Machine Learning for prediction of violent behaviors in schizophrenia spectrum disorders: a systematic review

Background

Schizophrenia spectrum disorders (SSD) can be associated with an increased risk of violent behavior (VB), which can harm patients, others, and properties. Prediction of VB could help reduce the SSD burden on patients and healthcare systems. Some recent studies have used machine learning (ML) algorithms to identify SSD patients at risk of VB. In this article, we aimed to review studies that used ML to predict VB in SSD patients and discuss the most successful ML methods and predictors of VB.

Methods

We performed a systematic search in PubMed, Web of Sciences, Embase, and PsycINFO on September 30, 2023, to identify studies on the application of ML in predicting VB in SSD patients.

Results

We included 18 studies with data from 11,733 patients diagnosed with SSD. Different ML models demonstrated mixed performance with an area under the receiver operating characteristic curve of 0.56-0.95 and an accuracy of 50.27-90.67% in predicting violence among SSD patients. Our comparative analysis demonstrated a superior performance for the gradient boosting model, compared to other ML models in predicting VB among SSD patients. Various sociodemographic, clinical, metabolic, and neuroimaging features were associated with VB, with age and olanzapine equivalent dose at the time of discharge being the most frequently identified factors.

Conclusion

ML models demonstrated varied VB prediction performance in SSD patients, with gradient boosting outperforming. Further research is warranted for clinical applications of ML methods in this field.

Enhancing Pre-Treatment Motivation Improves Forensic Mental Health Outcomes

Despite evidence that psychological treatments benefit from pre-treatment intervention, there remains no published research on the value of including a pre-treatment intervention in forensic mental health settings. The present study aimed to address this gap by examining the effects of adding a brief motivational preparatory program (MPP) to standard forensic psychiatric care. The MPP was based on hope theory and motivational interviewing within a cognitive-behavioral therapy approach. MPP participants and a waitlist control group completed a battery of self-report measures of hope and motivation to change, which were compared with respect to risk, demographic, offence history, and outcome variables. There was a significant increase in client motivation for change after completing the MPP. Additionally, those who completed the MPP evidenced modest reductions in aggressive behavior, but significantly increased engagement in subsequent forensic treatment and programming.

The HARM models: Predicting longitudinal physical aggression in patients with schizophrenia at an individual level

The prediction and prevention of aggression in individuals with schizophrenia remains a top priority within forensic psychiatric settings. While risk assessment methods are well rooted in forensic psychiatry, there are no available tools to predict longitudinal physical aggression in patients with schizophrenia within forensic settings at an individual level. In the present study, we used evidence-based risk and protective factors, as well as variables related to course of treatment assessed at baseline, to predict prospective incidents of physical aggression (4-month, 12-month, and 18-month follow-up) among 151 patients with schizophrenia within the forensic mental healthcare system. Across our HARM models, the balanced accuracy (sensitivity + specificity/2) of predicting physical aggressive incidents in patients with schizophrenia ranged from 59.73 to 87.33% at 4-month follow-up, 68.31-80.10% at 12-month follow-up, and 46.22-81.63% at 18-month follow-up, respectively. Additionally, we developed separate models, using clinician rated clinical judgement of short term and immediate violent risk, as a measure of comparison. Several modifiable evidence-based predictors of prospective physical aggression in schizophrenia were identified, including impulse control, substance abuse, impulsivity, treatment non-adherence, mood and psychotic symptoms, substance abuse, and poor family support. To the best of our knowledge, our HARM models are the first to predict longitudinal physical aggression at an individual level in patients with schizophrenia in forensic settings. However, it is important to caution that since these machine learning models were developed in the context of forensic settings, they may not be generalisable to individuals with schizophrenia more broadly. Moreover, a low base rate of physical aggression was observed in the testing set (6.0-11.6% across timepoints). As such, larger cohorts will be required to determine the replicability of these findings.

The Ontario Forensic Mental Health System: A Population-based Review

OBJECTIVE

The purpose of this study was to increase the understanding of the Canadian forensic psychiatry system by providing a population-based overview of the Ontario forensic mental health system.

METHODS

Data were collected on 1,240 accused adults who were subject to the jurisdiction of the Ontario Review Board (ORB) between 2014 and 2015. Archival data were retrieved from annual ORB hearing hospital reports for accused supervised by all nine adult forensic psychiatry facilities across Ontario.

