The objective of this study was to determine the variables that predict severe intimate partner violence (S-IPV) according to the typology of abusers. The data were derived from 1,610 police reports on intimate partner violence (IPV) in Catalonia (Spain) between 2016 and 2017 obtained through the Police Risk Assessment Questionnaire. The study has compared a group of antisocial aggressors ( n = 613) with a group of family-only perpetrators ( n = 997). The chi-square test shows significant differences between antisocial and family-only groups for most of the variables analyzed. To determine the predictive variables of S-IPV in both groups, binary regression analyses were performed. In the antisocial group, death threats and degrading treatment by the aggressor significantly increased the probability of S-IPV, as did the victim’s minimization or justification of the abuse, living together with the aggressor, isolation, and drug or alcohol abuse. In the family-only group, an increase in the severity of the abuse and death threats against partners significantly increased the likelihood of perpetrating S-IPV. For the victims, being abused by a previous partner and fear for her physical integrity were found to increase the probability of suffering S-IPV. On the other hand, having filed a prior complaint appears to protect women from S-IPV, but only when the victims have antisocial perpetrators. The findings show that S-IPV risk factors are common regardless of the sociocultural context. Modifying the weighting of the factors that make up the risk assessment tools according to the typology of the abuser is suggested, as well as improving knowledge of these factors to increase the accuracy of the estimated risk. Finally, adapting supervision and monitoring measures according to the type of aggressor and taking into consideration the woman’s own perception of the danger she is in are also suggested.