Am J of Geriatric Psychiatry 28:1 (2020) 99−107

Available online at www.sciencedirect.com

ScienceDirect
journal homepage: www.ajgponline.org

Regular Research Article

Depression Symptoms Declining

Among Older Adults: Birth Cohort
Analyses From the Rust Belt
Kevin J. Sullivan, Ph.D., Anran Liu, M.S., Hiroko H. Dodge, Ph.D.,
Carmen Andreescu, M.D., Chung-Chou H. Chang, Ph.D., Mary Ganguli, M.D.

ARTICLE INFO ABSTRACT

Article history: Objectives: To investigate potential birth cohort effects in depression symptoms
Received March, 14 2019 in older adults. Design: Population-based prospective cohort. Setting: Small-
Revised June, 3 2019 town communities in Pennsylvania. Participants: Three thousand two hundred
Accepted June, 5 2019 and twenty seven older adults (average baseline age = 71.6) born between 1902
and 1941. Measurements: Four decade-long birth cohorts were the primary pre-
Key Words: dictors in this study: 1902−1911, 1912−1921, 1922−1931, and 1932−1941.
Depression The outcome was symptoms of depression assessed at baseline and follow-up
birth cohort study visits using a modified Center for Epidemiologic Studies Depression Scale
epidemiology (mCES-D). The depression outcome was operationalized as: 1). A binary outcome
subsyndromal depression of having greater than equal to 5 depression symptoms on the total mCES-D at
any study visit, and 2). A continuous outcome of four factor-analyzed component
scores of the mCES-D including depressed mood, anergia/hopelessness, with-
drawal, and poor self-esteem. All analyses were jointly modeled with attri-
tion and adjusted for age, sex, education, Mini Mental State Examination
score, antidepressant medications, and total prescription medications.
Results: Participants from more recently born cohorts were significantly
less likely to have a study visit in which they reported greater than or equal
to 5 depression symptoms, controlling for attrition. Specifically, in compari-
son to the 1902−1911 referent cohort, the 1912−1921 birth cohort was 43%
less likely (odds ratio [OR] = 0.566, 95% confidence interval [CI]: 0.341
−0.939), the 1922− 1931 birth cohort was 63% less likely (OR = 0.0369,
95% CI: 0.215−0.632), and the 1932−1941 cohort was 79% less likely
(OR = 0.205, 95% CI: 0.106−0.399). The cohort effect was most evident in the

From the Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA; Department of Biostatis-
tics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA; Department of Neurology, Michigan Alzheimer’s Disease Cen-
ter, University of Michigan, Ann Arbor, MI; Department of Neurology, Layton Aging and Alzheimer’s Disease Center, Oregon Health &
Science University, Portland, OR; Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, PA; Department of Med-
icine, University of Pittsburgh School of Medicine, Pittsburgh, PA; and the Department of Neurology, University of Pittsburgh School of Medi-
cine, Pittsburgh, PA. Send correspondence and reprint requests to Kevin J Sullivan, Ph.D., Department of Epidemiology, University of
Pittsburgh Graduate School of Public Health, 130 N Bellefield Ave, Rm. 344, Pittsburgh, PA 15213. e-mail: KJS152@pitt.edu
Previous Meeting Presentation: American Association of Geriatric Psychiatry, 3/2/19, Atlanta, GA.
© 2019 American Association for Geriatric Psychiatry. Published by Elsevier Inc. All rights reserved.
https://doi.org/10.1016/j.jagp.2019.06.002

Am J Geriatr Psychiatry 28:1, January 2020 99

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
Depression Symptoms Declining Among Older Adults

depressed mood and anergia/hopelessness symptom composites.

Conclusion: Reduced rates of depression symptoms observed in successive
birth cohorts of older adults may reflect compression of morbidity or other sec-
ular trends. (Am J Geriatr Psychiatry 2020; 28:99−107)

