 95 to 100 years. In: J.-M. 
Robine, J. Vaupel, B. Jeune and M. Allard, Editors, Longevity: to the limits 
and beyond, Springer-Verlag, Berlin (1997).

5. J.F. Fries, Aging, natural death, and the compression of morbidity. N Engl J 
Med 303 3 (1980), pp. 130–135. Abstract-EMBASE | Abstract-MEDLINE | $Order 
Document

6. K.G. Manton, E. Stallard and H.D. Tolley, Limits to human life expectancy: 
evidence, prospects, and implications. Popul Dev Rev 17 (1991), pp. 603–637. 
Abstract-EconLit   | $Order Document

7. J.W. Vaupel and H. Lundstrom, The future of mortality at older ages in 
developed countries. In: W. Lutz, Editor, The future population of the world. 
What can we assume today?, International Institute for Applied Systems Analysis 
[IIASA], Laxenburg, Austria (1994), pp. 295–315.

8. J. Oeppen and J.W. Vaupel, Demography. Broken limits to life expectancy. 
Science 296 5570 (2002), pp. 1029–1031. Abstract-EMBASE | Abstract-GEOBASE | 
Abstract-Elsevier BIOBASE | Abstract-MEDLINE   | $Order Document | Full Text 
via CrossRef

9. S.J. Olshansky, B.A. Carnes and C. Cassel, In search of Methuselah: 
estimating the upper limits of human longevity. Science 250 (1990), pp. 
634–640. Abstract-EMBASE | Abstract-GEOBASE | Abstract-MEDLINE   | $Order 
Document

10. S.J. Olshansky, B.A. Carnes and A. Desesquelles, Demography. Prospects for 
human longevity. Science 291 5508 (2001), pp. 1491–1492. Abstract-EMBASE | 
Abstract-Elsevier BIOBASE | Abstract-MEDLINE   | $Order Document | Full Text 
via CrossRef

11. J.W. Vaupel, J.R. Carey, K. Christensen, T.E. Johnson, A.I. Yashin, N.V. 
Holm, I.A. Iachine, C. Kannisto, A.A. Khazaeli, P. Liedo, V.D. Longo, Y. Zeng, 
K.G. Manton and J.W. Cartsinger, Biodemographic trajectories of longevity. 
Science 280 5365 (1998), pp. 855–860. Abstract-EMBASE | Abstract-Elsevier 
BIOBASE | Abstract-MEDLINE   | $Order Document | Full Text via CrossRef

12. J.R. Wilmoth, The future of human longevity: a demographer's perspective. 
Science 280 5362 (1998), pp. 395–397. Full Text via CrossRef

13. W.J. Nusselder and J.P. Mackenbach, Lack of improvement of life expectancy 
at advanced ages in The Netherlands. Int J Epidemiol 29 1 (2000), pp. 140–148. 
Abstract-EMBASE | Abstract-GEOBASE | Abstract-Elsevier BIOBASE | 
Abstract-MEDLINE   | $Order Document | Full Text via CrossRef

14. F. Janssen, W.J. Nusselder, C.W.N. Looman, J.P. Mackenbach and A.E. Kunst, 
Stagnation in mortality decline among elders in The Netherlands. Gerontologist 
43 (2003), pp. 722–734. Abstract-MEDLINE | Abstract-PsycINFO | $Order Document

15. G. Caselli, Future longevity among the elderly. In: G. Caselli and A. 
Lopez, Editors, Health and mortality among elderly populations, Clarendon 
Press, Oxford (1996), pp. 235–265.

16. G.C. Myers, Comparative mortality trends among older persons in developed 
countries. , Clarendon Press, Oxford (1996).

17. T. Valkonen and F. van Poppel, The contribution of smoking to sex 
differences in life expectancy. Four Nordic countries and The Netherlands 
1970–1989. Eur J Public Health 7 (1997), pp. 302–310. Abstract-EMBASE   | 
$Order Document

18. J.J. Barendregt, C.W.N. Looman and H. Brønnum-Hansen, Comparison of cohort 
smoking intensities in Denmark and the Netherlands. Bull World Health Organ 80 
1 (2002), pp. 26–32. Abstract-EMBASE | Abstract-GEOBASE | Abstract-MEDLINE   | 
$Order Document

19. A.D. Lopez, N.E. Collishaw and T. Piha, A descriptive model of the 
cigarette epidemic in developed countries. Tob Control 3 3 (1994), pp. 242–247.

20. N.J. Wald and A.K. Hackshaw, Cigarette smoking: an epidemiological 
overview. Br Med Bull 52 1 (1996), pp. 3–11. Abstract-EMBASE | Abstract-MEDLINE 
| $Order Document

21. D.R. McNeil, T.J. Trussell and J.C. Turner, Spline interpolation of 
demographic data. Demography 14 2 (1977), pp. 245–252. Abstract-EconLit | 
Abstract-MEDLINE   | $Order Document

22. J. Vallin and F. Meslé, Les causes de décès en France de 1925 à 1978. , 
INED, PUF, Paris (1988).

23. J. Vallin and F. Meslé, Comment suivre l'evolution de la mortalite par 
cause malgre les discontinuites de la statistique? Le cas de la France de 1925 
a 1993. , INED, Paris (1996).

24. H. Kesteloot, Evolution of all-cause and cause-specific mortality in the 
age-class 75–84 years during the period 1970–1996. A worldwide overview. Verh K 
Acad Geneeskd Belg 63 5 (2001), pp. 405–430 (discussion, 431) .

