Testing measurement invariance of the GHQ-28 in stroke patients.

Abstract

In order to combine self-reported measures data from multiple studies to conduct an integrated data analysis, the construct measured must have the same meaning across the studies. This study investigated the measurement invariance of the General Health questionnaire (GHQ-28) in two stroke studies before combining the data for an integrative data analysis. The study used data from the Stroke Outcomes Study 1 (SOS1, n = 448) and second Stroke Outcomes Study (SOS2, n = 585). The initial analysis was a confirmatory factor analysis (CFA) for each study separately to confirm the four-factor structure of GHQ-28 questionnaire. Multi-group confirmatory factor analysis (MG-CFA) was used to assess the measurement invariance of the GHQ-28 questionnaire in the two stroke cohorts. Measurement invariance at configural invariance (same items associated with same factor across groups); factor loading invariance (equal factor loadings across groups) and scalar invariance (equal intercepts across groups) was examined. CFA supported all three invariances measured. Results showed that the GHQ-28 questionnaire has comparable measurement properties in the SOS1 and SOS2 stroke studies. Strong measurement invariance was established, and based on the results from this study, integrative data analysis of GHQ-28 scores from the two stroke studies is merited.