population age-distribution continues to shift higher,

better characterizing these population health trends
INTRODUCTION in older adults should be an essential prerequisite for
public health planning efforts. Changing trends in
M ajor Depressive Disorder (MDD) is among the
most common mental illnesses diagnosed in
the United States, with prevalence reportedly lower in
depressive disorders are well established for younger
individuals, with evidence suggesting that prevalence
is increasing, or at least stable, in adolescent and mid-
adults over the age of 65 compared to younger ages.1
life age groups.19−21 We are aware of no previous
Many older adults with subsyndromal depression,
studies that investigated cohort trends in depression
who do not meet the clinical diagnostic criteria for
symptoms specifically in older adults.
MDD,2 may still experience symptoms of depression
We investigated birth cohort trends in depression
that can negatively affect their quality of life, family
symptoms in four successive decade-long birth
well-being, health outcomes, and psychiatric morbid-
cohorts born between 1902 and 1941 in southwestern
ity.3,4 Both MDD and depression symptoms in older
Pennsylvania. Based on cohort trends evidencing
adults are often comorbid with chronic disease, cogni-
decreasing medical and cognitive morbidity,16,22 we
tive impairment, and disability.4 MDD diagnostic cri-
hypothesized that prevalence of depression symp-
teria require that symptoms not be “attributable to”
toms would decrease in the more recently born
another medical condition,2 a particularly challenging
cohorts.
attribution to disprove in the presence of chronic
comorbidity among older adults. Consequently, the
prevalence of subsyndromal depression in older adults
is estimated to be between 2 and 3 times higher than
MDD.3 METHODS
Depressive symptoms in older adults increase risk
Participants
of nonadherence to medical treatment,5 worse health
outcomes of other medical illness,6 cognitive decline,7 Data were pooled from two large prospective
Alzheimer’s disease,8 suicide,9 and overall mortal- cohort studies conducted between 1987 and the pres-
ity.10 Additionally, even in older adults with subsyn- ent in the Monongahela Valley region of southwestern
dromal depression, direct health care costs have been Pennsylvania. The Monongahela Valley Independent
reported to be almost 50% higher than in older adults Elders Survey (MoVIES), focused on dementia, ran
free from depressive symptoms, independent of med- from 1987 until 2001 with biennial assessments of an
ical comorbidities.11 From a public health and policy initial sample of 1681. The Monongahela-Youghiog-
perspective, it would be valuable to characterize time heny Healthy Aging Team (MYHAT) study, focused
trends in depression symptoms in older adults as the on mild cognitive impairment, has been ongoing since
U.S. population ages.12 2006 with annual assessments of an initial sample of
Birth cohort analyses have indicated several chang- 1982. Both samples were recruited using age-stratified
ing patterns in older adults born more recently, as random sampling from voter registration lists. Inclu-
compared to older adults born closer to the start of sion criteria were similar in both studies: age 65+, no
the 20th century. More recently born cohorts have significant vision or hearing impairment, not institu-
been observed to have a lower incidence of dementia tionalized at study entry, and having decisional
and less cognitive impairment in older age than those capacity. Additionally, the MoVIES study required
born in earlier decades.13−16 Additionally, these birth participants to have at least a 6th grade education
cohorts are more likely to be nonsmokers and have and be fluent in English. While not explicitly
higher educational attainment.17,18 As the worldwide required for study inclusion, all participants in the

100 Am J Geriatr Psychiatry 28:1, January 2020

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
 Sullivan et al.

MYHAT study met these same education and flu- present study as greater than or equal to 5 endorsed
ency criteria. MYHAT also required participants to items on the mCES-D, which we have previously
score at least 21 on an age-education corrected established as our threshold.31 We used two depres-
Mini-Mental State Exam (MMSE).23,24 Further details sion outcomes in the models; 1. an mCES-D score
regarding recruitment and assessment procedures higher than equal to 5 and 2. the number of symp-
have been reported previously for MoVIES25,26 and toms endorsed in each of the four factors derived
MYHAT.27,28 Participants were reassessed biennially from the factor analysis described next.
in MoVIES and annually in MYHAT, and were fol-
lowed for an average of 5.72 years. Both studies
were approved by the University of Pittsburgh Insti- Statistical Analyses
tutional Review Board and all participants provided Descriptive statistics
written informed consent.
Across the two studies we identified four birth cohorts We calculated descriptive statistics in the overall
of sufficient size: 1902−1911, 1912−1921, 1922−1931, and sample and by each birth cohort. For categorical
1932−1941. We excluded participants born before 1902 variables, we report frequencies and percentages.
(n = 14), born after 1941 (n = 48), and participants with For continuous variables, we report mean and
no depression data at any visit (n = 338). Compared to standard deviation for baseline age, or median,
the analysis sample, the 338 participants excluded from 25% percentile, and 75% percentile for baseline
the analysis for missing data were more likely to be from MMSE and total number of prescription medica-
earlier born cohorts, be less educated, and had a larger tions. Significant differences between birth cohorts
proportion of men (Supplementary Table S1). for categorical variables were tested using Pear-
son’s x2 Test (or Fisher’s Exact Test when any cell
count <5). Significant differences between birth
Predictor variables cohorts for continuous variables were tested using
Birth cohort was the primary predictor variable. one-way analysis of variable when normally dis-
Other covariates included the following: Demo- tributed, and Kruskal-Wallis Rank Sum Test when
graphics, including baseline age, sex, race, and not normally distributed.
self-reported education level categorized as less
than high school (<HS), graduated HS, or some
Factor analysis
college education or higher (>HS); MMSE score
from each study visit; total number of prescription To find the latent subgroups of the individual
drugs, as a measure of overall morbidity29,30; and symptoms of the mCES-D, we used principal compo-
use of antidepressant drugs. nents factor analysis and the varimax rotation. Apply-
ing the Kaiser criterion (i.e., eigenvalues ≥1), we
identified four subgroups of individual symptoms
Outcome variables
using baseline mCES-D data. The four factors
At each study visit (excluding the baseline MoVIES explained 46% of the variance of the original matrix.
visit), participants completed a modified Center for All items included in the four subgroups achieved
Epidemiologic Studies Depression Scale (mCES-D).31 loading greater than 0.4. The items in each factor, as
The modified scale includes all 20 depression symp- well as factor loadings, are displayed in Table 1. Factor
toms in the original CES-D,32 but rather than recalling 1 (depressed mood) accounted for 28% of the variance
the number of days during the past week that they of the original matrix. Factor 2 (anergia/hopelessness)
experienced each symptom, participants instead accounted for 7% of the variance of the original matrix.
report whether or not they experienced that symptom Factor 3 (withdrawal) accounted for 6% of the variance
over most of the preceding week. Each symptom is of the original matrix. Factor 4 (poor self-esteem)
scored as absent/present (0/1) with a possible maxi- accounted for 5% of the variance of the original matrix.
mum score of 20, so that the total mCES-D score Three items failed to load and therefore were omitted
represents number of symptoms. Having at least sub- when factors were used as outcomes in the models.
syndromal depression was operationalized for the The identified factors, based on the current analysis

