Structured clinical assessment and management of risk of violent recidivism in mentally disordered offenders

Background: The assessment of risk of violence among mentally disordered offenders has been a controversial but well-researched area in forensic psychology and psychiatry during the last decades. The main focus of this research has been on the predictive validity of various risk factors and methods of combining risk factors to gain the highest possible predictive accuracy. In the present thesis, risk assessment is defined more broadly than predictive accuracy, and also includes process factors, risk-management, and the communication and decision-making associated with risk of violence. The overall aim of this thesis was to explore the process of structured risk assessment in its naturalistic clinical setting. Method: Four different samples were across 5 studies. A guideline for structured clinical risk assessment, Historical-Clinical-Risk assessment (HCR-20), was used in the first 4 studies. The first study used a 6 raters x 6 patients design to establish inter-rater reliability and validity of the HCR-20. 54 forensic patients were followed over time and monitored for inpatient violence and violence after discharge during three risk-management conditions in study 2. A sample of 40 nurses, assessing the same 8 patients, was included in studies 3 and 4. Finally study 5 included a sample of 88 decisionmakers, divided into 3 groups; Clinicians, Criminal law professionals and Controls. Results: The HCR-20 was found to have reasonable reliability and validity in study 1. The main finding in study 2 was that the predictive accuracy of the HCR-20 was influenced by the intensity of risk management (AUC .64 compared to .82). In study 3 it was found that structured clinical risk assessment was not "immune" to emotional bias in the assessment process. 43% of the variance in risk-scores could be attributed to the assessors' emotions towards the patient. The information utilised to make the assessment, and how the assessor values it, also influenced the assessments in study 4. Placing value on personal interaction was more associated with inpatient violence than with recidivism. In study 5 we found that the inclination of making release decisions was greatly influenced (eta2=.58 ) by the prospect of making false negative error of judgement. Conclusions: Structured clinical risk assessments can be undertaken in a reliable and valid way in forensic clinical settings. Attention needs to be paid to factors that might influence the outcome of the assessments and the risk-management decisions that are the consequence of risk assessment. These factors can be emotional biases, evaluation of different kinds of information that form the basis for the process. There needs to be an awareness of other factors than probabilities that influence decisions about risk. It is suggested that future descriptive, as opposed to prescriptive, research is needed on the processes and influences on risk assessments, as they are actually conducted by clinicians in actual forensic, psychiatric and correctional settings, and not by researchers or trained research assistants