Pickard et. al.

< - - Back to     Measurement matrix         MEGs Search

Quick Links:   Sample   Response Rate   Internal Consistency    DIF Method    DIF Evidence    DIF Impact    Expert Opinion

Pickard et. al.   ( see abstract )

Center for Epidemiologic Studies-Depression Scale (CES-D)

Name/ Reference

Pickard AS, Dalal MR, Bushnell DM. A Comparison of Depressive Symptoms in Stroke and Primary Care: Applying Rasch Models to Evaluate the Center for Epidemiologic Studies-Depression Scale. Value in Health, (2006) 9(1):59-64.

Source contact info

A. Simon Pickard, College of Pharmacy, University of Illinois at Chicago, Room 164, MC 886, 833 South Wood Street, University of Illinois At Chicago, Chicago, IL 60612, USA.
E-mail: pickard1@uic.edu

Availability (private or public)

Public

Conceptual framework

Authors stated that IRT-based models may be used to inform clinicians and outcomes researchers about whether differences exist between primary care- and stroke-based depression based on patterns of responses to the CES-D. “The study utilized IRT-based models to compare responses to symptom-related items on the CES-D by depressed stroke and primary-care patients to: 1) illustrate the use of IRT-based (Rasch) models for comparing scale functioning across different patient subgroups; and 2) inform clinicians and outcome researchers about CES-D scale functioning and depressive symptomatology in stroke- compared with primary care-based depression.”

Purpose of measure & application (clinical, research, survey, screening)

depression screening in the general population

Sample characteristics

Of the 101 stroke patients with complete item responses to the CES-D at 3 months (three had incomplete data), 32 patients (32%) were defined as depressed according to CES-D scores.

Depressed Stroke patients: mean (sd) age = 69 years (14) and 59% were female.

Depressed primary care-based patients: mean (sd) age= 42 years (15), and the 67% were female.

Recruitment methods

Secondary data from two sources:

1) Stroke patient data were collected as part of a study of health outcomes after stroke (see Pickard et al 2004 below). “Patient assessments of the CES-D collected at 3 months postrecruitment were analyzed. Patients who were diagnosed with ischemic stroke were recruited with 2 weeks of stroke before discharge from two large teaching hospitals in Edmonton , Canada . Participants had to be 18 years of age or older. Patients were not eligible if they had receptive aphasia, were cognitively impaired, unable to comprehend the English language, or had a very poor prognosis, based on consultation with attending clinicians” (pg 60).

2) “Data on USA-based primary-care patients with depression were obtained from the Longitudinal Investigation of Depression Outcomes (LIDO) study. The LIDO project was a longitudinal study of depressive symptoms, quality of life and health services use among primary-care patients in Barcelona (Spain), Be’er Sheva (Israel), Melbourne (Australia), Porto Alegre (Brazil), St. Petersburg (Russia), and Seattle (the United States). At each site, consecutive adult (age 18– 75) visitors were invited to participate in the study. After providing informed consent, participants completed a screening questionnaire that included the CESD”. (See Herrman et al 2002; Bech et al 2003. below) (pg 60)

Data collection method

questionnaires

1) Stroke patients (see Pickard et al 2004 below)

2) USA-based primary-care patients with depression (See Herrman et al 2002; Bech et al 2003. below)

Response rate

Not provided

Format & design (readability, # of items, time to complete, response categories)

“The CES-D is a 20-item scale designed to measure depressive symptoms experienced in the past week. Responses are interpreted based on a simple summary score, calculated by summing the item responses. Patients were categorized as depressed if CES-D scores were 16 or higher, a threshold used in community-based studies and in stroke. Higher scores are associated with more frequent depressive symptoms.”

Type of measurement (nominal, ordinal, interval, ratio)

ordinal

Scoring (range, direction, rules, missing data)

Not Provided

Availability of translations & source

Not Provided

Psychometric Properties:

Scale construction

Not Provided

Basic summary statistics

CES-D scores for Depressed Stroke patients: mean (sd) age = 69 years (14); for Depressed primary care-based patients: mean (sd) age= 42 years (15).

