On the Obsoleteness of Violence Risk Assessment Tools: Where are we, evidence-based clinical practice wise?
Are violent risk assessment tools more helpful or harmful in today’s forensic clinical practice?
Violence risk assessment tools are widely used in forensic settings, especially in Western countries such as the US, UK, Sweden, and New Zealand (Schwalbe, 2008; Khiroya et al., 2009; Hanson & Morton-Bourgon, 2009; Vess, 2008). Their usage remains controversial due to the many implications of the results. In the UK, such tools determine how inpatients can access resources, eligibility for detention, parole, and probation, and the length of supervision required as they reintegrate into community settings (Singh et al., 2011). Therefore, it is crucial to scrutinise how such tools predict offending behaviours and whether they are still reliable or obsolete in forensic clinical practice.
To illustrate, vast systematic reviews and meta-analyses have examined these tools’ predictive validity rates (Singh et al., 2011; Monahan & Skeem, 2014; Hurducas et al., 2014). Although it has been acknowledged that there are more than 120 of these risk assessment tools (Hurducas et al., 2014), studies mainly look at the eleven most frequently used tools in forensic settings (Ramesh et al., 2018): the HCR-20 (Webster et al., 1997), the LSI-R (Andrews & Bonta, 1995), the PCL-R (Hare, 2003), the SORAG (Quinsey et al., 2006), the SVR-20 (Boer et al., 1997), the SARA (Kropp et al., 1999), the Static-99 (Harris et al., 2003), the SAVRY (Borum et al., 2003), the VRAG (Quinsey et al., 2006), the BVC (Woods & Almvik, 2002), and the DASA (Simmons et al., 2023). Hence, the present essay shall focus on these tools, which are the most frequently used, accessible, and impactful in UK forensic clinical practice. It will also consider the pros and cons of each tool while discussing the implications for future research and usage.
First, we shall look at the different contexts in which these tools are used and categorise them by their nature of prediction, long-term vs imminent, respectively. For the long-term ones, the HCR-20 (Webster et al., 1997) was designed to assess individual risk in forensic, criminal justice, and civil psychiatric settings and comprises three domains: historical and clinical risk factors and risk management factors. Secondly, the LSI-R (Andrews & Bonta, 1995) was explicitly designed to predict the likelihood of recidivism amongst adult offenders. Hence, it is used to assist professionals in making decisions for supervision and treatment and includes domains such as criminal history, education/employment, finances, and attitudes/orientation — among others. Thirdly, the PCL-R (Hare, 2003) is designed to diagnose psychopathy that was operationally defined by Cleckley (1941) and comprises two domains: (1) selfish, callous, and remorselessness use of others, and (2) unstable and antisocial lifestyle. Fourth, the SORAG (Quinsey et al., 2006) is used to assess the likelihood of violent recidivism (including sexual) among previously convicted sex offenders.
Fifth, the SVR-20 (Boer et al., 1997) is explicitly administered to predict violence risk (including sexual) in sex offenders, and it has four domains: historical, social/contextual, and individual risk factors, as well as protective factors. Sixth, the SARA (Kropp et al., 1999) is utilised to predict violence in men arrested for spousal assault, and it looks at general and spousal violence. Seventh, the Static-99 (Harris et al., 2003) is used to predict the longer-term probability of sexual recidivism in adult male offenders who have previously committed a sexual offence. Eighth, the SAVRY (Borum et al., 2003) is administered to assess the risk of violence perpetration, specifically in adolescents, and it looks at domains such as historical, social/contextual, individual risk factors, and protective factors. Last or ninth, the VRAG (Quinsey et al., 2006) is designed to predict violence risk amongst previously violent, mentally disordered offenders. Whereas for the imminent ones, the BVC (Woods & Almvik, 2002) is designed to look at the short-term risk assessment within 24 hours and considers domains such as confusion, irritability, boisterousness, verbal threats, and attacks on objects, and similarly — the DASA (Simmons et al., 2023) is also used to predict violent behaviours within 24-hour and it comprises of seven items and domains such as irritability, impulsivity, unwillingness to follow instructions, sensitivity to perceived provocation, easily angered when requests are denied, negative attitudes, and verbal threats.
