Risk assessment tools in criminal justice and forensic psychiatry: The need for better data

Artificial intelligence and forensic mental health in Africa: a narrative review

This narrative review examines the integration of Artificial Intelligence (AI) tools into forensic psychiatry in Africa, highlighting possible opportunities and challenges. Specifically, AI may have the potential to augment screening in prisons, risk assessment/management, and forensic-psychiatric treatment, alongside offering benefits for training and research purposes. These use-cases may be particularly advantageous in contexts of forensic practice in Africa, where there remains a need for capacity building and service improvements in jurisdictions affected by distinctive sociolegal and socioeconomic challenges. However, AI can also entail ethical risks associated with misinformation, privacy concerns, and an overreliance on automated systems that need to be considered within implementation and policy planning. Equally, the political and regulatory backdrop surrounding AI in countries in Africa needs to be carefully scrutinised (and, where necessary, strengthened). Accordingly, this review calls for rigorous feasibility studies and the development of training programmes to ensure the effective application of AI in enhancing forensic-psychiatric services in Africa.

Evaluation of whether commonly used risk assessment tools are applicable to women in forensic psychiatric institutions

OBJECTIVE

By providing a structured assessment of specific risk factors, risk assessment tools allow statements to be made about the likelihood of future recidivism in people who have committed a crime. These tools were originally developed for and primarily tested in men and are mainly based on the usual criminological background of men. Despite significant progress in the last decade, there is still a lack of empirical research on female offenders, especially female forensic psychiatric inpatients. To improve prognosis in female offenders, we performed a retrospective study to compare the predictive quality of the following risk assessment tools: PCL-R, LSI-R, HCR-20 v3, FAM, and VRAG-R.

METHOD

Data were collected from the information available in the medical files of 525 female patients who had been discharged between 2001 and 2017. We examined the ability of the tools to predict general and violent recidivism by comparing the predictions with information from the Federal Central Criminal Register.

RESULTS

Overall, the prediction instruments had moderate to good predictive performance, and the study confirmed their general applicability to female forensic psychiatric patients.

CONCLUSION

The LSI-R proved to be particularly valid for general recidivism, and both, LSI-R and HCR-20 v3, for violent recidivism.

Facilitating a safe and caring atmosphere in everyday life in forensic mental health wards - a qualitative study

PURPOSE

To explore healthcare professionals' experiences with facilitating a safe and caring atmosphere in patients' everyday lives in forensic mental health wards.

METHODS

This qualitative study employed interviews with 16 healthcare professionals working shifts in two forensic mental healthcare wards in Norway. Data were analysed using phenomenological hermeneutic analysis.

RESULTS

The findings are presented in terms of two themes. The first theme is "Creating a calming atmosphere" and includes the subthemes "Creating caring surroundings with safety, comfort and trust" and "Balancing everyday life activities". The second theme is "Facilitating risk assessments and care" and includes the subthemes "Acting as a team", "Becoming aware of the meaning in signs" and "Becoming aware of vulnerability and the window of tolerance".

CONCLUSIONS

Involvement in patients' history and lived lives is important both for understanding general social behaviour as well as for assessing signs, symptoms, and changes in patients' conditions; furthermore, it provides valuable information that allows healthcare professionals to become aware of the underlying meanings in signs, which can facilitate examinations and treatment. Acting as a team is essential to solve issues in a calm and safe way when signs of violence occur. In addition, our participants highlighted the need to be aware of individual patients' vulnerability and windows of tolerance to obtain a deeper understanding of patients' lived lives as a whole in the context of providing therapy and care to patients.

Evidence-based sentencing and scientific evidence

Evidence-based sentencing (EBS) is a new name for an aspiration that has deep roots in criminal law: to apply the sentence most appropriate to each offender's risk of reoffending, in order to reduce that risk as far as possible. This modern version of the traditional sentencing goals of rehabilitation and incapacitation fits into the broader approach of so-called "evidence-based public policy." It takes the view that the best existing evidence for reducing reoffending are modern structured risk assessment tools and claims to be able to achieve several goals at once: reducing reoffending, maintaining high levels of public safety, making more efficient use of public resources, and moving criminal policy away from ideological battles by basing it on the objective knowledge provided by the best available scientific evidence. However, despite the success of this approach in recent years, it is not clear to what extent it succeeds in correctly assessing the risk of individual offenders, nor whether it achieves its intended effect of reducing recidivism. This paper aims to critically examine these two issues: the quality of the scientific evidence on which EBS is based, and the available data on the extent to which it achieves (or does not achieve) its intended goals.

Violence risk assessment instruments in forensic psychiatric populations: a systematic review and meta-analysis

BACKGROUND

Although structured tools have been widely used to predict violence risk in specialist mental health settings, there is uncertainty about the extent and quality of evidence of their predictive performance. We aimed to systematically review the predictive performance of tools used to assess violence risk in forensic mental health, where they are routinely administered.

METHODS

In our systematic review and meta-analysis, we followed PRISMA guidelines and searched four databases (PsycINFO, Embase, Medline, and Global Health) from database inception to Nov 1, 2022, to identify studies examining the predictive performance of risk assessment tools in people discharged from forensic (secure) mental health hospitals. Systematic and narrative reviews were excluded from the review. Performance measures and descriptive statistics were extracted from published reports. A quality assessment was performed for each study using the Prediction Model Risk of Bias Assessment Tool. Meta-analysis was conducted on the performance of instruments that were independently externally validated with a sample size greater than 100. The study was registered with PROSPERO, CRD42022304716.

FINDINGS

We conducted a systematic review of 50 eligible publications, assessing the predictive performance of 36 tools, providing data for 10 460 participants (88% men, 12% women; median age [from 47 studies] was 35 years, IQR 33-38) from 12 different countries. Post-discharge interpersonal violence and crime was most often measured by new criminal offences or recidivism (47 [94%] of 50 studies); only three studies used informant or self-report data on physical aggression or violent behaviour. Overall, the predictive performance of risk assessment tools was mixed. Most studies reported one discrimination metric, the area under the receiver operating characteristic curve (AUC); other key performance measures such as calibration, sensitivity, and specificity were not presented. Most studies had a high risk of bias (49 [98%] of 50), partly due to poor analytical approaches. A meta-analysis was conducted for violent recidivism on 29 independent external validations from 19 studies with at least 100 patients. Pooled AUCs for predicting violent outcomes ranged from 0·72 (0·65-0·79; I2=0%) for H10, to 0·69 for the Historical Clinical Risk Management-20 version 2 (95% CI 0·65-0·72; I2=0%) and Violence Risk Appraisal Guide (0·63-0·75; I2=0%), to 0·64 for the Static-99 (0·53-0·73; I2=45%).

INTERPRETATION

Current violence risk assessment tools in forensic mental health have mixed evidence of predictive performance. Forensic mental health services should review their use of current risk assessment tools and consider implementing those with higher-quality evidence in support.

FUNDING

Wellcome Trust.

