Intimate Partner Violence Risk Assessment: Predictive Ability & Application to Social Work

Background: In their lifetimes, one-third of US women experienced physical violence, rape, and/or stalking by an intimate partner, and one-quarter experienced severe physical violence (Black et al., 2011). Given the prevalence and negative sequelae of intimate partner violence (IPV) (e.g., Campbell, 2002; Coker et al., 2002), social workers are likely to encounter survivors across a number of practice settings. The use of validated IPV risk assessment instruments can be incorporated into practice as evidence-based intervention. Assessment scores can aid in safety planning, advocacy, and counseling (Campbell, 2002, 2004; Kress, 2008); help determine who is appropriate for batterers’ treatment (Morgan & Gilchrist, 2010; Jones & Gondolf, 2001); inform police, prosecutorial, and judicial responses to IPV (Bennett, Goodman & Dutton, 2000; Hilton et al., 2004; Roehl & Guertin, 2000); and facilitate communication between social workers and professionals in other systems (Kropp, 2004; Shepherd et al., 2002). The purpose of this research is to determine the IPV risk assessment instrument with the greatest predictive accuracy and provide social workers with guidelines for application depending on practice setting. 

Methods: To date, multiple research studies have examined predictive ability of the Danger Assessment (DA), the Domestic Violence Screening Inventory (DVSI), the Kingston Screening Instrument for Domestic Violence (K-SID), the Spousal Assault Risk Assessment (SARA), and the Ontario Domestic Assault Risk Assessment (ODARA). The average predictive validity weighted by sample size was calculated using the Receiver Operating Characteristic Area Under the Curve (AUC).

Results: The ODARA has the highest average weighted AUC (=.667, k=4) followed, from most to least predictive, by the SARA (AUC=.628, k=6), DA (AUC=.618, k=4), DVSI (AUC=.582, k=3), and K-SID (AUC=.537, k=2). The effect size for average AUCs for IPV risk assessment instruments is small, with the exception of a medium effect size for the ODARA (Rice & Harris, 2005).

Conclusions: Given the average AUC for these instruments, and consideration of their intended uses, it is suggested that social workers working with law enforcement officers or with offenders at the time of arrest use the ODARA. The DVSI is most appropriate for social workers working within the court system to assist with determinations about pre-trial release and may be useful for communicating risk to judges, prosecutors, and probation officers. The SARA has broader applicability for social workers who straddle criminal justice and victim services settings. The DA should be used by social workers employed in victim advocacy settings or those coming into contact with survivors in their professional capacity. The SARA and DA are more useful than the ODARA or the DVSI in determining targets for intervention. Social workers are in the unique position of identifying survivors’ risk from a variety of practice settings and can use risk assessment results as a means to communicate safety concerns and information about risk factors to IPV survivors as well as other actors within similar or complementary intervention systems.