© International Epidemiological Association 1999
International Journal of Epidemiology 1999;28:1110–1118
Printed in Great Britain
Trends in dementia-free life expectancy
among elderly members of a large health
maintenance organization
Catherine Sauvaget,a Ichiro Tsuji,a Mary N Haanb and Shigeru Hisamichia
Background This study examined the secular trends of life expectancy without dementia
among elderly American members of a health maintenance organization, and
observed if an increased life expectancy is accompanied by an increase in the
duration of life with dementia.
Methods
The data derived from two chronological 9-year prospective cohort studies of
members of the Kaiser Permanente Medical Care Program of Northern California.
The first and second cohorts included 2702 and 2926 people aged >65 years free
from dementia at baseline. Life expectancy without dementia or dementia-free
life expectancy (DemFLE) is defined as the average number of years a person
is expected to live without dementia. Total life expectancy is equal to the sum of
DemFLE and life expectancy with dementia. Estimations of DemFLE were based
on mortality data and incidence of dementia, using double-decrement life tables.
Results
Between the first and second cohorts, all-cause mortality rates declined, while
the incidence of dementia remained constant in both men and women. Among
the males, total life expectancy increased at a higher rate than DemFLE. Consequently, the duration of life with dementia was extended in the second cohort.
Conversely, among the females DemFLE increased at a higher rate than total life
expectancy, thus the duration of life with dementia decreased in the second cohort.
The median age of dementia onset was postponed by 2–3 years in the second
cohort for females, and did not show any specific difference between the two
cohorts in males.
Conclusion
The trends of health expectancies suggest an extension of the duration of life
with dementia for males and a compression of dementia for females. A decreased
incidence of risk factors for dementia among females in the second cohort such
as stroke may explain these trends.
Keywords
Health expectancy, dementia, incidence, compression and extension of morbidity
Accepted
29 April 1999
In recent decades, a decline in the mortality rate in the US has
led to an increase in life expectancy. With a rapid extension of
life expectancy, the number of elderly people suffering from
dementing illnesses is escalating. Cognitive impairment and
mental diseases are among the most common morbid conditions in later life. The prevalence of dementia among elderly
Americans aged >65 years is estimated to be between 5% and
a Department of Public Health, Tohoku University School of Medicine,
Sendai, Japan.
b Department of Epidemiology and Preventive Medicine, University of
California School of Medicine, Davis, CA, USA.
Reprint requests to: Catherine Sauvaget, Department of Public Health,
Tohoku University School of Medicine, Sendai 980–8575, Japan. E-mail:
sauvaget@mail.cc.tohoku.ac.jp
10%.1–8 The number of elderly people suffering from dementia
and living in the US was estimated at 3.1 million people in
1990; this number is projected to increase 2.5- to 4-fold by the
year 2040.9
The question that arises is whether longevity is a reward.
Nobody would appreciate the added years of life if they were
accompanied by disability or dementia. In order to answer this
question, a new health indicator that takes both morbidity
and mortality into consideration has emerged.10 Healthy life
expectancy is defined as the duration an individual at a given
age is expected to live without any specific morbid condition. A
comparison of healthy life expectancies between different times
allows us to appraise the quality of increased years of life. Three
scenarios are possible. The first argues that life expectancy increases faster than healthy life expectancy, thus the duration to
1110
TRENDS IN DEMENTIA-FREE LIFE EXPECTANCY
be spent with poor health increases. At an extreme limit, life
expectancy increases while healthy expectancy remains the
same; this figure leads to ‘the pandemic of mental disorders and
chronic diseases’.11 Second, a more optimistic concept implies
a stagnation in life expectancy concomitant with an increase of
healthy expectancy; this theory is known as ‘the compression of
morbidity’.12,13 The third pattern, called ‘dynamic equilibrium’,
describes a stability of morbidity and health status levels, thus life
expectancy and healthy expectancy increase at the same rate.14
Dementia-free life expectancy (DemFLE) is defined as the
average number of years an individual at a given age is expected
to live without dementia. This estimation is based on mortality
data and on the prevalence or incidence of dementia. To date,
DemFLE has been estimated for European countries, Australia
and Japan.15–20 In no previous studies has DemFLE been
measured on similar populations at different periods. Thus, it
is still unknown whether an increased life expectancy is associated with an extension or a compression of the duration to be
lived with dementia.
