摘要：Spanish police makes an extensive use of intimate partner violence (IPV) risk assessment on a daily basis. Improved prediction procedures have encouraged the search for greater refinement of IPV predictors by adjusting to specific targets, such as lethal outcomes or potential victimization of children. This paper describes the evolution of the VPR5.0 tool (VioGén System Police Risk Assessment) as an algorithm aimed at improving predictability of intimate partner homicides (IPH). A sample of 2,159 records was used, 159 of whom were IPH victims. The sample was divided into two comparable groups of cases (IPH) and controls (N-IPH) to validate the results. The results showed that 13 out of 35 risk factors were significantly related to IPH with an effect size different to that of general N-IPH (with OR values ranging between 1.507 and 8.087). Binary logistic regression showed six significant factors that correctly classified 86.3% of the IPH. The new H-Scale performance parameters were comparable to those obtained in studies with the same objective (sensitivity 84%, specificity 60%, OR = 8.130, AUC = .80, PPV = .19 and NPV = .97).
