This study was designed to investigate the base rate of violent behavior, the predictive validity of the Classification of Violence Risk (COVR) software, and specific risk factors for violence among nonforensically involved psychiatric patients in Sweden.
On discharge from two psychiatric hospitals in Stockholm, 331 patients were interviewed. Telephone interviews with the patients and supportive others, as well as data from a national criminal register, were used to measure violent behavior 20 weeks after discharge. After the baseline interview, patients were assigned to different risk groups by the COVR software. Predicted risk was compared with the occurrence of actual acts of violence during the follow-up.
Gender differences in base rates of violent behavior among the general psychiatric population were not found during the 20 weeks of follow-up after discharge. Violent behavior was significantly predicted by young age of males and by level of anger, violent thoughts, and victimization of females. The predictive validity of the COVR software was comparable between females (area under the curve [AUC]=.78) and males (AUC=.76).
Violent behavior was uncommon for all patients. Although several risk factors were significantly associated with violence by each gender, the COVR software could predict violence equally well for both genders.