Am J Geriatr Psychiatry 28:1, January 2020 101

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
Depression Symptoms Declining Among Older Adults

TABLE 1. Factor-Analyzed Composites of the mCES-D in MoVIES and MYHAT Sample (n = 3,227)

Factors
Anergia and
mCES-D Item Depressed Mood Hopelessness Withdrawal Poor Self-Esteem
I felt I could not shake off the blues even 0.588a 0.202 0.266 0.227
with help from family and friends
I felt depressed 0.729a 0.248 0.182 0.074
I was happy 0.574a 0.377 0.047 0.036
I felt lonely 0.497a 0.272 0.131 0.167
I had crying spells 0.667a 0.108 0.275 0.049
I felt sad 0.765a 0.138 0.198 0.156
I felt that everything I did was an effort 0.113 0.709a 0.270 0.040
I felt hopeful about the future 0.256 0.596a -0.109 0.055
I enjoyed life 0.464 0.481a 0.002 0.181
I could not get going 0.157 0.676a 0.288 0.001
I was bothered by things that don’t usually bother me 0.207 0.095 0.690a 0.055
I did not feel like eating; my appetite was poor 0.157 0.290 0.421a 0.038
I had trouble keeping my mind on what I was doing 0.051 0.286 0.609a 0.058
I talked less than usual 0.233 0.101 0.481a 0.229
I thought my life had been a failure 0.165 0.191 0.092 0.485a
People were unfriendly 0.124 0.045 0.073 0.788a
I felt that people dislike me 0.044 0.013 0.080 0.801a
I felt that I was as good as other people 0.087 0.366 0.161 0.232
I felt fearful 0.352 0.255 0.044 0.197
My sleep was restless 0.225 0.312 0.327 0.082
Factor % Variance Explained 28.0% 7.0% 6.0% 5.0%

Notes: mCES-D: Modified Center for Epidemiologic Studies Depression Scale. Cells contain factor loadings, with the highest factor loading
≥0.400 highlighted under the factor name. Three items failed to load onto any factor: I felt that I was as good as other people, I felt fearful, and My
sleep was restless.
a
Factor loading ≥0.400.

sample comprised of both MYHAT and MoVIES, are generalized linear mixed model with logit link
similar to a previous factor analysis of the mCES-D adjusted for age, sex, education, antidepressant
performed in only the MoVIES participants,33 and are usage, MMSE, and total prescription medications
fairly consistent with reported meta-analyses factor was used. These covariates were adjusted for on the
analyzing the original CES-D.34,35 basis that they were associated both with the pri-
mary predictor (birth cohort) and outcome (symp-
toms of depression). The submodel of time to
Models
attrition was a Weibull proportional hazards model
In modeling, the effect of birth cohort on the adjusted for age, sex, and education. A shared ran-
presence of more than equal to 5 depression symp- dom effect term was used to link these two submo-
toms or one of the four mCES-D factors, we recog- dels. In joint models of the four mCES-D profiles,
nized the need to account for possibly nonrandom the submodel of the main event was a linear mixed
attrition and mortality rates, as higher rates of model adjusted for age, sex, education, antidepres-
both are associated with depression. 10,36 There- sant usage, MMSE, and total prescription medica-
fore, we used a joint modeling approach with tions; and the submodel of time to attrition was a
shared random effects to simultaneously model Weibull proportional hazards model adjusted for
the time-varying main events (depression symp- age, sex, and education. A shared random effect
toms greater than or equal to 5 or the four mCES-D term was used to link the two submodels. Sex and
factors) and time to attrition to assess the associa- education were time-invariant covariates. Age,
tion between the main event and birth cohort antidepressant medication use, MMSE, and total
while adjusting for attrition. In the submodel of number of prescription medications were time-
greater than or equal to 5 depression symptoms, a varying covariates. R version 3.5.1 and SAS 9.3

102 Am J Geriatr Psychiatry 28:1, January 2020

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
 Sullivan et al.

