Assessing the cognitive contributors to violence and their potential to improve the predictive accuracy of contemporary risk assessments

Research on risk assessments has made considerable progress in recent years, with the development of long- and short-term predictive tools, as well as those which are specific to different types of offending including violent, sexual, and general offending. However, while there are many assessments, they differ little in their content and, as a result of the repetitive use of certain risk factors, researchers have suggested that contemporary risk assessments have hit a glass ceiling, as many have only moderate predictive accuracy. Risk assessments are further limited by the use of self-report, collateral information, and file reviews to assess clinical and risk-related factors, rather than validated performance measures, hindering their predictive accuracy. In parallel with these findings, converging evidence from neuropsychology and neurobiology has underlined the brain regions associated with violent behaviour and subsequent research has further demonstrated this through observational studies. The link between neuropsychological functioning and violence risk raises the potential for behaviourally measured neuropsychological risk factors to improve the predictive accuracy of existing risk assessments as well as aiding in the identification and management of these risk factors. The overarching aim of this thesis was to develop a risk assessment tool, for use alongside existing risk assessments, to increase their predictive accuracy, and to aid clinicians in the assessment and management of neuropsychological risk factors for violence. The development of the proposed tool followed the example of the National Institute of Health’s development of the MATRICS Consensus Cognitive Battery (MCCB) which sought to identify not only cognitive domains identified by the research literature, but also those that, in a relatively new field, may yet to be examined empirically, through expert consensus. Following the introductory and background chapters which provide context on the aetiology of cognitive impairments in violent populations, and a critical evaluation of the methodological limitations in violence research and violence risk assessments, are a cohesive set of studies which build upon each other to lay the foundation for the development of a structured risk assessment tool. The first step in this process was to critically examine the state of the literature and to identify neuropsychological measures which differentiate violent from non-violent offenders, and then to examine their association with violent outcomes in predictive and correlational studies. A systematic review and meta-analysis was undertaken to quantitatively summarise the association between neuropsychological measures and violent outcomes. After acknowledging that existing reviews in this area have largely focused on executive functions and specific diagnostic groups, the review adopted a broader approach, first examining factors which differentiate violent from non-violent offenders (part 1), followed by separately analysing the neuropsychological correlates of violence (part 2). Forty-two studies were included in the analyses, and 12 individual neuropsychological domains were examined in part 1, and five in part 2. The findings from this study revealed a large range of effect sizes with wide confidence intervals, highlighting significant heterogeneity due to methodological differences between studies, making it difficult to draw definitive conclusions. However, measures of impulsivity, inattention, and lack of insight boasted significant correlations with prospectively measured violent outcomes, revealing their potential to add a small amount of incremental validity to existing risk assessments. In parallel to the review and meta-analysis, a Delphi study was conducted to achieve an expert consensus on the neuropsychological risk factors thought to be essential for inclusion in a violence risk assessment. The Delphi method consisted of three rounds, in which an international sample of researchers and clinicians participated. A consensus was achieved on a list of 10 cognitive domains, ranked in order of importance, including, poor inhibitory control, behavioural impulsivity, risk taking, affective empathy, response monitoring, affect recognition, cognitive insight, reasoning, cognitive empathy, and clinical insight. Taking the results from the meta-analysis and the Delphi study, a neuropsychological battery was developed consisting of measures of inattention, response inhibition, reasoning, response monitoring, risk taking, affect recognition, cognitive empathy, and clinical insight, as well as a measure of intelligence. A prospective pilot study was designed to test the feasibility of implementing a neuropsychological battery into routine practice to aid in the assessment and management of risk, and the battery was piloted in n= 32 mentally ill offenders from high, medium, and low secure settings within NHS Scotland, and n=31 violent offenders who are living in the community and are service users of Criminal Justice Social Work in Edinburgh, Scotland. The findings of this study evidenced the pervasive cognitive impairments seen in these populations with a history of violence, regardless of their setting, however, the potential to improve the predictive accuracy of existing risk assessments appeared limited. The final study was a pilot case-series which sought to re-consent inpatients from the high secure setting to examine neuropsychological measures which explain the variance in short-term risk of aggression. The study was designed with a focus on how the major limitations seen in the violence literature may be addressed, such as small samples and low base rates of observational outcomes. Although an effort was made to tackle these limitations, a low base rate of recorded incidents persisted. However, the study demonstrated how a repeated measures outcome can help to minimally increase power in the presence of small samples, and the use of a novel statistical analysis demonstrated how inferences can be drawn on continuous outcomes even when there is a low base rate of incidents, increasing the specificity of results. Taken together, the findings of this thesis demonstrate that neuropsychological measures have limited potential to improve the predictive accuracy of risk assessments, however, they bring focus to the often-profound cognitive impairments seen in violent offenders and the necessity of assessing for cognitive impairment in parallel with the assessment of risk. The robust assessment of cognitive strengths and weaknesses has the potential to enhance the formulation of risk and inform its management and treatment, optimising treatment outcomes and the management of violence risk.