Back to - - > Measurement matrix
|
Table 1a: Studies Assessing Cross-Cultural Use of CES-D
as Continuous Variable with Older Adults |
|||||
|
Author |
Study Site |
Sample Characteristics |
Study Methodology |
Response
Rate |
Results |
|
Baker
et al., 1996 |
|
N
= 96 African Americans age 60+ drawn from home health agency in |
Assess
presence of depressive symptoms, using the 20-item CES-D, in African-
American community-dwelling elders. Multiple
regression analysis examined the association of medical illness,
medication use, social network, and functional impairment on depressive
symptoms. |
87% |
19.8%
scored $16 on CES-D. Persons with 6 + chronic illnesses and those using 4 +
prescription medications had higher scores, as did urban dwellers.
Strongest predictors of depressive symptoms were medical illness and
social networks. |
|
Barón
et al., 1989 |
|
N
= 314 American Indians age 45+ from 4 |
Principal
Components Factor Analysis with varimax rotation
|
38% |
Cronbach's
alpha = .86. Four-factor model explained 49% of total variance. Factor
loadings did not correspond to previous research. No specific somatic
factor was identified and the positive affect factor separated into two
factors. |
|
Blazer
et al., 1998 |
|
Data
are from Duke site of the Established Populations for Epidemiologic
Studies of the Elderly (EPESE). N=3041
community-dwelling |
Confirmatory
factor analysis and LISREL used to confirm four-factor structure;
bivariate analysis used to determine prevalence of individual symptoms by
race; LISREL performed to control for potential confounding demographic
and social risk factors. |
|
Four-factor
structure replicated. In bivariate analysis, Blacks more likely to report
less hope about future, poor appetite, trouble concentrating, more effort
for usual activities, less talking, and feeling people are unfriendly. In
multi-variate analysis, only the racial difference in interpersonal
relations persisted. |
|
Boey,
1999 |
|
N=554
community-dwelling Chinese elders in N=
31 patients from a psychogeriatric assessment clinic of a voluntary agency
participated in small-scale validation study.
|
Test
the reliability of the 10-item short form of the CES-D cross-sectionally
and longitudinally. Examined associations between CES-D-10 and ADLs, Life
Satisfaction Scale (LLS), Lubben Social Network Scale and self-rated
health. |
|
Cronbach’s
alpha of CES-D-10 =.79. consistency over 3 years (r = .44, p <.01);
accuracy of classification of cases with depressive symptoms comparable to
20-item form (Kappa =.84); significant correlations with ADLs, life
satisfaction, social support, and self-rated health.
|
|
Callahan
& Wolinsky, 1994 |
|
N=3047
Urban primary care setting. Overall mean age 69, range 60-102; 30% male;
65% Black, 34.5% White Black
women (mean age 69.8; £
8 yrs
educ, 40%; £ $800/mo income, 85%) White
women (mean age 67.2, £
8 yrs
educ, 43%; £
$800/mo income, 84%) Black men (mean age 69.5; £
8 yrs educ,
54%; £
$800/mo income, 72%) White
men (mean age 67.4; £
8 yrs educ,
51%; £
$800/mo income, 66%) |
Separate
principal components factor analysis performed for each of the four
race/gender groups and for respondents with imputed values for missing
data (37%) and those who answered all items. Imputed values were most
common for the positive affect construct. Cognitive impairment,
alcoholism, and chronic illnesses also assessed. |
92% |
Cronbach’s
alpha=.85 (both male groups, .80, Black women, .84, white women, .87).
Factor structures for race/gender groups differed markedly, as did groups
with and without imputed data. Disparities were not resolved by removing
persons with low education, cognitive impairment, alcoholism, or varying
assumptions for imputation. But were resolved by eliminating 5 items,
suggesting a need to modify the CES-D in these populations. |
|
Chapleski
et al., 1997
|
|
N=277
American Indian elders age 55+ in eastern |
Confirmatory
factor analysis (LISREL-8) used to examine factor structure of CES-D and
test alternate models for the full sample and for the three residential
sub-groups. The 20-item and
12-item versions were evaluated. |
73% |
12-item
version of Liang et al. (1989) for Mexican Americans provided superior fit
over 20-item version for overall sample. Factor structure and factor
loadings were similar across residential strata. Cronbach’s alpha for
the 12-item scale = .83. |
|
Chi
& Boey, 1993
|
Hong Kong |
N=91
Chinese elders in 1.
