American Journal of Epidemiology
ª The Author 2007. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: [email protected].
Vol. 166, No. 4
DOI: 10.1093/aje/kwm100
Advance Access publication June 7, 2007
Original Contribution
Cohort Differences in Self-Rated Health: Evidence from a Three-Decade,
Community-Based, Longitudinal Study of Women
Henian Chen1,2,3, Patricia Cohen1,3,4, and Stephanie Kasen1,3
1
Department
Department
3
Department
4
Department
2
of
of
of
of
Epidemiology, New York State Psychiatric Institute, New York, NY.
Biostatistics, New York State Psychiatric Institute, New York, NY.
Psychiatry, College of Physicians and Surgeons, Columbia University, New York, NY.
Epidemiology, Mailman School of Public Health, Columbia University, New York, NY.
Received for publication November 20, 2006; accepted for publication February 22, 2007.
Despite the fact that life expectancy has nearly doubled over the past century, the US public has become
increasingly preoccupied with issues of health and illness. In this study, the authors investigated cohort differences
in self-rated health between women born in 1935–1944 (preboomers) and women born in 1945–1954 (baby
boomers). A randomly selected, community-based sample of 618 mothers, 314 preboomers, and 304 baby
boomers was interviewed. Over three decades, self-rated health was assessed in 1975, 1983, 1985–1986,
1991–1994, and 2001–2004. An individual growth model showed a linear decline (0.61 per year, p < 0.001) in
self-rated health from mean ages 31–59 years combined, with a quadratic age effect (0.03, p < 0.001). Baby
boomers reported lower self-rated health (mean difference, 5.30; p < 0.001) and more rapid decline per year
(slope difference, 0.52; p < 0.001) than did preboomers of overlapping ages; those differences remained after
adjusting for demographics, socioeconomic variables, personality factors, health behaviors, chronic illness, and
depression symptoms. Study findings have important implications with regard to the potential growing burden on
the nation’s health care system, suggesting that generational changes in health evaluations and expectations may
continue to increase demand for medical care.
cohort effect; health; women
Abbreviations: SD, standard deviation; SRH, self-rated health.
The baby-boomer cohort includes over 65 million individuals and thus represents close to one fourth of the population of the United States (1). In contrast to their more
conservative parents, baby boomers belong to the generation
of protest and activists, expressed in part by less conventional dress, recreational drug use, and a more individualist
focus. As the first cohort to grow up in an affluent and
largely peaceful postwar society characterized by unprecedented educational and employment opportunities and
social change, especially for women, baby boomers
may be considered privileged in comparison to earlier
generations.
Nevertheless, baby-boomer women in midlife are reported to be significantly more depressed and have lower
self-esteem compared with preboomer women at equivalent
ages (2). Moreover, studies have found a higher incidence of
arthritis, obesity, diabetes, and heart disease and more illicit
drug use among members of the baby-boomer generation
than among their predecessors (3–7). As such, baby
boomers will likely demand more medical services and intervention than past generations. Recent projections show
that the aging of baby boomers will result in a doubling of
the number of persons aged 65 years or older with arthritis
or chronic joint symptoms by 2030, when the last of the
Correspondence to Dr. Henian Chen, Columbia University/NYSPI, 100 Haven Avenue, 31F, New York, NY 10032 (e-mail: chenhen@
pi.cpmc.columbia.edu).
439
Am J Epidemiol 2007;166:439–446
440 Chen et al.
baby-boom generation will be 65 years of age (8). Because
of the size of this generation, their aging has profound societal implications relative to previous and subsequent generations (9).
One of the most frequently used measures of health status
is self-rated health (SRH), a single question asking people to
rate their overall health on a scale from excellent to poor.
There is widespread agreement that this simple global question provides a useful summary of how people perceive their
overall health status (10). SRH predicts subsequent mortality (11), as well as functional decline (12), later chronic
disease (13), and recovery from major medical events
(14). Recent studies (15, 16) have shown an association
between SRH and frequently used biomarkers and endocrine measures. SRH has been found to have good test-retest
reliability (17) and concurrent validity (18). SRH has been
recommended for population health and health monitoring
by the World Health Organization, the Centers for Disease
Control and Prevention in the United States, and the European Union Commission (19–22). In the current study, we
examined SRH trajectories over 28 years in a representative
sample of women and compared SRH in two cohorts of
women, preboomers (born between 1935 and 1944) and
baby boomers (born between 1945 and 1954), assessed at
comparable ages. We drew on longitudinal data obtained
from a representative sample of 758 mothers, assessed prospectively regarding SRH in 1975, 1983, 1986, 1993, and
2003, when they were mean ages 31, 39, 42, 49, and 59
years, respectively. Information on personality traits,
chronic illness, employment status, marital status, socioeconomic status, smoking, alcohol consumption, weight gain,
and depression symptoms was obtained in all but the first
interview. The research questions we addressed are as
follows:
1. What is the natural history of SRH over the adult years
in this community sample of women from a 28-year
longitudinal follow-up?