(SAS Institute) were used for data analysis, includ- In the joint model of greater than or equal to 5
ing the NLMIXED procedure for the joint models. depression symptoms, we observed a significant over-
all birth cohort effect (F(3,3220) = 9.27, p <0.0001). Specif-
ically, participants from more recently born cohorts
were less likely than the 1902−1911 referent birth
RESULTS
cohort to have a study visit in which they endorsed
Descriptive statistics are provided for the overall greater than or equal to 5 mCES-D symptoms (Table 3).
analysis sample (n = 3,227) and by birth cohort in In testing additional pairwise comparisons of each birth
Table 2. The proportion of each birth cohort that had at cohort to each earlier born cohort, each subsequent birth
least one study visit with greater than or equal to 5 cohort was less likely than each earlier born cohort to
mCES-D symptoms was lower in the more recently have a study visit with greater than or equal to 5 depres-
born cohorts than in the earlier born cohorts (x2 = 12.19, sion symptoms (1922−1931 versus 1912−1921: odds ratio
df = 3, p = 0.007), a trend that was generally consistent [OR] = 0.652, 95% confidence interval [CI] OR: 0.464
in each age group, excluding age 85+ in which event −0.915; 1932−1941 versus 1912−1921: OR = 0.363, 95%
counts, and to some extent prevalence, were low. Addi- CI OR: 0.231−0.571; 1932−1941 versus 1922−1931:
tionally, the earlier born cohorts had lower education OR = 0.557, 95% CI OR: 0.350−0.886). However, after
(x2 = 459.32, df = 6, p <0.001), lower MMSE scores applying a Bonferroni correction of the type I error
(Kruskal-Wallis H = 217.71, df = 3, p <0.001), were less rate accounting for all six birth cohort comparisons
likely to take antidepressant medications (x2 = 105.21, (a = 0.05/6 = 0.0083), only the 1932−1941 versus
df = 3, p <0.001), and took fewer total number of 1902−1911, 1932−1941 versus 1912−1921, and 1922
prescription medications (Kruskal-Wallis H = 239.42, −1931 versus 1902−1911 comparisons were statisti-
df = 3, p <0.001). Lastly, and not unexpectedly, the ear- cally significant.
lier born cohorts came primarily from the earlier con- In joint models of the four mCES-D factor outcomes,
ducted MoVIES study (x2 = 1738.3, df = 3, p <0.001), we observed a significant birth cohort effect in three
and there was slightly lower average years of follow-up of the four factors: Depressed mood (F(3,3220) = 9.16,
in the birth cohorts that were older at baseline (F(1,3223) = p <0.001), anergia/hopelessness (F(3,3220) = 59.75,
17.75, p <0.001), particularly the 1902−1911 cohort. p <0.001), and withdrawal (F(3,3220) = 3.85, p = 0.009)

TABLE 2. Descriptive Statistics by Birth Cohort and Total Sample

1902−1911 1912−1921 1922−1931 1932−1941 Total p
Analysis sample, N 305 1202 1051 669 3227 —
Baseline age, mean (SD) 80.6 (4.2) 75.7 (8.2) 79.0 (5.5) 70.2 (2.9) 76.1 (7.1) <0.001a
Female sex, N (%) 204 (66.9) 727 (60.5) 634 (60.3) 400 (59.8) 1965 (60.9) 0.160b
<HS education, N (%) 173 (56.7) 420 (34.9) 152 (14.5) 42 (6.3) 787 (24.4) <0.001b
HS education, N (%) 52 (17.1) 484 (40.3) 497 (47.3) 296 (44.3) 1329 (41.2)
>HS education, N (%) 80 (26.2) 298 (24.8) 402 (38.3) 331 (49.5) 1111 (34.4)
Baseline antidepressant use, N (%) 7 (2.3) 51 (4.2) 113 (10.8) 112 (16.8) 283 (8.8) <0.001b
MYHAT study, N (%) 12 (3.9) 323 (26.9) 926 (88.1) 669 (100.0) 1930 (59.8) <0.001b
Baseline MMSE, median (Q1-Q3) 23.5 (21.8-26.0) 27.0 (26.0-28.0) 28.0 (27.0-29.0) 29.0 (28.0-30.0) 28.0 (27.0-2.09) <0.001c
Baseline number of Rx 2 (1-3) 2 (0-4) 4 (2-6) 4 (2-6) 3 (1-5) <0.001c
Medications, median (Q1-Q3)
Person-years, Mean (SD) 4.60 (3.64) 6.13 (4.08) 5.39 (3.62) 6.03 (3.46) 5.72 (3.80) <0.001a
Overall at least one visit with mCES-D ≥5, N (%) 70 (23.0) 229 (19.1) 168 (16.0) 102 (15.2) 569 (17.6) 0.007b
Baseline age between 65-74 (n = 1,540) - 153 (19.51) 15 (11.72) 97 (15.45) 265 (17.21 0.031d
Baseline age between 75−84 (n = 1,288) 64 (22.94) 25 (26.04) 148 (16.97) 5 (12.20) 242 (18.79) 0.025d
Baseline age 85+ (n = 399) 6 (23.08) 51 (15.84) 5 (9.80) - 62 (15.54) 0.326d