Out of their minds? Externalist challenges for using AI in forensic psychiatry

Harnessing the power of machine learning (ML) and other Artificial Intelligence (AI) techniques promises substantial improvements across forensic psychiatry, supposedly offering more objective evaluations and predictions. However, AI-based predictions about future violent behaviour and criminal recidivism pose ethical challenges that require careful deliberation due to their social and legal significance. In this paper, we shed light on these challenges by considering externalist accounts of psychiatric disorders which stress that the presentation and development of psychiatric disorders is intricately entangled with their outward environment and social circumstances. We argue that any use of predictive AI in forensic psychiatry should not be limited to neurobiology alone but must also consider social and environmental factors. This thesis has practical implications for the design of predictive AI systems, especially regarding the collection and processing of training data, the selection of ML methods, and the determination of their explainability requirements.

Predictive validity on clinical item-level of the HKT-R divided into clinical patient classes

BACKGROUND

Because of the heterogeneity of forensic groups, latent class analysis (LCA) can allow for the formation of stronger homogeneous patient classes, which can improve the predictive validity of forensic risk assessment tools, such as the Historical Clinical Future - Revised (HKT-R), which was used in this study. In particular, dynamic clinical risk and protective items are important in treatment and are obligatory assessed annually for every forensic patient with a TBS measure in the Netherlands. Therefore, this study investigated the predictive validity of the HKT-R at clinical item-level per patient class.

METHOD

A cohort of 332 forensic patients, who were discharged from highly secured Forensic Psychiatric Centers/Clinics (FPCs) in the Netherlands between 2004 and 2008, was followed. LCA was performed to cluster this group of patients based on psychopathology and criminal offenses. The predictive validity of the HKT-R clinical items by class was assessed with official reconviction data two and five years after discharge as outcome measure.

RESULTS

Four classes were identified. The predictive validity of the HKT-R clinical items showed differences between and within classes on admission or discharge, and for predicting violent reoffending after two or five years after discharge.

DISCUSSION

Different risk/protective factors of the HKT-R may play a role for different subgroups of patients. Therefore, this heterogeneity should be considered for any measure or intervention.

Development and Validation of a Prediction Tool for Reoffending Risk in Domestic Violence

Importance

Current risk assessment tools for domestic violence against family members were developed with small and selected samples, have low accuracy with few external validations, and do not report key performance measures.

Objective

To develop new tools to assess risk of reoffending among individuals who have perpetrated domestic violence.

Design, Setting, and Participants

This prognostic study investigated a national cohort of all individuals arrested for domestic violence between 1998 and 2013 in Sweden using information from multiple national registers, including National Crime Register, National Patient Register, Longitudinal Integrated Database for Health Insurance and Labour Market Studies Register, and Multi-Generation Register. Data were analyzed from August 2022 to June 2023.

Exposure

Arrest for domestic violence.

Main Outcomes and Measures

Prediction models were developed for 3 reoffending outcomes after arrest for domestic violence: conviction of a new violent crime (including domestic violence), conviction of any new crime, and rearrest for domestic violence at 1 year, 3 years, and 5 years. The prediction models were created using sociodemographic factors, criminological factors, and mental health status-related factors, linking data from multiple population-based longitudinal registers. Cox proportional hazard multivariable regression was used to develop prediction models and validate them in external samples. Key performance measures, including discrimination at prespecified cutoffs and calibration statistics, were investigated.

Results

The cohort included 27 456 individuals (mean [SD] age, 39.4 [11.6] years; 24 804 men [90.3%]) arrested for domestic violence, of whom 4222 (15.4%) reoffended and were convicted for a new violent crime during a mean (SD) follow-up of 26.5 (27.0) months, 9010 (32.8%) reoffended and were convicted for a new crime (mean [SD] follow-up, 22.4 [25.1] months), and 2080 (7.6%) were rearrested for domestic violence (mean [SD] follow-up, 25.7 [30.6] months). Prediction models were developed with sociodemographic, criminological, and mental health factors and showed good measures of discrimination and calibration for violent reoffending and any reoffending. The area under the receiver operating characteristic curve (AUC) for risk of violent reoffending was 0.75 (95% CI, 0.74-0.76) at 1 year, 0.76 (95% CI, 0.75-0.77) at 3 years, and 0.76 (95% CI, 0.75-0.77) 5 years. The AUC for risk of any reoffending was 0.76 (95% CI, 0.75-0.77) at 1 year and at 3 years and 0.76 (95% CI, 0.75-0.76) at 5 years. The model for domestic violence reoffending showed modest discrimination (C index, 0.63; 95% CI, 0.61-0.65) and good calibration. The validation models showed discrimination and calibration performance similar to those of derivation models for all 3 reoffending outcomes. The prediction models have been translated into 3 simple online risk calculators that are freely available to use.

Conclusions and Relevance

This prognostic study developed scalable, evidence-based prediction tools that could support decision-making in criminal justice systems, particularly at the arrest stage when identifying those at higher risk of reoffending and screening out individuals at low risk of reoffending. Furthermore, these tools can enhance treatment allocation by enabling criminal justice services to focus on modifiable risk factors identified in the tools for individuals at high risk of reoffending.

A clinical risk prediction tool for identifying the risk of violent offending in severe mental illness: A retrospective case-control study

BACKGROUND

Individuals with severe mental illness are at a higher risk of violence than the general population. However, there is a lack of available and simple tools to screen for the risk of violent offending in clinical settings. We aimed to develop an easy-to-use predictive tool to assist clinicians' decision-making to identify risk of violent offences in China.

METHODS

We identified 1157 patients with severe mental illness who committed violent offending and 1304 patients who were not suspected of violent offending in the matched living areas. We used stepwise regression and Lasso's method to screen for predictors, built a multivariate logistic regression model, and performed internal validation with the 10- fold cross-validation to develop the final prediction model.

RESULTS

The risk prediction model for violence in severe mental illness included age (beta coefficient (b) = 0.05), male sex (b = 2.03), education (b = 1.14), living in rural areas (b = 1.21), history of homeless (b = 0.62), history of previous aggression (b = 1.56), parental history of mental illness (b = 0.69), diagnosis of schizophrenia (b = 1.36), episodes (b = -2.23), duration of illness (b = 0.01). The area under curve for the predictive model for the risk of violence in severe mental illness was 0.93 (95% CI: 0.92-0.94).

CONCLUSIONS

In this study, we developed a predictive tool for violent offending in severe mental illness, containing 10 items that can be easily used by healthcare practitioners. The model was internally validated and has the potential for assessing the risk of violence in patients with severe mental illness in community routine care, although external validation is necessary.

Neurotechnology to reduce recidivism: Ethical and legal challenges

Crime comes with enormous costs, not only financial but also in terms of loss of mental and physical health and, in some cases, even loss of life. Recidivism is responsible for a considerable percentage of the crimes, and therefore, society deems reducing recidivism a priority. To reduce recidivism, several types of interventions can be used, such as education and employment-focused rehabilitation programs which are intended to improve psychological and social factors. Another way to prevent reoffending is to influence the offender's brain functions. For example, medication can be offered to treat delusions or to diminish sexual drive. In the near future, innovative neurotechnologies are expected to improve prediction and prevention of reoffending. Potential positive effects of such neurotechniques include a safer society and earlier release of prisoners who are no longer "at high risk" to relapse into criminal behavior. Meanwhile, employing these neurotechniques in the criminal justice system raises fundamental concerns, for example, about autonomy, privacy and mental integrity. This chapter aims to identify some of the ethical and legal challenges of using neurotechnologies to reduce recidivism.

