Sexual recidivism: A review of static,dynamic and actuarial predictors

The need for accurate risk assessment of sexual offence recidivism has never been greater. It is widely accepted that actuarial risk instruments outperform clinical judgement and the literature has recently witnessed a surge of empirically derived actuarial measures. However, in spite of the increased levels of predictive accuracy, actuarial measures have been criticized as being unrepresentative, lacking specificity, and being heavily reliant on static risk factors without taking into account dynamic risk, psychological emotional states and treatment effects. Rather than offering a critique of the actuarial movement, this paper offers a summary of static and dynamic risk factors associated with sexual offence recidivism as identified from the literature. Implications of incorporating dynamic factors into risk assessments and actuarial measures are discussed.     

Methodological development: structured outcome assessment and community risk monitoring (SORM)

This paper describes an effort to develop a clinical tool for the continuous monitoring of risk for violence in forensic mental health clients who have left their institutions and who are dwelling in the community on a conditional release basis. The model is called Structured Outcome Assessment and Community Risk Monitoring (SORM). The SORM consists of 30 dynamic factors and each factor in SORM is assessed in two ways: The current absence, presence or partial och intermittent presence of the factors, which is an actuarial (systematized and 'objective') assessment. Secondly, the risk effect, i.e. whether the presence/absence of factors currently increases, decreases or is perceived as unrelated to violence risk, is a clinical (or impressionistic) assessment. Thus, the factors considered via the SORM can be coded as risk factors or protective factors (or as factors unimportant to risk of violence) depending on circumstances that apply in the individual case. Further, the SORM has a built-in module for gathering idiographical information about risk-affecting contextual factors. The use of the SORM and its potential as a risk monitoring instrument is illustrated via preliminary data and case vignettes from an ongoing multicenter project. In this research project, patients leaving any of the 9 participating forensic hospitals in Sweden is assessed at release on a variety of static background factors, and the SORM is then administered every 30 days for 2 years.     

A review of dynamic variables and their relationship to risk prediction in sex offenders

In recent years, a good deal of attention has been paid to the question of recidivism in sex offenders, with an emphasis on the pool of identified static variables associated with risk. The paper aims to review the developing evidence-base for dynamic (changeable) variables, describing their clinical and research characteristics. It also considers the relationship between dynamic domains and apparently fixed developmental variables, such as attachment and victimization experiences. Three main models are presented, and examined in terms of their ability to enhance static prediction models. Results would suggest that there are four dynamic domains, which can predict sexual recidivism independently of static risk prediction, and that these domains should be of central consideration in the continuing development of treatment programmes.     

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.     

A comparison of actuarial methods for identifying repetitively violent patients with mental Illnesses

This is a progress report on the development of practical methods for the actuarial prediction of violence. The literature indicates that actuarial prediction is more accurate than clinical prediction, but in practice actuarial methods seem to be used rarely. Here we address two obstacles to the clinical use of actuarial prediction methods. First, clinicians may be averse to actuarial methods that require calculations. To remedy this, we developed a regression tree screen that presents actuarial information about violence in a series of yes/no questions. Second, using actuarial methods to identify the small minority of violent patients in a general psychiatric population may be too costly. To remedy this, we developed a method to prescreen patients for intensive evaluation using an inexpensive assessment. We evaluated regression trees and two-stage screening by comparing their accuracies against conventional actuarial methods. The results showed that actuarial predictions based on regression trees and two-stage screens were as accurate as regression-based methods in identifying repetitively violent patients. These easier-to-use methods may therefore be useful techniques for actuarial predictions.     

Workplace Assessment of Targeted Violence Risk: The Development and Reliability of the WAVR‐21

This study describes the development of the WAVR‐21, a structured professional judgment guide for the assessment of workplace targeted violence, and presents initial interrater reliability results. The 21‐item instrument codes both static and dynamic risk factors and change, if any, over time. Five critical items or red flag indicators assess violent motives, ideation, intent, weapons skill, and pre‐attack planning. Additional items assess the contribution of mental disorder, negative personality factors, situational factors, and a protective factor. Eleven raters each rated 12 randomly assigned cases from actual files of workplace threat scenarios. Summary interrater reliability correlation coefficients (ICCs) for overall presence of risk factors, risk of violence, and seriousness of the violent act were in the fair to good range, similar to other structured professional judgment instruments. A subgroup of psychologists who were coders produced an ICC of 0.76 for overall presence of risk factors. Some of the individual items had poor reliability for both clinical and statistical reasons. The WAVR‐21 appears to improve the structuring and organizing of empirically based risk‐relevant data and may enhance communication and decision making.     

Exploring the validity of the Level of Service Inventory-Revised with Native American offenders

The use of actuarial risk/need assessment tools is an increasingly important part of the correctional landscape. Actuarial tools ideally will provide a valid, dynamic assessment of an offender's overall risk/need level, and will identify their most prevalent criminogenic needs. What results is typically a number or score that can be used to assign an offender to a risk level that is associated with an assumed likelihood of recidivism. Testing the predictive validity of actuarial risk/need assessment tools is of paramount concern, particularly when they are utilized with new (and under-researched) populations. The current study assessed the predictive validity of the Level of Service Inventory-Revised using a sample of Native American and White offenders in a northern midwestern state. Results showed the instrument to have modest predictive validity utilizing the entire sample of offenders, with varying results for subsequent subgroups.     

What is grievance thinking and how can we measure this in sexual offenders?