In the study reported here, data were derived from two
chronological cohorts of subjects aged >65 years. These subjects
were all members of the Kaiser Permanente Medical Care Program of Northern California (KPMCP-NC), one of the largest
prepaid health maintenance organizations in the US.21 The
KPMCP-NC provides comprehensive medical services to more
than two million members of all ages. In order to analyse trends
in life expectancies free of dementia, two cohorts of approximately
3000 elderly subjects each were followed for 9 years. For both
cohorts in 1971 and 1980, all-cause mortality and onset of
dementia were recorded. Identical study designs and methods
were applied to each cohort. Estimations of DemFLE were based
on mortality and the incidence of dementia during the followup period.
The objective of the study is to examine the changes over
time of DemFLE among two populations of elderly Americans
members of the KPMCP-NC, and to observe whether an increase
in longevity is accompanied by an increase in the duration to be
spent with dementia, i.e. a ‘longer life associated with worsening health’.22
Method
Data collection
The study design has been published elsewhere.23 In brief, the
data were derived from medical records of two cohorts of elderly
Americans who were members of the KPMCP-NC between
1971 and 1979, and between 1980 and 1988. Members of the
first cohort were >65 years in 1971, and those of the second
cohort were >65 years in 1980. The 1971 cohort included 2877
people, while the 1980 cohort included 3113 people. The sample
selection was stratified to attain equal numbers of observations
(1000 in each) in three sex-age cells for each cohort: 65–69,
70–79, >80 years. Outpatient records of the study subjects were
abstracted for 9 years from the baseline year (January 1971 and
January 1980).
A mortality search was performed using the California Automated Mortality Linkage (CAMLIS). The CAMLIS uses vital
statistics tapes obtained from the California Department of Health
Services to produce a weighted value that reflects the level
of probability that a subject’s death and vital statistics match.
1111
All borderline values were reviewed individually to determine
whether the match was correct. During the follow-up period,
onset of dementia, death, or withdrawal were recorded. In the
1971 and 1980 cohorts, 7.1% and 8.6% of the subjects withdrew from the membership of KPMCP-NC before the end of the
follow-up period. Comparison of the withdrawal populations
in the two cohorts did not show any significant differences in
terms of sex, age, race, and education level.
Calculation of incidence of dementia and mortality
Incident cases of dementia were identified by chart review. The
KPMCP-NC did not perform a screening programme to detect
dementia. In order to be diagnosed as having dementia, patients
had to visit clinics. Under these circumstances, less severe cases
of dementia or cases referred directly to a nursing home could
have been missed, and thus the number of cases could be underestimated. Diagnosis was made by a primary care physician
when the person visited the KPMCP-NC and, in most cases, was
confirmed by a neurologist. Dementia was diagnosed according
to the International Classification of Diseases-9th revision (ICD-9)
criteria. Incident cases of dementia were defined as those
diagnosed after the baseline date. Severity of dementia was not
considered in this study since such information was not available for some cases.
Cases of dementia present at or before baseline were considered as prevalent cases. Prevalent dementia was identified by
reviewing the medical charts for 4 years before the beginning of
the follow-up period. In the first cohort, on 1 January 1971, 175
prevalent cases of dementia were identified and excluded from
the analyses; for the second cohort, on 1 January 1980, 187
prevalent cases were excluded. The 1971 cohort included 2702
subjects (males: 1320, females: 1382) who were free of dementia
at baseline, and the 1980 cohort included 2926 such subjects
(males: 1459, females: 1467).
The age- and sex-specific incidence rates of dementia and the
age- and sex-specific mortality rates during the follow-up period
were calculated with the person-years method. In the first
and second cohorts, the person-years at risk was the time from
either 1 January 1971 or 1 January 1980 to the earliest date for
incidence of dementia, death, withdrawal from KPMCP-NC, or
either to 31 December 1979 or 31 December 1988. In the 1971
cohort, 17 454 person-years were observed for the incidence of
dementia (males: 8090 person-years, females: 9364 personyears), and in the 1980 cohort, this figure was 19 387 personyears (males: 9306 person-years, female: 10 081 person-years).
Calculation of DemFLE and total life expectancy
For estimation of DemFLE, we performed a double-decrement
life table analysis, as proposed by Katz et al.24 The probability of
living free from dementia in the DemFLE calculations was based
on the following formula:
d
dm
1px = 1 – 1px – 1px
where:
1px = probability of living free from dementia during the
interval between age x and age x + 1
d
1px = probability of dying during the interval between age x
and age x + 1
dm = probability of becoming demented during the
1px
interval between age x and age x + 1
1112
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
The estimations of the total life expectancy (TLE) were based on
the probability of dying, according to the following formula:
d
1px = 1 – 1px
where:
1px = probability of living during the interval between age x
and age x + 1
d
1px = probability of dying during the interval between age x
and age x + 1
d
dm were obtained by the person-years method as
1px and 1px
described above. The probabilities were estimated for each year
of age and each sex.
