Grayson et. al.

< - - Back to      Measurement matrix    MEGs Search

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

Grayson et. al.   (see abstract)

Center for Epidemiologic Studies Depression Scale (CES-D)

Name/ Reference

Grayson DA, Mackinnon A, Jorm AF, Creasey H, Broe GA. Item Bias in the Center for Epidemiologic Studies Depression Scale: Effects of Physical Disorders and Disability in an Elderly Community Sample. Journal of Gerontology: Psychological Sciences. (2000) 55B(5): P273–P282

Source contact info

D.A. Grayson, 14 Poplar Grove, Lawson, NSW 2783 Australia.

E-mail: dgrayson@med.usyd.edu.au

Availability (private or public)

public

Conceptual framework

“Debate has centered on whether the inclusion of somatic symptoms may artifactually raise depression scores because of higher rates of physical disorder and effects of medications among older persons. This study examines if variations in responses to the CES-D with physical health reflect an actual association with depression and the extent to which they reflect item-level artifacts.”

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

measure of depressive symptomatology

Sample characteristics

Participants in this analysis were part of the Sydney Older Persons’ Study. 630 participants aged 75 or older, living in the community (not in hostels or nursing homes) in the inner western suburbs of Sydney , Australia . Only individuals n=506 with no missing data were included in analyses.

The average age of participants was 80.9 years (SD=4.17, range=75–97.8), 52% were men. The average education level (1–10) =4.61 (s.d.=2.05); 58% were widowed. The mean CES-D score was 8.76

Recruitment methods

Half the participants were selected with an area-based random probability sample; the other half were randomly drawn using Department of Veteran Affairs lists of entitled veterans living in the study area (including women veterans and war widows).

Data collection method

The data analyzed were collected from a random probability sample of elderly persons living in the community. Participants underwent a physical examination by a geriatrician; medical diagnoses were used. Sociodemographic data on all 630 participants were obtained by a social scientist interviewer at the respondent’s home. [Education status: 10-point scale, ranging from 1 (no schooling) to 10 (completed tertiary). Marital status: four (dichotomous) categories, married, widowed, never married, and divorced/separated.]

Responses to the 20-item CES-D were obtained from 604 participants at this interview. The CES-D was either self-administered or administered orally to the visually impaired.

“A second interview was conducted in the participant’s home by a physician experienced in geriatric medicine. At this interview, data were obtained from 527 individuals on disability: mobility (5-point scale, 0, 1/4, 1/2, 3/4, 1”; hi=complete immobility; incontinence (4-point scale, 0–1, hi= bladder and bowel incontinence); self care (ADL); 10-point scale, scored 0 to 3, hi= maximal disability with bathing, feeding, and dressing; and IADL; 7-point scale, scored 0 to 3, hi= maximal disability with cooking, house-cleaning, and shopping. The range of chronic diseases common among elderly individuals rated by interviewing physicians (4-point scale, well, mild, moderate, or severe—0, 1/3, 2/3, 1) was obtained on 522 individuals.

Response rate

Not provided

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

CES-D contains 20 item, each scored 0–3; with 3 indexing higher depression.

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

ordinal

Scoring (range, direction, rules, missing data)

The analyses reported were undertaken on the 506 individuals for whom complete relevant data were available. A small number of cases had missing data imputed (imputation method not discussed).

Theoretical range=0-60.

Availability of translations & source

Not provided

Psychometric Properties:

Scale construction

Not provided

Basic summary statistics

CES-D total mean= 8.76, sd= 7.90 for this sample

Variability

Not provided

Test-retest reliability

“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these]

Interrater reliability

Not provided

Internal consistency

Cronbach’s alpha for the 20 items for this sample was 0.85.

There was a 0.995 correlation between the score obtained by weighting each item with its loading on the single common factor score and the usual CES-D total score.

