Employing the CES-D within a population of the elderly: dimensional structure and measurement artifacts

Employing the Center for Epidemiological Studies Depression Scale (CES-D) we are confronted with very specific artifacts resulting from the confounding of one aspect of the construct ("well being") with the direction of the item wording. We investigated the structure and applicability of the CES-D by means of a sample of 663 individuals over 75 years of age within the "Leipzig Longitudinal Study in the Aged." Since the assumptions of a partial-credit Rasch model turned out to be too restrictive, one 3- and one 4-class latent-class model for ordered categories were calculated. Three classes of the latter were ordered on a latent continuum "depression," one class comprises those respondents who exhibit response patterns indicating the artifact mentioned above. Depression is obviously overestimated for these respondents. Connections with age and the Mini-Mental State Examination Score are presented. Results show that care should be taken if the CES-D is employed within a population of the elderly and the sumscore should not be used as a statistic for depression.