Risk assessment of intimate partner violence in a police setting : reliability and predictive accuracy

Sammanfattning: The Swedish Police Authority conducts violence risk assessments in cases of intimate partner violence (IPV) using specific assessment tools. Such assessments are conducted in order to identify high-risk offenders and thereafter implement suitable interventions to prevent repeat IPV. In this thesis, two different risk assessment tools have been evaluated: The Police Screening Tool for Violent Crimes (PST-VC) and the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER, Kropp, Hart, & Belfrage, 2005; 2010). The overall aim has been to contribute to improving the knowledge on police employees’ violence risk assessment and management, specifically with regard to the predictive validity and inter-rater reliability of such assessments. In the first study, we evaluated whether the PST-VC can be used by police employees to identify high-risk cases of repeat IPV. In addition, the preventive effects of the recommended crime preventive and victim protective actions were discussed and also whether these create a confounding problem with respect to predictive validity. The results showed that the predictive accuracy of the tool was fairly weak. Further, the assessors recommended a higher level of interventions in high-risk cases, but these did not reduce the rate of repeat IPV. Study II aimed to examine the inter-rater reliability of the PST-VC and the BSAFER. Police employees conducted pairwise assessments of IPV cases using one of these tools. The tools were evaluated separately and the cases used for the assessments were different for each tool. This means that the consistency of the assessments could not be compared head-to-head across the tools. The results were nonetheless rather similar for both tools; the inter-rater reliability for the individual items was low for most of the individual factors, but was relatively high for the global risk assessments. A suggested explanation for this was that the assessors may have used their tacit knowledge, rather than the individual items, in their global risk assessments and that they shared this tacit knowledge, at least to some extent. The third study focused on the B-SAFER tool, and on the predictive accuracy of the individual items and the global risk assessments in relation to repeat IPV. The study also aimed to examine to what extent the recommended crime preventive and victim protective actions were implemented and whether these interventions had a preventive effect on repeat IPV. The predictive accuracy of the individual B-SAFER items and the global risk assessments was low overall. The majority of the recommended interventions were not implemented, and they did not prevent repeat IPV. The final study (IV) took the form of a systematic literature study with the aimof evaluating the predictive accuracy of IPV risk assessments conducted bypractitioners in different settings, with IPV recidivism as the outcome measure.The number of studies that fulfilled the inclusion criteria was small (N= 11).One of these studies was conducted in a treatment setting, while all the otherswere conducted in criminal justice settings. The predictive accuracy for theglobal risk assessments ranged from low to medium, and the role of treatmentor other interventions to prevent repeat IPV had been analyzed in one way orthe other in eight of the studies. However, there was no consistency withregard to the importance of the interventions for repeat IPV.In summary, the predictive accuracy of the police employees’ IPV riskassessments was rather low, and the same applied to the inter-rater reliabilityfor most of the individual items included in the tools. The level of consistencywas higher, however, for the global risk assessments. The IPV preventiveinterventions were not effective in preventing repeat IPV. The predictivevalidity of IPV risk assessments conducted in other settings was found to besimilar, but results regarding the potential mediating role of interventions weremixed.