Variability

Not provided

Test-retest reliability

Not provided

Interrater reliability

Not provided

Internal consistency

Not provided

Content validity

Not provided

Construct validity

Not provided

Concurrent validity

Not provided

Predictive validity

Not provided

Sensitivity to change

Not provided

Differential Item Functioning (DIF):

Variable studied (e.g., groups)

Stroke and primary-care groups

Sample size

Stroke n=32; primary-care n=366

DIF method used

(e.g., MH, IRT, Logistic regression, MIMIC, other factor analysis)

The 1-parameter Rasch model.

“The model appropriate for ordered response categories, the Rasch Rating Scale model, was used to evaluate item hierarchy and item fit statistics on the CES-D. Item fit was evaluated using goodness-of-fit statistics, reported as infit mean-squares (MNSQ). Rasch analysis was conducted with Winsteps”

DIF was defined as lack of equivalence in performance across groups. DIF was assessed using t-tests on differences in mean item calibration between the stroke and primary care groups. The t-test was the ratio of difference between item difficulty estimates and the square root of the sum of the squared standard errors at 95% confidence interval. (pg 61)

Test of model assumptions

Item fit was examined using the INFIT mean-squares (ratio of observed to predicted variance for an item). The possible multidimensionality of the CES-D based on other factor analyses of the CES-D was discussed as a possible violation of the unidimensionality assumption.

“Item hierarchies based on mean logit calibrations were strongly correlated between stroke and primary care-based groups (Spearman’s rank order coefficient, r=0.75). Items relating to interpersonal disruption feelings (“I felt that people disliked me” and “people were unfriendly”) were hardest to endorse in both groups. Misfitting items, that is, MNSQ higher than 1.40, in PSD included “my sleep was restless,” “I had crying spells,” “people were unfriendly,” and “I felt just as good as other people.” No items misfit the scale in the primary care-based depression group”.

Purification

Not performed

Evidence of uniform DIF

“In comparing item functioning between the two groups, four items demonstrated statistically significant DIF: “my sleep was restless,” “I felt that people disliked me,” “I did not feel like eating,” and “I had crying spells.” Each of these items identified with statistically significant DIF demonstrated a logit difference of approximately 0.5 or more across the two groups.”

“The CES-D scale functioning was found to be quite impressive in primary care-based depression, with no items misfitting the scale, although only three items slightly misfit the scale in stroke. DIF observed between depressed stroke and primary-care patients may imply that slightly different clusters of depressive occur in stroke compared with primary-care patients, but this is conjectural given the small size of the stroke sample. In addition, the same items have been previously associated with bias in studies of large nonstroke samples.”

Evidence of non-uniform DIF

Not applicable due to method used.

Magnitude of DIF

A criterion of .5 logits between item calibrations was shown. The logit differences ranged form .77 to .03. Items with the largest differences were: disliked (.77), appetite (.65), restless (.61), crying (.48), as good as others (.48).

Impact of DIF

Given that only one item was identified as uniquely psychometrically problematic in the stroke subgroup, stroke-specific changes to the CES-D scale are not recommended.

Limitations according to authors:

1. Authors noted the importance of the unidimensionality assumption in scale-related issues identified using Rasch models. They also noted that the factor structure of the CES-D in nonstroke patients has been described as having four underlying factors: depressive affect, positive affect, somatic and interpersonal disruption.

2. Authors recognized limitations in the generalizability of study results due to sampling issues:

a) “Generalizability of the results to stroke patients may be limited by the timing of the assessment, which was assessed at approximately 3 months poststroke. The nature of PSD may change over the course of recovery and PSD does not remain constant throughout the poststroke period.” (pg 63).

b) “Because of the small stroke sample size, the present study has been described as a preliminary investigation”. (pg 63).

Limitations according to expert review:

1. The limitations of this study include the small n upon which the analyses were based (32 in the stroke group). This may have lead to low power to detect DIF. Additionally, the Rasch model does not allow detection of non-uniform DIF. Finally, tests of unidimensionality were not performed and the authors concede that the measure has been found to be multidimensional in the same analyses.

2. It is unclear whether or not the method used for assessing DIF was the most appropriate. Relevant covariates could not be introduced into model given method used for DIF analysis.