2. Is there a cohort difference regarding SRH between
baby boomers and preboomers assessed at overlapping
ages? If yes, is this cohort difference independent of
demographic and socioeconomic variables related to
SRH? If so, is the difference accounted for by chronic
illness, depression symptoms, or health-related behaviors?
MATERIALS AND METHODS
Sample
Participants were mothers drawn from a sample of 976
families, randomly selected in 1975 for a study of child
behavior on the basis of residing in one of two Upstate
New York counties and having at least one child between
1 and 10 years of age living at home. These women were
first interviewed in 1975, at a mean age of 31 years, about
their health and family background. In 1983 (at a mean age
of 39 years), 718 of these mothers were located and reinterviewed. To replace the disproportionate number of poor,
urban women lost to follow-up, we randomly sampled
mothers from 54 new families from poor, urban neighbor-
hoods in the same two counties to achieve a total of 772 in
the 1983 sample. Follow-up at mean ages of 42 (1986), 49
(1993), and 59 (2003) years included SRH, health behaviors, personality characteristics, depression symptoms, and
social role factors. Detailed information regarding study
methodology is available from previous reports ((23–25);
also refer to the following website: http://www.nyspi.cpmc.
columbia.edu/childcom).
The women in the sample were 91 percent White; resided
in urban, suburban, and rural areas; spanned the full socioeconomic status range; and were representative of the northeastern region of the United States from which they were
sampled. Retention rates were 95 percent in 1986 and 90
percent in 1993. In the recent 2003 follow-up, 609 women
were reinterviewed and 75 women had died, thus accounting
for 90 percent of the 1983 sample; those remaining were not
reinterviewed because of refusal to participate (n ¼ 31),
serious illness (n ¼ 17), study time constraints (n ¼ 16),
and failure to be located (n ¼ 12).
After selection by birth years from 1935 to 1954, 618
women remained for analysis in this study. Preboomers
(n ¼ 314, 50.8 percent) were defined as those born between
1935 and 1944 (10 years). Their mean ages at assessment
were 34.42 (standard deviation (SD), 2.77), 42.79 (SD,
2.86), 45.39 (SD, 2.86), 51.21 (SD, 2.80), and 61.36 (SD,
2.82) years. Eighty-one women were lost from the study at
2003, including 35 who had died. Baby boomers (n ¼ 304,
49.2 percent) were defined as those born between 1945 and
1954 (10 years). Their mean ages at assessment were 26.63
(SD, 2.51), 35.11 (SD, 2.54), 37.66 (SD, 2.55), 43.66 (SD,
2.59), and 53.83 (SD, 2.60) years. Forty-one women were
lost at 2003, including nine who had died.
Measures
Outcome variable: SRH. At each assessment, participants completed a five-point Likert-response item to rate
their health from excellent to poor (Would you say your
health in general is excellent, very good, good, fair, or
poor?). To improve the intrinsic meaning of the SRH score
(26), we used a linear transformation such that SRH score ¼
[(observed score – minimum possible score (1))/(maximum
possible score (5) – minimum possible score (1))] 3 100.
Thus, the lowest possible score was defined as zero and the
maximum possible score as 100, with higher scores indicating better health.
Predictors of SRH. We used demographic variables, personality factors, health behaviors, and depression and psychiatric symptoms assessed during each of the last four
follow-ups in these analyses.
Demographic variables. Race, socioeconomic status in
1983 (the standardized sum of standardized measures of
years of one’s own/spouse’s educational level, occupational
status, and family income), time-varying marital status, and
time-varying employment status were the demographic variables assessed.
Personality factors. We carried out a factor analysis to
consolidate the 39 items assessing maternal personality
originating from a range of scales (27) used in each interview. Three factors include maladaptation (defiance,
Am J Epidemiol 2007;166:439–446
Cohort Differences in Health
impulsivity, alienation; 15 items, reliability coefficient a ¼
0.82), empowered/goal directed (self-reliant, confident, assertive; 13 items, reliability coefficient a ¼ 0.80), and
socialized/warm (gentle, compassionate, rule abiding). The
third factor did not relate to SRH and was not used in the
following analyses.
