The predictive validity of youth violence risk assessment tools: a systematic review

ABSTRACT

This paper reports the findings of a systematic search of published literature which reports the predictive validity of violence risk assessment tools specifically designed for use with youth. A total of 38 studies, involving 9,307 participants, reported data for six different tools; the most common of which were the SAVRY and the YLS/CMI. Each of the tools demonstrated at least moderate levels of predictive validity, with the predictive validity of several newer assessment tools yet to be established. The results provide an up-to-date overview of the state of knowledge in an area in which practitioners make choices about which tools to use on an almost daily basis. It is important that practitioners are aware of the strength of evidence that is available to support the choice of violence risk assessment tools and the interpretation of results.     

Examining the Science and Practice of Violence Risk Assessment with Female Adolescents

While the field of violence risk assessment among adult males has progressed rapidly, several questions remain with respect to the application of forensic risk assessment tools within other populations. In this article, we consider the empirical evidence for the assessment, prediction, and management of violence in adolescent girls. We discuss limitations of generalizing violence risk assessment findings from other populations to adolescent girls and point out areas where there is little or no empirical foundation. Critical issues that must be addressed in research prior to the adoption or rejection of such instruments are delineated. Finally, we provide practice guidelines for clinicians currently involved with adolescent females within risk assessment contexts.     

Predictive Validity of the SVR-20 and Static-99 in a Dutch Sample of Treated Sex Offenders

In this retrospective study, the interrater reliability and predictive validity of 2 risk assessment instruments for sexual violence are presented. The SVR-20, an instrument for structured professional judgment, and the Static-99, an actuarial risk assessment instrument, were coded from file information of 122 sex offenders who were admitted to a Dutch forensic psychiatric hospital between 1974 and 1996 (average follow-up period 140 months). Recidivism data (reconvictions) from the Ministry of Justice were related to the risk assessments. The base rate for sexual recidivism was 39%, for nonsexual violent offenses 46%, and for general offenses 74%. Predictive validity of the SVR-20 was good (total score: r = .50, AUC = .80; final risk judgment: r = .60, AUC = .83), of the Static-99 moderate (total score: r = .38, AUC =.71; risk category: r = .30, AUC = .66). The SVR-20 final risk judgment was a significantly better predictor of sexual recidivism than the Static-99 risk category.     

Risk assessment for intimate partner violence: how can the police assess risk?

ABSTRACT

The use of risk assessment tools by frontline police for intimate partner violence has the potential to make a difference to policing. In this paper, the key aspects of intimate partner violence risk assessment are outlined critically with a particular emphasis on how they can be used in practice. Two, evidence-based, exemplars are reviewed. These are the Ontario Domestic Abuse Risk Assessment (ODARA), an example of the actuarial approach, and the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER), an example of the structured professional judgement approach. In addition, the victim giving his or her own appraisal of risk is discussed. All three approaches have some validity when administered properly but practical factors reduce this validity. The content of the risk assessment tools are outlined and practical concerns such as training, time to administer, reliability, validity, and the overlap of intimate partner violence with other forms of offending are discussed. A balanced overview of the strengths, weaknesses and future potential of intimate partner violence risk assessment is provided.     

Static screening instruments for risk of minor and major violence

Background: Screens for violent convictions that are simple, accessible and parsimonious are needed, as a first stage in identifying those at high risk for further assessment. Aims: To construct and validate screening tools for minor and major violence convictions for released prisoners. Methods: Internal validation sample of 1647 serious offenders and an external validation of 46,704 general prisoners. The outcomes were binary indicators for having at least one conviction for minor and major violence. Risk factors were convictions for violence and age. Results: In the external validation sample, the instrument for risk of minor violence (PMIV) identified correctly 60.8% of male and 66.2% of female general prisoners. For risk of major violence, the instrument (PMAV) identified correctly 68.0% of male and 79.3% of female prisoners. Conclusions: The PMIV and PMAV will efficiently assist practitioners in a first stage of screening before in-depth clinical assessment of risk for future violent convictions.     