Because the observed probabilities were somewhat unsteady
due to a smaller sample size at higher ages, a smoothing method
with logistic regression was used. In this method, the observed
age-specific probabilities were fit to the exponential function:
Mx = e(α + β ln x)
where Mx represents the predicted probability at age x, α and
β the regression coefficients, and ln x the natural logarithm of
age x. In the data presented here, all regression coefficients
were significant (P , 0.001) and R2 of each model was between
0.84 and 0.91.
The DemFLE corresponds to the number of years a person
can expect to live without dementia. Dementia life expectancy
(DemLE) is defined as the time to be spent with dementia. The
TLE is the sum of these two durations.
A survival analysis was applied to the data. The association
between dementia onset and potential risk factors of dementia
was examined using the Cox proportional hazards regression
model. The dependent variable was the number of years from
baseline of each cohort (1971 and 1980) to the point of the
development of dementia. Independent variables included age,
sex, education level, smoking status, and prevalence of specific
medical conditions, respectively.
Results
Table 1 presents the characteristics of the study subjects according to the cohort. Sex, age distribution, and education level
were similarly distributed in the two study populations. The
frequency of outpatient visits did not differ by cohorts. However,
Whites were significantly less represented in the second cohort
while Blacks were more numerous (χ2, P , 0.001). Prevalence
and incidence of dementia in the 1980 cohort did not show
any significant difference between the two races (prevalence
of dementia in Whites: 6.5%, Blacks: 4%, χ2 P = 0.077;
incidence of dementia in Whites: 1.8%, Blacks: 1.6%, χ2
P = 0.469).
Table 1 Frequency distribution of socio-demographic variables
according to the cohort
1971 cohort
Variables
1980 cohort
No.
%
No.
%
Men
1320
48.9
1459
49.9
Women
1382
51.1
1467
50.1
65–69
606
22.5
664
22.7
70–74
366
13.6
381
13.0
75–79
938
34.8
933
31.9
80–84
672
24.9
699
23.9
85+
116
4.3
249
8.5
White
2351
92.4
2410
84.2
Black
103
4.1
312
10.9
Asian
35
1.4
68
2.4
Hispanic
16
0.6
41
1.4
Other
40
1.5
33
1.1
2380
88.1
2553
87.3
322
11.9
373
12.8
>1 visit
2662
98.5
2894
98.9
No visit
40
1.5
32
1.1
P-value
Sex
0.449
Age (years)
0.406
Racea
0.001
Educational level
Low-Middleb
Highc
0.344
Outpatient visitsd
0.197
a Unspecified race was deleted.
b Elementary and grade, high school, Trade, 1–2 years college.
c 3–4 years college, graduate college.
d Outpatient visits in Medicine, Surgery, Home Health, and Nursing Home.
subtypes was prevented since all types of dementia were
included in the analyses.
The crude incidence rate of dementia in the 1971 cohort was
2.21 per 100 person-years (2.35 for males, 2.09 for females); in
the 1980 cohort, the crude rate was 2.36 (2.33 for males, 2.39
for females). The standardized incidence ratio (SIR) of the 1980
cohort relative to the 1971 cohort was 1.00 for males and 1.09
for females. There was no significant difference in the incidence
rate of dementia between the two cohorts (P . 0.05 in both
sexes).
Table 2 presents the age-specific incidence rates of dementia
by sex in the two cohorts. Similar rates were observed for both
sexes in the 1971 and 1980 cohorts, except for men in the age
group 70–74 years (threefold higher in the 1980 cohort) and for
women aged >85 years (1.5 times higher in the 1980 cohort).
Mortality
Incidence of dementia
Among the first cohort (n = 2702), 386 subjects were newly
diagnosed as having dementia. In the second cohort (n = 2926),
457 incident cases were detected during the follow-up period.
In the 1971 cohort, most of the cases were coded as non-specific
dementia (non-specific dementia: 97%, Alzheimer’s dementia:
2%), while in the 1980 cohort, the percentage of Alzheimer cases
increased tenfold (non-specific dementia: 75%, Alzheimer’s
dementia: 20%). Bias due to the misclassification of dementia
In the 1971 cohort, 1057 people died during the follow-up
period; and in the 1980 cohort, 1071 deaths were recorded.
Table 3 shows the age-specific mortality rates by sex in the two
cohorts.