Content validity

Not provided

Construct validity

“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these]

It correlates well with clinical ratings of depression (see Roberts & Vernon, 1983; Weissman, Sholomskas, Pottenger, Prusoff, & Locke, 1977 listed below)

Concurrent validity

“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these]

Predictive validity

Not provided

Sensitivity to change

Not provided

Differential Item Functioning (DIF):

Variable studied (e.g., groups)

Sociodemographic:

Age (years) mean(sd)= 80.86(4.17)

Education level (1–10) mean(sd)=4.61(2.05)

Gender 52% male

Marital status: Married 27%

                     Widowed 58%

                      Never married 9%

                      Divorced/Separated 6%

Disability: Mobility mean(sd)=0.12 (0.16)

                 Incontinence mean(sd)=0.06 (0.15)

                 ADL mean(sd)=0.04 (0.22)

                 IADL mean(sd)=0.56 (0.75)

Physical disorders: Heart disease mean(sd)=0.17(0.21)

                               Stroke mean(sd)= 0.06(0.15)

                               Peripheral vascular disease mean(sd)=0.05(0.14)

                              Chronic lung disease mean(sd)=0.08(0.19)

                               Bone and joint disease mean(sd)= 0.29(0.22)

                               Other systemic disease mean(sd)= 0.14(0.20)

                               Gait instability mean(sd)=0.20(0.23)

                               Gait slowing mean(sd)=0.11(0.25)

                               Obesity mean(sd)=0.05(0.14)

Cognitive impairment mean(sd)= 0.28(0.36)

Sample size

The analyses included 506 individuals (with no missing data).

DIF method used

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

The Multiple Indicators, Multiple Causes (MIMIC) Model as applied to the CES-D “consists of a measurement model for the CES-D, with all 20 items loading on a single factor representing the latent depression variable. The demographic, disability, and physical disorder covariates that are investigated as potential sources of bias in this study are included in the model as predictors of the latent variable. This part of the model may be viewed as a multiple regression of the latent variable against the covariates. In addition to estimating the effect of the covariates on the latent variable ((), the model allows the effects of the covariates on each of the CES-D items to be assessed directly.” (pg 274)

Test of model assumptions

First, the acceptability of a single common factor model in representing the scale was evaluated using confirmatory factor analysis.

A single common factor model was fitted to the 20 CES-D items. This model provided an adequate fit to the data (GFI=0.908, TLI=0.836, RMSEA=0.062). The Scmid-Leiman multifactor model was also fitted, “which posits a single Depression factor loading, to all 20 items, and four residual factors which correspond to the original four CES-D factors Somatic, Well-being, Interpersonal, and Affective. This model fit the data better (GFI=0.949, TLI=0.929, RMSEA=0.041), although at the cost of substantially reduced parsimony.” (pg 276)

The single Depression factor accounted for 93% of the common factor variance on the CES-D total score.

Purification

“In all models examining item bias, Item 6 (“felt depressed”) was used as a reference item for which the direct paths from predictor variables were set to 0. A constraint of this form is necessary to achieve identification. Thus, bias in other items was assessed relative to this reference.” The authors discuss that the selection of the item for identification and for reference to the other items is assumed to be unbiased. This selection was arbitrary. The construct of depression on which the effect of various disability/disorder predictors is evaluated is defined in this study to be that for which the item “felt depressed” has no bias with respect to these predictors. ” (pg 281)

Evidence of uniform DIF

“These bias effects are relative to Item 6 (“felt depressed”), which was constrained for identification purposes to have zero bias. Of the 20 analyses using single predictors, only the results for 17 analyses are presented, as Education, Divorced, and Obesity did not yield estimates with critical ratios exceeding either 1.5 on the predictor loading to the CES-D Depression factor or 2.0 for any of the 19 bias loadings. These three predictors are not examined further. For all 17 models reported in Table 2, the GFIs ranged between 0.911 and 0.915, the TLIs between 0.817 and 0.833, and the RMSEAs between 0.060 and 0.062, all indicating satisfactory fit.” (pg 276)

“None of the sociodemographic variables was significantly associated with the CES-D Depression factor, although numerous item-specific effects were found: Older participants reported being more bothered by things and less hopeful about the future; men found things less of an effort, were less fearful, slept better, and reported crying less. Being widowed was associated with feeling at least as good as others and with more fear and loneliness. None of these effects were associated with elevated depression. Of the disability variables, the increases on the depression factor were associated with disability of any sort other than ADL.