Extensive research has examined the reliability and predictive abilities of the previously discussed risk assessment tools. To summarise, it seems that most of these tools vary widely — indicating that when used alone, they may not be very reliable in predicting violent behaviours (Singh et al., 2011; Monahan et al., 2014; Ramesh et al., 2018). Nevertheless, these results need to be highlighted: (1) in general, tools that are specifically designed for specific populations perform better in terms of their predictive abilities (e.g., the SAVRY), and (2) imminent risk assessment tools (within 24-hour period) (e.g., BVC; DASA) are generally more accurate than the long-term ones; in particular, imminent tools had an AUC value of 0.83, indicating good accuracy, while long-term tools (e.g., HCR-20; PCL-R; VRAG) had a median AUC of 0.68, suggesting moderate accuracy (Ramesh et al., 2018). For the former, Singh et al. (2011) meta-analysis emphasised the need to focus on tailoring risk assessment tools to specific demographics or groups of offenders to acquire more reliable results. This is especially interesting as the research also found that the most frequently used long-term assessment tools, such as the HCR-20, LSI-R, and PCL-R, produced the lowest rates of predictive validity.
Regarding imminent versus long-term risk assessment tools, Ramesh et al. (2018) recent meta-analysis revealed that the most widely used tool — HCR-20 — produced only moderate accuracy for forensic inpatients’ violence prediction. Additionally, the PCL-R and VRAG did not perform exceptionally well for forensic inpatients despite seemingly producing promising results in community settings. In contrast, minimally researched and utilised imminent tools recommended by NICE (2015), such as the BVC and DASA, were found to be more accurate in predicting violence risk. Therefore, it becomes crucial to research these imminent tools in predicting violence, as opposed to merely focusing on the most frequently used long-term risk assessment tools. In line with this, it should be noted that the imminent tools may be more accurate due to the clinical versus actuarial (or statistical) dichotomy within the discourse on forensic risk assessment tools.
Touching on said discourse, clinical approaches to risk assessment tools comprise these themes: they tend to be less structured, more individualised, and flexible. Hence, it allows more room for nuance and contextual interpretation of what “risk” entails (Wertz et al., 2023). Hence, imminent tools may be more in line with this approach. On the other hand, actuarial approaches focus more on highly structured statistical methods and put heavier emphasis on empirically validated risk assessment tools (Ramesh et al., 2018). Therefore, long-term tools are more suited to this category. Nevertheless, seemingly dichotomous categories should not be interpreted at face value. Much research has talked about how these approaches exist within a continuum — where, in practice, clinicians implement both approaches. For instance, the HCR-20, in its practical sense, incorporates both processes, where risk is assessed through structured means, but it also considers Structured Professional Judgment (SJP) (Pedersen et al., 2010). Therefore, it combines the empirical strengths of actuarial or statistical methods but also respects the nuanced understanding of individual cases provided by clinical judgment.
It is evident that the case for risk assessment tools is a complex one. Hence, it is imperative to discuss the previous findings and implications for future usage critically. For one, despite the many controversies surrounding risk assessment tools’ accuracy, it only partially negates their usefulness in forensic settings. For one, they enhance informed decision-making processes in forensic settings, such as providing care, supervision, and intervention strategies (McNiel et al., 2003; Carroll, 2007). Through an effective one, they can provide community-based care plans that provide a safer environment for reintegration (Green et al., 2010). Other than that, they also help with identifying individuals who pose a high risk of violence (Monahan et al., 2014). Granted, as previously discussed, these results should not be interpreted in a tight and literal sense, but they help give direction for clinical interventions (Carroll, 2007). Therefore, it also guides clinicians in creating more effective care plans and risk management strategies. These can potentially address numerous underlying issues, such as comorbidities and other identified risks (Whiting & Fazel, 2020).
