NCJ Number
84394
Date Published
1977
Length
22 pages
Annotation
This study proposes a method for reporting findings that combines true positives and true negatives to measure total accuracy of predictions.
Abstract
A retrospective study of dangerous behavior among criminally insane offenders in Colorado confirmed the findings of Cocozza and Steadman (1974) regarding the predictive validity of the Legal Dangerousness Scale (LDS). The Baxstrom patients on whom LDS was based were released over 10 years ago, and replication findings were remarkably similar. The sample was composed of younger, less institutionalized offenders with an average length of stay in hospital of four-and-a-half years. Of 111 male patients with two to four years at risk, less than three percent have been recommitted as legally insane. Over half were involved in significant posthospital disruption, and the behavior of 33 was classified as dangerous. In addition to four LDS items of criminal record, eight other high risk indicators in offender histories were found to be related to posthospital behavior. Guttman-type properties of ordinality and summation of unequal weights were not confirmed with the Colorado sample, although cumulative scale properties were suggested. The LDS was found to be more reliable for older offenders, and contrary to usual findings, the youngest age group accounted for proportionately less dangerous behavior. (Resources in Education (ERIC) abstract)