Mobility, ADL, and IADL also had direct effects on several items, in particular poor appetite, finding everything an effort, restless sleep, and inability to get going. Physical disorder variables also influenced the CES-D Depression factor and items directly. Heart disease, stroke, any other systemic disease, gait instability, and cognitive impairment were all associated with a genuine rise in depression, although they all show other effects on the CES-D. The items “good as others,” “talked less,” “people unfriendly,” “enjoyed life,” “crying spells,” “felt sad,” and “people dislike me” all showed significant negative direct loadings, indicating that individuals with accompanying physical disorders underreport on these items for reasons unassociated with depression. As with disability, items “poor appetite,” “everything an effort,” and “inability to get going” had higher endorsement levels in individuals with particular diseases for reasons other than a disease-related elevation in depression.” (pg 276)

“Individuals with a disability, bone and joint disease, and stroke were more likely to report that “everything is an effort” above and beyond levels of depression, and again, so it may be. The items “good as others,” “talked less,” “people unfriendly,” “enjoyed life,” “crying spells,” “felt sad,” and “people dislike me” had only negative associations with particular disorders. That is, participants with more severe physical disorder respond to these items in a less extreme manner than expected for given levels of depression.” (pg 279)

Evidence of non-uniform DIF

Cannot be tested with the MIMIC model.

Magnitude of DIF

Item-level magnitude could be estimated by examination of the $_ior the path from the studied covariate to the item.

“Bias effects of age on CES-D of 0.06 arise from the only significant loadings, which are on the items “bothered by things” (0.02) and “hopeful about the future” (0.04); whereas age has no loading on the Depression factor.”

Impact of DIF

The bias effect (impact) of the items on the total score is estimated by the sum of all the direct loadings (G $_i) from the studied variable to the 20 items. The genuine effect of the predictors on the latent variable is estimated as the sum of the loadings from the depression factor to the items (G 8_i) multiplied by the path from the predictors to the depression factor (γ).

“The regression of age on CES-D total score yields a beta weight of 0.09: Each increase of 1 year in age is associated with an increase of 0.09 in CES-D total.” “The bias effects range from negligible (widowed) to over seven times the magnitude (bone and joint disease) of the effects on the Depression factor (averaging 157%) and are frequently in the opposite direction. The situation does not markedly improve when the conservative principle of statistical significance is adopted to distill the relative contributions: Of 17 predictors, 9 show only nondepression effects on the CES-D, and only 1 (incontinence) supports the use of the CES-D as an unbiased measure of depression; with the remaining 7 predictors showing joint contributions, the bias component ranges from 4% (heart disease) to 64% (mobility) of the magnitude of the genuine depression component.” (pg 278)

Limitations according to authors:

1. “The detail of the reported results turns very much on the choice for identification purposes of the item “felt depressed” as unbiased. The factor of depression then becomes that for which this item is unbiased, and biases on other items are in relation to this particular factor. The choice was made on the grounds of face validity and gains support by the empirical plausibility of the detailed results which followed; for instance, in the physical disorders “bone” and “stroke” have no effect on “genuine” depression but cause overendorsement on the symptom “everything an effort.” This makes empirical sense. However, had we chosen instead “everything an effort” to be unbiased, the corresponding factor of depression would have absorbed these effects, showing positive associations between these physical disorders and the (new) factor of depression, whereas the symptom “felt depressed” would now show biases indicating underreporting. But even with the unavoidable arbitrariness introduced by this identification issue, the presence per se of substantial item biases in the CES-D is unambiguous”. (pg P279) (It is noted by the reviewer that the necessity and the effects of “purification” of the item selected for identification has not been studied and is an area of controversy).

Additional limitations by reviewer:

1. Sample sizes for the different variables was not reported, however, based on overall n, sample size might have not been appropriate for the analyses performed.

2. Multicollinearity among covariates was not discussed.

3. Non-uniform DIF cannot be examined.

Strengths according to authors:

1. Authors highlighted the comprehensive medical evaluation utilized in the study, to argue that this study is the first to determine the contribution of specific physical disorders to bias in the CES-D.

Additional strengths by the reviewer:

This is an innovative parametric method that uses latent variable models to examine the DIF effects as well as the magnitude and impact of DIF.

Key references:

1. Weissman, M. M., Sholomskas, D., Pottenger, M., Prusoff, B. A., & Locke, B. Z. (1977). Assessing depressive symptoms in five psychiatric populations: A validation study. Journal of Epidemiology, 106, 203–214.