Within a decade, men have presented a decline in mortality
rates at all ages, except after the age of 85 years in the 1980
cohort. Women also showed lower mortality rates before the
age of 85 years in the second cohort. The standardized mortality
ratio (SMR) in the second cohort relative to the first cohort,
TRENDS IN DEMENTIA-FREE LIFE EXPECTANCY
1113
Table 2 Incidence rates for dementia per 100 person-years according to sex, age group, and cohort
Males
Females
1971–1979
Rate (95% CI)
1980–1988
Rate (95% CI)
1971–1979
Rate (95% CI)
65–69
0.22 (0.19–0.25)
0.30 (0.27–0.33)
0.48 (0.45–0.51)
0.00
70–74
0.32 (0.30–0.34)
0.89 (0.88–0.90)
0.67 (0.65–0.69)
0.68 (0.66–0.70)
75–79
1.66 (1.61–1.71)
1.76 (1.71–1.81)
1.65 (1.61–1.69)
1.39 (1.36–1.42)
80–84
3.01 (2.91–3.11)
2.84 (2.75–2.93)
2.97 (2.88–3.06)
2.68 (2.60–2.76)
85+
6.10 (5.80–6.40)
5.52 (5.28–5.76)
3.82 (3.65–3.99)
5.68 (5.45–5.91)
Age group (years)
1980–1988
Rate (95% CI)
Table 3 Mortality rates per 100 person-years according to sex, age group, and cohort
Males
Females
1971–1979
Rate (95% CI)
1980–1988
Rate (95% CI)
1971–1979
Rate (95% CI)
1980–1988
Rate (95% CI)
65–69
1.95 (1.86–2.04)
1.48 (1.43–1.53)
1.43 (1.38–1.48)
1.47 (1.42–1.52)
70–74
4.04 (3.87–4.21)
2.97 (2.86–3.08)
1.92 (1.86–1.98)
2.00 (1.94–2.06)
75–79
6.30 (6.05–6.55)
4.57 (4.41–4.73)
2.89 (2.98–3.07)
2.80 (2.71–2.89)
80–84
8.79 (8.47–9.11)
7.57 (7.30–7.84)
5.46 (5.28–5.64)
3.92 (3.80–4.04)
11.95 (11.35–12.53)
12.36 (11.83–12.89)
8.98 (8.58–9.38)
9.08 (8.73–9.43)
Age group (years)
85+
Table 4 Total life expectancy (TLE), dementia-free life expectancy (DemFLE), dementia life expectancy (DemLE) and per cent of life without
dementia of the total life expectancy (%DemFLE) for selected ages according to sex and cohort
1971 cohort
Age (years)
1980 cohort
TLE
DemFLE
DemLE
%DemFLE
TLE
DemFLE
DemLE
%DemFLE
65
16.71
15.92
0.79
95.27
18.11
16.93
1.18
93.48
70
13.30
12.58
0.72
94.59
14.47
13.38
1.09
92.47
75
10.38
9.72
0.66
93.64
11.29
10.31
0.98
91.32
80
7.95
7.35
0.60
92.45
8.61
7.75
0.86
90.01
85
5.97
5.46
0.51
91.46
6.43
5.69
0.74
88.49
90
4.41
3.97
0.44
90.02
4.69
4.08
0.61
86.99
95
3.20
2.84
0.36
88.75
3.36
2.86
0.50
85.12
65
21.12
18.82
2.30
89.11
21.48
19.63
1.85
91.39
70
17.33
15.19
2.14
87.65
17.47
15.74
1.73
90.10
75
13.98
12.00
1.98
85.84
13.87
12.30
1.57
88.68
80
11.05
9.29
1.76
84.07
10.73
9.36
1.37
87.23
85
8.59
7.04
1.55
81.96
8.11
6.94
1.17
85.57
90
6.56
5.24
1.32
79.88
5.97
5.01
0.96
83.92
95
4.94
3.84
1.10
77.73
4.30
3.53
0.77
82.09
Males
Females
was 0.87 for men (P , 0.01), and 0.91 for women (P , 0.05).
In other words, mortality rates were significantly lower in the
second cohort for both males and females.
Health expectancies
Table 4 shows the duration of TLE, DemFLE, DemLE and the
percentage of life without dementia within the TLE on every
5-year age interval for each gender in both cohorts, and Figure 1
shows the trends of the three expectancies in both genders.
Over the decade between the 1971 and 1980 cohorts, TLE
has increased among males, especially at younger ages. In males
aged 65 years old, TLE was 16.7 years in the 1971 cohort, and
18.1 years in the 1980 cohort. After the age of 85 years, TLE in
both cohorts tended to be similar. The TLE for females younger
than 80 years did not change. For females aged 65 years, TLE
was 21.1 years in the first cohort and 21.5 in the second cohort.
After the age of 75 years, TLE in the second cohort tended to be
lower than in the first cohort.
As compared with the 1971 cohort, DemFLE in the 1980
cohort increased in both males and females at younger ages.