Prediction of violence in male schizophrenia using sMRI, based on machine learning algorithms

BACKGROUND

Violent behavior in patients with schizophrenia (SCZ) is a major social problem. The early identification of SCZ patients with violence can facilitate implementation of targeted intervention.

METHODS

A total of 57 male SCZ patients were recruited into this study. The general linear model was utilized to compare differences in structural magnetic resonance imaging (sMRI) including gray matter volume, cortical surface area, and cortical thickness between 30 SCZ patients who had exhibited violence and 27 SCZ patients without a history of violence. Based on machine learning algorithms, the different sMRI features between groups were integrated into the models for prediction of violence in SCZ patients.

RESULTS

After controlling for the whole brain volume and age, the general linear model showed significant reductions in right bankssts thickness, inferior parietal thickness as well as left frontal pole volume in the patients with SCZ and violence relative to those without violence. Among seven machine learning algorithms, Support Vector Machine (SVM) have better performance in differentiating patients with violence from those without violence, with its balanced accuracy and area under curve (AUC) reaching 0.8231 and 0.841, respectively.

CONCLUSIONS

Patients with SCZ who had a history of violence displayed reduced cortical thickness and volume in several brain regions. Based on machine learning algorithms, structural MRI features are useful to improve predictive ability of SCZ patients at particular risk of violence.

A critical ethnographic perspective on risk and dangerousness in forensic psychiatry

In the Canadian forensic psychiatric context, the concepts of risk and dangerousness interact, intersect, and morph into the notion of significant threat to the safety of the public. Stemming from the results of a critical ethnography of the Ontario Review Board, this article unpacks the central role of forensic psychiatric nursing, as an example of a 'psych' discipline (e.g., psychiatry and psychology), in a system that is built to produce risky persons and to legitimize their detention and supervision. By using excerpt of interviews conducted with nurses, ethnographic observations of Review Board hearings, and other documentary artifacts, the findings illustrate how rationalizations of risk and dangerousness are contingent on space, time, and observer. Depending on the time of the assessment or on the health-care professional who performs it, different elements including, but not limited to, mental illness, interpersonal relationships, financial instability, and sexual vulnerability, are relied upon in very fluid, interchangeable, and discretionary ways to justify findings of dangerousness. Such a dynamic expands the reach of psychiatry's legitimacy at identifying risky conduct and controlling risky persons to domains very loosely associated with the notion of dangerousness. The work of Foucault and Castel provides the theoretical backdrop on which rests the discussion and the implications for nursing.

Predicting crime during or after psychiatric care: Evaluating machine learning for risk assessment using the Danish patient registries

BACKGROUND

Structural changes in psychiatric systems have altered treatment opportunities for patients in need of mental healthcare. These changes are possibly associated with an increase in post-discharge crime, reported in the increase of forensic psychiatric populations. As current risk-assessment tools are time-consuming to administer and offer limited accuracy, this study aims to develop a predictive model designed to identify psychiatric patients at risk of committing crime leading to a future forensic psychiatric treatment course.

METHOD

We utilized the longitudinal quality of the Danish patient registries, identifying the 45.720 adult patients who had contact with the psychiatric system in 2014, of which 474 committed crime leading to a forensic psychiatric treatment course after discharge. Four machine learning models (Logistic Regression, Random Forest, XGBoost and LightGBM) were applied over a range of sociodemographic, judicial, and psychiatric variables.

RESULTS

This study achieves a F1-macro score of 76%, with precision = 57% and recall = 47% reported by the LightGBM algorithm. Our model was therefore able to identify 47% of future forensic psychiatric patients, while making correct predictions in 57% of cases.

CONCLUSION

The study demonstrates how a clinically useful initial risk-assessment can be achieved using machine learning on data from patient registries. The proposed approach offers the opportunity to flag potential future forensic psychiatric patients, while in contact with the general psychiatric system, hereby allowing early-intervention initiatives to be activated.

The predictive performance of criminal risk assessment tools used at sentencing: Systematic review of validation studies

Although risk assessment tools have been widely used to inform sentencing decisions, there is uncertainty about the extent and quality of evidence of their predictive performance when validated in new samples. Following PRISMA guidelines, we conducted a systematic review of validation studies of 11 commonly used risk assessment tools for sentencing. We identified 36 studies with 597,665 participants, among which were 27 independent validation studies with 177,711 individuals. Overall, the predictive performance of the included risk assessment tools was mixed, and ranged from poor to moderate. Tool performance was typically overestimated in studies with smaller sample sizes or studies in which tool developers were co-authors. Most studies only reported area under the curve (AUC), which ranged from 0.57 to 0.75 in independent studies with more than 500 participants. The majority did not report key performance measures, such as calibration and rates of false positives and negatives. In addition, most validation studies had a high risk of bias, partly due to inappropriate analytical approach used. We conclude that the research priority is for future investigations to address the key methodological shortcomings identified in this review, and policy makers should enable this research. More sufficiently powered independent validation studies are necessary.

Towards a More Evidence-Based Risk Assessment for People in the Criminal Justice System: the Case of OxRec in the Netherlands

Risk assessment tools are widely used throughout the criminal justice system to assist in making decisions about sentencing, supervision, and treatment. In this article, we discuss several methodological and practical limitations associated with risk assessment tools currently in use. These include variable predictive performance due to the exclusion of important background predictors; high costs, including the need for regular staff training, in order to use many tools; development of tools using suboptimal methods and poor transparency in how they create risk scores; included risk factors being based on dated evidence; and ethical concerns highlighted by legal scholars and criminologists, such as embedding systemic biases and uncertainty about how these tools influence judicial decisions. We discuss the potential that specific predictors, such as living in a deprived neighbourhood, may indirectly select for individuals in racial or ethnic minority groups. To demonstrate how these limitations and ethical concerns can be addressed, we present the example of OxRec, a risk assessment tool used to predict recidivism for individuals in the criminal justice system. OxRec was developed in Sweden and has been externally validated in Sweden and the Netherlands. The advantages of OxRec include its predictive accuracy based on rigorous multivariable testing of predictors, transparent reporting of results and the final model (including how the probability score is derived), scoring simplicity (i.e. without the need for additional interview), and the reporting of a wide range of performance measures, including those of discrimination and calibration, the latter of which is rarely reported but a key metric. OxRec is intended to be used alongside professional judgement, as a support for decision-making, and its performance measures need to be interpreted in this light. The reported calibration of the tool in external samples clearly suggests no systematic overestimation of risk, including in large subgroups.

