Psychometric Properties of the Center for Epidemiologic Studies Depression (CES-D) with Older African Americans in the South

Introduction

The Center for Epidemiologic Studies Depression Scale (CES-D) has been widely used to measure the frequency of recent experienced depressive symptoms in clinical and research settings.  However, four reverse-scored items (e.g., good, hopeful, happy, enjoyed) were associated with measurement problems among older adults (e.g., Carlson et al., 2011). Exploratory factor analyses have shown anywhere from two to seven factors for the CES-D scale and provided strong evidence for racial/ethnic difference (Kim et al., 2011). Thus, it is important to determine whether these racial/ethnic group differences in the CES-D reflect true differences in depressive symptoms or are instead due to measurement variance.

The purpose of this study is to investigate overall psychometric properties of the CESD-20.

Methods

The Jackson Heart Study is a population-based study designed to examine the factors related to the increased CVD morbidity and mortality among African Americans residing in the South. 1,488 valid cases over 55 years at the baseline examination are selected. With the Cronbach alphas and intra-factor correlation coefficients, both exploratory and confirmatory factor analyses are used to identify the best factor structure and unique items’ loading patterns to each factor.

Results

 The 20-item reliability (alpha = .771) is less internally consistent than any prior studies.  When excluding reversed items, alphas are significantly improved to .855 (total sample) & .872 (+ 55 ~ 65 older).  Two-way ANOVA found that women (M=10.98) are significantly more depressed than men (M=9.39) with no age difference. Cronbach alphas of four subscales are fairly acceptable in depressed (.807), somatic (.706), positive (.711), with the two item interpersonal factor (.512) as an exception. Positive affect is not significantly correlated with depressed affect and interpersonal factor, but weakly negatively associated with somatic complaints (r = -.058, p= .023). Using the standard cutoff score of >16, 15.2% of sample would be moderately depressed, 4.6% (>25) might be diagnosed with severe depression with CESD-20 item scale. The CFA found that both 3-factor and 4-factor models are well-fitted with overall fit indices (e.g., CFI= .95, SRMR = .043~.050). However, EFA revealed that items like “good”, “hopeful”, “talk less” are defiantly or weakly loaded to overall scale. Interpersonal dimension is highly loaded by five-items (unfriendly, talk less, dislike, failure, and fearful)

 

Discussion  

This study found that the CESD has relatively less solid psychometric properties when used with older African Americans due to problematic reverse-score items. As depression is a somewhat subjective and abstract concept, this study indicates that the differences in conceptualization of depressive symptoms in racial/ethnic groups could impact the validity of a depressive diagnosis as a standard measurement. Thus, it suggests practitioners to use the CES-D composite score with extra caution when applied to older minority population even as a screening tool. Researchers should be aware that to simply measure the variance between aggregate scores of racial/ethnic groups does not provide the most valid data to analyze differences between those groups because of guarded or disguised expression of depression among older African Americans.