The psychometric properties of the Colorado Symptom Index

Abstract

The Colorado Symptom Index (CSI: Shern et al. 1994, Milbank Quart 72:123-148) is widely used in research as a self-report measure of psychiatric symptomatology, yet little information exists regarding the scale's psychometric properties. Additionally, the CSI has no cut-off score denoting the need for further psychiatric assessment. This study examined the CSI's psychometric properties and established a cut-off score. Analyses were based on 3,874 adult Florida Medicaid respondents. The CSI had excellent internal consistency (.92) and test-retest reliability (.71). Evidence of the CSI's validity was strong; CSI scores distinguished among individuals with and without mental health services needs and were significantly correlated with functioning. Results using a contrasting groups approach indicate that 30 is a reasonable "clinical" cut-off score. At this score, the CSI had respectable sensitivity (.76) and specificity (.68) and a ROC curve analysis suggests that the CSI is "fair to good" discriminator of individuals with psychiatric disabilities.