Consulting to residential programmes for sexually aggressive delinquents

Abstract

Delinquent adolescent boys, especially those looked after in residential settings, often have a long history of coercive antisocial behaviour. Many have a history of aggression and violence, including serious sexual offences such as rape and indecent assault. In light of the Criminal Justice Act 1991 the residential care sector will come under more pressure to accommodate sexually aggressive recidivistic delinquents. However, few staff who work in residential settings with delinquents have received training in the management and treatment of sexually aggressive behaviour. Time spent in residential care settings may therefore do little to reduce the risk of sexual aggression. This paper proposes that the assessment and management of risk is central to the effective rehabilitation of this group of young people. Based on work carried out at Glenthorne Centre a risk assessment strategy is presented consisting of several different strands: an analysis of organisational decision-making; identification of sources and types of information to inform decisions; the development and implementation of a risk assessment protocol; and staff training. A case example is used to illustrate the use of the protocol at Glenthorne.     

Risk Technology in Sentencing: Testing the Promises and Perils (Commentary on Hannah-Moffat, 2011)

The target article is a critique of the movement toward using structured risk assessment tools to inform decisions about sentencing. In this commentary, I analyze (a) the conditions under which it may be more or less fair to use well-validated risk assessment tools in this manner and (b) the extent to which doing so is likely to exacerbate, ameliorate, or have no effect on existing racial and other biases in sentencing. I recommend a policy-relevant research agenda that would specifically test whether and how adding well-validated risk assessment tools to the routine sentencing process alter the severity or nature of sentences. This agenda would also evaluate the extent to which these tools are implemented in “real world” settings faithfully enough to bridge the usual divide between science and practice.     

Criminal Justice Contact,Stressors, and Depressive Symptoms Among Black Adults in the United States

Criminal justice contact—defined as lifetime arrest, parole, or incarceration is associated with higher odds of experiencing depressive symptoms. Findings from this study suggest that there is a mechanism that links family, neighborhood, and financial stressors among Black adults with criminal justice contact to depressive symptoms. Using the National Survey of American Life (NSAL), modified poisson regression analyses were used to determine the association between criminal justice contact, stressors, and depressive symptoms among a national sample of Black adults (n = 3570). In the full model, the odds of experiencing depressive symptoms for Black adults who had criminal justice contact was reduced (PR: 1.14 to PR: 1.07). Black adults who reported experiencing family stressors (PR: 1.10, 95% CI: 1.06, 1.16), neighborhood stressors (PR: 1.09, 95% CI: 1.04, 1.13), and financial stressors (PR: 1.13, 95% CI: 1.06, 1.19) were statistically significant and higher than those who reported not experiencing any of these stressors. Stressors partially mediates the relationship between criminal justice contact and depressive symptoms (OR: 1.05, Bias-corrected 95% CI: 1.04, 1.07); causing the effect of criminal justice contact to reduce to 1.06, leaving a significant indirect effect of 1.05. The total effect is 1.93 times larger than the direct effect, and 48.3% of the total effect is due to stressors. These findings emphasize the need to further explore the family, neighborhood, and financial stressors for Black adults with criminal justice contact in order to further our understanding of their depressive symptoms.     

Static and Dynamic Patient Characteristics as Predictors of Criminal Recidivism: A Prospective Study in a Dutch Forensic Psychiatric Sample

If clinicians in forensic psychiatry want to reduce risk of reoffending in their patients, they require insight into dynamic risk factors, and evidence that these add predictive power to static risk indicators. Predictors need to be evaluated under clinically realistic circumstances. This study aimed to validate dynamic and static variables as predictors of reconviction in a naturalistic outcome study. Data on static and dynamic risk factors were collected for 151 patients discharged from Dutch forensic psychiatric hospitals. Community follow-up was prospective, with a 5.5 year minimum. A prediction model was developed using Cox regression analysis. The magnitude of the predictive power of this model was estimated using receiver operating characteristic (ROC) analysis. The final prediction model contained four static and no dynamic predictors. The model's ROC area under the curve was .79 (95% CI .69–.89). Clinical risk ratings were non-predictive. Post hoc analyses exploring the influence of subgroups of patients did not yield better models. It is concluded that a small set of static predictors yielded a good estimate of future reconvictions; inclusion of dynamic predictors did not add predictive power.     