Violence Risk Assessment and Risk Management: Case-Study of Filicide in an Italian Woman

BACKGROUND

At an international level, the risk assessment and management process of violent offenders follows a standard method that implies well-defined theoretical models and the use of scientifically validated tools. In Italy, this process is still highly discretionary. The aim of this study is to highlight the advantages deriving from the use of risk assessment tools within the framework of a single case study; Methods: Recidivism risk and social dangerousness of an Italian woman perpetrator of filicide were assessed through the administration of the Level of Service/Case Management Inventory (LS/CMI) instrument supported by Historical Clinical Risk-20 Version 3 (HCR-20 V3); Results: The administration of LS/CMI showed that, in this single case, the subcomponents represent a criminogenic risk/need factor are: Family/Marital, Companions, Alcohol and Drug Problem and Leisure; while constituting strengths: employment and the absence of a Pro-criminal Orientation and an Antisocial Pattern; Conclusions: Data collected through LS/CMI indicated life areas of a single case, which should be emphasised not only to assess the risk of re-offending and social dangerousness but also for a social rehabilitation programme more suited to the subject. This study demonstrates that the LS/CMI assessment tool is suitable for the Italian context.

Validation and recalibration of OxMIV in predicting violent behaviour in patients with schizophrenia spectrum disorders

Oxford Mental Illness and Violence (OxMIV) addresses the need in mental health services for a scalable, transparent and valid tool to predict violent behaviour in patients with severe mental illness. However, external validations are lacking. Therefore, we have used a Dutch sample of general psychiatric patients with schizophrenia spectrum disorders (N = 637) to evaluate the performance of OxMIV in predicting interpersonal violence over 3 years. The predictors and outcome were measured with standardized instruments and multiple sources of information. Patients were mostly male (n = 493, 77%) and, on average, 27 (SD = 7) years old. The outcome rate was 9% (n = 59). Discrimination, as measured by the area under the curve, was moderate at 0.67 (95% confidence interval 0.61–0.73). Calibration-in-the-large was adequate, with a ratio between predicted and observed events of 1.2 and a Brier score of 0.09. At the individual level, risks were systematically underestimated in the original model, which was remedied by recalibrating the intercept and slope of the model. Probability scores generated by the recalibrated model can be used as an adjunct to clinical decision-making in Dutch mental health services.

The Prediction and Influential Factors of Violence in Male Schizophrenia Patients With Machine Learning Algorithms

BACKGROUND

Early to identify male schizophrenia patients with violence is important for the performance of targeted measures and closer monitoring, but it is difficult to use conventional risk factors. This study is aimed to employ machine learning (ML) algorithms combined with routine data to predict violent behavior among male schizophrenia patients. Moreover, the identified best model might be utilized to calculate the probability of an individual committing violence.

METHOD

We enrolled a total of 397 male schizophrenia patients and randomly stratified them into the training set and the testing set, in a 7:3 ratio. We used eight ML algorithms to develop the predictive models. The main variables as input features selected by the least absolute shrinkage and selection operator (LASSO) and logistic regression (LR) were integrated into prediction models for violence among male schizophrenia patients. In the training set, 10 × 10-fold cross-validation was conducted to adjust the parameters. In the testing set, we evaluated and compared the predictive performance of eight ML algorithms in terms of area under the curve (AUC) for the receiver operating characteristic curve.

RESULT

Our results showed the prevalence of violence among male schizophrenia patients was 36.8%. The LASSO and LR identified main risk factors for violent behavior in patients with schizophrenia integrated into the predictive models, including lower education level [0.556 (0.378-0.816)], having cigarette smoking [2.121 (1.191-3.779)], higher positive syndrome [1.016 (1.002-1.031)] and higher social disability screening schedule (SDSS) [1.081 (1.026-1.139)]. The Neural Net (nnet) with an AUC of 0.6673 (0.5599-0.7748) had better prediction ability than that of other algorithms.

CONCLUSION

ML algorithms are useful in early identifying male schizophrenia patients with violence and helping clinicians take preventive measures.

Protocols for Psychological Assessment of Recidivism and Dangerousness: a systematic review of the Brazilian scientific production

Abstract The psychological assessment of recidivism and dangerousness aims to provide subsidies to legal operators on the possibility of an individual to repeat criminal offenses. In the present study, a systematic review of Portuguese-language articles was carried out in the “SciELO”, “Lilacs” and “Periódicos Capes” databases to identify available instruments to carry out this assessment in Brazilian populations and their predictive capacity. It was found that the Brazilian scientific production is too scarce, only nine empirical studies have been published on the subject and only one instrument with this objective is suitable for use in forensic practice. Six other instruments have been studied, but none are suitable for practical implementation. These results point to the need for scientific production on psychological assessment of recidivism and dangerousness to provide psychologists with the necessary instruments for their performance in Legal Psychology.

Forensic Brain-Reading and Mental Privacy in European Human Rights Law: Foundations and Challenges

A central question in the current neurolegal and neuroethical literature is how brain-reading technologies could contribute to criminal justice. Some of these technologies have already been deployed within different criminal justice systems in Europe, including Slovenia, Italy, England and Wales, and the Netherlands, typically to determine guilt, legal responsibility, or recidivism risk. In this regard, the question arises whether brain-reading could permissibly be used against the person's will. To provide adequate legal protection from such non-consensual brain-reading in the European legal context, ethicists have called for the recognition of a novel fundamental legal right to mental privacy. In this paper, we explore whether these ethical calls for recognising a novel legal right to mental privacy are necessary in the European context. We argue that a right to mental privacy could be derived from, or at least developed within in the jurisprudence of the European Court of Human Rights, and that introducing an additional fundamental right to protect against (forensic) brain-reading is not necessary. What is required, however, is a specification of the implications of existing rights for particular neurotechnologies and purposes.

Innovations in violence risk assessment: What aviation can teach us about assessing and managing risk for rare and serious outcomes

This paper describes several ongoing challenges in the field of violence risk assessment (VRA), particularly with respect to establishing acceptable levels of measurement reliability and validity of commonly used risk assessment instruments, and demonstrating their ability to reduce risk and avert harmful outcomes. Drawing on analogous concepts from the risk assessment and management process in the aviation industry, several key lessons and aspirational principles for research and practice in the field of VRA are described. It is argued that significantly more attention is required to evaluate the ability of VRA tools to generate effective risk management plans that measurably lower risk and rates of violent outcomes. Three propositions for advancing common VRA research designs are discussed: (1) improved operationalization of risk management plans and their ability to reduce violence; (2) improved measurement of change in risk status over prospective follow-up periods, and (3) a stronger emphasis on short-term assessments with closer temporal proximity between risk factors and outcomes. Collectively, these advancements may enhance the validity and utility of VRA instruments by permitting better specification of the conditions under which risk factors exert effects, and the development of effective risk management plans that join together explanatory frameworks for the causes of violence with strategies to avoid their recurrence.

Predicting offenses among individuals with psychiatric disorders - A machine learning approach

BACKGROUND

Actuarial risk estimates are considered the gold-standard way to assess whether psychiatric patients are likely to commit prospective criminal offenses. However, these risk estimates cannot individually predict the type of criminal offense a patient will subsequently commit, and often simply assess the general likelihood of crime occurring in a group sample. In order to advance the predictive utility of risk assessments, better statistical strategies are required.