Strengths:

1. The authors provide a discussion of the DIF results that is anchored in findings from earlier studies.

Key references:

1. Parikh RM, Eden DT, Price TR, et al. The sensitivity and specificity of the Center for Epidemiologic Studies Depression Scale in screening for post-stroke depression. Int J Psychiatry Med 1988;18:169–81

2. Radloff LS. The CES-D Scale: a self-report depression scale for research in general population. Appl Psych Meas 1977;1:385–401.

3. Shinar D, Gross CR, Price TR, et al. Screening for depression in stroke patients: the reliability and validity of the Center for Epidemiologic Studies Depression Scale. Stroke 1986;17:241–5.

4. Stommel M, Given BA, Given CW, et al. Gender bias in the measurement properties of the center for epidemiologic studies depression scale (CES-D). Psychiatry Res 1993;49:239–50.

5. Pickard AS, Johnson JA, Feeny DH, et al. Agreement between self- and proxy assessment in stroke: a comparison of generic HRQL measures. Stroke 2004;35:607–12.

6. Herrman H, Patrick DL, Diehr P, et al. Longitudinal investigation of depression outcomes in primary care in six countries: the LIDO study. Functional status, health service use and treatment of people with depressive symptoms. Psychol Med 2002;32:889–902.

7. Bech P, Lucas R, Amir M, et al. Association between clinically depressed subgroups, type of treatment and patient retention in the LIDO study. Psychol Med 2003;33:1051–9.

( see abstract )

