Psychometric Properties of the GAD-7 Among College Students: Reliability, Validity, Factor Structure, and Measurement Invariance

The GAD-7 (seven-item Generalized Anxiety Disorder Scale) is a brief, freely available questionnaire measuring anxiety, which is a common mental health problem among college students. The current study examined the psychometric properties of the GAD-7 among undergraduate students from a public midwestern university (n = 582; M = 19.0 +/- 1.0 years old; 79.4% women; 81.6% White) who completed the GAD-7, State-Trait Anxiety Inventory-Trait subscale (STAI-T), and Behavioral Inhibition and Activation Scale-Reward subscale (BIAS-R). The psychometric properties of two GAD-7 models were examined: a one-factor model, equivalent to a single total score, and a two-factor model, including cognitive-emotional and somatic tension factors. The GAD-7 total score had excellent reliability (both alpha and omega = .91) and evidence for convergent (r with STAI-T = 0.70, p < .001) and discriminant validity (r with BIAS-R = -0.04, p = .350). The cognitive-emotional factor had excellent reliability (alpha = .90; omega = .91), whereas the somatic tension factor had adequate reliability (alpha = .76; omega = .77). Both factors had evidence for convergent validity (cognitive-emotional: r with STAI-T = 0.70, p < .001; somatic tension: r with STAI-T = 0.60, p < .001) and discriminant validity (cognitive-emotional: r with BIAS-R = -0.05, p = .227; somatic tension: r with BIAS-R = -0.02, p = .700). Confirmatory factor analysis supported both the one-factor model (root-mean-square error of approximation [RMSEA] = 0.098, comparative fit index [CFI] = 0.994) and the two-factor model (RMSEA = 0.069, CFI = 0.997), albeit with a high interfactor correlation (r = .91). Measurement invariance was supported across gender, race/ethnicity, and college class for both models. The findings support using the GAD-7 total score among U.S. college students, as is common in practice, but also provide evidence for the two-factor model in this population. Considering its parsimony, the total score would be preferable in applied settings for a quick measurement of anxiety symptom severity.