Normals (N=31; age 60-69, 45%, 70-79, 39%, 80+, 16%; 42% male, 48%
married; 53% no formal educ) 2.
Normals matched to clinical group (N=30; age 60-69,17 %, 70-79, 53%, 80+,
30%; 53% male, 53% married; 43% no formal educ) 3.
Clinical (N=31; age 60-69, 16%, 70- 79, 45%, 80+, 26%; 26% male, 29%
married; 81% no formal educ). |
Evaluated
validity and reliability of four geropsychiatric screening instruments:
CES-D, Life Satisfaction Index-A form, General Health Questionnaire, and
Short Portable Mental Status Questionnaire |
|
CES-D:
Cronbach's alpha =.80, split-half=.74. Test-retest not performed.
Discriminated well between matched normal (mean = 28.1) and clinical (mean
= 38.0) group (p <.005). High
correlation with clinical assessments of psychiatrists. |
|
Davidson,
Feldman, & Crawford, 1994 |
|
N
= 303 frail elders in New York City area senior housing (mean age 79.8;
15% male; 15% married; 47% White and 53% non-white-- 26% identified as
Hispanic)
|
Confirmatory
factor analysis used to test 4-factor structure of the CES-D and the
existence of a single underlying second-order factor. Also examined
contribution of somatic factor to total symptom score and associations of
age, race, functional status and health with the four factors. |
|
Overall
mean score 18.6 (whites, 21.6; non-whites, 15.9). Four-factor model
replicated and one underlying second-order factor found. Somatic items did
not unduly impact total scores. Age and health did not affect somatic
subscale more than other subscales. Functional limitation was associated
with higher somatic scores. |
|
Krause
& Liang, 1992 |
|
Data
are from three nationwide surveys. 1)
N = 1094 American whites (mean age= 70 ±
7.4; 33% male; yrs educ 11.2 ±
3.3) N=464 American Blacks (mean age=69.8
± 7.4; 32% male; yrs educ 8.4 ± 3.9). 2)
N = 2041 Japanese (mean age 69.0±
6.7; 45% male; yrs educ 8.7 ± 2.8). 3)
N=3865 Taiwanese (mean age = 68.6 ±
6.4; 43% male; yrs educ 4.0 ±
4.6) |
Using
an 11-item CES-D, ANOVA and ANCOVA (adjusting for age, sex, and education)
used to estimate the prevalence of depressive symptoms in four culturally
diverse groups (overall scale) and to evaluate cultural variations in the
ways symptoms are manifest (sub-scales). |
91.8% |
Japanese
elders had lowest overall levels of depressive symptoms, then Taiwanese,
American whites, and American Blacks. Hypotheses that Japanese elders
express depressive symptoms as interpersonal complaints and Taiwanese as
somatic symptoms were not supported. Americans had higher scores than
Asians on all three symptom clusters (depression, somatic, interpersonal).
No major cross-cultural variations in symptom manifestation. |
|
Krause
& Liang, 1993 |
|
N
= 2721 Chinese in |
LISREL-7
was used to estimate a model of relationships among stress, social
support, and depressive symptoms. The latter was assessed by 3 latent
constructs: depressed affect, somatic symptoms, and positive affect.