Back to   TOP    Measurement matrix         MEGs Search

Center for Epidemiologic Studies-Depression Scale (CES-D)
Name/ Reference	Pickard AS, Dalal MR, Bushnell DM. A Comparison of Depressive Symptoms in Stroke and Primary Care: Applying Rasch Models to Evaluate the Center for Epidemiologic Studies-Depression Scale. Value in Health, (2006) 9(1):59-64.
Source contact info	A. Simon Pickard, College of Pharmacy, University of Illinois at Chicago, Room 164, MC 886, 833 South Wood Street, University of Illinois At Chicago, Chicago, IL 60612, USA. E-mail: pickard1@uic.edu
Availability (private or public)	Public
Conceptual framework	Authors stated that IRT-based models may be used to inform clinicians and outcomes researchers about whether differences exist between primary care- and stroke-based depression based on patterns of responses to the CES-D. “The study utilized IRT-based models to compare responses to symptom-related items on the CES-D by depressed stroke and primary-care patients to: 1) illustrate the use of IRT-based (Rasch) models for comparing scale functioning across different patient subgroups; and 2) inform clinicians and outcome researchers about CES-D scale functioning and depressive symptomatology in stroke- compared with primary care-based depression.”
Purpose of measure & application (clinical, research, survey, screening)	depression screening in the general population
Sample characteristics	Of the 101 stroke patients with complete item responses to the CES-D at 3 months (three had incomplete data), 32 patients (32%) were defined as depressed according to CES-D scores. Depressed Stroke patients: mean (sd) age = 69 years (14) and 59% were female. Depressed primary care-based patients: mean (sd) age= 42 years (15), and the 67% were female.
Recruitment methods	Secondary data from two sources: 1) Stroke patient data were collected as part of a study of health outcomes after stroke (see Pickard et al 2004 below). “Patient assessments of the CES-D collected at 3 months postrecruitment were analyzed. Patients who were diagnosed with ischemic stroke were recruited with 2 weeks of stroke before discharge from two large teaching hospitals in Edmonton , Canada . Participants had to be 18 years of age or older. Patients were not eligible if they had receptive aphasia, were cognitively impaired, unable to comprehend the English language, or had a very poor prognosis, based on consultation with attending clinicians” (pg 60). 2) “Data on USA-based primary-care patients with depression were obtained from the Longitudinal Investigation of Depression Outcomes (LIDO) study. The LIDO project was a longitudinal study of depressive symptoms, quality of life and health services use among primary-care patients in Barcelona (Spain), Be’er Sheva (Israel), Melbourne (Australia), Porto Alegre (Brazil), St. Petersburg (Russia), and Seattle (the United States). At each site, consecutive adult (age 18– 75) visitors were invited to participate in the study. After providing informed consent, participants completed a screening questionnaire that included the CESD”. (See Herrman et al 2002; Bech et al 2003. below) (pg 60)
Data collection method	questionnaires 1) Stroke patients (see Pickard et al 2004 below) 2) USA-based primary-care patients with depression (See Herrman et al 2002; Bech et al 2003. below)
Response rate	Not provided
Format & design (readability, # of items, time to complete, response categories)	“The CES-D is a 20-item scale designed to measure depressive symptoms experienced in the past week. Responses are interpreted based on a simple summary score, calculated by summing the item responses. Patients were categorized as depressed if CES-D scores were 16 or higher, a threshold used in community-based studies and in stroke. Higher scores are associated with more frequent depressive symptoms.”
Type of measurement (nominal, ordinal, interval, ratio)	ordinal
Scoring (range, direction, rules, missing data)	Not Provided
Availability of translations & source	Not Provided
Psychometric Properties:
Scale construction	Not Provided
Basic summary statistics	CES-D scores for Depressed Stroke patients: mean (sd) age = 69 years (14); for Depressed primary care-based patients: mean (sd) age= 42 years (15).
Variability	Not provided
Test-retest reliability	Not provided
Interrater reliability	Not provided
Internal consistency	Not provided
Content validity	Not provided
Construct validity	Not provided
Concurrent validity	Not provided
Predictive validity	Not provided
Sensitivity to change	Not provided
Differential Item Functioning (DIF):
Variable studied (e.g., groups)	Stroke and primary-care groups
Sample size	Stroke n=32; primary-care n=366
DIF method used (e.g., MH, IRT, Logistic regression, MIMIC, other factor analysis)	The 1-parameter Rasch model. “The model appropriate for ordered response categories, the Rasch Rating Scale model, was used to evaluate item hierarchy and item fit statistics on the CES-D. Item fit was evaluated using goodness-of-fit statistics, reported as infit mean-squares (MNSQ). Rasch analysis was conducted with Winsteps” DIF was defined as lack of equivalence in performance across groups. DIF was assessed using t-tests on differences in mean item calibration between the stroke and primary care groups. The t-test was the ratio of difference between item difficulty estimates and the square root of the sum of the squared standard errors at 95% confidence interval. (pg 61)
Test of model assumptions	Item fit was examined using the INFIT mean-squares (ratio of observed to predicted variance for an item). The possible multidimensionality of the CES-D based on other factor analyses of the CES-D was discussed as a possible violation of the unidimensionality assumption. “Item hierarchies based on mean logit calibrations were strongly correlated between stroke and primary care-based groups (Spearman’s rank order coefficient, r=0.75). Items relating to interpersonal disruption feelings (“I felt that people disliked me” and “people were unfriendly”) were hardest to endorse in both groups. Misfitting items, that is, MNSQ higher than 1.40, in PSD included “my sleep was restless,” “I had crying spells,” “people were unfriendly,” and “I felt just as good as other people.” No items misfit the scale in the primary care-based depression group”.
Purification	Not performed
Evidence of uniform DIF	“In comparing item functioning between the two groups, four items demonstrated statistically significant DIF: “my sleep was restless,” “I felt that people disliked me,” “I did not feel like eating,” and “I had crying spells.” Each of these items identified with statistically significant DIF demonstrated a logit difference of approximately 0.5 or more across the two groups.” “The CES-D scale functioning was found to be quite impressive in primary care-based depression, with no items misfitting the scale, although only three items slightly misfit the scale in stroke. DIF observed between depressed stroke and primary-care patients may imply that slightly different clusters of depressive occur in stroke compared with primary-care patients, but this is conjectural given the small size of the stroke sample. In addition, the same items have been previously associated with bias in studies of large nonstroke samples.”
Evidence of non-uniform DIF	Not applicable due to method used.
Magnitude of DIF	A criterion of .5 logits between item calibrations was shown. The logit differences ranged form .77 to .03. Items with the largest differences were: disliked (.77), appetite (.65), restless (.61), crying (.48), as good as others (.48).
Impact of DIF	Given that only one item was identified as uniquely psychometrically problematic in the stroke subgroup, stroke-specific changes to the CES-D scale are not recommended.