Domestic Violence Offenders: Predictors of Attrition and Evaluation of Recidivism Effect in a Hybrid Diversion Program

Background and Purpose: Attrition in batter treatment interventions is an established problem.  Yet, its connection with recidivism remain unclear. Dropouts may be at greater risk of recidivating than untreated offenders (McMurran & Thedosi, 2007). However, Olver and colleagues (2011) caution that common risk factors for attrition and recidivism may confound interpretation. Differences in court and treatment approaches may also be relevant. The present study aimed to 1) identify predictors of attrition, 2) evaluate relative recidivism risk for non-completers, and 3) examine that risk while controlling for predictors of attrition in a misdemeanor court using a CBT-Duluth Model hybrid.

Methods:          

Completers and dropouts were the groups in this cohort observational study.  A proportional random sample (n=221) was drawn from the two-year population of offenders referred to the program between 7/1/09 and 6/30/11 (N=1180). The sample was stratified based on treatment year, completion status, and gender.   The sample was 75% male, young (median age=29), and ethnically diverse.  Motivation was evaluated by the Safe at Home instrument and connection to employment or school was assessed. Criminal follow-up was conducted by the prosecutor’s office to identify one-year recidivism for any new offense and domestic violence. The analysis had three steps: 1) multiple logistic regression was used to identify predictors of attrition, 2) relative risk ratios were computed to evaluate the bivariate treatment effect of program completion on recidivism, and 3) multivariate predictors of recidivism were evaluated through logistic regression. 

Results: Attrition was significantly predicted by a model with age, gender, race/ethnicity, work/school engagement, and the motivational scales.  The following variables were significant: age (OR=.96), work/school (OR=3.55), and the preparation/action scale (OR=.51). The absolute risk for any recidivism was 11% and for a DV charge was 5%. Program completion significantly predicted lower rates of any recidivism (RR = .33) and DV recidivism (RR=.16).  Any recidivism was not significantly predicted by a model that included program completion, age, gender, race/ethnicity, work/school and the motivational scales. As the omnibus test was marginal, it is worth noting that program completion (OR=3.20) and less contemplation (OR=1.65) were significant predictors of recidivism.  DV recidivism was significantly predicted by a model that included program completion, age, gender, race/ethnicity, work/school and the motivational scales.  Within this model, program completion was the only significant predictor (OR=6.42) Note: if accepted, the presentation will contain complete descriptions of all analyses.

Conclusions and Implications: The study supports the importance of age, engagement in work or school and motivation in predicting program completion by DV offenders. When compared with established research, the above rates of recidivism are lower and the treatment effect of program completion is larger. This study offers support for program completion lowering future domestic violence recidivism when controlling for the same demographic, socio-economic and motivational variables that predict program completion with misdemeanor offenders receiving a hybrid of CBT and the Duluth Model The design is unable to evaluate if dropping out had an iatrogenic effect.  Implications for future research, policy and practice will be explored.