Nevertheless, previous findings on the variability of the predictive abilities of these tools raise some fundamental questions regarding their obsoleteness in the context of clinical practice. The most glaring one is the risk of false positives and negatives (Ramesh et al., 2018), which affect the results and interpretation of individual cases. For example, assessments of data from a specific population should not be generalised to other populations, especially when it can lead to labelling and stigmatising already heavily segregated and marginalised populations, such as those coming from distinct cultural or socio-economic backgrounds (Forrester et al., 2018; Ramesh et al., 2018). These “high risk” labels stick with individuals for a lifetime — and can hinder their reintegration into society and access to employment or social support, hence exacerbating said “risks” generated by the previous tools (Forrester et al., 2018).
To synthesise the previous discussion, the limitations to risk assessment tools are extensively researched and debated as (1) there are inherent difficulties in predicting long-term human behaviour, especially in the context of violent acts. Many variables contribute to offending behaviours and cannot be interpreted cross-sectionally in a vacuum. For instance, a person’s likelihood to offend can be influenced by a personal crisis or provocation, and traditional risk assessments might not be able to capture it (Cooke & Michie, 2010); (2) there is a lack of consideration for contextual factors, and most risk assessment tools merely look at the history of individual’s violence and substance misuse, without accounting for the structural issues in which the patients were going through. For example, much research (Monahan & Skeem, 2014; Yang et al., 2010) has established the importance of social support, access to employment, and socio-economic factors that affect offending behaviours; (3) there is an overreliance on static factors and historical data, whereby aspects of an individual’s history or characteristics are immediately associated with the potential risk of offending (Cooke & Michie, 2010). This sort of outlook ignores the potential for rehabilitation and other factors that may skew the results of risk assessment; and (4) there are many challenges in assessing the dynamic nature of risk assessment; variables such as emotional states, stress levels, behavioural and environmental changes are essential in understanding individual’s tendency to offend (Skeem et al., 2011; Monahan & Skeem, 2014). However, these are less likely to be captured via traditional risk assessment tools.
How do we go about addressing the previous issues? The most important thing here is to look at existing tools and look for ways to improve their performance, mainly through these aspects: (1) incorporating technological advances such as artificial intelligence and machine learning to predict risk. There is some evidence that machine learning algorithms and their ability to identify risk factors are usually overlooked by traditional risk assessment tools (Rizer, 2021). However, more research still needs to scrutinise its findings (Werth, 2019), (2) expounding research on neurobiological and genetic processing; this is so clinicians can better understand the gene x environment interplay in the offending pathway, thus generating more reliable results beyond risk assessment tools (Mitjans et al., 2019), (3) focusing more on individualised risk profiles (van Schendel, 2024), as well as, (4) integrating the complex associations between environmental and situational factors to reoffending (Monahan et al., 2014; Werth, 2019), (5) emphasising on collaborative and multidisciplinary approaches that involve psychologists, psychiatrists, social workers, and other professionals involved in the criminal justice system, and incorporating patient-centred perspectives and lived experiences (Gerace et al., 2013; Ramesh et al., 2018). The latter is essential for gaining a holistic assessment process involving various experts from different fields so individualised care plans can not only be implemented but reviewed meaningfully, decreasing the likelihood of stigmatisation, segregation, and other structural barriers to the reintegration process.
Ultimately, violence risk assessments are historically pivotal in determining the reintegration process of institutionalised offenders or inpatients. The history is a complex one, and its use remains something that needs to be continuously scrutinised and improved as it can easily segregate marginalised offenders even further. In particular, there needs to be a more holistic approach to risk assessment tools, and a multidisciplinary understanding of care plans’ efficacy is needed. It might be too soon to establish its obsoleteness for the time being, but remaining critical is given if these tools are still widely used within clinical practice.
