Invariance of the Brief Symptom Inventory in Schizophrenia

Maanse Hoe
Montgomery Ross Fisher Building, Rm 102L
School of Social Work University of  Southern California
669 West 34 th St.
Los Angeles CA 90089
USA
Phone: 213  740-2011
FAX: 213  740-8905
Email: hoe@usc.edu

Purpose: In recent years, the effectiveness of community-based treatment programs has been demonstrated in reducing relapse, rehospitalization, and enhancing clinical outcomes. However, existing studies to measure the effectiveness of community-based treatment programs have not sufficiently addressed the limitations of the measurements. Thus, it is unclear whether the measurements used to assess clinical outcomes appropriately measure the symptoms and whether the constructs of the measures are invariant across time. The specific aim of this study is to test factorial invariance of the Brief Symptom Inventory (BSI; Derogatis & Melisaratos, 1983) across four study groups and across time with data from community-based treatment programs.

Methods: There has been limited analysis of items in instruments because of methodological problems. The factor analysis of Likert items has been problematic due to a restricted range in the response format and the presence of skewness in many items. Other studies have simply ignored the problems and conducted factor analysis anyway. However, some recent advances in structural equation modeling have made it possible to address these problems. This study uses a multi-group confirmatory factor analysis to overcome the aforementioned problems. The data from Brekke’s two projects at University of Southern California (N=172, N=165), as well as two other data sets from Brekke's colleagues at the University of Wisconsin (N=135, N=185).

Results: The results of cross-group analysis show strong validation of the Derogotis’ nine-factor model. The results of cross-time analysis show that factor loadings and factor variances were invariant over time.

Implications for practice: These results indicate that the BSI measures the same construct across the four study groups and over time. The BSI is an appropriate instrument to measure clinical outcomes in community-based treatment programs. However, future research should examine whether or not the BSI is invariant in practice settings, not in research settings.