Implementing violence and incident reporting measures on a forensic mental health unit

The assessment of risk and prediction of violence in mental health units can play a large role in creating a safer environment for both the staff and the patients. Nurses in forensic units are in a unique position in regards to assessment of violence as they spend a great deal of time with the patients. Nurses on a forensic mental health unit scored the Brøset Violence Checklist (BVC) twice daily for 12 weeks for all patients either resident on or admitted to the unit (N?=?46). The Staff Observation Aggression Scale-Revised (SOAS-R) was used to report any adverse incidents (N?=?51). Data were examined at the both the item and scale level. Main results showed the area under the curve values of the BVC score, slide rule, and the sum of BVC and slide rule score in turn demonstrated strong predictive accuracy for inpatient aggression (0.68–0.73). Through logistic regression analyses the BVC uniquely predicted inpatient aggression but adding the slide rule did not improve prediction. Predictive accuracy was found across three diagnostic groups – dementia, psychosis and substance use disorders. These results provide further support on the predictive accuracy of the BVC for short-term violence in forensic mental health settings.     

Long-term follow-up of criminal recidivism in young sex offenders: Temporal patterns and risk factors

Abstract

Recent studies indicate that risk for criminal re-offending in identified adult sex offenders may persist decades after the index offence, calling for inquiry into these issues also for younger sex offenders. The author studied temporal patterns for criminal reconvictions and the validity of 16 empirically motivated risk factors for recidivism among all young sex offenders (15–20 years of age) subjected to pre-sentence forensic psychiatric evaluations in Sweden during 1980–1995 (N = 126). The Ss were followed from release and for an average of 115 months. One-hundred-and-seventeen (115 male, 2 female) Ss were available for follow-up with a mean time-at-risk of 80 months. Base rates for sexual and violent non-sexual reconvictions were 30% and 42%, respectively. Using time-at-risk-adjusted hazard ratios derived from Cox regression, characteristics indicative of deviant sexual interest (any previous sex offending behaviour, an index sex offence in a public area, involving a stranger victim, offending on two or more offence occasions, and two or more victims) were found to increase the risk for sexual recidivism. Markers of antisocial lifestyle (early signs of DSM-IV Conduct Disorder and any prior violent conviction), sex offence-related characteristics indicating aggression proneness (use of threats or force and physical victim injury), and victim penetration, increased the risk for violent non-sexual reconvictions. Survival analyses indicated that the rate of sexual reconvictions as a function of time at risk decreased distinctly after 6–7 years whereas the rate for violent non-sexual reconvictions remained largely the same.     

Risk and the Female Offender: An Analysis of Classification and Supervision Issues

Risk is a defining feature of punishment in the United States. This is evidenced by the profusion of actuarial risk assessment instruments, now entering their 4th generation. Although the literature on risk assessment specific to female offenders is growing, certain questions remain unresolved. In particular, it is still largely unknown how gender might influence classification and supervision for offenders under community supervision. This study reports findings from a survey of community corrections officers. The perceived level of risk, the decision to override a classification score, and the perceived importance of risk and need factors are discussed.     