Depression symptoms. A nine-item scale based on the
Hopkins Symptom Checklist-90 Depression Scale was used
to assess depressive symptoms at follow-ups from 1983 to
2003 (a ¼ 0.85, a ¼ 0.87, a ¼ 0.88, and a ¼ 0.85). The
validity of this measure has been supported in prior work on
this sample of women (23, 24).
Chronic illness. Chronic disease presence was defined
as any autoimmune, cancer, diabetes, musculoskeletal, cardiovascular, gastrointestinal, neurologic, or respiratory disease that interfered with role function or for which regular
medication was taken (set at 0 ¼ absent, 1 ¼ present).
Smoking and alcohol consumption. Included were
scaled measures of amount and frequency of cigarette smoking and alcohol consumption.
Weight gain. Weight was self-reported in 1975 and
2003. We calculated weight gain by using a woman’s
2003 weight minus her 1975 weight. We imputed weight
in 1975 for the women recruited in 1983 based on the weight
and age information from the entire sample.
Statistical analysis
An individual growth model was used to fit the SRH age
trajectory (28) with the PROC MIXED procedure from the
SAS statistical software package (29). The first step in these
analyses was to determine the best basic model fit for the
relation between SRH and age. Thus, we first examined
between-participant differences (‘‘random effects’’) in the
mean level of SRH. The estimated variance of the mean
SRH was 166.91 (standard error, 13.18; p < 0.001), meaning that the 618 participants differed significantly in their
average SRH. Adding age-related linear change in SRH
(age), quadratic change (age2), and cubic change (age3) in
the model sequentially showed significant model improvement (v2 ¼ 470.5, p < 0.001) for the linear age effect and
further significant improvement for the quadratic age effect
(v2 ¼ 113.2, p < 0.001). Therefore, these linear and quadratic age effects were retained in the random model for the
SRH analyses using unstructured covariance (the best one
for these data), showing that both linear and quadratic
changes varied significantly across the study participants.
The results from the quadratic model are complex. Thus,
to simplify interpretation of these trends when we moved
from the above ‘‘random’’ to the ‘‘fixed’’ or average effect
level, we also fit a spline model with knots at ages 40 and 50
years (30). The spline model estimates the linear changes in
SRH before age 40 years, between ages 40 and 50 years, and
after age 50 years and may be considered an alternative to
the combined linear and quadratic age changes.
An initial growth model was estimated for the entire
sample as well as separately for preboomer women and
baby-boomer women over the 28 years covered by the five
assessments. Subsequent models added cohort and cohortby-age interaction variables to the model to test the mean
Am J Epidemiol 2007;166:439–446
441
and trajectories differences in SRH at comparable ages.
Final models added all potential predictors to the model
estimates. Dummy variables indicating women who were
newly recruited in 1983 and women who were lost from
the study at the 2003 assessment were tested as potentially
necessary control variables but were removed because they
were not statistically significant. Residual diagnostics were
used to assess the adequacy of the fitted models. Histograms
of residuals did not indicate discernable skew, and normal
quantile plots displayed no systematic departure from
a straight line. Thus, the normal residual assumption is tenable in our SRH data.
RESULTS
Estimating the trajectory of SRH across 28 years
Quadratic model. The basic growth model of ‘‘fixed’’ or
average effects on SRH included linear and nonlinear (quadratic) age change in SRH for the entire sample (table 1).
The average SRH score estimated at age 42 years was 79.94
percentage points, with a significant linear decline of 0.61
percentage point per year from mean ages 31 to 59 years
combined with a 0.03 percentage point quadratic age effect.
When the analyses were confined to women born before
1945, the average SRH score estimated at age 42 years
was 82.60, with a significant linear decline of 0.32 percentage point per year combined with a 0.03 percentage point
quadratic age effect (table 1). For the women in the more
recent cohort, the average SRH score estimated at age 42
years was 77.30, with a significant linear decline of 0.84
percentage point per year combined with a 0.03 percentage
point quadratic age effect. Thus, baby-boomer women reported more than a 5 percentage points lower SRH estimated
at age 42 years and a 0.52 percentage point more rapid decline per year compared with preboomer women (table 1).
