Reliability and Construct Validity of the LSI for Colorado Inmates

Maureen L. O'Keefe, Colorado Department of Corrections

ABSTRACT
The Level of Supervision Inventory (LSI) measures offenders' criminal risk across 10 needs domains, lending itself as a potential diagnostic assessment in corrections settings. Absent from the literature are LSI reliability and validity findings with American offenders. The present study examined the reliability and construct validity of LSI subtotal scores and explored the relationship of the LSI total score to Colorado's inmate classification tool and the Millon Clinical Multiaxial Inventory-111. The sample included 652 male and female incarcerated felons. The reliability of the subscales was found to be modest using Cronbach's alpha (OC = .65) and subtotal-total correlations (I = .22 to .67). Subtotal scores were compared to subjective diagnostic ratings, obtained independently of the LSI, in seven overlapping domains. In 4 of the 7 areas, convergent validity estimates were significant and larger than the divergent validity correlations. The relationship of the LSI total score to the inmate classification instrument was weak (r = .22); however, a multiple regression analysis revealed a stronger relationship of the LSI to six MCMI scales (R = .47). These exploratory analyses suggest that the LSI is tapping into domains beyond risk, yet only subscales with more items may adequately evaluate needs domains.

(Return to Program Resources)

Updated 05/20/2006