Cited works
Andrews, D. A., & Bonta, J. (1995). LSI-R: The Level of Service Inventory-Revised. Toronto, ON: Multi-Health Systems.
Boer, D. P., Hart, S. D., Kropp, P. R., & Webster, C. D. (1997). Manual for the Sexual Violence Risk — Professional guidelines for assessing risk of sexual violence. Burnaby, BC: Simon Fraser University, Mental Health, Law, and Policy Institute.
Carroll, A. (2007). Are violence risk assessment tools clinically useful? Australian & New Zealand Journal of Psychiatry, 41(4), 301–307. https://doi.org/10.1080/00048670701213237
Cleckley, H. (1941). The mask of sanity. St. Louis, MO: C.V. Mosby.
Cooke, D. J., & Michie, C. (2010). Limitations of diagnostic precision and predictive utility in the individual case: A challenge for forensic practice. Law and Human Behavior, 34(4), 259–274. https://doi.org/10.1007/s10979-009-9176-x
Forrester, A., Till, A., Simpson, A., & Shaw, J. (2018). Mental illness and the provision of mental health services in prisons. British Medical Bulletin, 127, 101–109. https://doi.org/10.1093/bmb/ldy027
Gerace, A., Curren, D., & Muir-Cochrane, E. (2013). Multidisciplinary health professionals’ assessments of risk: How are tools used to reach consensus about risk assessment and management? Journal of Psychiatric and Mental Health Nursing, 20(6), 557–563. https://doi.org/10.1111/jpm.12049
Green, B., Carroll, A., & Brett, A. (2010). Structured risk assessment in community forensic mental health practice. Australasian Psychiatry, 18(6), 538–541. https://doi.org/10.3109/10398562.2010.498513
Hanson, R. K., & Morton-Bourgon, K. (2009). The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1–21.
Hare, R. D. (1991). The Hare Psychopathy Checklist-Revised (PCL-R). North Tonawanda, NY: Multi-Health Systems.
Harris, A. J. R., Phenix, A., Hanson, R. K., & Thornton, D. (2003). Static-99 coding rules: Revised 2003. Ottawa, ON: Public Safety Canada.
Hurducas, C. C., Singh, J. P., de Ruiter, C., & Petrila, J. (2014). Violence risk assessment tools: A systematic review of surveys. International Journal of Forensic Mental Health, 13(3), 181–192. https://doi.org/10.1080/14999013.2014.942923
Khiroya, R., Weaver, T., & Maden, T. (2009). Use and perceived utility of structured violence risk assessments in English medium secure forensic units. Psychiatrist, 33, 129–132.
Kropp, P. R., Hart, S. D., Webster, C. D., & Eaves, D. (1999). Spousal Assault Risk Assessment guide (SARA). Toronto, ON: Multi-Health Systems.
McNiel, D. E., Gregory, A. L., Lam, J. N., Binder, R. L., & Sullivan, G. R. (2003). Utility of decision support tools for assessing acute risk of violence. Journal of Consulting and Clinical Psychology, 71, 945–953.
Mills, J. F. (2017). Violence risk assessment: A brief review, current issues, and future directions. Canadian Psychology / Psychologie canadienne, 58(1), 40–49. https://doi.org/10.1037/cap0000100
Mitjans, M., Seidel, J., Begemann, M., et al. (2019). Violent aggression predicted by multiple pre-adult environmental hits. Molecular Psychiatry, 24, 1549–1564. https://doi.org/10.1038/s41380-018-0043-3
Monahan, J., & Skeem, J. L. (2014). The evolution of violence risk assessment. CNS Spectrums, 19(5), 419–424. https://doi.org/10.1017/S1092852914000145
National Institute for Health and Care Excellence. (2015). Violence and aggression: Short-term management in mental health, health and community settings (NG10). https://www.nice.org.uk/guidance/ng10/chapter/1-recommendations
Pedersen, L., Rasmussen, K., & Elsass, P. (2010). Risk assessment: The value of structured professional judgments. International Journal of Forensic Mental Health, 9(2), 74–81. https://doi.org/10.1080/14999013.2010.499556
Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. A. (1998). Violent offenders: Appraising and managing risk. Washington, DC: American Psychological Association.