Notes: HS: high school; MYHAT: Monongahela Youghiogheny Healthy Aging Team; MMSE: Mini-Mental State Exam; mCES-D: Modified Center for
Epidemiologic Studies Depression Scale; SD: Standard Deviation; Q: Quartile; Rx: Prescription. p-values reflect tests for significant differences
between birth cohorts using the following tests:
a
ANOVA.
b
Pearson’s x2 test.
c
Kruskal-Wallis rank sum test.
d
Fisher’s Exact Test.

Am J Geriatr Psychiatry 28:1, January 2020 103

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
Depression Symptoms Declining Among Older Adults

0.161
0.117
0.088
<0.001
0.044
0.067
0.014
0.129
<0.001
0.086

Notes: Reference groups are born 1902−1911, Male Sex, <HS Education, and no antidepressant use. CI: Confidence Interval; Rx: Prescription; HS: high school; mCES-D: Modified Center
TABLE 3. Covariate Effects on Depression Symptoms (≥5

—
p
mCES-D)
OR 95% CI OR p

Poor Self-Esteem

0.005 to 0.001

0.011 to 0.001

0.002 to 0.001
0.015 to 0.003
0.015 to 0.002
0.017 to 0.001

0.000 to 0.006
0.001 to 0.000

0.001 to 0.001

0.000 to 0.001
Born 1902−1911 (referent) — — —

95% CI b
Born 1912−1921 0.566 0.341−0.939 0.028

—
Born 1922−1931 0.369 0.215−0.632 <0.001
Born 1932−1941 0.205 0.106−0.399 <0.001
Age 0.708 0.542−0.926 0.012
Female sex 2.658 1.963−3.600 <0.001
HS education 0.435 0.307−0.617 <0.001

0.006
0.007
0.008
0.003
0.003
0.005
0.006
0.004
0.001
0.000
>HS education 0.336 0.229−0.491 <0.001

—
b
Antidepressant use 2.670 1.882−3.787 <0.001
MMSE 0.856 0.824−0.890 <0.001
Number of Rx medications 1.152 1.104−1.203 <0.001

0.390
0.334
0.019
<0.001
<0.001
0.005
0.003
<0.001
<0.001
<0.001
—
p
Notes: Reference Groups are Born 1902−1911, Male Sex, <HS Educa-
tion, and no antidepressant use. CI: confidence interval; HS: high

0.033 to 0.003
0.010 to 0.003

0.020 to 0.003
0.021 to 0.004

0.005 to 0.003
school; mCES-D: Modified Center for Epidemiologic Studies Depression

0.020 to 0.008
0.021 to 0.007

0.016 to 0.027

0.013 to 0.034

0.001 to 0.003
Withdrawal

Model: Linear mixed submodel and Weibull proportional hazards time to attrition submodel joint with shared random effect terms.
Scale; MMSE: Mini-Mental State Exam; OR: odds ratio; Rx: Prescription.

95% CI b
Degrees of Freedom for all terms = 3220.

—
Model: Generalized linear mixed submodel with logit link and Wei-
bull proportional hazards time to attrition submodel joint with shared
random effect terms.

0.006
0.007
0.018
0.007
0.021
0.012
0.012
0.023
0.004
0.002
(Table 4). The birth cohort effect in each of these fac-

—
b
tors was fairly consistent with the primary analysis,
with participants in more recently born cohorts
<0.001
<0.001
<0.001
0.002
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
endorsing fewer symptoms in comparison to the 1902 —
p

−1911 birth cohort. The birth cohort effect was most

Anergia/Hopelessness

evident for depressed mood and anergia/hopeless-

0.109 to 0.053
0.160 to 0.105
0.179 to 0.121
0.016 to 0.004

0.045 to 0.015
0.055 to 0.025

0.001 to 0.006
0.011 to 0.030

0.012 to 0.045

0.007 to 0.010
ness. Only the 1932−1941 cohort reported fewer symp-
95% CI b

toms of withdrawal compared to the 1902−1911

—

referent cohort. We observed no cohort effect for poor

self-esteem (F(3,3220) = 1.00, p = 0.390). Figure 1 displays

for Epidemiologic Studies Depression Scale; MMSE: Mini-Mental State Exam.

trajectories in each mCES-D factor by birth cohort.
0.081
0.132
0.150
0.010
0.020
0.030
0.040
0.028
0.008
0.008
—
b

DISCUSSION
0.524
0.009
0.003
0.233
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
—
TABLE 4. Covariate Effects on mCES-D Factor Scores

The presented analysis investigated depression in

two ways: a categorical cutoff of greater than or equal
Depressed Mood

0.045 to 0.006
0.054 to 0.011

0.042 to 0.016
0.049 to 0.024

0.006 to 0.003
0.026 to 0.013

0.008 to 0.002
0.024 to 0.039

0.017 to 0.047

0.002 to 0.004

to 5 symptoms of depression on the total mCES-D, and

95% CI b

four factor analyzed components of the mCES-D.