Exploratory factor analysis of the CES-D |
83% |
Original
four-factor solution replicated. Financial strain was associated with
economic support but also related to high levels of somatic symptoms, low
levels of positive self-evaluation, and high levels of depressive affect. |
Mackinn-on
et al., 1998 |
|
Data
are from the WHO South East Asian Regional Office (SEARO) |
Confirmatory
factor analysis (LISREL-8.2) was used to test fit of four-factor model,
relative importance of four sub-scales and adequacy of a single
second-order factor. Used 17 of the 20 CES-D items (3 were deemed
culturally inappropriate) and a 3-point response scale. |
|
Original
four-factor solution replicated for all countries. A single-factor model
adequately fit the data at all sites and sub-scales added little
additional information. Comparable performance of CES-D in these countries
and North American and European cultures. There was no evidence that
somatization predominated in these countries, but items with high loadings
on well-being loaded low on general factor. |
|
Mahard,
1988 |
(NYC) |
N
= 60 Puerto Ricans aged 55-82 in |
Assessed
internal consistency; ability to discriminate clinical and non-clinical
populations; construct validity (in relation to stress, coping, and social
desirability). |
|
Cronbach’s
alpha=.87 in patient and non- patient groups.
CES-D discriminates well between patients (mean 29.5 ± 12.8) and nonpatients (mean 16.2 ±10.6) at p <.001. Higher rates of
depressive symptoms were related to increased levels of stress and fewer
coping resources. Scores may be affected by social desirability.
|
|
McCall-um
et al.,
1995 |
|
N
= 2805. Prospective community study of Australians. Age 60 +;
Study population generally representative of Australian-born
elders in the Australian population (44% male) |
Confirmatory
factor analysis (LISREL 7.20) used to test the four-factor model; assess
the adequacy of a single second-order depression factor; and examine
associations of factors with age, gender, and education. 16-item CES-D and
3 response categories used. |
73% |
Original
four-factor solution replicated. Aligns with research on North American
populations and confirms the exceptional functioning of well-being scale
in |
|
Miller,
Markides, & Black, 1997
|
|
Data
are from the Hispanic EPESEa
N=2536
Hispanic Americans (mean age=
73; 42.5% male; 44% born in N=330
Three Generation Study of Mexican Americans in |
Exploratory
factor analysis to explore viable factor structures for the CES-D.
Confirmatory factor analysis (MIMIC) to test predictive strength of
background vars. (age, gender, education, language of interview) and
dimensionality of CES-D. |
86%
in H-EPESE |
Two
factors identified: Depression (a=
.90) and Well-being (a=.88). Affect, somatic, and interpersonal
scales highly correlated; well-being not as highly correlated with others.
Goodness-of-fit lower for
Three Generation Study data. |
|
Somervell
et al., 1992 |
|
N=120 Northwest Coast American Indian elders. |
Exploratory factor analysis
|
|
Factor
structure differed from that in the literature. No clear distinction
between depressed affect and somatic factors |
|
Tran,
1997 |
|
Data
are from 1986 National Survey of Americans’ Changing Lives. Black
females (N=438 aged 24-59, yrs educ 11.8 ±
2.7, 34.5% married; N = 340 aged 60+, yrs educ 8.6 ±
3.7, 27.6% married; White females (N = 652 aged 24-59, yrs educ 12.7 ±
2.6, 64.7% married; N = 764 age 60+, yrs educ 11.2 ±3.1, 51.2% married. |
Exploratory
factor analysis with oblique rotation used to evaluate equivalence of
factor structure of 11-item CED-D across groups. A four-factor restriction
for each group was used to compare the factor structure of this short
version with the original 20-item version. |
|
Factor
structures for the 11-item CES-D differed among the four age by race
categories. Factor structure for the 11-item version also differed from
that of the 20-item scale. |
|
Table 1b: Studies Assessing Cross-Cultural Use of CES-D
as Screening Instrument with Elderly Persons |
|||||
|
Author |
Study
Site |
Sample Characteristics |
Study Methodology |
Response
Rate |
Results |
|
Baker
et al., 1995 |
|
N = 39 psychiatric patients age 50+ (49 % African American; 61% White; 51% 70-92 years old, 77% female; 51% widowed. |
All patients had diagnosis of affective disorder confirmed by SCID interview. |
|
Sensitivity=71% for Blacks and 85% for Whites. Specificity not reported. |
|
Barón
et al., 1989 |
|
N
= 314 American Indians age 45+ from 4 |
Criteria:
DSM-III and RDC diagnoses. Receiver Operating Characteristics (ROC)
procedures used to determine optimal cut-point. |
38% |
Sensitivity
= 100% for both criteria; specificity 73% and 71% relative to DSM-III and
RDC respectively. ROC analysis
suggests cut-point of 24 to
maintain high sensitivity and reduce false positive rate. |
|
Beekman
et al., 1997 |
|
N
= 487 Longitudinal Aging Study, |
Baseline
interview included CES-D; those scoring ³
16 received adapted version of DIS including DSM-III affective and anxiety
disorders sections to establish criterion validity. Other background,
cognitive, and physical health data included. |
81.7% |
Weighted
sensitivity 100%; specificity 88%; positive predictive value 13.2%. False
positives not more likely among elders with physical illness, cognitive
decline, or anxiety. |
|
Madianos
Gournas, & Stefanis, 1992 |
|
N
= 251 residents of two boroughs of |
Affective
disorders estimated by a clinical examination with semi-structured
psychiatric interview (PEF) supplemented by DSM-III criteria.
