Structured professional judgement approach to risk assessment: generalisability across patient groups for the prediction of adverse outcomes in secure mental health care

This is a redacted version of the the final thesis. Copyright material has been removed to comply with UK Copyright Law.

This thesis comprises a rigorous and coherent body of work related to the use of the HCR-20 and the START to inform risk assessment and management of secure mental health inpatients. The thesis contributes significant theoretical and applied knowledge by: 1) investigating the extent to which these tools can be generalised beyond restricted validation samples to the full range of individuals in contact with secure services, 2) determining whether they can aid assessment and management of adverse outcomes beyond aggression, and 3) offering practical, empirically-derived advice for clinicians regarding management strategies that may reduce the occurrence of adverse events. This collection of papers has used considerably novel methods, such as rocreg analysis in risk assessment of behavioural outcomes, and high quality, routinely collected data to gain a more realistic representation of what occurs in clinical practice. Further, the papers draw on larger sample sizes than have previously been reported in this area, allowing for more complex statistical analysis. This thesis has helped clarify the contexts in which these instruments perform effectively and therefore has important implications for clinical risk assessment in inpatient settings. Specifically, there is evidence that the HCR-20 and the START may aid assessment and management of aggression for the majority of groups examined, and that both tools have some efficacy for predicting self-harm among female populations. However, the HCR-20 should not be used to inform prediction and management of aggression and self-harm for individuals with developmental and organic disorders and is unlikely to be beneficial at informing risk management strategies targeting self-harm among males; the START should not be used to inform prediction and management of substance abuse, victimisation, or unauthorised leave. Finally, this thesis highlights a number of directions for future research to continue advancement in this area.