As shown in figure 1, these significant SRH differences not
only were reflected at age 42 years but also actually increased from ages 30 to 61 years, the ages at which members
of both cohorts overlapped. At age 61 years, the babyboomer cohort reported their health status as an average
of about 50, that is, ‘‘good,’’ whereas the earlier cohort
had reported their health status as more than a half scale
unit higher, close to ‘‘very good.’’
Spline model. The spline models examined the linear
annual changes in SRH prior to age 40 years, from age 40
to 50 years, and after age 50 years (table 1). For the entire
sample, the annual linear changes were 0.19 percentage
point (p < 0.001), 0.90 percentage point (p < 0.001), and
1.09 percentage points (p < 0.001) over these three age
ranges, respectively. When the analyses were confined to
women born before 1945, the linear annual changes were
0.08 percentage point (p ¼ 0.52), 0.89 percentage point
(p < 0.001), and 0.96 percentage point (p < 0.001) over
ages of less than 40 years, between ages 40 and 50 years,
and after age 50 years. For the women in the more recent
cohort, these changes were 0.25 percentage point (p <
0.001), 1.18 percentage points (p < 0.001), and 1.38
percentage points (p < 0.001) for age less than 40 years,
between ages 40 and 50 years, and after age 50 years. Thus,
442 Chen et al.
TABLE 1. Changes in self-rated health status (per year) from 1975 to 2003, Upstate New Yorky
Total
(n ¼ 618)
Preboomers
(n ¼ 314)
Baby boomers
(n ¼ 304)
Cohort differences
(n ¼ 618)
Estimate
95% CIz
Estimate
95% CI
Estimate
95% CI
Estimate
95% CI
Self-rated health at age 42 years
79.94**
78.67, 81.21
82.60**
80.82, 84.39
77.30**
75.50, 79.09
5.30**
7.82, 2.78
Linear annual change (age)
0.61** 0.68, 0.53 0.32** 0.45, 0.19 0.84** 0.95, 0.73 0.52**
0.69, 0.35
Quadratic model
2
Quadratic age change (age )
0.03** 0.03, 0.02 0.03** 0.04, 0.02 0.03** 0.04, 0.02
0.004
0.008, 0.016
Spline model (age in years)
<40
0.19** 0.31, 0.08
40– 50
0.90** 1.24, 0.57 0.89** 1.29, 0.49 1.18** 1.79, 0.58 0.29
0.99, 0.42
>50
1.09** 1.22, 0.95 0.96** 1.12, 0.81 1.38** 1.62, 1.14 0.41**
0.68, 0.13
0.08
0.17, 0.33
0.25** 0.39, 0.12 0.34*
0.62, 0.06
* p < 0.05; **p < 0.01.
y All parameter entries are maximum likelihood estimates fitted by individual growth models by using the PROC MIXED procedure from the
SAS statistical software package (SAS Institute, Inc., Cary, North Carolina). Age was centered at 42 years (intercept should be interpreted as
average self-rated health at age 42 years). Cohort was coded 0 ¼ preboomer, 1 ¼ baby boomer (cohort difference ¼ boomer finding – preboomer
finding). Refer to the Materials and Methods section of the text for a definition of preboomers and baby boomers.
z CI, confidence interval.
the cohort differences were 0.34 percentage point (p ¼
0.02), 0.42 percentage point (p ¼ 0.25), and 0.41 percentage point (p ¼ 0.004) for these three age groups (table 1).
Predictors of SRH
For preboomer women, the most important predictors of
SRH were chronic illness, depression, empowered/goaldirected personality, employment, maladaption personality,
and socioeconomic status. One standard deviation increase
in chronic illness, depression, and maladaption lowered
SRH significantly by 7.01 percentage points, 4.34 percentage points, and 1.64 percentage points, respectively. One
standard deviation increase in empowered/goal-directed
personality, employment, and socioeconomic status raised
SRH significantly by 2.34 percentage points, 1.87 percentage points, and 1.41 percentage points, respectively.
For baby-boomer women, the most important predictors
of SRH were chronic illness, socioeconomic status, depression, weight gain, smoking, and maladaption. One standard
FIGURE 1. Change in self-rated health status from 1975 to 2003, by birth cohort, Upstate New York. Refer to the Materials and Methods section of
the text for an explanation of the health-status point scale used on the y-axis and a definition of baby boomers and preboomers.
Am J Epidemiol 2007;166:439–446
Cohort Differences in Health
deviation increase in chronic illness, depression, weight
gain, smoking, and maladaption lowered SRH significantly
by 5.17 percentage points, 2.94 percentage points, 2.52 percentage points, 1.37 percentage points, and 1.32 percentage
points, respectively. One standard deviation increase in socioeconomic status raised SRH significantly by 3.94 percentage points.
