Purpose: Asian American ethnic groups differ in their culture and health/mental health behavior. Despite the heterogeneity of Asian American ethnic groups, there is a lack of systematic approach to understand the differences and how these differences may impact on social work practice. Human services agencies employ mental health assessment tools to decide the type, intensity, and the length of mental health services offered to individuals. While ethnicity and race information are obtained, it is unclear whether the measures are culturally sensitive or if this type of information is considered when treatment plans are developed. This research focused on the measurement issue, evaluating whether the underlying assumptions of measurement invariance would hold up across three different groups of Asian Americans.

Method: The data were collected from 1995 to 2001 from an ethnic mental health agency in a multi-cultural metropolitan city. A total of 470 Asian Americans from three Asian American ethnic groups (Chinese, Miens, and Vietnamese) were incorporated in this study. The primary mental health assessment instrument measured four mental health areas of the client's status: emotional distress, psychotic symptoms, anti-social behaviors, and capacity for community living. Confirmatory Factor Analyses (EQS 6.1) using structural equation modeling techniques (maximum likelihood estimation) were used (EQS, Bentler, 1994) to test the measurement model. Once an adequate fit of the general measurement model for the overall Asian American sample was established, multiple-group analyses were used to determine the consistency of the model across the different ethnic groups (Farrel, 1994).

Findings: Findings rejected measurement invariance across three Asian American ethnic groups. The unconstrained model of multi-group analysis indicated an adequate fit to the data (c²(6)=15.83, CFI=.96, RMSEA=0.059, and BIC=.20), while fit indices for the fully constrained model were c²(13) =44.70, CFI= .88, RMSEA =.072, and BIC = 9.96. The unconstrained model fit the data significantly better than the constrained model, in which the difference in c² was 28.87 and significant (df=7, p< .001) and BIC increased by 9.96. Follow-up Lagrangian Multiplier tests (Bentler, 1995) were used to evaluate each constraint. Releasing the constraints for both Psychotic Symptoms and Anti-social Behaviors among the three groups significantly improved the overall fit. The fit indices for the partially constrained model were c²(10) =17.05, CFI= .97, and RMSEA =.017, and BIC = -9.67. The partially constrained model was also a significantly better model than the fully constrained model based on various fit indices.

Implications: These findings suggested that the weights of Psychotic Symptoms and Anti-social Behavior in the measurement structure differed across Asian American ethnic groups. The results underscore the needs 1) to develop culturally appropriate assessment tools for diverse groups and to pay attention to cultural sensitivity of measurement for distinctive ethnic/cultural groups and 2) to scrutinize ethnic/racial categories in mental health research with any seemingly monolithic ethnic/racial groups including Asian Americans. Diverse ethnic groups respond differently to each area of mental health symptoms and functioning. In practice, it is important for practitioners to be aware of the differences to better serve their clients who have different needs.