BACKGROUND: The depression module of the Patient Health questionnaire (PHQ-9) is a wide-spread self-report instrument for the assessment of depression with compelling psychometric characteristics when relying on classical test theory assumptions. This study aimed at evaluating whether the PHQ-9 may be interpreted as a dimensional scale measuring depression severity in the elderly general population using Rasch analysis with special emphasis on its unidimensional structure and differential item functioning (DIF) due to gender, age, and the presence of somatic multimorbidity. METHODS: A representative sample of the elderly German general population (N=1631; age 60-85 years, 53.5% female) filled in the PHQ-9, a questionnaire about chronic medical conditions and a demographic data sheet. Unidimensionality and psychometric properties of the PHQ-9 were ascertained applying confirmatory factor analysis (CFA) and Rasch analysis. RESULTS: Results revealed substantial violations of the unidimensionality of the scale: item 8 (retardation or agitation) had to be eliminated and multiple residual correlations were added. Gender-related DIF emerged for two items, and three items showed insufficient Rasch model fit. LIMITATIONS: The large sample leads to high statistical power that might technically increase the probability of detecting model misfit or DIF. The sampling procedure leads to a possible underestimation of morbidity due to the exclusion of those elderly patients living in nursing homes. CONCLUSIONS: Results suggest that - when applied in the elderly general population - the PHQ-9 should be interpreted in terms of a diagnostic algorithm for classificatory decisions about a DSM-IV based probable diagnosis of depression rather than as a dimensional scale.