Findings regarding the primary demographic confounders, socioeconomic status and employment status, were interesting. Socioeconomic status was a predictor of SRH in
both groups but showed a greater influence in the more recent cohort. On the other hand, employment was not statistically significant in the more recent cohort, potentially
reflecting the larger association of health with employment
for women of older ages as well as the likely higher ageadjusted employment rates across the younger cohort, given
historical employment trends. Chronic illness showed
a slightly, although not significantly higher relation with
SRH in the preboomer cohort. The remaining time-varying
predictors of SRH were used explicitly in an attempt to
explain differences in SRH trajectories between the cohorts.
Cigarette smoking, alcohol consumption, and weight gain
all increased and may have mediated cohort differences.
Cohort differences in depression, empowerment, and maladaption may also account for differences in SRH.
We added those nine time-varying predictors to the basic
cohort-difference model (n ¼ 618). Compared with the basic cohort-difference model, the final model showed that the
nine time-varying predictors accounted for the following
proportions of total variance: 50.4 percent of the mean
of SRH, 30.1 percent of the linear change in SRH, and
15.9 percent of the quadratic change in SRH in the entire
sample. The key finding in this model was that cohort differences in SRH, both mean level and decline with age,
remained significant after chronic illness, depression symptoms, personality factors, health behaviors, and demographic and socioeconomic variables were statistically
controlled (table 2).
DISCUSSION
These findings from a community-based, longitudinal
study of women provide initial support for the following:
1) The average SRH score estimated at age 42 years was
79.94, with a significant linear decline of 0.61 percentage
point per year from mean ages 31 to 59 years combined with
a 0.03 percentage point quadratic age effect for the entire
sample. 2) Baby-boomer women reported a 5.30 percentage
points lower SRH and a 0.52 percentage point more rapid
decline per year than did preboomer women at the same
ages. 3) Changes in demographic variables, socioeconomic
status, personality factors, health behaviors, chronic illness,
and depression symptoms over cohort and age did not explain the significant cohort difference, although they did
reduce the remaining mean difference about 40 percent from
the original average age-adjusted level of 5.30 percentage
points to 2.86 percentage points. The cohort difference in
annual age-related decline, however, remained essentially
the same following statistical consideration of these possible
cohort differences, from an overall level of about 0.42 perAm J Epidemiol 2007;166:439–446
443
centage point per year in the original model without
‘‘explanatory variables’’ to 0.43 percentage point in the final
model. 4) SRH was related to chronic illness, depression
symptoms, socioeconomic status, and maladaption for both
preboomer and baby-boomer women. Personality factors
and employment were significantly related to the SRH of
preboomer women, and health behaviors (smoking, weight
gain) were significantly related to SRH for the baby-boomer
cohort.
The present findings are consistent with previous findings
indicating that SRH declines with age (22, 31–33) and extend them by specifying the nature of these age-related declines. The curvilinear trend shows that, although there is
a net decline in SRH during the adult life course, this net
linear decline actually belies a more complex pattern of
relative stability in early adulthood up to about age 40 years,
followed by a more precipitous drop thereafter. The present
findings are of interest because they confirm that midadult to
older baby-boomer women report poorer SRH than preboomer women do. Among 2,587 community residents
who were age 65 years or older between 1982 and 1988,
cohorts born between 1900–1905 and prior to 1900 had
better SRH than cohorts born between 1906–1911 and
1912–1917 (34). To our knowledge, our study is the first
to report the cohort difference in SRH among women, even
after controlling for several relevant potential explanatory
variables.
A part of this cohort difference in SRH, but by no means
all, may be due to the higher age-adjusted and self-reported
prevalence of arthritis, obesity, diabetes, heart disease, depression, illicit drug use, and lower self-esteem among more
recent cohorts of women than among their predecessors
(2–7). Such a cohort difference may occur because knowledge about health and medical care has changed rapidly in
this century (33, 34), and the concept of health might change
over time (35). Older cohorts may have different interpretations of and expectations about health and health care and
may therefore rate their health status differently (36).