At 65 years old, men in the first cohort presented a DemFLE of
15.9 years, while in the second cohort at the same age, DemFLE
1114
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Figure 1 Trends of total life expectancy (TLE), dementia-free life expectancy (DemFLE),
and dementia life expectancy (DemLE) according to sex and cohort
TLE 80–88
DemFLE 80–88
DemLE 80–88
TLE 71–79
DemFLE 71–79
DemLE 71–79
was 16.9 years. The DemFLE for 65-year-old men in the second
cohort was longer than TLE itself for the same age in the first
cohort. Women aged of 65 years in the first cohort had a
DemFLE of 18.8 years, and in the second cohort a DemFLE of
19.6 years. After 80 years old, in both sexes, DemFLE in the
1971 and 1980 cohorts did not change.
As shown in Figure 1 for males, trends of DemFLE in the
second cohort and TLE in the first cohort overlapped. Within a
decade, the DemFLE of the 1980 cohort reached the duration of
TLE of the 1971 cohort.
The percentage of life free from dementia within the total life
(% DemFLE) decreased between the two cohorts among males,
TRENDS IN DEMENTIA-FREE LIFE EXPECTANCY
1115
Figure 2 Median age of dementia onset according to sex, year of follow-up, and cohort
—–■—–
– –▲– –
1971 Cohort
1980 Cohort
and increased among females. For males 65 years old in the
1971 cohort, DemFLE represented 95% of the TLE, and 93% in
the 1980 cohort. For females at the same age in the first cohort,
DemFLE represented 89% of the TLE, and in the second cohort
91%.
Between the 1971 and 1980 cohorts, the duration of life with
dementia presented an opposite trend between both genders:
increasing in men and decreasing in women. For men aged 65
years in the first cohort, DemLE was 0.8 years, and in the
second cohort DemLE reached 1.2 years. For women at 65 years
old in the 1971 cohort, DemLE was 2.3 years, this duration
decreased to 1.8 years in the 1980 cohort.
Age of dementia onset
The median age of dementia onset for each year of follow-up
according to sex and cohort is presented in Figure 2.
Because the age composition of the cohorts varied every year,
a comparison of the onset age for each year of follow-up is more
accurate than a comparison of the summarized onset age of
the whole cohort. If TLE remains constant with time, a delay
of the onset age observed in the second cohort is an indirect
indicator of a compression of morbidity. Among males, no
specific trend was noticed: a constant or delayed age of onset
was observed for some years of follow-up, as well as an earlier
age of onset in the second cohort for some other years. For
1116
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 5 Relation between the number of years free from dementia and potential predictors of development of dementia determined by Cox
proportional hazards model
1971–1979 cohort
Variables
No.
1980–1988 cohort
RR (95% CI)a
RR (95% CI)b
No.
RR (95% CI)a
RR (95% CI)b
Sex
Female
1382
Male
1320
1467
1.00 (0.92–1.08)
1459
1.18 (1.09–1.29)
1276
1.01 (0.93–1.09)
Age (years)
>75
1467
,75
1235
1650
1.18 (1.09–1.28)
Education level
Low
2380
High
322
2553
0.99 (0.87–1.13)
373
1.04 (0.93–1.17)
Smoking status
Current or past
Never
753
1949
1924
0.98 (0.90–1.08)
1002
0.99 (0.91–1.07)
Stroke history
Yes
163
No
2539
209
1.30 (1.07–1.57)
1.27 (1.04–1.54)
1.06 (0.97–1.16)
1.03 (0.95–1.13)
2717
1.21 (1.02–1.42)
1.18 (1.00–1.40)
1.09 (1.00–1.18)
1.06 (0.98–1.16)
CVDc history
Yes
831
No
1871
1055
1871
Hypertension
Yes
835
No
1867
1209
1.00 (0.91–1.09)
1717
1.02 (0.89–1.17)
2676
1.26 (0.89–1.80)
2885
1.04 (0.88–1.21)
2715
1.05 (0.97–1.13)
Diabetes mellitus
Yes
288
No
2414
250
1.06 (0.92–1.23)
Parkinson’s disease
Yes
46
No
2656
41
1.20 (0.83–1.72)
Depression
Yes
201
No
2501
211
1.08 (0.92–1.26)
a (95%CI): crude relative hazard and 95% CI.
b (95%CI): Relative hazard adjusted by sex and age and 95% CI.
c Cardiovascular disease.
females, the median age of dementia onset was postponed by
2–3 years.