Dynamic risk and sexual offending: the conundrum of assessment

ABSTRACT

The potential to reduce sexual victimisation, promote community safety, and decrease incarceration costs has resulted in considerable progress in terms of how we understand and predict sexual recidivism. And yet, the past decade has seen a degree of fragmentation emerge as research attention has shifted away from relative risk prediction (with its focus on static risk factors) to the identification of factors capable of reducing risk through intervention (i.e. dynamic risk). Although static and dynamic risk are often treated as orthogonal constructs [Beech, A. R., & Craig, L. A. (2012). The current status of static and dynamic factors in sexual offender risk assessment. Journal of Aggression, Conflict and Peace Research, 4(4), 169–185. doi:10.1108/17596591211270671], there are arguments to support a claim that the two are in fact functionally related [see Ward, T. (2015). Dynamic risk factors: Scientific kinds or predictive constructs. Psychology, Crime & Law (in 22(01–02), 2–16); Ward, T., & Beech, A. R. (2015). Dynamic risk factors: A theoretical dead-end? Psychology, Crime & Law, 21(2), 100–113. This discussion clearly affects how we assess dynamic risk. This review considered several commonly used methods of assessment and the evidence offered for their predictive accuracy. Of note were differences in the predictive accuracy of single psychometric measures versus composite scores of dynamic risk domains and the conventions used for establishing effect sizes for risk assessment tools.     

Application of operational psychology to assessment of inmates

Abstract

The newly developed actuarial risk instruments, Static-2002 (Hanson & Thornton, 2003a) and Risk Matrix 2000 (RM2000) (Thornton et al., 2003), were cross-validated and compared with Static-99 (Hanson & Thornton, 2000) in a retrospective follow-up study of 304 forensic psychiatrically evaluated sexual offenders (mean time-at-risk=14.8 years). All instruments predicted any sexual, non-sexual violent and any violent recidivism moderately for child molesters [area under the curve (AUC) of the receiver operating characteristic (ROC)=0.67 (95% CI=0.57–0.77) to 0.76 (95% CI=0.68–0.85)] but poorly for rapists [AUCs=0.64 (95% CI=0.51–0.70) to 0.68 (95% CI=0.59–0.77)]. The instruments also predicted severe sexual recidivism moderately for child molesters [AUCs=0.74 (95% CI=0.63–0.86) to 0.79 (95% CI=0.70–0.89)] but were of no value in informing on this outcome for rapists [AUCs=0.58 (95% CI=0.47–0.68) to 0.63 (95% CI=0.53–0.73)]. AUC estimates were generally higher for Static-2002 and RM2000 than for Static-99, but no single instrument demonstrated statistical superiority. The study provides support for implementing actuarial instruments in applied risk assessments of child molesters.     

Socially desirable responding and psychometric assessment of dynamic risk in sexual offenders against children

Does socially desirable responding (SDR) represent a threat to the validity of dynamic risk assessment in sex offenders’ self-reports? We studied a sample of men (N?=?218) who completed a psychometric assessment battery while enrolled at Kia Marama, a prison-based treatment program for sexual offenders against children. SDR, as measured by the Marlowe-Crowne scale, was elevated compared to non-offender samples, increased from pre- to post-treatment, and was negatively correlated with dynamic risk (rs?=??.30 to ?.40), consistent with previous research. Measures of dynamic risk factors derived from the psychometric battery were correlated with sexual recidivism, but correlations were little changed when variance associated with SDR was removed using a regression procedure [Mills, J. F., &; Kroner, D. G. (2006). Impression management and self-report among violent offenders. Journal of Interpersonal Violence, 21(2), 178–192.]. These results suggest that there is a substantial component of the variance in psychometric self-reports of sexual offenders that is associated with SDR (approximately 10%), but that it does not compromise the predictive or construct validity of dynamic risk measures derived from these reports.     

Risk,reconviction and “sexually risky behaviour” in sex offenders

Abstract

The Challenge Project is a community assessment and treatment programme in southeast London. The Project has been evaluated extensively in relation to community failure, with consideration of static risk prediction tools and the role of key developmental variables. This current paper is an extension of earlier evaluations, considering an extended range of outcomes with a longer-term follow-up period of more than nine years at risk in the community. The official sexual reconviction rate for this sample of sex offenders (of whom 275 were at risk in the community) was 12%. A wider definition of “sexually risky behaviours” as an outcome variable, drawn from a range of police intelligence, suggested a “truer” sexual re-offending rate of 20%. Although static tools were moderately predictive of community failure, the accuracy of risk prediction was enhanced if static tools were considered in conjunction with developmental variables.     