|
82% |
27.1%
dysphoric or depressive symptoms. Cut-off
³16, sensitivity and specificity = 83.4% and
85.9%. Construct validity: lower SES, widowed, high stress and live alone
associated with more depressive symptoms. Test-retest reliability =.76.
|
|
Papassoti-ropoulos
& Heun, 1999 |
|
N=287
Stratified random sample drawn of community-dwelling persons age 60+ in
|
Compares
the performance of the CES-D and the General Health Questionnaire-12 using
DSM-III-R criteria. |
|
Prevalence
of depression was 3.5%. Both discriminated well between depressed and
nondepressed subjects, but also high rates of false positives (CES-D,
90.1%). ROC analysis showed optimal cut-off for case identification = 9 /
10, with corresponding sensitivity and specificity 75% and 72%. |
|
Somervell
et al., 1993 |
|
N=120 Northwest Coast American Indian elders. |
Criterion was DSM-III-R diagnosis derived from the Lifetime version of the Schedule for Affective Disorders and Schizophrenia. |
|
Sensitivity
for major depression 100%; specificity, 82.1%. For broad category of
depressive disorders, sensitivity = 77.8%;
specificity, 84.7%. ROC derived cut points did not improve
performance of CES-D. |
Table 2a: Studies Assessing Cross-Cultural Use of
GDS as Continuous Variable with Older Adults
|
|||||
Authors
|
Study Site |
Sample
Characteristics
|
Study Methodology |
Response Rate |
Results |
|
Al-Shammari
& Al-Subaie, 1999 |
|
N=7,970
elders, age 60+ in Saudi Arabia sampled from primary health care centers (PHCs)
for the first-stage sampling and family health records in a PHC catchment
area for the second-stage sampling, (overall mean age 68.8 ± 7.7, male
69.1 ± 7.7, female 67.7 ± 7.7; urban 65.9%, rural 31.7%, remote 2.5%;
single 9.4%, married 70.1%, divorced/widowed 20.5%; illiterate 79.3%; not
working 55.1%) |
GDS
long form Bivariate
analyses by socio-demographic characteristics, housing, financial status,
diagnoses and medication, living arrangements, loss of close relative,
recreational activities, health perceptions, ADL and other health
problems. ICD-9 used to
classify clinical diagnoses via history, physical
examinations, and appropriate lab tests. |
Male 98.8% Female
79.8% |
Prevalence
rate for severe depressive symptoms was 8.4%.
Correlates of depression were poor education, unemployed, old age,
being female, divorced or widowed, and living in remote rural area (social
association). Also, more
common among those who had poor housing arrangements and limited privacy. |
|
Black
et al., 1995 |
|
N=47
Danish geriatric patients |
GDS
long form Using
Danish GDS, each patient was tested by two physicians and one nurse.