A study based on a US national sample from 1993–2001
showed that the percentage of individuals reporting fair or
poor SRH increased from 13.4 percent in 1993 to 15.5 percent in 2001, or about 1.2 percent per year (95 percent confidence interval: 0.9 percent, 1.6 percent) (23). Hoeymans
et al. (33) studied 513 men aged 70 years or older from 1990
to 1995 and found that those aged 75–79 and 85–89 years in
1995 had significantly worse SRH than men in those same
age groups in 1990. The proportion of men who rated themselves as healthy declined by 3.3 percent (p < 0.001) per
year from 1990 to 1995, supporting a time effect (time
change) on SRH. Our study shows that SRH declined
0.46 percentage point per year of time and declined 0.30
percentage point per year of age among women (table 3);
moreover, a greater decline in SRH per year of time was
found for baby-boomer women (0.35 percentage point)
compared with preboomer women (0.05 percentage
point).
Over the past decade, the US population has enjoyed
rapidly declining mortality rates at all ages. Early diagnosis
and treatment of life-threatening chronic diseases are the
major factors. Clearly, age-adjusted mortality rates are
444 Chen et al.
TABLE 2. Standardized predictors of self-rated health status from 1983 to 2003 in two cohorts, Upstate
New Yorky
Preboomers
(n ¼ 314)
Estimate
Baby boomers
(n ¼ 304)
95% CIz
Estimate
Total
(n ¼ 618)
95% CI
Estimate
95% CI
Chronic illness
7.01** 8.13, 5.89 5.17** 6.23, 4.11 5.99** 6.77, 5.22
Depression symptoms
4.34** 5.72, 2.96 2.94** 4.16, 1.72 3.60** 4.53, 2.68
Goal-directed personality
2.34**
1.05, 3.62*
0.70
0.51, 1.91
1.42**
0.53, 2.30
Employment
1.87**
0.84, 2.91*
0.68
0.32, 1.68
1.35**
0.62, 2.08
Maladaption personality
1.64*
Socioeconomic status
1.41*
Alcohol consumption
0.97
2.96, 0.32 1.32* 2.60, 0.05 1.47** 2.39, 0.54
0.02, 2.80
3.94**
0.11, 2.06
0.08
2.46, 5.41
1.15, 0.99
2.65**
0.41
1.62, 3.67
0.36, 1.18
No. of cigarettes smoked daily
0.87
2.11, 0.36
1.37** 2.53, 0.21 1.12** 1.97, 0.27
Weight gain
0.72
2.28, 0.86
2.52** 3.89, 1.16 1.97** 3.00, 0.94
Cohort difference
2.86** 5.17, 0.54
Cohort difference in linear age change
0.43** 0.70, 0.17
* p < 0.05; **p < 0.01.
y All parameter entries are maximum likelihood estimates fitted by individual growth models by using the PROC
MIXED procedure from the SAS statistical software package (SAS Institute, Inc., Cary, North Carolina). Cohort was
coded 0 ¼ preboomer, 1 ¼ baby boomer. All predictors were standardized for the entire sample measured over time
by subtracting mean and dividing by standard deviation to facilitate comparison of magnitudes of regression
coefficients by using a one standard deviation change. Refer to the Materials and Methods section of the text for
a definition of preboomers and baby boomers.
z CI, confidence interval.
declining. Studies show that the US population is experiencing ‘‘longer life but worsening health’’ in part because of
increasing survival time for chronically ill, functionally limited, and disabled persons (37, 38). However, in these analyses, cohort differences in functionally significant chronic
illness did not account for cohort differences in SRH.
Mortality rates are often used as summary measures of
current health status when comparing different populations.
However, these measures have the limitation of measuring
only the very extreme of ill health. Alternatively, SRH as an
approach to quantifying the health of a population tends to
be regarded as flawed because of its subjectivity. SRH is
a personal, subjective summary of all the information that
people have available about anything they understand to be
elements of ‘‘health’’ (16). Our findings show that SRH is
related to chronic illness and depression symptoms for both
preboomer and baby-boomer women, and they support
a view that SRH reflects both physical and psychological
health.