Ramesh, T., Igoumenou, A., Vazquez Montes, M., & Fazel, S. (2018). Use of risk assessment instruments to predict violence in forensic psychiatric hospitals: A systematic review and meta-analysis. European Psychiatry, 52, 47–53. https://doi.org/10.1016/j.eurpsy.2018.02.007
Rizer, A. (2021). Artificial intelligence and risk assessment tools: Problems and solutions. Washburn Law Journal, 60(3), 495–510.
Rogers, R. (2000). The uncritical acceptance of risk assessment in forensic practice. Law and Human Behavior, 24(5), 595–605. https://doi.org/10.1023/A:1005575113507
Schwalbe, C. S. (2008). A meta-analysis of juvenile justice risk assessment instruments: Predictive validity by gender. Criminal Justice and Behavior, 35, 1367–1381.
Simmons, M. L., Maguire, T., Ogloff, J. R. P., Gabriel, J., & Daffern, M. (2023). Using the Dynamic Appraisal of Situational Aggression (DASA) to assess the impact of unit atmosphere on violence risk assessment. Journal of Psychiatric and Mental Health Nursing, 30(5), 942–951. https://doi.org/10.1111/jpm.12913
Singh, J. P., Grann, M., & Fazel, S. (2011). A comparative study of violence risk assessment tools: A systematic review and meta-regression analysis of 68 studies involving 25,980 participants. Clinical Psychology Review, 31(3), 499–513. https://doi.org/10.1016/j.cpr.2010.11.009
Skeem, J. L., Manchak, S., & Peterson, J. K. (2011). Correctional policy for offenders with mental illness: Creating a new paradigm for recidivism reduction. Law and Human Behavior, 35(2), 110–126. https://doi.org/10.1007/s10979-010-9223-7
Van Schendel, S. (2024). Regulating Risk Profiling by Law Enforcement: A task for data protection law, nondiscrimination law, and criminal procedural law.
Vess, J. (2008). Sex offender risk assessment: Consideration of human rights in community protection legislation. Legal and Criminological Psychology, 13, 245–256.
Werth, R. (2019). Risk and punishment: The recent history and uncertain future of actuarial, algorithmic, and evidence-based penal techniques. Sociology Compass, 13, Article e12659. https://doi.org/10.1111/soc4.12659
Wertz, M., Schobel, S., Schiltz, K., & Rettenberger, M. (2023). A comparison of the predictive accuracy of structured and unstructured risk assessment methods for the prediction of recidivism in individuals convicted of sexual and violent offense. Psychological assessment, 35(2), 152–164. https://doi.org/10.1037/pas0001192
Whiting, D., & Fazel, S. (2020). Epidemiology and Risk Factors for Violence in People with Mental Disorders. In B. Carpiniello, A. Vita, & C. Mencacci (Eds.), Violence and Mental Disorders (Comprehensive Approach to Psychiatry, Vol. 1). Springer, Cham. https://doi.org/10.1007/978-3-030-33188-7_3
Woods, P., & Almvik, R. (2002). The Brøset violence checklist (BVC). Acta Psychiatrica Scandinavica. Supplementum, (412), 103–105. https://doi.org/10.1034/j.1600-0447.106.s412.22.x
Yang, M., Wong, S. C. P., & Coid, J. (2010). The efficacy of violence prediction: A meta-analytic comparison of nine risk assessment tools. Psychological Bulletin, 136(5), 740–767. https://doi.org/10.1037/a0020473
