Background/Purpose: The National Health and Nutrition Examination Survey (NHANES) Depression Screener (i.e., the Patient Health Questionnaire-9 [PHQ-9]) is widely used. Although measurement properties of the PHQ-9 are well-established among patient populations, its performance has not been evaluated in general populations. In addition, items on the PHQ-9 have not been investigated using differential item functioning (DIF). This study used Rasch analysis to evaluate the PHQ-9 for (a) dimensionality and item fit; and (b) DIF attributes relative to gender and race/ethnicity in a large U.S. sample.
Method: Demographic information and PHQ-9 responses were obtained from the 2009-2010 NHANES. An un-weighted sample of 3,175 male and 3,352 female adult respondents (28.9% Latino [LAT], 47.2% non-Hispanic White [NHW] and 18.3% non-Hispanic Black [NHB]) was included. Rasch analysis assessed the dimensionality, model fit and DIF due to respondent's gender and race/ethnicity. DIF was also evaluated using a generalized Mantel-Haenszel approach.
Analysis/Results: All items on the PHQ-9 demonstrated good item fit and measured a single construct, with high internal consistency (Cronbach's Alpha = .84). Item difficulty ranged from -1.07 logits to 1.36 logits. Both Rasch analysis and the generalized Mantel-Haenszel approach identified three DIF items based on gender, five DIF items when comparing LAT and NHW, and three DIF items when comparing NHW and NHB. However, the magnitudes of DIF were negligible (DIF size < 0.42 logits).
Conclusions: Rasch and the generalized Mantel-Haenszel methods provide powerful tools for DIF detection. The PHQ-9 is an effective tool for assessing intensity and symptoms of depression among diverse U.S. populations.