Do multiple outcomes and conditional factors improve prediction of batterer reassault?

Researchers and practitioners recognize the need to develop better prediction of abuse and reassault among men referred to batterer programs. Previous approaches have produced relatively weak predictions for primarily dichotomized "reassault versus no reassault" outcomes. The primary objective of this research was to improve prediction using multinomial logistic regressions with multiple outcomes and conditional factors. We used a unique data set from a multisite batterer program evaluation with a large sample size (n = 840), a longitudinal 15-month follow-up, and high response rates (70% of the women). Multiple outcomes included no abuse, verbal abuse or controlling behavior, threats, one-time reassault, and repeated reassaults. Using multiple outcomes improved prediction with risk markers over prediction of a dichotomous outcome and suggests that "repeat reassault" should be distinguished from "one-time reassault." Conditional factors did not substantially improve prediction. The low "correct classification" suggests, however, the difficulty of identifying the most dangerous batterers.