RESULTS

The sample included not criminally responsible (NCR; 91.6%) and unfit to stand trial (UST; 8.4%) accused. The majority of the sample was male (85.7%), single (70.1%), unemployed (63.6%), with a high school education (48.8%). Most were on a detention order (78.5%) and almost half were living in the community at the time of the report (48.8%). The majority had prior contact with psychiatric services (83.1%) and/or the criminal justice system (70.6%) before entering the forensic system. A history of elopement (31.5%) and inpatient aggression was high (60.6%). Most had a psychotic spectrum disorder (81.6%) and over half had a substance use disorder (57.2%) in the reporting year. A range of index offences was observed (69.9% violent, 20.3% general, 9.8% sexual), and the majority of the sample (61.0%) had an index offence that resulted in no injury or a minor injury to the victim.

CONCLUSION

The Canadian forensic psychiatry system is comprised of a unique subset of justice-involved individuals. This study provides a detailed examination of accused who are subject to the jurisdiction of the ORB and provides key insight into risk factors associated with offending behaviour in this population. The results of this study will provide a framework for future studies examining the association between mental disorder and violence and the treatment trajectories for those in the forensic psychiatry system.

Absconsion in forensic psychiatric services: a systematic review of literature

While serious concerns are often raised when patients abscond or leave unauthorized from psychiatric services, there is limited knowledge about absconsion in forensic psychiatric services. Following the preferred reporting items for systematic reviews and meta-analyses guideline, we searched Medline/PubMed, PsycINFO, EMBASE, CINAHL, Scopus, and Web of Science through May 2020 for eligible reports on absconsion in forensic patients with no language limits. The search string combined terms for absconsion, forensic patients, and psychiatry in various permutations. This was supplemented by snowball searching for additional studies. Of the 565 articles screened, 25 eligible studies, including two interventional, seven cross-sectional, and 16 case-controlled studies spanning five decades were included. Absconsion and re-absconsion rates ranged from 0.2% to 54.4% and 15% to 71%, respectively, albeit higher rates trended with less secure psychiatric units. Previous absconsion, aggression, substance use, high Historical Clinical Risk Management-20 score, anti-sociality, psychiatric symptoms, sexual offending, and poor treatment adherence were the factors reported with a degree of predictive value for absconsion. However, the construct of absconsion was heterogeneous in the included studies and the quality of evidence on the predictors of absconsion was limited. Serious risky behaviors including re-offending, violence, self-harm, suicide, rape, and manslaughter were perpetrated by patients during unauthorized leave. Nevertheless, the rates of re-offending were generally low in the included studies (highest recidivism rate = 0.11). There is need for standardized assessment and documentation of absconsion to improve risk analysis and management. Furthermore, it is necessary to develop a structured guideline for defining absconsion, and to create a protocol that operationalizes all absconsion-related behaviors/events to promote reliable assessment and comparative analysis in future studies.

Mistakes in interpersonal perceptions: Social cognition in aggressive forensic psychiatry patients

BACKGROUND

While there is an established link between untreated psychosis and aggression, an enhanced understanding of the role of social cognition is still needed.

AIMS

To examine social cognitive functioning among patients in a specialist forensic mental health service who had been deemed not criminally responsible for acts of violence due to a psychotic disorder. It was hypothesised, first, that such patients would show reduced social cognitive functioning compared with healthy, nonviolent comparison participants and, second, that those who continued to be aggressive while inpatients would demonstrate significant reductions compared to the now nonaggressive group.

METHODS

The study samples were of 10 recently aggressive and 15 not-recently aggressive patients and 20 healthy, nonviolent comparison participants. Each completed the Toronto Empathy Questionnaire (TEQ), the Reading the Mind in the Eyes Test-Revised (RMET) and the Interpersonal Perception Task-15 (IPT-15).

RESULTS

There was no significance between group differences on the RMET and TEQ. The patient group as a whole, however, showed significant interpersonal misperceptions, with specific misperceptions on IPT-15 deception and kinship subscales, while at the same time lacking self-awareness of their errors. Misperceptions on the IPT-15 competition subscale were unique to recently aggressive patients.

CONCLUSIONS

Select aspects of reduced social cognitive functioning were found among not criminally responsible patients with psychosis who had committed violent acts and who continued to act aggressively while forensic inpatients. These findings enhance our understanding of the role of social cognition in predisposing toward violence and the potential importance of incorporating interventions which improve social cognition directly. We suggest also the potential for future research using virtual reality technologies in treatment.