2. Gatz, M., & Hurwicz, M.-L. (1990). Are old people more depressed? Crosssectional data on CES-D factors. Psychology and Aging, 5, 284–290.

3. Hertzog, C., Van Alstine, J., Usala, P. D., Hultsh, D. F., & Dixon , R. (1990). Measurement properties of the Center for Epidemiological Studies Depression Scale (CES-D) in older populations. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 64–72.

4. Mackinnon, A. J., McCallum, J., Andrews, G., & Anderson, I. (1998). The Center for Epidemiologic Studies Depression Scale in older community samples in Indonesia, North Korea, Myanmar, Sri Lanka, and Thailand. Journal of Gerontology: Psychological Sciences, 53B, P343– P352.

5. McCallum, J., Mackinnon, A., Simons, L., & Simons, J. (1995). Measurement properties of the Center for Epidemiologic Studies Depression Scale: An Australian community study of aged persons. Journal of Gerontology: Social Sciences, 50B, S182–S189.

6. Radloff, L. S. (1977). The CES-D scale: A self report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401.

7. Radloff, L. S., & Teri, L. (1986). Use of the Center for Epidemiologic Studies

8. Depression Scale with older adults. Clinical Gerontology, 5, 119–136.

9. Roberts, R. E., & Vernon , S. W. (1983). The Center for Epidemiologic Studies Depression Scale: Its use in a community sample. American Journal of Psychiatry, 140, 41–46.

(see abstract)