—

Degrees of Freedom for all terms = 3220.

Across the four birth cohorts we examined, we report

significantly lower endorsement of depression symp-
toms on the mCES-D in more recently born cohorts.
0.006
0.025
0.033
0.003
0.032
0.029
0.037
0.032
0.005
0.003

Specifically, in comparison to the earliest born 1902

—
b

−1911 cohort, we observed 79% lower odds of having

greater than or equal to 5 depression symptoms in the
Antidepressant use
Born 1902−1911
Born 1912−1921
Born 1922−1931
Born 1932−1941

1932−1941 birth cohort, 63% lower odds in the 1922

>HS education

medications
Number of Rx

−1931 birth cohort, and 43% lower odds in the 1912

HS education
Female sex

−1921 birth cohort. From a public health and popula-

MMSE

tion trends standpoint, these reported effect sizes sug-

Age

gest a meaningful reduction in depression symptom

104 Am J Geriatr Psychiatry 28:1, January 2020

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
 Sullivan et al.

FIGURE 1. Predicted Average Number of Symptoms in Each mCES-D Factor by Birth Cohort and Age.

endorsement among older adults across successive birth younger age groups.19,21 While no previous studies
cohorts. Additionally, each subsequent birth cohort had have examined cohort effects in depression symp-
lower odds of having greater than or equal to 5 depres- toms among older adults, the Epidemiologic Catch-
sion symptoms when compared to each preceding birth ment Area (ECA) examined trends in MDD in
cohort, suggesting a continually decreasing endorse- cohorts born in the years 1905−1964, which overlaps
ment of depression symptoms across the examined partially with the birth years represented in the MoV-
cohorts, although this observation was not fully IES/MYHAT cohort.19 However, the ECA study
statistically significant with adjustment for multiple took place when the participants were considerably
comparisons. When considering the factor analyzed younger.37 To illustrate, the cumulative incidence of
components of the mCES-D, the declining trend was major depression in the ECA 1935−1944 birth cohort
reflected most in items involving depressed mood and was calculated up to age 44. In the MoVIES/MYHAT
anergia/hopelessness. birth cohort with the same birth decade (1932−1941),
Our reported declining rate of symptoms of we prospectively measured symptoms of depression
depression in older adults stands in contrast to the in participants who had an average age of 70 at base-
increasing or stable rates of MDD reported from line. The ECA investigators further indicated that the

Am J Geriatr Psychiatry 28:1, January 2020 105

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
Depression Symptoms Declining Among Older Adults

increasing depression trend appeared to be slowing investigate other secular trends, which might help to
in the most recently born cohorts. explain declining trends in depression symptoms.
In the present study, the cohort effect was unex- Future population research focused on depressive dis-
plained by changing patterns in education and antide- orders should include diagnostic assessment to detect
pressant medication usage, which were higher in the these disorders and determine whether there are also
more recently born cohorts. It is possible that lower declining trends in major depression in older adults.
reporting of depression symptoms is related to com-
pression of morbidity, with more recently born older
adults living longer into older age free from functional
CONCLUSION
impairment than earlier born older adults.22 Symptoms
of depression are more common in individuals with As life expectancy increases in the population,
medical conditions.38 However, we observed a signifi- reduced rates of depression symptoms in older adults
cant birth cohort effect even with adjustment for total might reflect compression of morbidity, or a declining
prescription medications, which we have used previ- trend in medical comorbidity. If confirmed in other
ously to represent medical morbidity.29,30 cohorts, the trend is a positive development that may
Given the reported associations between cognitive also influence other health outcomes.
impairment and depression,39 the observed cohort
trend may be related to trends evidencing compres- The authors would like to thank all of the participants and
sion of cognitive morbidity.13−16 However, while staff of the Monongahela Valley Independent Elders Survey,
lower MMSE was related to higher overall mCES-D and the Monongahela-Youghiogheny Healthy Aging Team.
symptoms and more endorsed items on each of the Conflict of Interests and Sources of Funding: KJS reports
four factors, including MMSE in our model did not grants from NIH, during the conduct of the study; AL
explain the observed birth cohort effect. reports grants from NIH, during the conduct of the study;
We have no obvious interpretation for the finding HHD reports grants from NIH, during the conduct of the
that there was a significant birth cohort effect in the study; CA reports grants from NIH, during the conduct of
expression of the Depressed Mood, anergia/hopeless- the study; CHC reports grants from NIH, during the conduct
ness, and withdrawal symptoms but not poor self- of the study; other from Alzheimer's Association, outside the
esteem symptoms. Potentially, it may suggest that submitted work; MG reports grants from the NIH, during
there is not only a quantitative change in depression the conduct of the study; personal fees and other from NIH
symptoms but a qualitative time trend in how older Center for Scientific Review, personal fees and other from Bio-
adults experience or express depression. This may be gen Pharma, personal fees and other from University of
of clinical relevance as different generations age into Southern California, other from Centre for Brain Research,
older adulthood and deserves further investigation. Indian Institute of Science, Bangalore, India, personal fees
and other from Dalhousie University, Halifax, NS, Canada,
other from Alzheimer's Association, other from Mt. Sinai
Strengths and limitations
Medical Center, Miami Beach, FL, other from University of
The two cohorts pooled in the present study were Texas Health Sciences at San Antonio, TX, outside the sub-
recruited from communities in economically depressed mitted work. This work was supported by the National Insti-
postindustrial regions and represent an underserved tute on Aging at the National Institutes of Health (R01
population rarely targeted for health research. The sta- AG023651, U01 AG06782, R01 AG07562, P30 AG053760,
ble population of the region facilitates longitudinal P30 AG008017 and T32 AG000181).
research. Both MoVIES and MYHAT were randomly
selected community-based samples which enhance their
external validity (generalizability) to the population at
SUPPLEMENTARY MATERIALS
large. As the present results are reported from four birth
cohorts that are predominantly white, our findings Supplementary material associated with this article
should be replicated in other cohorts with larger repre- can be found in the online version at https://doi.org/
sentations of ethnic minorities. Further analyses should 10.1016/j.jagp.2019.06.002.

