Assessing the risk of domestic violence recidivism

Abstract

Aim: To determine what factors independently predict violent DV-related re-offending among a cohort of people convicted of a(ny) DV offence and given a non-custodial penalty.

Method: Data from the NSW Bureau of Crime Statistics and Research (BOCSAR) Reoffending Database were used to examine violent DV-related reconviction. A cohort of DV offenders convicted in 2011-12 was first identified using domestic violence lawpart codes, and followed up for two years. To identify the best fitting model we first examined bi-variate relationships between explanatory variables and the dependent variable. We then estimated a multivariate logistic regression model to determine which variables independently predicted reconviction. Finally, we tested the predictive validity of the model using a range of cross-validation strategies.

Results: Among the cohort of adult offenders (n = 14,660), 8% were reconvicted of a violent DV-related offence within two years of the index conviction. Eleven explanatory variables were found to best predict reconviction – representing offender demographic, index offence, and criminal history characteristics. The resulting model showed acceptable levels of predictive validity.

Conclusion: To the extent that they direct appropriate interventions, risk assessment tools could be one part of a more complete community safety strategy aimed at violent DV recidivism. Limitations of the current study are discussed.