The empirical evidence base for the use of the HCR-20: A narrative review of study designs and transferability of results to clinical practice

The HCR-20, a widely used method of assessing and managing risk, relies on the structured professional judgement approach. This paper reports a narrative literature review of the HCR-20 studies to explore the applicability of the study results to the use of the HCR-20 in clinical practice. From a literature search using terms "HCR-20" and "HCR 20", 206 papers were included. Of studies using the HCR-20 version 2 (n = 191), 92% (n = 176) relied on variables based on scores derived by adding item scores, and 50% (n = 95) tested the HCR-20 using predictive validity methodology. Of the HCR-20 version 3 studies (n = 21), the "presence of risk factors" step was the most commonly examined (n = 18, 86%), but 2 of the 7 steps ("scenario planning" and "management") were not examined at all. Amongst those studies whose primary focus was on the HCR-20, 67% (n = 64/95) did so by assessing the predictive validity of the tool. Only one employed a design to test whether the use of the HCR-20 affected violence rates. The predominant study design provides support for the use of the HCR-20 as an actuarial tool, and there is limited empirical evidence in support of its effectiveness as a structured professional judgement approach to the assessment and management of the risk of violence.

Short-term clinical risk assessment and management: Comparing the Brockville Risk Checklist and Hamilton Anatomy of Risk Management

The current article aims to examine the performance of two brief, dynamic risk measures - the Brockville Risk Checklist (BRC4) and one of two versions of the Hamilton Anatomy of Risk Management [HARM-FV and electronic HARM-FV (eHARM-FV)] - scored at regular clinical case conferences for forensic psychiatric patients in two different settings. The eHARM represents a first-in-class dynamic risk assessment tool using data analytics. Two studies are presented from two forensic psychiatric hospitals in Ontario, Canada. The first study compared the HARM-FV, scored by trained research staff, with the BRC4, scored concurrently by clinical teams, on 36 forensic inpatients. In the second study, trained research staff scored both the BRC4 and the eHARM-FV on 55 forensic inpatients. Both studies demonstrated that the BRC4 and both HARM-FV tools were moderately and positively correlated with each other, with higher agreement for similar domains and items. In both samples, the risk measures performed better at identifying individuals who engaged in repeated or more serious problematic behavior. The HARM-FV and eHARM-FV produced higher area under the curve values for subsequent behavior compared with the BRC4. All three tools were effective at detecting future aggression and adverse incidents. We did not directly compare the HARM-FV and eHARM-FV.

A preliminary evaluation of the effectiveness of dialectical behaviour therapy in a forensic psychiatric setting

Dialectical behaviour therapy (DBT) is a therapy model incorporating elements of Eastern philosophies and cognitive behavioural principles. Originally designed for people struggling with chronic suicidality and borderline personality disorder (BPD), it has been adapted to treat complex, multi-diagnostic presentations, such as those in forensic mental health settings. To date, there has been little evaluation when the primary diagnosis is of psychosis. To explore the effectiveness of DBT, with patients, with multiple comorbidities, including psychosis, in a forensic psychiatric inpatient setting. A descriptive outcome study with a cohort of offender-patients in one specialist forensic mental health unit. Before and after treatment change scores were compared on anger, aggression, hopefulness, coping abilities, emotional intelligence, insight and subjective symptom severity scales, as well as staff-rated risk, and length of stay. Nine men and five women residents in one Canadian secure hospital completed a standard DBT programme, and self-ratings, over about 1 year. Scale scores indicated significantly increased insight and acknowledgment of problems. Apparently increased anger and vengeance scores were clinically associated. Independent staff ratings indicated reductions in risk and most patients achieved early release. This study provides support for extension of the use of DBT to offender-patients with psychosis among the complex mix in their presentation. It suggests that a randomised controlled trial with cost-benefit analysis is warranted, as well as further work, to promote understanding of mechanisms of effectiveness.

Implementation of risk assessment tools in psychiatric services

Violence remains a major risk management concern in psychiatric services with implications on the safety and well-being of patients, staff, and the public. Serious physical and psychological consequences of violence involving property damage, bodily injuries, and threat to life have been reported in mental health services. Risk assessment tools are important safeguard measures; however, research on clinical implementation is presently limited. Structured professional judgment (SPJ) risk management tools that incorporate professional discretion with analytical understanding of evidence-based risk factors are widely accepted for risk assessment. However, clinical utility is suboptimal due to several barriers, including those related to the tool, the clinical setting, and resistance from health professionals. To better understand the challenges militating against optimal implementation of risk assessment tools, we reviewed and presented some lessons from the implementation of clinical practice guidelines on a general scale and our experience implementing an SPJ tool called Hamilton Anatomy of Risk Management across a variety of psychiatric services. In summary, the clinical utility of risk assessment tools improves if the tool is psychometrically sound, concise, consensus rated, time efficient, and practical for planning risk management. User feedbacks on the tool utility are also important to sustain implementation.