106 Am J Geriatr Psychiatry 28:1, January 2020

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.
 Sullivan et al.

References
1. Hasin DS, Goodwin RD, Stinson FS, et al: Epidemiology of major 21. Twenge JM: Time period and birth cohort differences in depressive
depressive disorder: results from the National Epidemiologic Sur- symptoms in the US, 1982-2013. Soc Indic Res 2015; 121:437–454
vey on Alcoholism and Related Conditions. Arch Gen Psychiatry 22. Fries JF: Aging, natural death, and the compression of morbidity.
2005; 62:1097–1106 N Engl J Med 1980; 303:130–135
2. American Psychiatric Association: Diagnostic and Statistical Man- 23. Folstein MF, Folstein SE, McHugh PR: “Mini-mental state”. A prac-
ual of Mental Disorders (DSM-5). American Psychiatric Pub, 2013 tical method for grading the cognitive state of patients for the cli-
3. Meeks TW, Vahia IV, Lavretsky H, et al: A tune in “a minor” can nician. J Psychiatr Res 1975; 12:189–198
“b major”: a review of epidemiology, illness course, and public 24. Mungas D, Marshall SC, Weldon M, et al: Age and education cor-
health implications of subthreshold depression in older adults. rection of Mini-Mental State Examination for English and Spanish-
J Affect Disord 2011; 129:126–142 speaking elderly. Neurology 1996; 46:700–706
4. Pietrzak RH, Kinley J, Afifi TO, et al: Subsyndromal depression in 25. Ganguli M, Dodge HH, Chen P, et al: Ten-year incidence of
the United States: prevalence, course, and risk for incident psy- dementia in a rural elderly US community population: the MoV-
chiatric outcomes. Psychol Med 2013; 43:1401–1414 IES Project. Neurology 2000; 54:1109–1116
5. DiMatteo MR, Lepper HS, Croghan TW: Depression is a risk fac- 26. Ganguli M, Belle S, Ratcliff G, et al: Sensitivity and specificity for
tor for noncompliance with medical treatment: meta-analysis of dementia of population-based criteria for cognitive impairment:
the effects of anxiety and depression on patient adherence. Arch the MoVIES project. J Gerontol 1993; 48:M152–M161
Intern Med 2000; 160:2101–2107 27. Ganguli M, Chang CC, Snitz BE, et al: Prevalence of mild cogni-
6. Alexopoulos GS: Depression in the elderly. Lancet 2005; tive impairment by multiple classifications: the Monongahela-
365:1961–1970 Youghiogheny Healthy Aging Team (MYHAT) project. Am J Ger-
7. Barnes DE, Alexopoulos GS, Lopez OL, et al: Depressive symp- iatr Psychiatry 2010; 18:674–683
toms, vascular disease, and mild cognitive impairment: findings 28. Ganguli M, Fu B, Snitz BE, et al: Mild cognitive impairment: inci-
from the Cardiovascular Health Study. Arch Gen Psychiatry dence and vascular risk factors in a population-based cohort.
2006; 63:273–279 Neurology 2013; 80:2112–2120
8. Green RC, Cupples LA, Kurz A, et al: Depression as a risk factor 29. Ganguli M, Dodge HH, Mulsant BH: Rates and predictors of mor-
for Alzheimer disease: the MIRAGE Study. Arch Neurol 2003; tality in an aging, rural, community-based cohort: the role of
60:753–759 depression. Arch Gen Psychiatry 2002; 59:1046–1052
9. Conwell Y, Duberstein PR, Caine ED: Risk factors for suicide in 30. Mulsant BH, Ganguli M, Seaberg EC: The relationship between
later life. Biol Psychiatry 2002; 52:193–204 self-rated health and depressive symptoms in an epidemiological
10. Blazer DG: Depression in late life: review and commentary. J Ger- sample of community-dwelling older adults. J Am Geriatr Soc
ontol A Biol Sci Med Sci 2003; 58:249–265 1997; 45:954–958
11. Ludvigsson M, Bernfort L, Marcusson J, et al: Direct costs of very 31. Ganguli M, Gilby J, Seaberg E, Belle S: Depressive symptoms and
