ABSTRACT
This prospective, naturalistic, cohort study was designed to report the one-year rate of violence among Swedish forensic psychiatric clients, during forensic psychiatric care, prison, and probation. Another purpose was to validate the risk assessment instruments COVR, LSI-R, HCR-20V3, and SAPROF in these settings. We assessed 200 detainees undergoing a forensic psychiatric investigation in Stockholm by means of the four instruments, 193 of them were followed up for one year. Violent acts were committed by 23.3% of the participants, and occurred more often at forensic psychiatric hospitals than correctional settings. According to the analyses, the predictive validity was small for COVR (AUC = .61), medium for LSI-R (AUC = .70), and large for HCR-20V3 (AUC = .79) and SAPROF (AUC = .78).