Understanding motives for aggression in forensic psychiatric patients: a preliminary study

The primary aim of this study was to explore motivations underpinning aggression among men detained within conditions of high security. Thirty men residing at a high secure psychiatric hospital completed self-report measures, including the Aggression Motivation Questionnaire, Revised EXPAGG and Barratt Impulsiveness Scale-IIr. The Historical items of the Historical, Clinical and Risk-Management (HCR-20) and the Psychopathy Checklist-Screening Version were rated. A subsample of participants agreed to complete a functional assessment on an aggressive incident that had occurred during their placement (n = 9). Increased psychopathy and impulsivity, and the presence of historical risk items were predicted to associate with higher levels of both aggression motivation and beliefs supportive of aggression. Young age at first violent incident and personality disorder related positively to aggression motivation. Thematic analysis conducted on the functional assessments identified social recognition, emotion regulation, communication and protection as functions underpinning aggression. Results are discussed with regards to their implication for violence treatment and assessment, with a focus on motivation recommended.     

Prediction of adolescent sexual reoffending: A meta-analysis of the J-SOAP-II, ERASOR, J-SORRAT-II, and Static-99

Several risk assessment tools, including the Juvenile Sex Offender Assessment Protocol-II (Prentky & Righthand, 2003), the Estimate of Risk of Adolescent Sexual Offense Recidivism (Worling & Curwen, 2001), the Juvenile Sexual Offense Recidivism Risk Assessment Tool-II (Epperson, Ralston, Fowers, DeWitt, & Gore, 2006), and the Static-99 (Hanson & Thornton, 1999), have been used to assess reoffense risk among adolescents who have committed sexual offenses. Given that research on these tools has yielded somewhat mixed results, we empirically synthesized 33 published and unpublished studies involving 6,196 male adolescents who had committed a sexual offense. We conducted two separate meta-analyses, first with correlations and then with areas under the receiver operating characteristic curve (AUCs). Total scores on each of the tools significantly predicted sexual reoffending, with aggregated correlations ranging from .12 to .20 and aggregated AUC scores ranging from .64 to .67. However, in many cases heterogeneity across studies was moderate to high. There were no significant differences between tools, and although the Static-99 was developed for adults, it achieved similar results as the adolescent tools. Results are compared to other meta-analyses of risk tools used in the area of violence risk assessment and in other fields. (PsycINFO Database Record (c) 2012 APA, all rights reserved).     

Red-teaming the panopticon* (mobilising adaptive change in secure and forensic settings)

In England and Wales, secure and forensic psychiatric institutions provide a high-cost, low-volume service that imposes significant restrictions upon detainees. Patients may be detained under the Mental Health Act in such settings for several years or even life, as they are deemed to present a significant risk to themselves or the public. Patients under s37/41 require the Home Office to approve any increase in their freedom. Best practice requires reoffending risk to be assessed before a patient is discharged. Evaluation of risk is an inexact actuarial science operating in a political arena, and research has indicated risk assessment tools have little positive predictive validity. There is concern amongst the wider psychiatric and judicial communities about the ethics of current practice. We examine these issues and consider means of improving risk assessment through red-teaming, increased collaboration between clinician and patient and a paradigm shift towards greater emphasis on patient self-agency.     

Actuarial risk assessment and recidivism in a sample of UK intellectually disabled sexual offenders

Abstract

This study examines the effectiveness of three risk assessment instruments: Static-99, Risk Matrix 2000 (RM2000) and the Rapid Risk of Sex Offender Recidivism (RRASOR), in predicting sexual recidivism among 27 intellectually disabled sex offenders. The overall sexual offence reconviction rate was 30%, while non-recidivists remained offence-free over 76 months of follow-up. Static-99 presented as performing as well as guided clinical judgements in mainstream population studies [area under the curve (AUC)=0.64] exceeding the performance of RM2000 (AUC=0.58) in predicting sexual recidivism. However, the results were not statistically significant. In contrast to previous findings, the RRASOR presented the worst level of prediction (AUC=0.42). These results highlight the need to investigate further with larger sample sizes and in conjunction with more dynamic measures of risk. Proposed relevant factors are discussed in detail.