Psychometric properties of the CES-D and four of its short-forms among South African primary care patients with a chronic condition

Background: To enhance practicality of routine depression screening in primary care, shorter versions of the 20-item CES-D have been created. However, providers lack evidence to guide their choice of short-form as their psychometric properties have not been systematically compared to the full scale. We compared the reliability, criterion validity, and dimensionality of four CES-D short-forms with the full CES-D in a South African primary care population. Methods: Using the CES-D, we screened 4010 adults receiving HIV and/or diabetes treatment in primary care, of which 882 were rescreened prior to enrolment in a psychological intervention trial. We assessed the internal consistency and temporal stability, of the CES-D-12, -10, -9, and -8, and evaluated each short-form’s criterion validity. Exploratory and confirmatory factor analysis tested multiple factor solutions for the full CES-D and each short-form. Results: All short-forms demonstrated high internal consistency (a > .85) and moderate temporal stability (intraclass correlations: .73-.75) comparable to the full CES-D. While all short-forms had excellent criterion validity (area under the receiver operator curve [AUROC] > .90), the CES-D-12 and CES-D-10 had superior criterion validity (AUROC = .99) and classification accuracy (93.1% and 93.7% respectively). The CES-D-12 emerged as the short-form with the closest conceptual alignment to the full scale, both optimally fitted by a three-factor model. Conclusions: Our findings suggest that the CES-D-10 and -12 are the best alternatives to the full CES-D. While the CES-D-10 offers efficiency advantages for case detection, the CES-D-12 may provide more insight into depression symptom profiles.