Predictors of aggression and self-harm: A pilot study of the validity of the HCR-20 and VRS within a high secure hospital Simon Draycott (Broadmoor Hospital), Emily Glorney (University of Surrey), Darragh O’Neill (University of Surrey) and Samantha McCullough (University of Surrey) Reliable and accurate risk assessments within a forensic environment aid the identification of those that pose a high risk of harm to the public, the staff, other patients and themselves (Daffern & Howells, 2007; Mills, 2005). It has been noted that approximately 60% of patients had engaged in at least one act of aggression (such as verbal, physical aggression against others, physical aggression against property and inappropriate sexual behaviour) (Nichollsa et al., 2009). Self-harming behaviour is also more prevalent within a forensic environment as 17% of convicted male adult offenders in the prison service and 19% of male patients in a secure hospital had engaged in self-harming behaviour during their sentence (Jackson, 2000; Maden et al., 2000). At present actuarial risk assessments are most commonly used within forensic settings, combining statistical accuracy and clinical judgement. However it is debated whether measures that employ more historical (static) risk factors or changing (dynamic) risk factors are more reliable at measuring risk. Widely used in forensic services, the HCR-20 is comprised of 10 historical items, 5 clinical items and 5 risk management items (Webster, Douglas, Eaves, & Hart, 1997). It is noted as a strong assessment tool with high predictive validity (Daffern & Howells, 2007). In comparison the Violence Risk Scale (VRS) is comprised of 20 dynamic risk factors such as lifestyle, clinical change and attitudes and 6 static items (Dolan & Fullam, 2007). The VRS has been shown to have good inter-rater reliability, internal consistency and high predictive validity for short term and long term violent recidivism (Wong & Gordon, 2006). Stone (2002) encouraged the use of dynamic risk factors as they allow a current assessment of the individual and their present risk. The HCR-20 and VRS were explored as the HCR-20 utilises a number of static factors and the VRS utilises a number of dynamic factors offering a greater comparison as well as both having strong validity between level of risk determined and aggressive and self-harming behaviours (Dolan & Fullam, 2007; Gray, et al., 2003). Within the Dangerous and Severe Personality Disorder (DSPD) unit at Broadmoor Hospital there is a clinical requirement to complete risk assessments at least annually. Therefore it is hypothesised that the VRS will be more predictive of aggression, violence and self-harming behaviour than the HCR-20. It is also predicted that the VRS will be more predictive over time due to the dynamic risk factors, with the second assessment showing a higher predictive validity in a later time period as it should have been adjusted for clinical change. Results For Time 1 the Spearman’s rho revealed a statistically significant relationship between self-harm incidents and both HCR-20 and VRS scores. For Time 2 there was a significant relationship between the total number of incidents and number of violent incidents and both HCR-20 and VRS scores. All other relationships were not established. The data for the MANOVA was transformed using “Rank Cases” and followed up with histograms. There were no significant effects noted from the MANOVA, but a stronger relationship between total incidents in Time 2 and the HCR-20 scores taken at Time 1, F (1, 6) = 5.518, p = .057, was found. Table 1 Statistical Results of the Spearman’s rho including significance and effect sizes for all variables Conclusions The results did not indicate that the VRS was more predictive of aggression, violence and self-harming behaviour, nor did they support the hypothesis that the VRS was more predictive over time. Due to the sample size available was limited ROC Curve analysis was not possible which is considered a reliable statistical analysis when measuring rates of aggression (Douglas & Webster, 1999). As psychopathy can be used to make accurate predictions of violence and many clinicians have expressed concerns over the treatment of psychopathy and patients’ therapeutic resistance this may further limit the results (Barbaree, 2005; Hart, 1998). However this has provided an opportunity to look at risk of violence and self-harm among the Personality Disorder population, where these is limited validity studies of risk assessments. It has also identified the need for further research examining the use of dynamic risk assessments. Methodology The study employs a correlational design, involving the total scores of the HCR-20 and VRS and the number of incidents reported within time periods, analysed following secondary data analysis procedures (Andersen, Prause & Cohen-Silver, 2011). Participants were all male patients within the DSPD Unit at Broadmoor Hospital. The average age of patients was 43.3 years and all had a primary mental health diagnosis of psychopathic personality disorder. Two individual time frames (T1 and T2) were identified using the available incident report data. Each case was matched to ensure they have a valid risk assessment score for both measures within each time period, and due to this many cases were excluded. There were departures from normal distribution thus a Spearman’s rho and MANOVA were used to assess the relationship and this over time. I would like to thank both my supervisors and Broadmoor Hospital for their support with this project.