Abstract
Risk assessment tools in forensic psychiatry have historically been validated on predominantly male offender populations, raising questions regarding their generalizability to female violent offenders. This study evaluated the 10-year predictive validity of the Historical Clinical Risk Management-20, Version 3 (HCR-20 V3) in a cohort of adult female violent offenders discharged from secure forensic psychiatric facilities. A sample of 284 female forensic patients evaluated at discharge between 2010 and 2014 was tracked prospectively over a 10-year follow-up period using national criminal and psychiatric registries. Receiver Operating Characteristic (ROC) analyses and Cox proportional hazards regression models were conducted to evaluate predictive accuracy for violent and general recidivism. Over the follow-up period, 22.5% of the sample engaged in violent reoffending and 38.0% in general reoffending. The HCR-20 V3 Total score demonstrated moderate predictive validity for violent recidivism (AUC = 0.71, 95% CI [0.64, 0.78]) and general recidivism (AUC = 0.74, 95% CI [0.68, 0.80]). Notably, dynamic subscales—the Clinical (C) subscale (AUC = 0.73) and Risk Management (R) subscale (AUC = 0.75)—outperformed the static Historical (H) subscale (AUC = 0.63) in predicting violent reoffending. Cox regression indicated that dynamic items reflecting emotional dysregulation, relationship instability, and substance misuse were strong independent predictors of time-to-violent reoffending. Findings support the application of the HCR-20 V3 for female forensic patients, while highlighting the heightened importance of dynamic, contextual risk factors in female violence risk management.