When examining the relation between potential predictors
of dementia and dementia onset with the Cox proportional
hazards model using older age group, low education level, and
presence of specific medical conditions as the reference category
(Table 5), we found that the absence of medical history of stroke
increased significantly the number of years free from dementia
(relative hazard [RH] = 1.30, P = 0.0083 in the 1971 cohort;
RH = 1.21, P = 0.0269 in the 1980 cohort). A significant relation
was also observed between dementia onset and the absence
of cardiovascular disease in the second cohort (RH = 1.09,
P = 0.0478). However, when adjusting for sex and age, the
significance with cardiovascular disease disappeared (RH = 1.06,
P = 0.1529). The other variables such as younger age, being
male, having a high education level, being free from hypertension,
or diabetes mellitus, or Parkinson’s disease, or depression,
respectively, did not show any significant relation with the age
of dementia onset in one or both cohorts.
Discussion
This paper presents chronological series of life expectancies
without dementia among elderly Americans. Estimates of TLE
and DemFLE were based on the longitudinal observation of two
cohorts of elderly members of the same health maintenance
organization. Trends in life expectancy and health expectancy
over a 9-year period were examined. The results indicated that
trends of increase in TLE and in DemFLE were quite different
between the genders. The DemFLE increased at a slower rate
than life expectancy among males. Consequently, DemLE increased in the second cohort and the %DemFLE decreased in the
TRENDS IN DEMENTIA-FREE LIFE EXPECTANCY
1980 cohort. Conversely, for females, with a faster progression
rate of DemFLE compared to TLE, the %DemFLE improved and
the duration with dementia decreased in the second cohort.
Before interpreting the findings, we should point out the three
major limitations of this study. First, while the two cohorts were
similar in terms of sex and age distribution, education level and
frequency of outpatient visits, they differed in terms of race. The
second cohort included less White subjects but more Black
elderly. Differences in the dementia prevalence due to race are
controversial. One study has reported a higher prevalence of
dementia among Blacks,25 while two others showed no difference.26,27 Studies on race and incidence of dementia are scarce.
The present report is consistent with a recent survey which
found no significant difference in the incidence of dementia
between the two races.27 Consequently, we may assume that
the difference in the race distribution between the 1971 and the
1980 cohorts has not biased the comparison over time of the
DemFLE.
Second, in the KPMCP-NC, the mental status of the members
was not routinely screened. The members had to visit the clinics
in order to receive a dementia diagnosis. Some patients would
be unable or reluctant to visit when their mental status declined;
also, some cases would have been referred directly to nursing
homes or kept at home without a diagnostic evaluation. Under
these circumstances, some of the dementia cases would have
been missed and, therefore, the incidence rate may have been
underestimated. As we will describe later, however, the incidence of dementia in the populations we studied was consistent
with those of two previous population-based studies among
elderly Americans.28,29
Third, the attitude of both physicians and patients to dementia
has changed over the decades; consequently, these changes may
have affected the estimate of the rate of incidence of dementia.
Diagnostic practices have varied between the 1970s and the
1980s. During the 1980s, the diagnostic criteria of dementia were
established internationally and the definition of senile dementia
appeared in DSM-III and ICD-9.30 Moreover, the development
of standardized diagnostic interviews and mental examinations
has improved the detection of dementing illnesses. The use of
computer tomography scans and other imaging tools has increased dramatically between the two cohorts.23 Improvements
in the detection of Alzheimer’s dementia are reflected in the
dementia distribution among each cohort.31 Moreover, physicians
have become more inclined to tell the patient and his/her
family the true diagnosis. In addition, families of demented
elderly patients are more aware of dementing illnesses, and seek
greater medical and social support. These changes in physicians’
and patients’ attitudes may have resulted in increased number
of diagnosis of dementia. However, the present data showed
that the incidence of dementia did not change throughout the
study period 1971–1988.
The strength of the present study lies in the utilization of the
same method among comparable populations. The calculation
methods used in the 1971 and 1980 cohorts were identical, taking
into consideration the incidence of dementia and mortality at
the same time.
The incidence of dementia in these two cohorts was consistent with the results of two previous reports of population-based
studies in the US. Both studies showed higher incidence rate with
older age. In the Rochester study, incidence rates increased from
1117
0.13 per 100 person-years for the age group 60–69 years to 2.18
for the elderly aged >80 years.29 In the Framingham study, the
rates were lower than in the present report because the severity
of dementia was restricted to moderate and severe cases.28
Between the first and the second cohorts, mortality decreased,
especially among the males, while the incidence of dementia
remained constant among both genders. Trends of increase in
TLE and DemFLE were quite different between the two sexes.
For males, although both TLE and DemFLE increased, the degree
of increase was more remarkable in the former than in the latter.