Back to   TOP    Measurement matrix    MEGs Search

Center for Epidemiologic Studies Depression Scale (CES-D)
Name/ Reference	Grayson DA, Mackinnon A, Jorm AF, Creasey H, Broe GA. Item Bias in the Center for Epidemiologic Studies Depression Scale: Effects of Physical Disorders and Disability in an Elderly Community Sample. Journal of Gerontology: Psychological Sciences. (2000) 55B(5): P273–P282
Source contact info	D.A. Grayson, 14 Poplar Grove, Lawson, NSW 2783 Australia. E-mail: dgrayson@med.usyd.edu.au
Availability (private or public)	public
Conceptual framework	“Debate has centered on whether the inclusion of somatic symptoms may artifactually raise depression scores because of higher rates of physical disorder and effects of medications among older persons. This study examines if variations in responses to the CES-D with physical health reflect an actual association with depression and the extent to which they reflect item-level artifacts.”
Purpose of measure & application (clinical, research, survey, screening)	measure of depressive symptomatology
Sample characteristics	Participants in this analysis were part of the Sydney Older Persons’ Study. 630 participants aged 75 or older, living in the community (not in hostels or nursing homes) in the inner western suburbs of Sydney , Australia . Only individuals n=506 with no missing data were included in analyses. The average age of participants was 80.9 years (SD=4.17, range=75–97.8), 52% were men. The average education level (1–10) =4.61 (s.d.=2.05); 58% were widowed. The mean CES-D score was 8.76
Recruitment methods	Half the participants were selected with an area-based random probability sample; the other half were randomly drawn using Department of Veteran Affairs lists of entitled veterans living in the study area (including women veterans and war widows).
Data collection method	The data analyzed were collected from a random probability sample of elderly persons living in the community. Participants underwent a physical examination by a geriatrician; medical diagnoses were used. Sociodemographic data on all 630 participants were obtained by a social scientist interviewer at the respondent’s home. [Education status: 10-point scale, ranging from 1 (no schooling) to 10 (completed tertiary). Marital status: four (dichotomous) categories, married, widowed, never married, and divorced/separated.] Responses to the 20-item CES-D were obtained from 604 participants at this interview. The CES-D was either self-administered or administered orally to the visually impaired. “A second interview was conducted in the participant’s home by a physician experienced in geriatric medicine. At this interview, data were obtained from 527 individuals on disability: mobility (5-point scale, 0, 1/4, 1/2, 3/4, 1”; hi=complete immobility; incontinence (4-point scale, 0–1, hi= bladder and bowel incontinence); self care (ADL); 10-point scale, scored 0 to 3, hi= maximal disability with bathing, feeding, and dressing; and IADL; 7-point scale, scored 0 to 3, hi= maximal disability with cooking, house-cleaning, and shopping. The range of chronic diseases common among elderly individuals rated by interviewing physicians (4-point scale, well, mild, moderate, or severe—0, 1/3, 2/3, 1) was obtained on 522 individuals.
Response rate	Not provided
Format & design (readability, # of items, time to complete, response categories)	CES-D contains 20 item, each scored 0–3; with 3 indexing higher depression.
Type of measurement (nominal, ordinal, interval, ratio)	ordinal
Scoring (range, direction, rules, missing data)	The analyses reported were undertaken on the 506 individuals for whom complete relevant data were available. A small number of cases had missing data imputed (imputation method not discussed). Theoretical range=0-60.
Availability of translations & source	Not provided
Psychometric Properties:
Scale construction	Not provided
Basic summary statistics	CES-D total mean= 8.76, sd= 7.90 for this sample
Variability	Not provided
Test-retest reliability	“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these]
Interrater reliability	Not provided
Internal consistency	Cronbach’s alpha for the 20 items for this sample was 0.85. There was a 0.995 correlation between the score obtained by weighting each item with its loading on the single common factor score and the usual CES-D total score.
Content validity	Not provided
Construct validity	“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these] It correlates well with clinical ratings of depression (see Roberts & Vernon, 1983; Weissman, Sholomskas, Pottenger, Prusoff, & Locke, 1977 listed below)
Concurrent validity	“The CES-D has demonstrated good internal consistency, test-retest reliability, concurrent validity on clinical and self-report criteria, and construct validity.” [no references provided for these]
Predictive validity	Not provided
Sensitivity to change	Not provided
Differential Item Functioning (DIF):
Variable studied (e.g., groups)	Sociodemographic: Age (years) mean(sd)= 80.86(4.17) Education level (1–10) mean(sd)=4.61(2.05) Gender 52% male Marital status: Married 27% Widowed 58% Never married 9% Divorced/Separated 6% Disability: Mobility mean(sd)=0.12 (0.16) Incontinence mean(sd)=0.06 (0.15) ADL mean(sd)=0.04 (0.22) IADL mean(sd)=0.56 (0.75) Physical disorders: Heart disease mean(sd)=0.17(0.21) Stroke mean(sd)= 0.06(0.15) Peripheral vascular disease mean(sd)=0.05(0.14) Chronic lung disease mean(sd)=0.08(0.19) Bone and joint disease mean(sd)= 0.29(0.22) Other systemic disease mean(sd)= 0.14(0.20) Gait instability mean(sd)=0.20(0.23) Gait slowing mean(sd)=0.11(0.25) Obesity mean(sd)=0.05(0.14) Cognitive impairment mean(sd)= 0.28(0.36)
Sample size	The analyses included 506 individuals (with no missing data).
DIF method used (e.g., MH, IRT, Logistic regression, MIMIC, other factor analysis)	The Multiple Indicators, Multiple Causes (MIMIC) Model as applied to the CES-D “consists of a measurement model for the CES-D, with all 20 items loading on a single factor representing the latent depression variable. The demographic, disability, and physical disorder covariates that are investigated as potential sources of bias in this study are included in the model as predictors of the latent variable. This part of the model may be viewed as a multiple regression of the latent variable against the covariates. In addition to estimating the effect of the covariates on the latent variable ((), the model allows the effects of the covariates on each of the CES-D items to be assessed directly.” (pg 274)
