Factorial Validity and Invariance of the GHQ-12 Among Clinical and Nonclinical Samples

• Published in: Assessment (Odessa, Fla.) Fla.), 2013-04, Vol.20 (2), p.219-229
• Main Authors: Fernandes, Helder Miguel, Vasconcelos-Raposo, José
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Aged, 80 and over; Anxiety Disorders - diagnosis; Anxiety Disorders - psychology; Brazil; Chronic Illness; College Students; Comparative Analysis; Cross-Cultural Comparison; Depressive Disorder - diagnosis; Depressive Disorder - psychology; Factor Analysis; Factor Structure; Female; Foreign Countries; Goodness of Fit; Health; Humans; Kidney Failure, Chronic - psychology; Kidney Failure, Chronic - therapy; Male; Mass Screening; Measurement Techniques; Middle Aged; Models, Statistical; Patients; Psychometrics - statistics & numerical data; Quality of Life - psychology; Questionnaires; Reference Values; Reliability; Renal Dialysis - psychology; Reproducibility of Results; Surveys and Questionnaires; Translating; Young Adult
• ISSN: 1073-1911; 1552-3489
• DOI: 10.1177/1073191112465768

The purpose of this study was to examine the internal reliability, factorial validity, and measurement invariance of a Brazilian–Portuguese version of the General Health Questionnaire–12 (GHQ-12) across clinical and nonclinical groups. The clinical sample consisted of 228 chronic hemodialysis patients (41.7% female), with a mean age of 48.23 (SD = 16.02) years, whereas the nonclinical sample comprised 350 university students (59.1% female), with a mean age of 21.15 (SD = 2.04) years. Initial model fit comparison of 16 factor structures reported in the literature, through confirmatory factor analysis (CFA), provided mixed results. Additional revisions based on CFA modification indices demonstrated that a reduced (8-item) one-factor model with three error covariances provided the best fit to the data in both samples (comparative fit index > .98 and root mean square error of approximation < .05) and acceptable internal reliability estimates (>.70). Pearson’s correlations between the original and the reduced scale was .96 for both samples. Multigroup analysis supported partial strong measurement equivalence across groups. Latent mean differences showed an expected higher factor mean for the clinical sample. Overall results demonstrated that the 8-item one-factor of the GHQ-12 showed good psychometric properties in both samples and provided preliminary support for meaningful comparisons of factor means across clinical and nonclinical groups.