The aim of this study was to develop a new assessment tool to predict intimate partner femicide and severe violence. The sample for this study consisted of 1,081 men who were reported to the police station (Basque Country, Spain), because of having committed intimate partner violence. First, the most significant differences between the severe violence group (n=269) and the less severe violence group (n=812) in sociodemographic variables were determined. Results showed that both the perpetrators and the victims of the severe violence group had a higher rate of immigration. Second, the proposed 20-item scale derived from a larger 58-item scale, where only the most discriminative items between severe and non-severe intimate partner violence were taken into account. Psychometric properties of reliability and validity were rather good. Cut-off scores have been proposed according to sensitivity and specificity. This structured professional judgment (an easy-to-use tool) appears to be suitable to the requirements of criminal justice professionals and is intended for use as the basis of safety planning. Implications of these results for further research are commented upon.