AIM

To develop a machine learning model to predict the type of criminal offense committed in a large transdiagnostic sample of psychiatry patients, at an individual level.

METHOD

Machine learning algorithms (Random Forest, Elastic Net, SVM), were applied to a representative and diverse sample of 1240 patients in the forensic mental health system. Clinical, historical, and sociodemographic variables were considered as potential predictors and assessed in a data-driven way. Separate models were created for each type of criminal offense, and feature selection methods were used to improve the interpretability and generalizability of our findings.

RESULTS

Sexual offenses can be predicted from nonviolent and violent offenses at an individual level with a sensitivity of 82.44% and specificity of 60.00%, using only 36 variables. Furthermore, in a binary classification model, sexual and violent offenses can be predicted at an individual level with 83.26% sensitivity and 77.42% specificity using only 20 clinical variables. Likewise, non-violent and sexual offenses can be individually predicted with 74.60% sensitivity and 80.65% specificity using 30 clinical variables.

CONCLUSION

The current results suggest that machine learning models can show greater accuracy than gold-standard risk assessment tools (AUCs 0.70-0.80). However, unlike existing risk tools, this approach allows for the prediction of cases at an individual level, which is more clinically useful. Despite this, it is important to note that a large subset of patients in the sample were involved in the criminal system in the past, prior to an official diagnosis. Therefore, many of the variables that predict offenses may be derived from the issues of prior offenses. Irrespective of this, the accuracy of prospective models is expected to only improve with further refinement.

Managing restricted patients in acute, non-secure in-patient services: clinical, ethical and resource implications of long waits for a response from the Ministry of Justice

AIMS AND METHOD

In-patients subject to Section 37/41 of the Mental Health Act 1983 (MHA) require permission from the Ministry of Justice (MoJ) for leave, transfer and discharge. This study aimed to quantify the time spent waiting for the MoJ to respond to requests, using data on restricted patients recalled to a non-forensic unit over 8 years.

RESULTS

Eleven admissions were identified. The mean total time waiting for response was 95 days per admission, with an estimated cost of £40 922 per admission.

CLINICAL IMPLICATIONS

Current procedures may contribute to considerable increases in length of stay. This goes against the principles of the MHA, as non-secure services rarely provide the range of interventions which justify prolonged admission. We suggest several ways to resolve this issue, including broadening the guidance for the use of voluntary admissions and civil sections, and allowing clinicians to make decisions on leave and transfer where there is little risk.

Violent and non-violent offending in patients with schizophrenia: Exploring influences and differences via machine learning

OBJECTIVES

The link between schizophrenia and violent offending has long been the subject of research with significant impact on mental health policy, clinical practice and public perception of the dangerousness of people with psychiatric disorders. The present study attempts to identify factors that differentiate between violent and non-violent offenders based on a unique sample of 370 forensic offender patients with schizophrenia spectrum disorder by employing machine learning algorithms and an extensive set of variables.

METHODS

Using machine learning algorithms, 519 variables were explored in order to differentiate violent and non-violent offenders. To minimize the risk of overfitting, the dataset was split, employing variable filtering, machine learning model building and selection embedded in a nested resampling approach on one subset. The best model was then selected, and the most important variables applied on the second data subset.

RESULTS

Ten factors regarding criminal and psychiatric history as well as clinical, developmental, and social factors were identified to be most influential in differentiating between violent and non-violent offenders and are discussed in light of prior research on this topic. With an AUC of 0.76, a sensitivity of 72% and a specificity of 62%, a correct classification into violent and non-violent offences could be determined in almost three quarters of cases.

CONCLUSIONS

Our findings expand current research on the factors influencing violent offending in patients with SSD, which is crucial for the development of preventive and therapeutic strategies that could potentially reduce the prevalence of violence in this population. Limitations, clinical relevance and future directions are discussed.

Is the Psychopathic Brain an Artifact of Coding Bias? A Systematic Review

Questionable research practices are a well-recognized problem in psychology. Coding bias, or the tendency of review studies to disproportionately cite positive findings from original research, has received comparatively little attention. Coding bias is more likely to occur when original research, such as neuroimaging, includes large numbers of effects, and is most concerning in applied contexts. We evaluated coding bias in reviews of structural magnetic resonance imaging (sMRI) studies of PCL-R psychopathy. We used PRISMA guidelines to locate all relevant original sMRI studies and reviews. The proportion of null-findings cited in reviews was significantly lower than those reported in original research, indicating coding bias. Coding bias was not affected by publication date or review design. Reviews recommending forensic applications—such as treatment amenability or reduced criminal responsibility—were no more accurate than purely theoretical reviews. Coding bias may have contributed to a perception that structural brain abnormalities in psychopaths are more consistent than they actually are, and by extension that sMRI findings are suitable for forensic application. We discuss possible sources for the pervasive coding bias we observed, and we provide recommendations to counteract this bias in review studies. Until coding bias is addressed, we argue that this literature should not inform conclusions about psychopaths' neurobiology, especially in forensic contexts.

The Rapid Risk of Violence Screen (RROVS): a Brief Violence Risk Screening Tool for People in a Community Behavioral Health Setting

The Rapid Risk of Violence Screen (RROVS) is a brief screening tool that is designed for use in community-based behavioral health service settings to identify people who may need comprehensive violence risk assessment. This study examined the association between the RROVS total score and future criminal justice involvement including violent offenses. Results from this study suggest that the RROVS screening tool has predictive validity as it is associated with later criminal justice involvement with a violent offense. The RROVS may be a helpful tool for community-based behavioral health providers to screen incoming clients for violence risk to inform whether additional assessment for violence potential is warranted.

Managing outbreaks of highly contagious diseases in prisons: a systematic review

BACKGROUND

There are reports of outbreaks of COVID-19 in prisons in many countries. Responses to date have been highly variable and it is not clear whether public health guidance has been informed by the best available evidence. We conducted a systematic review to synthesise the evidence on outbreaks of highly contagious diseases in prison.

METHODS

We searched seven electronic databases for peer-reviewed articles and official reports published between 1 January 2000 and 28 July 2020. We included quantitative primary research that reported an outbreak of a given contagious disease in a correctional facility and examined the effects of interventions. We excluded studies that did not provide detail on interventions. We synthesised common themes using the Synthesis Without Meta-analysis (SWiM) guideline, identified gaps in the literature and critically appraised the effectiveness of various containment approaches.

RESULTS

We identified 28 relevant studies. Investigations were all based in high-income countries and documented outbreaks of tuberculosis, influenza (types A and B), varicella, measles, mumps, adenovirus and COVID-19. Several themes were common to these reports, including the public health implications of infectious disease outbreaks in prison, and the role of interagency collaboration, health communication, screening for contagious diseases, restriction, isolation and quarantine, contact tracing, immunisation programmes, epidemiological surveillance and prison-specific guidelines in addressing any outbreaks.

