Bridging Disciplinary Boundaries (January 11 - 14, 2007)

The purpose of this study was to test the measurement invariance of the Brief Symptom Inventory (BSI; Derogatis & Melisaratos, 1983) across three ethnic groups (Afro-American, Caucasian, and Latino) within the United States, and across the three nations (USA, Japan, and Korea) for the population with severe and persistent mental illness.

The measurement invariance of the BSI had not been tested across culturally different groups for the population with severe and persistent mental illness in clinical settings and in research as well. Thus, the question of whether or not the BSI was a valid instrument across ethnically or culturally different groups remained unanswered, although it was a prevalent psychological symptom measure in clinical settings and in research.

Method.

The study used secondary data, and the total sample size is 2,268 adults with persistent and severe mental illness from three nations: the United States (N = 1,247), Japan (N = 802), and Korea (N = 219). Selecting three countries was based on the framework of the most different system design of comparative research, and sampling method was convenience and purposive sampling (Van de Vijver & Leung, 1997).

Two competing hypotheses of the factor structure of the BSI, which were nine-factor model (Derogatis & Meliseratos, 1983) and one-factor model (Piersma, Boes, & Reaume, 1994, were hypothesized. These two factor models were examined by using multiple-group CFA in terms of the four types of measurement invariance (configural invariance, weak metric invariance, phi invariance, and measurement error invariance) across the three countries and across four ethnic groups with the United States.

Results.

The one-factor model fitted into the data very well in all types of invariance: configural invariance (GFI = .984 across the three ethnic groups; .987 across the three nations), weak metric invariance (GFI = .979; .981), phi invariance (GFI = .962; .936) and measurement error invariance (GFI = .961; .931). By contrast, the hypothesized nine-factor model did not fit the data in any type of invariance.

Implication.

This study was the first invariance study of the BSI for the population with severe and persistent mental illness. The results implied that the BSI had only one factor structure of general psychological distress, and the one-factor structure was invariant across three ethnic groups within the U.S. sample and across cultural groups from three nation samples as well.

It was the major implication of the study that the BSI was confirmed as a valid psychological distress measure in comparing symptom level differences among ethnic and cultural groups in the population with severe and persistent mental illness. Specifically, when researchers and clinicians would utilize the BSI to assess psychological distress of the population with severe and persistent mental illness, the results suggested that the nine subscales in the BSI should not be used separately, but the global psychological distress score (GSI) should be used. In other words, the nine sub-symptoms in the BSI might not have discriminant validity with each other. Therefore, the BSI profile analysis should be limited in research and clinical settings.