An argument can be put forward to explain the independent association of smoking and weight gain with SRH of
baby-boomer women: Baby boomers may have adopted
ideas of preventive medicine and health promotion about
health risk factors and health behaviors and therefore may
TABLE 3. Longitudinal changes in self-rated health status from 1975 to 2003 in two
cohorts, by time, Upstate New Yorky
Model 1
Self-rated health at a mean age of
42 years and 1993
Model 2
Estimate
95% CIz
Estimate
95% CI
77.27**
75.22, 79.31
81.98**
77.50, 86.46
Per year of time from 1975 to 2003
0.46**
0.70, 0.21
0.05
0.45, 0.35
Per year of age from 20 to 75 years
0.30*
0.56, 0.05
0.58**
0.98, 0.17
Time 3 age
0.03**
0.04, 0.03
0.03**
0.04, 0.03
Cohort difference
5.08*
9.11, 1.05
Cohort difference by time
0.31**
0.43, 0.18
* p < 0.05, **p < 0.01.
y All parameter entries are maximum likelihood estimates fitted by individual growth models by
using the PROC MIXED procedure from the SAS statistical software package (SAS Institute, Inc.,
Cary, North Carolina). Time was centered at 1993; age was centered at 42 years. Cohort was
coded 0 ¼ preboomer, 1 ¼ baby boomer. Refer to the Materials and Methods section of the text
for a definition of preboomers and baby boomers.
z CI, confidence interval.
Am J Epidemiol 2007;166:439–446
Cohort Differences in Health
use these ideas when assessing their own health (36, 39). It
is undeniably true that more recent generations are exposed
to much more health-related literature and advertising, all
tending to reinforce a view of great personal vulnerability,
regardless of the sanitary and medical strides achieved over
recent generations.
Several limitations of the current study require discussion. Because it was conducted with a sample of primarily
White women, the findings cannot be readily generalized to
non-White or male samples. A second limitation of our
study was that all women in this sample were mothers, so
findings may not generalize to women who are childless.
Our findings of a cohort difference in SRH may be important for medical policy and care in the United States. If
this difference continues to increase, more recent generations will continue to amplify the demand for medical care.
Our findings have important policy implications because
they suggest that the nation’s overall health goals (40) are
not being met. Identifying and addressing the cause of this
cohort difference in SRH are essential. Continued population surveillance of SRH offers promise for contributing to
our understanding of the broader determinants of women’s
health and for evaluating progress toward national health
goals.
ACKNOWLEDGMENTS
This study was supported by Office of Child Development grant OCD-CB-18 (1975); National Institute of Mental
Health grants MH-36971 (1983), MH-38916 (1986), and
MH-49191 (1992); and National Institute of Child Health
and Human Development grant HD-40685 (2001).
Conflict of interest: none declared.
REFERENCES
1. Rogerson PA, Kim D. Population distribution and redistribution of the baby-boom cohort in the United States: recent
trends and implications. Proc Natl Acad Sci U S A 2005;102:
15319–24.
2. A tale of expected consequences: USC longitudinal study of
generations. (http://www.usc.edu/dept/gero/research/4gen/
norella.htm). Accessed June 12, 2006.
3. Leveille SG, Wee CC, Lezzoni LI. Trends in obesity and
arthritis among baby boomers and their predecessors,
1971–2002. Am J Public Health 2005;95:1607–13.
4. Flegal KM, Carroll MD, Ogden CL, et al. Prevalence and
treads in obesity among US adults, 1999–2000. JAMA 2002;
288:1723–7.
5. Flegal KM, Williamson DF, Pamuk ER, et al. Estimating
deaths attributable to obesity in the United States. Am J Public
Health 2004;94:1486–9.
6. Gfroerer J, Penne M, Pemberton M, et al. Substance abuse
treatment need among older adults in 2020: the impact of the
aging baby-boom cohort. Drug Alcohol Depend 2003;69:
127–35.
7. Bonneux L, Looman CWN. High coronary heart disease rates
among Dutch women of the baby boom, born 1945–1959. Eur
J Public Health 2003;13:226–9.
Am J Epidemiol 2007;166:439–446
445
8. Public health and aging: projected prevalence of self-reported
arthritis or chronic joint symptoms among persons age>65
ages—United States, 2005–2030. MMWR Morb Mortal Wkly
Rep 2003;52:489–91.
9. Kinsella K, Velkoff VA. An aging world: 2001. Washington,
DC: US Census Bureau, 2001. (Series P95/01-1).
10. Fayers PM, Sprangers MAG. Understanding self-rated health.
Lancet 2002;359:187–9.
11. Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav
1997;38:21–37.
12. Ferraro KF, Farmer MM, Wybraniec JA. Health trajectories:
long-term dynamics among black and white adults. J Health
Soc Behav 1997;38:38–54.
13. Shadbolt B. Some correlates of self-rated health for Australian
women. Am J Public Health 1997;87:951–6.
14. Wilcox VL, Kasl SV, Ider EL. Self-rated health and physical
disability in elderly survivors of a major medical event.
J Gerontol 1996;51B:S96–104.
