Assessment of the Center for Epidemiological Studies Depression Scale factor structure among middle-aged workers in Japan


Norio Sugawara, MD, PhD, Department of Psychiatry, Hirosaki-Aiseikai Hospital, 1-6-2 Kitazono, Hirosaki, Aomori 036-8151, Japan. Email:


Our aim was to assess the internal consistency and structural/construct validity of the Center for Epidemiological Studies Depression Scale among middle-aged employees in Japan. We conducted a cross-sectional study of 7284 workers, aged 49.0 ± 6.3 (mean ± SD) years old. Structural/construct validity was assessed by confirmatory factor analysis. The 4-factor structure reported in the general population was replicated, and a second-order model with an overarching depression factor fitted well. These findings indicate that the Center for Epidemiological Studies Depression Scale is a valid and reliable measure of depressive symptoms for middle-aged workers in Japan.

IN JAPAN, THE annual number of suicides markedly increased in 1998 and has exceeded 30 000 per year thereafter. During this period, suicide cases have increased, especially among middle-aged working individuals.1

The Center for Epidemiological Studies Depression Scale (CES-D) is a widely used self-reported questionnaire that is simple to use and quick to implement.2 Previous factor analysis studies have found the 20-item CES-D to be a multidimensional instrument with as many as four correlated but distinct factors.3 The four underlying factors were labeled by Radloff as follows: depressed affect, somatic symptoms (lack of) positive affect, and interpersonal problems.2 In Japanese, only one study has examined the factor structure of CES-D, and it extracted four factors by the principal component analysis.4 However, the imbalance of sex ratio (male: n = 1870; female: n = 142) might influence the CES-D factor structure. Furthermore, sex difference of the CES-D factor structure has not been clarified in the Japanese population.

The objective of this study was to investigate the reliability of internal consistency and structural/construct validity of the CES-D by confirmatory factor analysis among employees in Japan.



A total of 7284 workers (4425 men and 2859 women), aged 49.0 ± 6.3 (mean ± SD) years old, completed the questionnaires on their demographic background. Informed consent was obtained from all subjects prior to the questionnaire.


The Japanese version of CES-D was administered to all participants.5 In accordance with previous studies,3,6 Radloff's four-factor model (model 4A) and Radloff's four-factor model with a second-order depression factor (model 4B) were analyzed by confirmatory factor analysis.

Statistical analysis

The reliability of internal consistency was evaluated using Cronbach's alpha. To assess the fit of the models for the data, practical fit indices were emphasized because χ2-tests of fit are highly sensitive to sample size and can lead to the rejection of well-fitting models. Three practical fit indices were used to evaluate the model fit: the root mean square error of approximation (RMSEA), the goodness of fit index (GFI), and the comparative fit index (CFI).6 RMSEA values of <0.05 indicate a good fit and are a better index for large samples. GFI values range between 0 and 1.00, with values near 1.00 indicating good fit. The CFI has an upper ceiling of 1.00 and values >0.90 indicate good fit. This index is independent of sample size and takes model complexity into consideration. The data were analyzed using the PASW Statistics-PC-software for Windows, Version 18.0.0 and Amos-PC-software for Windows, Version 17.0 (SPSS Inc., Chicago, IL, USA).


The mean CES-D score was 14.6 ± 7.7 (median 13.0). The overall scale was found to be reliable (alpha = 0.84). Corrected item-total correlations for individual items ranged from −0.05 (item 4, ‘good’) to 0.67 (item 6, ‘depressed’). Coefficient alphas were also very good for each of the four CES-D factors: 0.83 for depressive affect, 0.85 for somatic symptoms, 0.73 for interpersonal problems, and 0.69 for (lack of) positive affect.

Model 4A had the fit indices: d.f. = 164, χ2 = 4534.31, RMSEA = 0.061, GFI = 0.933, and CFI = 0.920 for all subjects (men: d.f. = 164, χ2 = 2716.84, RMSEA = 0.060, GFI = 0.934, CFI = 0.928; women: d.f. = 164, χ2 = 2020.00, RMSEA = 0.063, GFI = 0.925, CFI = 0.906). Model 4B had the fit indices: d.f. = 166, χ2 = 4548.21, RMSEA = 0.060, GFI = 0.933, and CFI = 0.920 for all subjects (men: d.f. = 166, χ2 = 2728.82, RMSEA = 0.059, GFI = 0.934, CFI = 0.928; women: d.f. = 166, χ2 = 2022.58, RMSEA = 0.063, GFI = 0.925, CFI = 0.906).

Factor loadings for model 4A are shown in Table 1. Intercorrelations between the four factors in model 4A ranged from −0.18 to 0.97 and were the lowest for correlations that included the positive affect factor.

Table 1.  Factor loadings for Center for Epidemiological Studies Depression Scale items by confirmatory factor analysis according to Radloff's 4-factor model
 Factor 1Factor 2Factor 3Factor 4
Bothered  0.710.65    
Appetite  0.580.55    
Mind  0.740.69    
Effort  0.770.71    
Sleep  0.620.57    
Talk  0.670.67    
Going  0.730.65    
Good    0.610.57  
Hopeful    0.690.69  
Happy    0.610.56  
Enjoy    0.520.52  
Unfriendly      0.810.77
Dislike      0.710.74
Interfactor correlations        
Factor 11.001.00      
Factor 20.970.941.001.00    
Factor 3−0.16−0.14−0.18−  
Factor 40.850.730.790.68−0.11−


This study evaluated the reliability and construct validity of the CES-D in Japanese middle-aged workers. Given that the coefficient alpha is influenced by the number of items in a scale, the internal consistency of CES-D factor subscales were also very strong (from 0.73 for two-item interpersonal problems, 0.69 for the four-item positive affect factor, and 0.85 for the seven-item somatic symptoms factor). The good fit of the second-order model supports the use of a total score of the CES-D as a global indicator for depressive symptoms. Although the positive affect scale, for instance, appears to be only weakly related to the overall depressive construct, the total score is considered as a valid measure.

In regards to interfactor correlations, except for the positive affect (which ranged from 0.68 to 0.97), the correlations between positive affect and other factors showed a quite different pattern (from −0.08 to −0.18). A previous study using principal component analysis showed a similar pattern of interfactor correlations between positive affect and other factors among 30–63-year-old subjects in Japan.4 The results of Western countries ranged from 0.3 to 0.7.6,7 Different response patterns on positive items between the Japanese and individuals in Western countries may affect our results.

There were some limitations to this study. First, all subjects were healthy employees. They may be healthier than community populations. Second, participants in the study consisted of a middle-aged population. Previous studies showed that correlations between factors were weaker among older people than younger ones.6

This is the largest study on the factor structure of the CES-D scale in Japan. The factor structure of the CES-D scale was assessed in enough participants, and showed a similar pattern in both sexes.