A Multiple Models Approach to Assessing Risk of Repeat DV

PROJECT SUMMARY FOR 2001-WT-BX-0500

This project will categorize cases from a newly available data source (the Chicago Women's Health Risk Study) to assess how diverse risk factors affect different types of domestic violence and how those risk factors interact to influence the likelihood of a person experience repeat, or severe repeat, intimate-partner violence (IPV). The applicant will use a recently developed risk assessment tool that features a classification-tree methodology to maximize distinctions among categories. The goal is to develop two screening tools-one for repeat violence and another for severe repeat violence-that will enable criminal-justice personnel and clinicians to categorize domestic-violence intimate terrorism (IT), violent resistance (VR), and mutual violent control (MVC). To assess the extent of repeat violence in each of the four categories, the applicant will calculate the base rate of partner violence, then use the multiple-models classification tree method to evaluate the likelihood of repeat violence and severe repeat violence for each category. Predictions will be derived from a series of separate models, then combined into a composite prediction score. The applicant will use CHAID (Chi-squared Automatic Interaction Detector) to assess the statistical significance of the bivariate association between eligible risk factors and the dichotomous out come measure.

Ca/ncf