25. H. Kesteloot, S. Sans and D. Kromhout, Evolution of all-causes and 
cardiovascular mortality in the age-group 75–84 years in Europe during the 
period 1970–1996; a comparison with worldwide changes. Eur Heart J 23 5 (2002), 
pp. 384–398. Abstract-EMBASE | Abstract-MEDLINE   | $Order Document

26. A.G. Condran, C.L. Himes and S.H. Preton, Old-age mortality patterns in 
low-mortality countries: an evaluation of population and death data at advanced 
ages, 1950 to the present. Popul Bull UN 30 (1991), pp. 23–60.

27. B. Forey, J. Hamling, P. Lee and N. Wald, Editors, International smoking 
statistics. A collection of historical data from 30 economically developed 
countries, Wolfson Institute of Preventive Medicine, Oxford University Press, 
London, Oxford (2002).

28. R. Doll, R. Peto, K. Wheatley, R. Gray and I. Sutherland, Mortality in 
relation to smoking: 40 years' observations on male British doctors. BMJ 309 
6959 (1994), pp. 901–911. Abstract-EMBASE | Abstract-Elsevier BIOBASE | 
Abstract-MEDLINE   | $Order Document

29. V. Kannisto, The advancing frontier of survival: life tables for old age. , 
Odense University Press, Odense, Denmark (1996).

30. E. Grundy, Demography and gerontology: mortality trends among the oldest 
old. Ageing Soc 17 6 (1997), pp. 713–725. Full Text via CrossRef

31. L. Bonneux, J.J. Barendregt, K. Meeter, G.J. Bonsel and P.J. van der Maas, 
Estimating clinical morbidity due to ischemic heart disease and congestive 
heart failure: the future rise of heart failure. Am J Public Health 84 1 
(1994), pp. 20–28. Abstract-EMBASE | Abstract-MEDLINE   | $Order Document

32. R.F. Schoeni, V.A. Freedman and R.B. Wallace, Persistent, consistent, 
widespread, and robust? Another look at recent trends in old-age disability. J 
Gerontol B Psychol Sci Soc Sci 56 4 (2001), pp. S206–S218. Abstract-EMBASE | 
Abstract-MEDLINE | Abstract-PsycINFO   | $Order Document

33. D.M. Cutler, Declining disability among the elderly. Health Aff (Millwood) 
20 6 (2001), pp. 11–27. Abstract-MEDLINE   | $Order Document | Full Text via 
CrossRef

34. I. Winblad, M. Jaaskelainen, S.L. Kivela, P. Hiltunen and P. Laippala, 
Prevalence of disability in three birth cohorts at old age over time spans of 
10 and 20 years. J Clin Epidemiol 54 10 (2001), pp. 1019–1024. SummaryPlus | 
Full Text + Links | PDF (68 K)

35. C. Sarti, D. Rastenyte, Z. Cepaitis and J. Tuomilehto, International trends 
in mortality from stroke, 1968 to 1994. Stroke 31 7 (2000), pp. 1588–1601. 
Abstract-EMBASE | Abstract-Elsevier BIOBASE   | $Order Document

36. K. Kuulasmaa, H. Tunstall-Pedoe, A. Dobson, S. Fortmann, S. Sans, H. 
Tolonen, A. Evans, M. Ferrario and J. Tuomilehto, Estimation of contribution of 
changes in classic risk factors to trends in coronary-event rates across the 
WHO MONICA Project populations. Lancet 355 9205 (2000), pp. 675–687. 
SummaryPlus | Full Text + Links | PDF (166 K)

37. R. Cooper, J. Cutler, P. Desvigne-Nickens, S.P. Fortmann, L. Friedman, R. 
Havlik, G. Hogelin, J. Marler, P. McGovern, G. Morosco, L. Mosca, T. Pearson, 
J. Stamler, D. Stryer and T. Thom, Trends and disparities in coronary heart 
disease, stroke, and other cardiovascular diseases in the United States: 
findings of the national conference on cardiovascular disease prevention. 
Circulation 102 25 (2000), pp. 3137–3147. Abstract-EMBASE | Abstract-Elsevier 
BIOBASE | Abstract-MEDLINE   | $Order Document

38. F. Meslé and J. Vallin, Mortality trends at older and oldest ages in France 
since 1950 (Evolution de la mortalite aux ages eleves en France depuis 1950). 
Dossiers et Recherches 68. , Institut National d'Etudes Demographiques [INED], 
Paris (1998).

39. T. Martelin, Trends in elderly mortality in the Nordic countries. Comp 
Gerontol [C] 1 (1987), pp. 39–48. Abstract-MEDLINE   | $Order Document

40. J.M. Sverre, A comparative study of trends in mortality rates of the ageing 
population in Norway, Sweden, Denmark, and Finland, 1966–1986. Scand J Soc Med 
23 4 (1995), pp. 227–232. Abstract-EMBASE | Abstract-MEDLINE   | $Order 
Document

41. S.L. Kivela, Changes in mortality among the elderly Finnish population 
1951–1979. Soc Sci Med 21 7 (1985), pp. 799–805. Abstract | Abstract + 
References | PDF (684 K)

42. E. Vartiainen, P. Puska, J. Pekkanen, J. Tuomilehto and P. Jousilahti, 
Changes in risk factors explain changes in mortality from ischaemic heart 
disease in Finland. BMJ 309 6946 (1994), pp. 23–27. Abstract-EMBASE | 
Abstract-MEDLINE   | $Order Document

Appendix I. The concordance table used for bridging five revisions of the 
International Classification of Diseases (ICD)

[Corresponding Author Contact Information] Corresponding author. Tel.: 
+31-(0)10-4087714; fax: +31-(0)10-4089449

1 The Netherlands Epidemiology and Demography Compression of Morbidity research 
group, which also includes J. Barndregt, L. Bonneux, C. de Laet, W. Nusselder, 
A. Peeters, A. Al Mamun, and F. Willekens.