Mental Health and Women on Probation or Parole: A Confirmatory Factor Analysis of the Brief Symptom Inventory (BSI-18)

Abstract

This study examined the robustness of the proposed factor structure of the Brief Symptom Inventory (BSI-18) in a sample of justice-involved women (N = 406). Efforts to assess accurately and intervene effectively regarding mental health issues (namely, somatization, depression, and anxiety) are critical due to growing proportions of women involved in the criminal justice system and the link between mental health and justice-system outcomes among women. Women on probation or parole provided responses to the BSI-18 as part of a larger study on victimization and substance use. Analyses included calculation of descriptive statistics and confirmatory factor and simulation analyses. Results show strong model–data fit indices reflecting that the BSI-18 is a robust and pragmatic self-report symptom inventory for women with criminal justice system involvement. Limitations related to cross-sectional design and a small, nonrandom sample are noted and implications for mental health assessment and intervention are discussed.

Keywords:

• Brief Symptom Inventory
• measurement
• factor structure
• confirmatory factor analysis
• justice-involved women
• mental health

Notes

Notes

1 Some may question why we did not perform an exploratory factor analysis and moved directly to a confirmatory factor analysis. The appropriate use of exploratory factor analysis occurs when the structure of scales is not hypothesized or not known. In this study, Derogatis (2000) stated the manner the scales should behave. The scales were hypothesized to form three factors and relate to a global measure of mental health. If they did not do so, the model fit would be poor. In other words, the current study tests the hypotheses of the three factor structure and global measure of mental health. Simply, exploratory factor analysis would be inappropriate in this study given the a priori hypotheses.

2 All of the coefficients are standardized factor loadings and are statistically significant at the .05 level.