The Total CES‐D score from the 14‐item version as a predictor of cognitive decline

Abstract Background Depressive symptomatology is associated with cognitive decline and greater risk for dementia. Studies have examined this association using self‐report measures such as the Centre for Epidemiologic Studies‐Depression Scale (CES‐D; Radloff, 1991). Although the CES‐D was originally designed as a 20‐item scale, certain items have been reported to misrepresent depressive symptoms due to sex and cultural differences, which are addressed in the 14‐item version of this measure (Carleton et al., 2013). The aim of this study was to examine how the CES‐D total score from the 14‐item version compared to the 20‐item version in predicting progression to cognitive decline from a cognitively unimpaired baseline, across genders. Method Data were extracted from the Wisconsin Registry for Alzheimer’s Prevention. A total of 1,055 participants were included who were cognitively unimpaired and stable at baseline and had at least one follow‐up assessment. Bivariate logistic regression analyses were conducted using follow‐up consensus diagnosis as outcome (cognitively unimpaired stable vs. cognitive unimpaired declining, mild cognitive impairment or dementia); baseline total CES‐D scores from either the 14‐item or 20‐item version were used as predictors in separate analyses; age at last follow‐up assessment, age difference between baseline and last follow‐up assessment, years of education, and polygenic risk score were covariates (see Table 1). Result In female participants, we observed that the logistic regression model with the total CES‐D score from the 14‐item version was statistically significant (χ2(5) = 54.674, p = .001) and explained 15.2% (Nagelkerke R 2 ) of the variance in predicting cognitive decline; this result was comparable to that with the 20‐item version, which was also statistically significant (χ2(5) = 53.276, p = .001) and explained 14.8% of the variance. In male participants, the model with the total score from the 14‐item version was statistically significant (χ2(5) = 14.846, p = .011) and explained 9.9% of the variance, whereas the model with the total score from the 20‐item version was statistically significant (χ2(5) = 14.070, p = .015) and explained 9.4% of the variance. Conclusion Our findings suggest that the total CES‐D score from the 14‐item version is comparable to, if not slightly better than, the score from the 20‐item version in predicting progression to a clinical level of cognitive decline in both males and females.