old persons with subsyndromal depression: a 5-year prospective associated factors in a rural elderly population: the MoVIES Proj-
study. Am J Geriatr Psychiatry 2018; 26:741–751 ect. Am J Geriatr Psychiatry 1995; 3:144–160
12. Cauley J: The demography of aging. In: AB N, Cauley J, eds. The 32. Radloff LS: The CES-D Scale: a self-report depression scale for
Epidemiology of Aging, Dordrecht: Springer, 2012:3–14 research in the general population. Appl Psychol Measure 1977;
13. Sullivan KJ, Dodge HH, Hughes TF, et al: Declining incident 1:385–401
dementia rates across four population-based birth cohorts. J Ger- 33. Andreescu C, Chang CC, Mulsant BH, Ganguli M: Twelve-year
ontol A Biol Sci Med Sci 2018;doi:10.1093/gerona/gly236 depressive symptom trajectories and their predictors in a commu-
14. Dodge HH, Zhu J, Lee CW, et al: Cohort effects in age-associated nity sample of older adults. Int Psychogeriatr 2008; 20:221–236
cognitive trajectories. J Gerontol A Biol Sci Med Sci 2014; 34. Shafer AB: Meta-analysis of the factor structures of four depres-
69:687–694 sion questionnaires: Beck, CES-D, Hamilton, and Zung. J Clin Psy-
15. Dodge HH, Zhu J, Hughes TF, et al: Cohort effects in verbal mem- chol 2006; 62:123–146
ory function and practice effects: a population-based study. Int 35. Carleton RN, Thibodeau MA, Teale MJ, et al: The center for epi-
Psychogeriatr 2017; 29:137–148 demiologic studies depression scale: a review with a theoretical
16. Langa KM, Larson EB, Karlawish JH, et al: Trends in the preva- and empirical examination of item content and factor structure.
lence and mortality of cognitive impairment in the United States: PLoS One 2013; 8:e58067
is there evidence of a compression of cognitive morbidity? Alz- 36. Lamers F, Hoogendoorn AW, Smit JH, et al: Sociodemographic and
heimers Dement 2008; 4:134–144 psychiatric determinants of attrition in the Netherlands Study of
17. Center for Disease Control: Ten great public health achieve- Depression and Anxiety (NESDA). Compr Psychiatry 2012; 53:63–70
ments - United States, 2001-2010. MMWR Morb Mortal Wkly Rep 37. Regier DA, Myers JK, Kramer M, et al: The NIMH epidemiologic
2011; 60:619–623 catchment area program. Historical context, major objectives,
18. Snyder TD: 120 Years of American Education: A Statistical Por- and study population characteristics. Arch Gen Psychiatry 1984;
trait. Washington, DC: United States Department of Education, 41:934–941
National Center for Education Statistics, 1993 38. Alexopoulos GS, Buckwalter K, Olin J, et al: Comorbidity of late
19. Wickramaratne PJ, Weissman MM, Leaf PJ, Holford TR: Age, period life depression: an opportunity for research on mechanisms and
and cohort effects on the risk of major depression: results from five treatment. Biol Psychiatry 2002; 52:543–558
United States communities. J Clin Epidemiol 1989; 42:333–343 39. Geda YE, Roberts RO, Knopman DS, et al: Prevalence of neuro-
20. Joyce PR, Oakley-Browne MA, Wells JE, Bushnell JA, Hornblow psychiatric symptoms in mild cognitive impairment and normal
AR: Birth cohort trends in major depression: increasing rates and cognitive aging: population-based study. Arch Gen Psychiatry
earlier onset in New Zealand. J Affect Disord 1990; 18:83–89 2008; 65:1193–1198

Am J Geriatr Psychiatry 28:1, January 2020 107

Downloaded for Janu IPD (ck01@fk.uns.ac.id) at Sebelas Maret University from ClinicalKey.com by Elsevier on June 03, 2020.
For personal use only. No other uses without permission. Copyright ©2020. Elsevier Inc. All rights reserved.