Test of model assumptions	First, the acceptability of a single common factor model in representing the scale was evaluated using confirmatory factor analysis. A single common factor model was fitted to the 20 CES-D items. This model provided an adequate fit to the data (GFI=0.908, TLI=0.836, RMSEA=0.062). The Scmid-Leiman multifactor model was also fitted, “which posits a single Depression factor loading, to all 20 items, and four residual factors which correspond to the original four CES-D factors Somatic, Well-being, Interpersonal, and Affective. This model fit the data better (GFI=0.949, TLI=0.929, RMSEA=0.041), although at the cost of substantially reduced parsimony.” (pg 276) The single Depression factor accounted for 93% of the common factor variance on the CES-D total score.
Purification	“In all models examining item bias, Item 6 (“felt depressed”) was used as a reference item for which the direct paths from predictor variables were set to 0. A constraint of this form is necessary to achieve identification. Thus, bias in other items was assessed relative to this reference.” The authors discuss that the selection of the item for identification and for reference to the other items is assumed to be unbiased. This selection was arbitrary. The construct of depression on which the effect of various disability/disorder predictors is evaluated is defined in this study to be that for which the item “felt depressed” has no bias with respect to these predictors. ” (pg 281)
Evidence of uniform DIF	“These bias effects are relative to Item 6 (“felt depressed”), which was constrained for identification purposes to have zero bias. Of the 20 analyses using single predictors, only the results for 17 analyses are presented, as Education, Divorced, and Obesity did not yield estimates with critical ratios exceeding either 1.5 on the predictor loading to the CES-D Depression factor or 2.0 for any of the 19 bias loadings. These three predictors are not examined further. For all 17 models reported in Table 2, the GFIs ranged between 0.911 and 0.915, the TLIs between 0.817 and 0.833, and the RMSEAs between 0.060 and 0.062, all indicating satisfactory fit.” (pg 276) “None of the sociodemographic variables was significantly associated with the CES-D Depression factor, although numerous item-specific effects were found: Older participants reported being more bothered by things and less hopeful about the future; men found things less of an effort, were less fearful, slept better, and reported crying less. Being widowed was associated with feeling at least as good as others and with more fear and loneliness. None of these effects were associated with elevated depression. Of the disability variables, the increases on the depression factor were associated with disability of any sort other than ADL. Mobility, ADL, and IADL also had direct effects on several items, in particular poor appetite, finding everything an effort, restless sleep, and inability to get going. Physical disorder variables also influenced the CES-D Depression factor and items directly. Heart disease, stroke, any other systemic disease, gait instability, and cognitive impairment were all associated with a genuine rise in depression, although they all show other effects on the CES-D. The items “good as others,” “talked less,” “people unfriendly,” “enjoyed life,” “crying spells,” “felt sad,” and “people dislike me” all showed significant negative direct loadings, indicating that individuals with accompanying physical disorders underreport on these items for reasons unassociated with depression. As with disability, items “poor appetite,” “everything an effort,” and “inability to get going” had higher endorsement levels in individuals with particular diseases for reasons other than a disease-related elevation in depression.” (pg 276) “Individuals with a disability, bone and joint disease, and stroke were more likely to report that “everything is an effort” above and beyond levels of depression, and again, so it may be. The items “good as others,” “talked less,” “people unfriendly,” “enjoyed life,” “crying spells,” “felt sad,” and “people dislike me” had only negative associations with particular disorders. That is, participants with more severe physical disorder respond to these items in a less extreme manner than expected for given levels of depression.” (pg 279)
Evidence of non-uniform DIF	Cannot be tested with the MIMIC model.
Magnitude of DIF	Item-level magnitude could be estimated by examination of the $_ior the path from the studied covariate to the item. “Bias effects of age on CES-D of 0.06 arise from the only significant loadings, which are on the items “bothered by things” (0.02) and “hopeful about the future” (0.04); whereas age has no loading on the Depression factor.”
Impact of DIF	The bias effect (impact) of the items on the total score is estimated by the sum of all the direct loadings (G $_i) from the studied variable to the 20 items. The genuine effect of the predictors on the latent variable is estimated as the sum of the loadings from the depression factor to the items (G 8_i) multiplied by the path from the predictors to the depression factor (γ). “The regression of age on CES-D total score yields a beta weight of 0.09: Each increase of 1 year in age is associated with an increase of 0.09 in CES-D total.” “The bias effects range from negligible (widowed) to over seven times the magnitude (bone and joint disease) of the effects on the Depression factor (averaging 157%) and are frequently in the opposite direction. The situation does not markedly improve when the conservative principle of statistical significance is adopted to distill the relative contributions: Of 17 predictors, 9 show only nondepression effects on the CES-D, and only 1 (incontinence) supports the use of the CES-D as an unbiased measure of depression; with the remaining 7 predictors showing joint contributions, the bias component ranges from 4% (heart disease) to 64% (mobility) of the magnitude of the genuine depression component.” (pg 278)