Screening for depression among first-visit psychiatric patients: comparison of different scoring methods for the...

The present paper aims to compare the ability of eight scoring methods for the Center for Epidemiologic Studies Depression Scale (CES-D), one of the most widely used self-report inventories for depression, to screen for a major depressive episode. The subjects were 591 patients who constituted representative samples of the first-visit patients to 23 psychiatric hospitals and clinics all over Japan. The criterion diagnoses were given by a semi-structured interview with established inter-rater reliability. The eight alternative scoring methods included the conventional Likert method, which gives the weights 0, 1, 2 and 3 to the four anchor points of the CES-D; the presence method, which gives the weights 0, 1, 1 and 1, respectively; the GHQ method, which gives the weights 0, 0, 1 and 1, respectively; the persistence method, which gives the weights 0, 0, 0 and 1, respectively; 10-item version, 5-item version, single-item version and the algorithmic method. On the basis of the receiver operating characteristic (ROC) analyses, it was found that the traditional Likert scoring method of the full CES-D performed best in detecting major depressive episodes among first-visit psychiatric patients. The presence method, the GHQ method and the 10-item version appeared to have a similar ability. The persistence method, the 5-item version and the single-item version resulted in significantly worse performance.