Gauging Measurement Features of the Brief Symptoms Inventory 18 (BSI-18)

In applications of the Brief Symptoms Inventory 18 (BSI-18), a well-known screening tool for psychiatric disorders, composite measures of the subscales are often used and scales reliabilities are examined using Cronbach's alphas. This study systematically examines the measurement features of the BSI-18. CFA models were used to gauge the factorial structure of the BSI-18 and estimate its item and scale reliabilities; and MIMIC model was used to test measurement invariance of the BSI-18 items. Using a sample of rural drug users in Ohio, our results show that the BSI-18 forms a hierarchic factorial structure with three first-order factors (i.e., somatization, depression, and anxiety) underlying a second-order factor (i.e., general severity). The CFA model-based item and scale reliabilities were calculated, and the total variance of each item explained by all the factors, and variances explained by the first- and second-order factors, respectively, were estimated using the Schmid-Leiman transformation. The MIMIC model results show that the only two of the 18 items display differential item functioning (DIF) for gender and education, indicating minor measurement non-invariance.