ارزیابی مجدد عاطفه مثبت در مرکز برای مقیاس افسردگی مطالعاتی اپیدمیولوژیک

In 2000, Schroevers and colleagues examined the reliability and validity of a two-factor structure for the Dutch version of the Center for Epidemiologic Studies-Depression (CES-D) scale in cancer patients and in a community sample. The authors concluded that a two-factor structure assessing Positive Affect (PA) and Depressed Affect was a better fit to the data than the standard four-factor structure. They argued further that the four reverse-scored items composing the PA scale should be dropped. Using similar measures and analyses as Schroevers and colleagues, we examined the factor structure and concurrent validity of the English version of the CES-D in university student and community samples. Across both samples the factor structure was more similar to Radloff's (1977) original four-factor structure than to the proposed two-factor structure. Additionally, our validity analyses indicated no problems with the PA items and suggested that PA might be more specifically related to depression than to other forms of psychopathology. We recommend that clinicians and researchers using the English version of the CES-D continue to use the full 20-item version.

The Center for Epidemiologic Studies-Depression (CES-D) scale is a self-report measure of depressive symptomatology for use in the general population (Radloff, 1977). The reliability and validity of the CES-D have been examined extensively (e.g., Hertzog et al., 1990, McCallum et al., 1995 and MacKinnon et al., 1998). In Radloff's original study, the CES-D demonstrated high internal consistency and acceptable test–retest reliability. Strong evidence for the validity of the CES-D scores was demonstrated by relations to observer and self-ratings of depression. To evaluate the factor structure of the CES-D, Radloff (1977) conducted a principal components analysis (PCA). A four-factor pattern (Depressed Affect, Positive Affect, Somatic and Retarded Activity, and Interpersonal) emerged consistently across three samples. Subsequent evaluations have demonstrated results similar to Radloff's findings (e.g., Fava, 1983 and Ross & Mirowsky, 1984) and strong evidence for the same factor structure across several age, cultural, and socioeconomic groups (Hertzog et al., 1990, McCallum et al., 1995, MacKinnon et al., 1998, Wong, 2000 and Boisvert et al., 2003). In 2006, Shafer conducted a meta-analysis of four depression questionnaires (i.e., Beck, CES-D, Hamilton, and Zung). Across the 21 articles (28 studies) and 22,340 participants included by Shafer, every CES-D item replicated into Radloff's original four-factor structure. In summarizing the results, Shafer wrote “Overall, the CES-D had relatively little variability across the factor analysis studies included in this meta-analysis compared to the other three depression tests. The results were clear and highly consistent with the initial factor analyses conducted by Radloff during the development of the CES-D” (p. 133–134). Not every study, however, has replicated the four-factor structure. For example, Thomas and Brantley (2004) found that three factors (Depressed Affect/Somatic, Positive Affect, and Interpersonal) provided the best fit to their CES-D data using a confirmatory factor analysis (CFA) in a sample of women with low-income. The authors hypothesized that minority groups might express depression more somatically than the general population. They further noted that their results were similar to the three-factor structure found by Guarnaccia et al. (1989) in a sample of Mexican-American women. In 2000, Schroevers and colleagues argued that a two-factor solution provided the best fit for their data using the Dutch version of the CES-D. The authors conducted a forced two-factor PCA and across a group of cancer patients and a nonclinical reference group, they found one factor consisting of the 16 negatively formulated items (Depressed Affect; DA) and a second factor consisting of the 4 positively formulated items (Positive Affect; PA). Next, they investigated the validity of the two factors by correlating each factor with several measures of psychosocial functioning. The authors determined that DA was strongly related (r > 0.40) to measures of anxiety, depression, and general psychological distress, while PA was more weakly correlated (r ≤ 0.40) with the same measures. Finally, they examined the utility of DA and PA in discriminating between the cancer patient and community reference groups on depressive symptoms. They found that the patients had significantly higher DA mean scores compared to the reference group, while there was no significant difference between groups on PA mean scores. Schroevers and colleagues concluded that there was weak support for the validity of the PA items as a measure of depressive symptomatology and argued that a summed score of the 16 DA items was a “more valid measure of depressive symptomatology, both in cancer patients and in a matched reference group of healthy individuals from the general population” (p. 1026). The first goal of the current study was to attempt to replicate some of Schroevers et al.'s (2000) empirical findings in two samples. Their results appear to contrast previous findings that have replicated the original four-factor structure across nations, age groups, and cultural groups. Moreover, previous work (e.g., Watson et al., 1988a and Clark & Watson, 1991) has indicated that low positive affect is an important factor in the assessment of depression. As such, we examined the factor structure of the English CES-D and the concurrent validity of the proposed DA and PA factors in university student and community participant samples. Both the student and community samples were previously collected for unrelated studies. From the measures available in these samples, we selected validity comparisons as similar as possible to those used by Schroevers and colleagues as well as those available across both of the previously collected samples to increase their comparability. We attempted to answer three questions: 1) Do the factor analyses in each sample replicate the findings of either Schroevers et al., 2000 and Radloff, 1977? 2) Are psychosocial variables related weakly to PA items but strongly to DA items? 3) In the student sample, we also investigated a) are there mean differences in DA and PA summed scores between participants with low, moderate, and high self-reported depression symptomatology? and b) do the DA and PA scores account for unique variance in several psychopathology scales?