Clinicians continue to need to make dangerousness determinations for legal and routine clinical purposes. Thus research needs to focus on developing methods for doing this task better. The proposed study does this by modeling two sets of relationships: 1) between case characteristics and clinicians' predictions of dangerousness and 2) between case characteristics and violence in mental patients. These models take an important step toward improving the prediction of dangerousness by providing a rich picture of what clinicians do, and should, take into account when assessing dangerousness. This study capitalizes on an opportunity to use previously collected data from two data sets gathered by the investigators. They have overlapping, rich sets of case characteristic information that will serve as a pool of potential predictors. The first study has clinical predictions of dangerousness on a sample of approximately 400 cases. The second has both clinical predictions and detailed self report and official accounts of patient violence on about 800 patients. Models for prediction at admission assessment (regarding both community and unit violence) and at discharge will be constructed using a variety of standard statistical approaches as well as classifications tree methodology. Models will be internally and externally cross-validated. The project will construct a preliminary method for structuring clinical assessments of dangerousness using the models of what clinicians actually do to alter these actuarial models. This decision aid will be a tangible and methodologically sound step toward systematizing the prediction of dangerousness.
Showing the most recent 10 out of 16 publications