As a result, the duration to be spent with dementia or DemLE
was prolonged and an extension of morbidity took place among
the males. For females at younger ages, DemFLE increased by
about one year under the relatively constant TLE. Thus, compression of morbidity was observed among females under 80
years of age. Women in their eighties presented a plateau in
DemFLE; while women aged 85 years and over showed a
decrease in both TLE and DemFLE, supporting an equilibrium.
However, estimations for the oldest ages should be considered
with the utmost caution since the probabilities of death and
dementia in the life table were predicted by a logistic regression
model according to the relationship between these events and
age. The discrepancy between the actual value and the predicted value increases with age, thus the estimate becomes less
reliable at older ages.
The results of the present study indicate that, especially among
females, DemFLE has grown in spite of a constant incidence rate
of dementia. This fact could be explained by a postponement
of dementia onset. In order to corroborate this trend, we have
calculated the median age of dementia onset for each year of
follow-up and have compared the results between both cohorts.
Males did not show any specific trend but the median age of
dementia onset was postponed by 2–3 years among females
in the second cohort. The compression of morbidity among
females was confirmed by the postponement of the median age
of dementia onset in the 1980 cohort.
A previous report on the same population has shown
decreased incidence rates of stroke for females in the second
cohort; while males up to 75 years of age presented increased
rates.23 A decreased incidence rate of stroke, one of the risk
factors for vascular dementia, may account for a later onset of
dementia and partly explain the sex difference observed.
The gender difference in the trends of DemFLE suggests to
us that sex-related factors may account for the risk and prevention of dementia. According to previous studies, oestrogen
use by post-menopausal women may postpone and decrease
the risk of Alzheimer’s dementia.32,33 However, the use of oestrogen in these two age/time cohorts is low. Moreover, the effect
of oestrogen replacement therapy on the risk of Alzheimer’s
disease needs to be fully explored, since the results are still
controversial.34,35
In line with previous studies on active life expectancy which
reported a pandemic of disabilities,36,37 the present report shows,
despite a stable incidence of dementia within a decade, a similar
trend for dementia among men (i.e. an extension of the
duration of life spent with dementia). A less pessimistic trend is
observed among women, supporting a compression of dementia
associated with a constant dementia incidence. The results
suggest that in women of the present cohorts, the onset of
dementia is delayed, although it is not fully prevented.
1118
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Acknowledgements
The authors are grateful to Dr Gary Friedman, Director of
the Kaiser Permanente Medical Care Program of Northern
California Division of Research, for his support in this study.
18 Ritchie K, Mathers C, Jorm A. Dementia-free life expectancy in
Australia. Aust J Public Health 1994;18:149–52.
19 Roelands M, van Oyen H, Baro F. Dementia-free life expectancy in
Belgium. Eur J Public Health 1994;4:33–37.
20 Sauvaget C, Tsuji I, Minami Y et al. Dementia-free life expectancy
among elderly Japanese. Gerontology 1997;43:168–75.
References
1 Bachman DL, Wolf PA, Linn R et al. Prevalence of dementia and
probable senile dementia of the Alzheimer type in the Framingham
Study. Neurology 1992;42:115–19.
2 Canadian Study of Health and Aging Working Group. Canadian Study
of Health and Aging: study methods and prevalence of dementia.
Can Med Assoc J 1994;150:899–913.
3 Eastwood MR, Rifat SL, Roberts D. The epidemiology of dementia in
North America. Eur Arch Psy Clin N 1991;240:207–11.
4 Evans DA, Funkenstein HH, Albert MS, Scherr PA, Cook NR, Chown
21 Haan MN, Selby JV, Quesenberry CP, Schmittdiel JA, Fireman BH,
Rice DP. The impact of aging and chronic disease on use of hospital
and outpatient services in a large HMO: 1971–1991. J Am Geriatr Soc
1997;45:667–74.
22 Verbrugge LM. Longer life but worsening health? Trends in health
and mortality of middle-aged and older persons. Milbank Mem Fund
Qly/Health Soc 1984;62:475–519.
23 Haan MN, Selby JV, Rice DP et al. Trends in cardiovascular disease
incidence and survival in the elderly. Ann Epidemiol 1996;6:348–56.
24 Katz S, Branch LG, Branson MH, Papsidero JA, Beck JC, Greer DS.
Active life expectancy. N Engl J Med 1983;309:1218–24.
MJ. Prevalence of Alzheimer’s Disease in a community population
of older persons. Higher than previously reported. JAMA 1989;262:
2551–56.
25 Heyman A, Fillenbaum G, Prosnitz B, Raiford K, Burchett B, Clark C.
5 Gauthier S, McDowell I, Hill G. Canadian Study of Health and Aging
26 Schoenberg BS, Anderson DW, Haerer AF. Severe dementia. Preval-
(CaSHA). Psychiatr J Univ Ottawa 1990;15:227–29.
