Improving measurement properties of the Patient Health Questionnaire-9 with rating scale analysis

The goal of this study was to explore the psychometric properties of the Patient Health Questionnaire-9 (PHQ-9; R. L. Spitzer, K. Kroenke, & J. B. W. Williams, 1999).

Factor analysis and Rasch rating scale analysis were used to examine the psychometric properties of the PHQ-9. The sample consisted of 202 adults with spinal cord injury (SCI).

The PHQ-9 items appear to form a usefully unidimensional scale. One "double-barreled" item, "Moving or speaking so slowly that other people could have noticed or being so fidgety or restless that you've been moving around a lot more than usual," misfit the Rasch model. Category probability curves indicate respondent difficulty in distinguishing between the 2 intermediate rating scale categories: several days and more than half the days. Combining these categories eliminated this problem and resulted in all items fitting the measurement model.

The measurement properties of the PHQ-9 can be improved by collapsing rating scale categories and by restructuring several double- and triple-barreled items. Adopting these changes may improve sensitivity in measuring depression after SCI.