DISCUSSION

Prisons are high-risk settings for the transmission of contagious diseases and there are considerable challenges in managing outbreaks in them. A public health approach to managing COVID-19 in prisons is required.

PROSPERO REGISTRATION NUMBER

CRD42020178827.

Exploring Similarities and Differences of Non-European Migrants among Forensic Patients with Schizophrenia

Migrants diagnosed with schizophrenia are overrepresented in forensic-psychiatric clinics. A comprehensive characterization of this offender subgroup remains to be conducted. The present exploratory study aims at closing this research gap. In a sample of 370 inpatients with schizophrenia spectrum disorders who were detained in a Swiss forensic-psychiatric clinic, 653 different variables were analyzed to identify possible differences between native Europeans and non-European migrants. The exploratory data analysis was conducted by means of supervised machine learning. In order to minimize the multiple testing problem, the detected group differences were cross-validated by applying six different machine learning algorithms on the data set. Subsequently, the variables identified as most influential were used for machine learning algorithm building and evaluation. The combination of two childhood-related factors and three therapy-related factors allowed to differentiate native Europeans and non-European migrants with an accuracy of 74.5% and a predictive power of AUC = 0.75 (area under the curve). The AUC could not be enhanced by any of the investigated criminal history factors or psychiatric history factors. Overall, it was found that the migrant subgroup was quite similar to the rest of offender patients with schizophrenia, which may help to reduce the stigmatization of migrants in forensic-psychiatric clinics. Some of the predictor variables identified may serve as starting points for studies aimed at developing crime prevention approaches in the community setting and risk management strategies tailored to subgroups of offenders with schizophrenia.

Applicability of two violence risk assessment tools in a psychiatric prison hospital population

The risk of violent behavior is known to be higher for patients who suffer from a severe mental disorder. However, specific prediction tools for clinical work in prison psychiatry are lacking. In this single-center study, two violence risk assessment tools (Forensic Psychiatry and Violence Tool, "FoVOx," and Mental Illness and Violence Tool, "OxMIV") were applied to a prison hospital population with a primary psychotic or bipolar disorder and subsequently compared. The required information on all items of both tools was obtained retrospectively for a total of 339 patients by evaluation of available patient files. We obtained the median and inter-quartile range for both FoVOx and OxMIV, and their rank correlation coefficient along with 95% confidence intervals (CIs)-for the full cohort, as well as for cohort subgroups. The two risk assessment tools were strongly positively correlated (Spearman correlation = 0.83; 95% CI = 0.80-0.86). Such a high correlation was independent of nationality, country of origin, type of detention, schizophrenia-spectrum disorder, previous violent crime and alcohol use disorder, where correlations were above 0.8. A lower correlation was seen with patients who were 30 years old or more, married, with affective disorder and with self-harm behavior, and also in patients without aggressive behavior and without drug use disorder. Both risk assessment tools are applicable as an adjunct to clinical decision making in prison psychiatry.

Neuroprediction and A.I. in Forensic Psychiatry and Criminal Justice: A Neurolaw Perspective

Advances in the use of neuroimaging in combination with A.I., and specifically the use of machine learning techniques, have led to the development of brain-reading technologies which, in the nearby future, could have many applications, such as lie detection, neuromarketing or brain-computer interfaces. Some of these could, in principle, also be used in forensic psychiatry. The application of these methods in forensic psychiatry could, for instance, be helpful to increase the accuracy of risk assessment and to identify possible interventions. This technique could be referred to as 'A.I. neuroprediction,' and involves identifying potential neurocognitive markers for the prediction of recidivism. However, the future implications of this technique and the role of neuroscience and A.I. in violence risk assessment remain to be established. In this paper, we review and analyze the literature concerning the use of brain-reading A.I. for neuroprediction of violence and rearrest to identify possibilities and challenges in the future use of these techniques in the fields of forensic psychiatry and criminal justice, considering legal implications and ethical issues. The analysis suggests that additional research is required on A.I. neuroprediction techniques, and there is still a great need to understand how they can be implemented in risk assessment in the field of forensic psychiatry. Besides the alluring potential of A.I. neuroprediction, we argue that its use in criminal justice and forensic psychiatry should be subjected to thorough harms/benefits analyses not only when these technologies will be fully available, but also while they are being researched and developed.

Assessing Risk Among Correctional Community Probation Populations: Predicting Reoffense With Mobile Neurocognitive Assessment Software

We seek to address current limitations of forensic risk assessments by introducing the first mobile, self-scoring, risk assessment software that relies on neurocognitive testing to predict reoffense. This assessment, run entirely on a tablet, measures decision-making via a suite of neurocognitive tests in less than 30 minutes. The software measures several cognitive and decision-making traits of the user, including impulsivity, empathy, aggression, and several other traits linked to reoffending. Our analysis measured whether this assessment successfully predicted recidivism by testing probationers in a large urban city (Houston, TX, United States) from 2017 to 2019. To determine predictive validity, we used machine learning to yield cross-validated receiver–operator characteristics. Results gave a recidivism prediction value of 0.70, making it comparable to commonly used risk assessments. This novel approach diverges from traditional self-reporting, interview-based, and criminal-records-based approaches, and can also add a protective layer against bias, while strengthening model accuracy in predicting reoffense. In addition, subjectivity is eliminated and time-consuming administrative efforts are reduced. With continued data collection, this approach opens the possibility of identifying different levels of recidivism risk, by crime type, for any age, or gender, and seeks to steer individuals appropriately toward rehabilitative programs. Suggestions for future research directions are provided.

Granting Leave to Patients in Bavarian Forensic-Psychiatric Hospitals: A Survey to Describe the Current Process and Develop Guidelines

Forensic-psychiatric patients reoffending or absconding during the leave granted to them (hereafter referred to as "granted leave") have gained increased attention by researchers and the general public. The patients' right to freedom on the one hand and the need for protection of the general public from serious harm on the other hand represent broadly discussed ethical issues. Thus, demands on quality regarding decisions on patients' granted leaves might be high. Despite such requirements, research on decision-making processes regarding granting leave in forensic psychiatry is very limited and focuses primarily on particular aspects. The present study aims at providing a first overview of the decision-making processes regarding granted leave in forensic psychiatry as a whole. Furthermore, the link between the particular steps of the process and absconding should be explored. In this way, the study results should contribute to provide a theoretical framework for the development of guidelines concerning granted leave in forensic psychiatry. A combination of qualitative and quantitative approaches will be used to collect data: information about risk assessment, decisions on granted leave, and documentation systems in forensic psychiatry will be collected via semi-structured interviews and quantified for further analyses using a checklist developed for this study; data on the implementation of risk assessment tools and documented patient information will be obtained via two self-constructed questionnaires; information about the absolute number of abscondences per hospital will be obtained from the Bavarian Authority for Forensic Commitment. The sample will include staff from all 13 forensic-psychiatric hospitals in Bavaria (Germany) comprising six professional groups: hospital directors, security officers, complementary therapists, psychiatrists, psychologists, social workers, and nursing staff. In each hospital, at least one member of each professional group should participate in the study. In total, 151 interviews will be held. As the study goals are descriptive, there are no pre-formulated hypotheses. Developing guidelines would be the first step towards further standardization of the granted leave decisional process in forensic psychiatry and to make it more transparent for patients, staff members, hospital directors, and the government.