6 Kramer M, German PS, Anthony JC, Von Korff M, Skinner EA.
Patterns of mental disorders among the elderly residents of Eastern
Baltimore. J Am Geriatr Soc 1985;33:236–45.
7 Robertson D, Rockwood K, Stolee P. The prevalence of cognitive
impairment in an elderly Canadian population. Acta Psychiat Scand
1989;80:303–09.
8 Weissman MM, Myers JK, Tischler GL et al. Psychiatric disorders
(DSM-III) and cognitive impairment among the elderly in a US urban
community. Acta Psychiatr Scand 1985;71:366–79.
9 Lilienfeld DE, Perl DP. Projected neurodegenerative disease mortality
in the United States, 1990–2040. Neuroepidemiology 1993;12:219–28.
10 Robine J-M, Ritchie K. Healthy life expectancy: evaluation of
global indicator of change in population health. Br Med J 1991;302:
457–60.
11 Kramer M. The rising pandemic of mental disorders and associated
chronic diseases and disabilities. Acta Psychiatr Scand 1980;62:
382–97.
12 Fries JF. Aging, natural death, and the compression of morbidity.
N Engl J Med 1980;303:130–35.
13 Fries JF. The compression of morbidity: near or far? Milbank Q 1989;
67:208–32.
14 Manton KG. Changing concepts of morbidity and mortality in the
elderly population. Milbank Q 1982;60:183–244.
15 Perenboom RJM, Boshuizen HC, Breteler MMB, Ott A, van de Water
HPA. Dementia-free life expectancy (DemFLE) in The Netherlands.
Soc Sci Med 1996;43:1703–07.
16 Ritchie K, Jagger C, Brayne C, Letenneur L. Dementia-free life
Estimated prevalence of dementia among elderly black and white
community residents. Arch Neurol 1991;48:594–98.
ence and clinical features in a biracial US Population. Arch Neurol
1985;42:740–43.
27 Fillenbaum GG, Heyman A, Huber MS et al. The prevalence and
3-year incidence of dementia in older black and white community
residents. J Clin Epidemiol 1998;51:587–95.
28 Bachman DL, Wolf PA, Linn RT et al. Incidence of dementia and prob-
able Alzheimer’s Disease in a general population: the Framingham
Study. Neurology 1993;43:515–19.
29 Schoenberg BS, Kokmen E, Okasaki H. Alzheimer’s Disease and other
dementing illnesses in a defined United States population: incidence
rates and clinical features. Ann Neurology 1987;22:724–29.
30 Ritchie K. International comparisons of dementia-free life expectancy:
a critical review of the results obtained. In: Mathers C, McCallum J,
Robine J-M (eds). Advances in Health Expectancies: Proceedings of the 7th
Meeting of the International Network on Health Expectancy (REVES),
Canberra, February 1994. Australian Institute of Health and Welfare:
AGPS, Canberra, 1994, pp.271–79.
31 Fox P. From senility to Alzheimer’s Disease: the rise of the Alzheimer’s
Disease movement. Milbank Q 1989;67:58–102.
32 Tang M-X, Jacobs D, Stern Y et al. Effect of oestrogen during meno-
pause on risk and age of onset of Alzheimer’s Disease. Lancet 1996;
348:429–32.
33 Paganini-Hill A, Henderson VW. Estrogen deficiency and risk of
Alzheimer’s Disease. Am J Epidemiol 1994;140:256–61.
34 Haskell SG, Richardson ED, Horwitz RI. The effect of estrogen
replacement on cognitive function in women: a critical review of the
literature. J Clin Epidemiol 1997;50:1249–64.
35 Yaffe K, Sawaya G, Lieberburg I, Grady D. Estrogen therapy in
postmenopausal women: effects on cognitive function and dementia.
JAMA 1998;279:688–95.
expectancy: preliminary calculations for France and the United
Kingdom. In: Robine J-M, Mathers C, Bone M, Romieu I (eds).
Calculation of Health Expectancy: Harmonization, Consensus Achieved and
Future Perspectives. Colloques INSERM/John Libbey Eurotext Ltd
1993;226:233–44.
36 Crimmins EM, Saito Y, Ingegneri D. Changes in life expectancy
17 Ritchie K, Robine J-M, Letenneur L, Dartigues J-F. Dementia-free life
37 van de Water HPA, Boshuizen HC, Perenboom RJM. Health expect-
expectancy in France. Am J Public Health 1994;84:232–36.
and disability-free life expectancy in the United States. Population and
Development Review 1989;15:235–67.
ancy in the Netherlands 1983–1990. Eur J Public Health 1996;6:21–28.