15. Jylha M, Volpato S, Guralnik JM. Self-rated health showed a
graded association with frequently used biomarkers in a
large population sample. J Clin Epidemiol 2006;59:465–71.
16. Lekander M, Elofsson S, Neve IM, et al. Self-rated health is
related to levels of circulating cytokines. Psychosom Med
2004;66:559–63.
17. Lundberg O, Manderbacka K. Assessing reliability of a
measure of self-rated health. Scand J Soc Med 1996;24:
218–24.
18. Rowan K. Global questions and scores. In: Jenkinson C, ed.
Measuring health and medical outcomes. London, United
Kingdom: Oxford University Press, 1994:54–76.
19. Manderbacka K, Lundberg O, Martikainen P. Do risk factors
and health behaviours contribute to self-ratings of health? Soc
Sci Med 1999;48:1713–20.
20. de Bruin A, Picavet HSJ, Nossikov A, eds. Health interview
surveys: towards international harmonization of methods and
instruments. Copenhagen, Denmark: WHO Regional Office
for Europe, 1996. (WHO regional publications, European
series, no. 58).
21. Hennessy CH, Moriary DG, Zack MM, et al. Measuring
health-related quality of life for public health surveillance.
Public Health Rep 1994;109:665–72.
22. Zack MM, Moriarty DG, Stroup DF, et al. Worsening trends in
adult health-related quality of life and self-rated health—
United States, 1993–2001. Public Health Rep 2004;119:
493–505.
23. Kasen S, Cohen P, Chen H, et al. Depression in adult women:
age changes and cohort effects. Am J Public Health 2003;93:
2061–6.
24. Kasen S, Cohen P, Berenson K, et al. Dual work and family
roles and depressive symptoms in two birth cohorts of women.
Soc Psychiatry Psychiatr Epidemiol 2005;40:300–7.
25. Kasen S, Chen H, Sneed J, et al. Social role and birth cohort
influences on gender-linked personality traits in women: a 20
year longitudinal analysis. J Pers Soc Psychol 2006;91:
944–58.
26. Cohen P, Cohen J, Aiken L, et al. The importance of units:
the circumstance for POMP. Multivar Behav Res 1999;34:
315–46.
27. Brook JS, Whiteman M, Nomura C, et al. Personality, family,
and ecological influences on adolescent drug use: a developmental analysis. J Chem Depend Treat 1988;1:123–61.
28. Chen H, Cohen P. Using individual growth model to analyze
the change in quality of life from adolescence to adulthood.
Health Qual Life Outcomes 2006;4:10.
446 Chen et al.
29. Littell RC, Milliken GA, Stroup WW, et al. SAS system
for fixed models. Cary, NC: SAS Institute, Inc,
1996.
30. Fitzmaurice GM, Laird NM, Ware JH. Applied longitudinal
analysis. Hoboken, NJ: John Wiley & Sons, Inc, 2004.
31. McCullough ME, Laurenceau JP. Gender and the natural history of self-rated health: a 59-year longitudinal study. Health
Psychol 2004;23:651–5.
32. McDonough P, Berglund P. Histories of poverty and selfrated health trajectories. J Health Soc Behav 2003;44:
198–214.
33. Hoeymans N, Feskens EJM, van den Bos GAM, et al.
Age, time, and cohort effects on functional status and
self-rated health in elderly men. Am J Public Health
1997;87:1620–5.
34. Idler EL. Age differences in self-assessments of health: age
changes, cohort differences, or survivorship? J Gerontol 1993;
48:S289–300.
35. Waidmann T, Bound J, Schoenbaum M. The illusion of failure:
trends in the self-reported health of the U.S. elderly. Milbank
Q 1995;73:253–87.
36. Conrad P. Wellness as virtue: morality and the pursuit of
health. Cult Med Psychiatry 1994;18:385–401.
37. Schneider EL, Brody JA. Aging, natural death and the compression of morbidity: another view. N Engl J Med 1983;309:
854–7.
38. Verbrugge LM. Longer life but worsening health? Trends in
health and mortality of middle-aged and older person. Milbank
Mem Fund Q Health Soc 1984;62:475–519.
39. Davison C, Davey SG, Frankel S. Lay epidemiology and the
prevention paradox: the implications of coronary candidacy
for health education. Soc Health Illness 1991;13:1–19.
40. US Department of Health and Human Services. Healthy people 2010: understanding and improving health. 2nd ed. 2
vols. Washington, DC: US Government Printing Office,
2000.
Am J Epidemiol 2007;166:439–446