Assessing risk for inpatient violence on high-security forensic psychiatric units

Abstract

While forensic psychiatric inpatient violence is a serious problem, research on risk assessment for this outcome is limited; the current research investigated the predictive validity of a number of structured risk/forensic instruments for inpatient violence. Research objectives included: 1) observing the profile of dynamic changes in violence risk detected by existing violence risk assessment instruments; 2) assessing whether existing violence risk assessment instruments could be used to assess risk for inpatient violence; 3) evaluating the contribution of dynamic risk measures to the prediction of inpatient violence; and 4) assessing the relationship between dynamic changes in risk and inpatient violence. Instruments included: the Historical Clinical Risk Management 20 - Version 3 (HCR-20V3), the Psychopathy Checklist Revised (PCL-R), the Short-Term Assessment of Risk and Treatability (START), the Revised Violence Risk Appraisal Guide (VRAG-R), and the Violence Risk Scale (VRS). Two studies were conducted on a maximum-security forensic psychiatric unit at Alberta Hospital Edmonton. Study 1 was a pseudo-prospective archival investigation (n = 99), while Study 2 was a prospective investigation (n = 19); all risk assessment scores were based on information available in institutional files.

Instruments designed to capture dynamic/clinical risk variables (HCR-20V3, START, VRS) detected dynamic changes in risk in this setting over longer follow-ups (i.e., between admission and discharge), but not over shorter follow-ups (i.e., 28 day periods). Predictive validity analyses indicated that specialized measures designed to capture relevant dynamic/clinical variables were significant predictors of inpatient violence; instruments that were not designed for this purpose (PCL-R and VRAG-R), did not demonstrate predictive validity for inpatient violence. Dynamic measures consistently demonstrated incremental predictive validity for inpatient violence, beyond the static measures. Additionally, change scores demonstrated incremental relationships with decreased inpatient violence, beyond pretreatment scores. Put another way, positive risk change was associated with decreased violence over the course of the patients’ stays in hospital.

Reliable and valid risk assessments are a necessary component of effective offender programming (Risk-Need-Responsivity Model) and the current results indicated that valid violence risk assessments for forensic psychiatric inpatient violence are possible. Implications for clinical practice and the reduction/mitigation of inpatient violence are discussed.