Samples stratified by Merck Manual Geriatrics diagnoses groups. |
62% |
Inter-rater
correlations ranged from .89 to .92. Data
suggested that Danish translation of the GDS has a high level of
inter-rater precision, independent of whether the rater was a physician or
nurse. |
|
Ferraro
et al., 1997 |
|
N=22
Native American elders from reservation in North Central Dakota who were
fluent in English (overall mean age 66.3±10.2,
average educational level 8.9±3.8) |
GDS
short form Other
demographic and psychometric information collected (sex, self-rated
health, types/numbers of medications currently taken, and performance on
the vocabulary sub-test of the WAIS-R). Correlation coefficient
calculated. |
100% |
Prevalence
rate 23% (elders scored between 6-15).
The result was higher than in previous studies that have obtained
GDS long form from non-Native American elders. |
|
Haller
et al., 1996 |
|
N=880
European elders living in different towns born between 1913 to 1918, ages
74 to 79 years old (male N =433, female N=452 |
GDS
short form Sample
stratified by age and sex. Mental
health, food intake and anthrometric data, blood samples and other data
were also collected. Bivariate
analyses and correlation coefficients examined. |
100% |
Prevalence
rate 11.6% for men and 27.5% for women.
Data suggested that the prevalence of depression was high in this
sample (one-fifth of the sample). Significant
differences in GDS scores according to the different types of education
and in women. |
|
Liu
et al., 1997 |
|
N=1,313
Chinese elders in rural area of Kinmen, an islet located west of Taiwan
and off the coast of mainland China, >
65 years of age in two of the four towns on this islet (male N=880, female
N=1175; no education 67%, less than 6 years of education 93%: farmers 47%) |
GDS
short form Retained
those who scored ³
5 on the GDS short form, then stratified by depressive symptoms.
Bivariate analyses tested. |
64% |
Sensitivity
or specificity was not computed. Based
on DSM III-R criteria, 13% were diagnosed as having depression, using
cutoff ³
5, 26%-screened positive on GDS. |
|
Mui,
1996a |
NYC |
N=50
immigrant Chinese elders in New York City who live in the community ages
62-91 (male N=25, female N=25; overall mean age 75.1±6.5;
most subjects finished 8 years of education; average length of stay in
U.S. 19 years, all born abroad; over 80% received less than $500/month
from SSI or Social Security) |
GDS
long form Principal
components analysis with varimax rotation to examine factor structure of
the new GDS short form. |
100% |
GDS
long form was reliable (alpha=.90; split half=.82), GDS short form was not
as reliable as long form (alpha=.72), GDS new short form (alpha=.89).
New short form may be culturally more sensitive. Two-factor
solution was identified. |
|
Zalsman
et al., 1998 |
|
N=27
clinically depressed inpatient elders ages 62-91 and N=21 normal healthy
elders ages 62-85 in |
GDS
short form |
100% |
Inter-rater
reliability was maximal (kappa=1.0), test-retest reliability (kappa=.88).
Neither sensitivity nor specificity was determined. |
Table 2b: Studies Assessing Cross-Cultural Use of
GDS as Depression Screening with Older Adults
|
|||||
Authors
|
Study Site |
Sample
Characteristics
|
Study Methodology |
Response Rate |
Results |
|
Abas
et al., 1998 |
|
N=164
African Caribbean primary care migrant elders, age 60+ in south London
(54% women, 46% men; overall mean age 68.3±5.9; 84%
born in Jamaica; Mean and median years living in UK 36±3.8) |
GDS
short form Standardized
psychiatric diagnosis of depression and specific diagnosis of cultural
specific “depressed/lost spirit” were used as criteria for validation
of GDS. |
71% |
Cutoff
³ 5, sensitivity 82%; specificity 62%. Modest
agreement between the medical and the culture-specific approaches to
diagnosing depression. Performance
was best at the lower cut-off of 4; African-Caribbean adults may be
significantly depressed when they admit to a small number of symptoms
regarding low mood. |
|
Baker
et al., 1993 |
|
N=58
African American elders living in senior citizen complex N=41
Mexican American age 55+ with affective disorder in-patients in |
GDS
short form Depressive
symptoms were stratified by DSM-IIIR diagnoses using bivariate analyses. |
African
Americans 94% Mexican
Americans 72% |
African
Americans: Cutoff score ³6
yields sensitivity of 35% and specificity 100%.