Malpractice Liability Due to Patient Violence

In an era of mass shootings and increasing public fear, psychiatrists are more often being asked to assess whether an individual is safe to return to school or work. In addition, assessment of the individual's risk of violence is required in daily clinical decisions regarding the need for hospital care. Given the inherent difficulty in predicting violence, mental health clinicians worry about potential liability that could result from their patient committing a violent act. This article provides an overview of malpractice liability for patient violence, violence risk factors, and principles of violence risk assessment. The authors also offer some practical risk management strategies to reduce clinicians' risk of liability for violent acts by patients.

Predictive Validity of the YLS/CMI in a Sample of Spanish Young Offenders of Arab Descent

This study was conducted to assess the predictive validity of the Youth Level of Service/Case Management Inventory (YLS/CMI) in young offenders of Arab descent, living in Spain. To address this subject, the Inventory was administered to a sample of Arab minor offenders (N = 116), and results were compared to a sample of non-Arab minor offenders (N = 140), who were all aged between 14 and 17 years. The charges filed after the date of the first assessment carried out by the Youth Offending Team were coded during the follow-up period (2012-2017). The Inventory showed a similar predictive validity for both groups. However, the values were always slightly higher in the non-Arab group than in the Arab group. With subtle cultural differences, the YLS/CMI seems to be a risk instrument capable of predicting recidivism among Arab young offenders.

Identifying the Added Value of Virtual Reality for Treatment in Forensic Mental Health: A Scenario-Based, Qualitative Approach

Background: Although literature and practice underline the potential of virtual reality (VR) for forensic mental healthcare, studies that explore why and in what way VR can be of added value for treatment of forensic psychiatric patients is lacking.

Goals: This study aimed to identify (1) points of improvements in existing forensic mental health treatment of in- and outpatients, (2) possible ways of using VR that can improve current treatment, and (3) positive and negative aspects of the use of VR for the current treatment according to patients and therapists.

Methods: Two scenario-based methods were used. First, semi-structured interviews were conducted with eight therapists and three patients to elicit scenarios from them. Based on these results, six scenarios about possibilities for using VR in treatment were created and presented to 89 therapists and 19 patients in an online questionnaire. The qualitative data from both methods were coded independently by two researchers, using the method of constant comparison.

Results: In the interviews, six main codes with accompanying sub codes emerged. Ideas for improvement of treatment were grouped around the unique characteristics of the forensic setting, characteristics of the complex patient population, and characteristics of the type of treatment. For possibilities of VR, main codes were skills training with interaction, observation of situations or stimuli without interaction, and creating insight for others into the patient. The questionnaire resulted in a broad range of insights into potential positive and negative aspects of VR related to the current treatment, the patient, the content of a VR application, and practical matters.

Conclusion: VR offers a broad range of possibilities for forensic mental health. Examples are offering training of behavioral and cognitive skills in a realistic context to bridge the gap between a therapy room and the real world, increasing treatment motivation, being able to adapt a VR application to individual patients, and providing therapists with new insights into a patient. These findings can be used to ground the development of new VR applications. Nevertheless, we should remain critical of when in the treatment process and for whom VR could be of added value.

Identifying Violent Behavior Using the Oxford Mental Illness and Violence Tool in a Psychiatric Ward of a German Prison Hospital

Background: Although there is evidence that individuals who suffer from severe mental disorders are at higher risk for aggressive behavior, only a minority eventually become violent. In 2017, Fazel et al. developed a risk calculator (Oxford Mental Illness and Violence tool, OxMIV) to identify the risk of violent crime in patients with mental disorders. For the first time, we tested the predictive validity of the OxMIV in the department of psychiatry at the prison hospital in Berlin, Germany, and presented findings from our internal validation. Materials and Methods: We designed a cohort study with 474 patients aged 16-65 years old who met the inclusion criteria of schizophrenia-spectrum or bipolar disorder and classified the patients into two groups: a violent group with 191 patients and a nonviolent group with 283 patients. Violence was defined as the aggressive behavior of a patient with the necessity of special observation. We obtained all the required information retrospectively through patient files, applied the OxMIV tool on each subject, and compared the results of both groups. Sensitivity, specificity, and positive/negative predictive values were determined. We used logistic regression including variable selection and internal validation to identify relevant predictors of aggressive behavior in our cohort. Results: The OxMIV score was significantly higher in the violent group [median 4.21%; Interquartile range (IQR) 8.51%] compared to the nonviolent group (median 1.77%; IQR 2.01%; p < 0.0001). For the risk of violent behavior, using the 5% cutoff for "increased risk," the sensitivity was 44%, and the specificity was 89%, with a positive predictive value of 72% and a negative predictive value of 70%. Applying logistic regression, four items were statistically significant in predicting violent behavior: previous violent crime (adjusted odds ratio 5.29 [95% CI 3.10-9.05], p < 0.0001), previous drug abuse (1.80 [1.08-3.02], p = 0.025), and previous alcohol abuse (1.89 [1.21-2.95], p = 0.005). The item recent antidepressant treatment (0.28 [0.17-0.47]. p < 0.0001) had a statistically significant risk reduction effect. Conclusions: In our opinion, the risk assessment tool OxMIV succeeded in predicting violent behavior in imprisoned psychiatric patients. As a result, it may be applicable for identification of patients with special needs in a prison environment and, thus, improving prison safety.

Violent criminal behavior in the context of bipolar disorder: Systematic review and meta-analysis

BACKGROUND

Despite the potential importance of understanding violent criminal behavior (VCB) in individuals suffering from bipolar disorder (BD), previous findings are conflicting. The aims of the present study are to clarify the association of VCB and BD in comparison to general population and other psychiatric conditions.

METHODS

A systematic review of literature from January 1st, 1980 through January 16th, 2017 from 3 electronic databases (MEDLINE/PubMed, EMBASE and PsycInfo), following the PRISMA and the MOOSE statements. Original peer-reviewed studies reporting data on VCB in BD were included. A random-effects meta-analysis was performed. Potential sources of heterogeneity were examined through subgroup and meta-regression analyses. The protocol was registered in PROSPERO, CRD42017054070.

RESULTS

Twelve studies providing data from 58,475 BD participants. The prevalence of VCB in BD was 7.1% (95%CI = 3.0‒16.5%; k = 4). The association of BD and VCB compared to general population was not significant (OR = 2.784; 95% CI, 0.687‒11.287, P = .152). The association was significant only in cross-sectional studies, in studies in which VCB was assessed through self-reported measures, and in studies conducted in the USA. BD was more likely to be associated with VCB when BD patients were compared to controls with depressive disorders, whilst it was found to be less associated with VCB when BD was compared to psychotic disorders.

LIMITATIONS

1. the methodological heterogeneity across the included studies. 2. causal inferences were precluded by the inclusion of cross-sectional studies.

