Predicting the Risk of Reassault Among Batterer Program Participants: Does Race Matter?

Alex Heckert, Indiana University of Pennsylvania
Edward W. Gondolf, Mid - Atlantic Addiction Training Inst.

ABSTRACT
There have been increasing efforts to predict reassault among batterer program participants. Prediction efforts using risk factors and risk assessment instruments has been rather weak. Researchers and practitioners have argued that prediction may be weakened because different risk factors are predictive among various racial and ethnic groups. We have demonstrated that prediction is improved by using a conditional prediction model with multiple outcomes. Distinguishing repeat reassaulters from one-time reassaulters and men who use threats without physical reassaults, improves prediction. Yet prediction is still far from ideal. We develop separate prediction equations for whites, AFrican-Americans, and Latinos, using a database of approximaately 500 cases drawn from 840 men and 688 female partners recruited at program intake. Using multinomial logistic regression to predict 15-month cumulative outcomes, we find that prediction of repeated reassault using risk markers is strongest for whites and stronger for whites than for the combined sample. Prediction for Latinos is the weakest. We also simulate three popular risk assessment iunstruments and find that the risk assessment instruments predict repeated reassault reasonably well among whites, modestly among African-Americans, and poorly among Latinos. We speculate that risk assessment uinstruments, which have been developed primarily using white samples, should be used with caution with minority batterers. When attempting to predict reassault, it appears that race does indeed matter, and future research should develop separate prediction by race.

(Return to Program Resources)

Updated 05/20/2006