Cut-off score ³ 4, sensitivity improved to 64% and
specificity 95%. Mexican Americans: Cutoff ³6,
sensitivity 64% and cutoff
³
4 improved sensitivity to 75%. Specificity
not calculated as total Mexican sample had affective disorder. |
|
Chan,
1996 |
|
N=461
Chinese psychiatric out-patients in Hong Kong 60+ (males N=167, females
N=290; overall mean age 70.2±7.27; married 54.3%, widowed 37.4%, other
8.2%; income < city median
HK$5,170 /mo. 91.4%, at or below public assistance levels 59.7%; no
education 42.1%) |
GDS
long form Discriminant
analysis using canonical discriminant functions (linear regression) tested
to compare original results using the Research Diagnostic Criteria. |
99% |
Sensitivity
70.6%, specificity 70.1%, false negative 29.4%, false positive 29.9% was
not satisfactory. Internal
consistency reliability .89 (alpha); test-retest reliability
.85. Criterion-related
(psychiatrist diagnosis) validity was good at .95 and concurrent validity
(with CES-D) was .96. |
|
Cwikel
& Ritchie, 1989 |
|
N=20
clinically depressed elders in outpatient psychiatric clinics matched with
N=20 normal elders from Jerusalem, Israel (age of samples ranged from
60-84; men N=30, female N=7; nationalities consisted of Russian N=9,
Romanian N=3, North African n=2, Western European n=4, and Israeli n=2; no
education n=1, 1-8 years of education n=7, 8+ years of education n=12) |
GDS
short form Stratified
by dementia and clinical status, which were checked by neurologist and
clinical psychologist. |
100% |
Cutoff
³ 7, sensitivity 70%, specificity 75%.
Thus, use of short GDS with heterogeneous population of elders, as
many as 30% who experienced depression may be missed and that among those
who were not clinically depressed, 25% may be wrongly diagnosed as
depressed. Low levels of sensitivity and specificity may be due to
cultural reasons. Jewish
elders may be reluctant to report feelings that might reflect a lack of
faith. |
|
Espino
et al., 1996 |
|
N=48
community-based, monolingual and bilingual Mexican American elders 65+,
recruited from a county-financed community-based psychiatric care clinic,
and the VA community-based outpatient psychiatric unit (overall mean age
75.24 ± 9.07, female 48%, hypertension 27%, diabetes mellitus 22.9%, arthritis
18.7%) |
GDS
long form Screened
elders with depression that were treated for 3+ weeks, severe dementia by
Mini Mental Status Exam, and unstable bipolar illness.
Principal components factor analysis with varimax rotation was
conducted and 5-factor solution was identified. |
73% |
Cutoff
³ 11, sensitivity 80%, specificity 61%.
Based on the clinical diagnosis, sensitivity 80% and specificity
50%. The factor structure
obtained is different from other studies on non-Mexican elders.
This suggests that these signs and symptoms of depression in older
Mexican Americans may be different than in other elderly Americans. |
|
Izal
& Montorio, 1993 |
|
N=60
elderly living in public residence attached to the National Institute of
Social Services in Spain (Overall mean age 84; female 61%, male 39%;
widowed 61%, married 23%, single 16%; average period of stay in residence
7 years) |
GDS
long form |
100% |
Cutoff
³ 11 showed prevalence of 50%, cutoff
³
14 showed prevalence of 35%. Alpha
coefficient .89, test-retest reliability .89 |
|
Lee
et al., 1993 |
|
N=113
normal Chinese elders and N=80 clinically depressed elders in |
GDS
short form Interviewed
and retained those without any significant medical and psychiatric
problems. Diagnoses set by DSM
III-R criteria. Discriminant analysis was used. |
100% |
Cutoff
³ 8, Sensitivity 96.3%, specificity 87.5%. |
|
Woo
et al., 1994 |
|
N=1,611
Chinese elders living in community in |
GDS
short form Univariate
analyses and stepwise logistic regression used. |
60% |
Cutoff
³ 8, sensitivity 96.3%, specificity 87.5%.
Prevalence rate 29.2% for males and 41.1% for females |
Back to TOP Measurement matrix