CONCLUSIONS

These findings might provide a more balance portrait of the association between BD and VCB to clinicians, law enforcement and general public.

Use of risk assessment instruments to predict violence in forensic psychiatric hospitals: a systematic review and meta-analysis

Background and Aims

Violent behaviour by forensic psychiatric inpatients is common. We aimed to systematically review the performance of structured risk assessment tools for violence in these settings.

Methods

The nine most commonly used violence risk assessment instruments used in psychiatric hospitals were examined. A systematic search of five databases (CINAHL, Embase, Global Health, PsycINFO and PubMed) was conducted to identify studies examining the predictive accuracy of these tools in forensic psychiatric inpatient settings. Risk assessment instruments were separated into those designed for imminent (within 24 hours) violence prediction and those designed for longer-term prediction. A range of accuracy measures and descriptive variables were extracted. A quality assessment was performed for each eligible study using the QUADAS-2. Summary performance measures (sensitivity, specificity, positive and negative predictive values, diagnostic odds ratio, and area under the curve value) and HSROC curves were produced. In addition, meta-regression analyses investigated study and sample effects on tool performance.

Results

Fifty-two eligible publications were identified, of which 43 provided information on tool accuracy in the form of AUC statistics. These provided data on 78 individual samples, with information on 6,840 patients. Of these, 35 samples (3,306 patients from 19 publications) provided data on all performance measures. The median AUC value for the wider group of 78 samples was higher for imminent tools (AUC 0.83; IQR: 0.71–0.85) compared with longer-term tools (AUC 0.68; IQR: 0.62-0.75). Other performance measures indicated variable accuracy for imminent and longer-term tools. Meta-regression indicated that no study or sample-related characteristics were associated with between-study differences in AUCs.

Interpretation

The performance of current tools in predicting risk of violence beyond the first few days is variable, and the selection of which tool to use in clinical practice should consider accuracy estimates. For more imminent violence, however, there is evidence in support of brief scalable assessment tools.

Prediction of violent crime on discharge from secure psychiatric hospitals: A clinical prediction rule (FoVOx)

BACKGROUND

Current approaches to assess violence risk in secure hospitals are resource intensive, limited by accuracy and authorship bias and may have reached a performance ceiling. This study seeks to develop scalable predictive models for violent offending following discharge from secure psychiatric hospitals.

METHODS

We identified all patients discharged from secure hospitals in Sweden between January 1, 1992 and December 31, 2013. Using multiple Cox regression, pre-specified criminal, sociodemographic, and clinical risk factors were included in a model that was tested for discrimination and calibration in the prediction of violent crime at 12 and 24 months post-discharge. Risk cut-offs were pre-specified at 5% (low vs. medium) and 20% (medium vs. high).

RESULTS

We identified 2248 patients with 2933 discharges into community settings. We developed a 12-item model with good measures of calibration and discrimination (area under the curve=0.77 at 12 and 24 months). At 24 months post-discharge, using the 5% cut-off, sensitivity was 96% and specificity was 21%. Positive and negative predictive values were 19% and 97%, respectively. Using the 20% cut-off, sensitivity was 55%, specificity 83% and the positive and negative predictive values were 37% and 91%, respectively. The model was used to develop a free online tool (FoVOx).

INTERPRETATION

We have developed a prediction score in a Swedish cohort of patients discharged from secure hospitals that can assist in clinical decision-making. Scalable predictive models for violence risk are possible in specific patient groups and can free up clinical time for treatment and management. Further evaluation in other countries is needed.

FUNDING

Wellcome Trust (202836/Z/16/Z) and the Swedish Research Council. The funding sources had no involvement in writing of the manuscript or decision to submit or in data collection, analysis or interpretation or any aspect pertinent to the study.

Selecting a risk assessment tool to use in practice:a 10-point guide

With the increase in the number of risk assessment tools and clinical algorithms in many areas of science and medicine, this Perspective article provides an overview of research findings that can assist in informing the choice of an instrument for practical use. We take the example of violence risk assessment tools in criminal justice and forensic psychiatry, where there are more than 200 such instruments and their use is typically mandated. We outline 10 key questions that researchers, clinicians and other professionals should ask when deciding what tool to use, which are also relevant for public policy and commissioners of services. These questions are based on two elements: research underpinning the external validation, and derivation or development of a particular instrument. We also recommend some guidelines for reporting drawn from consensus guidelines for research in prognostic models.

Identification of low risk of violent crime in severe mental illness with a clinical prediction tool (Oxford Mental Illness and Violence tool [OxMIV]): a derivation and validation study

BACKGROUND

Current approaches to stratify patients with psychiatric disorders into groups on the basis of violence risk are limited by inconsistency, variable accuracy, and unscalability. To address the need for a scalable and valid tool to assess violence risk in patients with schizophrenia spectrum or bipolar disorder, we describe the derivation of a score based on routinely collected factors and present findings from external validation.

METHODS

On the basis of a national cohort of 75 158 Swedish individuals aged 15-65 years with a diagnosis of severe mental illness (schizophrenia spectrum or bipolar disorder) with 574 018 patient episodes between Jan 1, 2001, and Dec 31, 2008, we developed predictive models for violent offending (primary outcome) within 1 year of hospital discharge for inpatients or clinical contact with psychiatric services for outpatients (patient episode) through linkage of population-based registers. We developed a derivation model to determine the relative influence of prespecified criminal history and sociodemographic and clinical risk factors, which are mostly routinely collected, and then tested it in an external validation. We measured discrimination and calibration for prediction of violent offending at 1 year using specified risk cutoffs.

FINDINGS

Of the cohort of 75 158 patients with schizophrenia spectrum or bipolar disorder, we assigned 58 771 (78%) to the derivation sample and 16 387 (22%) to the validation sample. In the derivation sample, 830 (1%) individuals committed a violent offence within 12 months of their patient episode. We developed a 16-item model. The strongest predictors of violent offending within 12 months were conviction for previous violent crime (adjusted odds ratio 5·03 [95% CI 4·23-5·98]; p<0·0001), male sex (2·32 [1·91-2·81]; p<0·0001), and age (0·63 per 10 years of age [0·58-0·67]; p<0·0001). In external validation, the model showed good measures of discrimination (c-index 0·89 [0·85-0·93]) and calibration. For risk of violent offending at 1 year, with a 5% cutoff, sensitivity was 62% (95% CI 55-68) and specificity was 94% (93-94). The positive predictive value was 11% and the negative predictive value was more than 99%. We used the model to generate a simple web-based risk calculator (Oxford Mental Illness and Violence tool [OxMIV]).

INTERPRETATION

We have developed a prediction score in a national cohort of patients with schizophrenia spectrum or bipolar disorder, which can be used as an adjunct to decision making in clinical practice by identifying those who are at low risk of violent offending. The low positive predictive value suggests that further clinical assessment in individuals at high risk of violent offending is required to establish who might benefit from additional risk management. Further validation in other countries is needed.

FUNDING

Wellcome Trust and Swedish Research Council.