Purpose. This study proposes and examines a three‐stage model of the grievance process, one of the dynamic risk factors related to risk of sexual recidivism, and aims to evaluate some of the existing measures of this construct. Methods. The research used a sample of 322 male sexual offenders who had completed a cognitive–behavioural programme that aims to reduce recidivism in higher‐risk sexual offenders (as measured using a static actuarial tool). Participants completed two questionnaires measuring aspects of grievance thinking pre‐ and post‐treatment. Results. The results indicated some support for stages 1 and 2 of the proposed model. Contrary to the study hypotheses, results indicated that both measures used have similar psychometric properties. Pre‐ to post‐change analyses suggest that the custodial treatment programme may be having some effect on grievance thinking. However, generally offenders' scores on both measures were low pre‐ and post‐treatment and as a result, according to individual change analyses, the majority did not demonstrate reliable or clinically significant change. Those who were classed as high scorers on either measure did, however, demonstrate such change. Conclusions. Further exploration of a three‐stage model of grievance is warranted. It appears that current measures of grievance in sexual offenders are not adequate to capture this concept fully.     

Forecasting Recidivism in Mentally Ill Offenders Released from Prison

Little research has focused on assessing the risk of mentally ill offenders (MIOs) released from state prisons. Here we report findings for 333 mentally ill offenders released from Washington State prisons. Logistic regression identified sets of variables that forecasted felony and violent reconviction as accurately as state-of-the-art risk assessment instruments. Sums of simple recoded versions of these variables predicted reoffense as well as complex logistic regression equations. Five of these 9 variables were found to be relative protective factors. Findings are discussed in terms of the value of stock correctional variables in forecasting risk, the need to base actuarial risk assessments on local data, the importance of protective factors in assessing MIO risk, and the need for dynamic, situational, and clinical variables that can further sharpen predictive accuracy of emergent risk in the community.     

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.     

Male victims and post treatment risk assessment among adult male sex offenders

The use of actuarial instruments to predict sex offender recidivism has gained increasing credibility in recent years. This paper is one in a series examining the impact of dynamic inpatient group therapy upon the predictive influence of static risk factors on recidivism among adult sex offenders. Successful completion of the Phoenix Program (Alberta Hospital Edmonton) has been shown to ameliorate the influence of static risk factors on sexual offense recidivism. Many studies have reported that sex offenders who have male victims are more likely to re-offend than those who do not have male victims. A sample of N=513 convicted adult male sex offenders was examined regarding the relationship between the static risk factor of having male victims, subsequent re-offense, and treatment impact. Interestingly, ever having had a male victim did not significantly correlate with sex offense recidivism, for either treatment completers, non-completers, or the combined group. However, having exclusively male victims was correlated with sex offense recidivism, but only among non-completers of the program (r=.155; p=.017). Analysis of a subset of 422 child molesters yielded a similar result, in that having male victims exclusively was only associated with sex offense recidivism among treatment non-completers (r=.189, p=.009).     

Applying a forensic actuarial assessment (the Violence Risk Appraisal Guide) to nonforensic patients

The actuarial Violence Risk Appraisal Guide (VRAG) was developed for male offenders where it has shown excellent replicability in many new forensic samples using officially recorded outcomes. Clinicians also make decisions, however, about the risk of interpersonal violence posed by nonforensic psychiatric patients of both sexes. Could an actuarial risk assessment developed for male forensic populations be used for a broader clientele? We modified the VRAG to permit evaluation using data from the MacArthur Violence Risk Assessment Study that included nonforensic male and female patients and primarily self-reported violence. The modified VRAG yielded a large effect size in the prediction of dichotomous postdischarge severe violence over 20 and 50 weeks. Accuracy of VRAG predictions was unrelated to sex. The results provide evidence about the robustness of comprehensive actuarial risk assessments and the generality of the personal factors that underlie violent behavior.     

The Influence of Actuarial Risk Assessment in Clinical Judgments and Tribunal Decisions about Mentally Disordered Offenders in Maximum Security

Research has shown that actuarial assessments of violence risk are consistently more accurate than unaided judgments by clinicians, and it has been suggested that the availability of actuarial instruments will improve forensic decision making. This study examined clinical judgments and autonomous review tribunal decisions to detain forensic patients in maximum security. Variables included the availability of an actuarial risk report at the time of decision making, patient characteristics and history, and clinical presentation over the previous year. Detained and transferred patients did not differ in their actuarial risk of violent recidivism. The best predictor of tribunal decision was the senior clinician's testimony. There was also no significant association between the actuarial risk score and clinicians' opinions. Whether the actuarial report was available at the time of decision making did not alter the statistical model of either clinical judgments or tribunal decisions. Implications for the use of actuarial risk assessment in forensic decision making are discussed.     

Beyond static and dynamic risk factors: the incremental validity of release planning for predicting sex offender recidivism

Both desistance research and strengths-based approaches to offender rehabilitation suggest that attempts to reduce sex offender recidivism should attend to an offender's release environment. Recent research has demonstrated that better quality release planning is associated with reduced recidivism; however, whether release planning contributes significant incremental validity in predicting recidivism over and above static and dynamic risk factors is unclear. In the present study, release planning was retrospectively assessed for a sample of child molesters (n = 196) who had been released into the community following completion of a prison-based treatment program and its relative contribution to recidivism risk prediction was investigated. The average follow-up period was 11.08 years, during which 13.3% of the sample were convicted of a new sexual offence. Hierarchical Cox regression analyses showed that release planning contributed additional predictive validity for sexual recidivism after controlling for static and dynamic risk factors. Findings suggest that assessment of release planning might improve accuracy of sex offender risk assessments and that improved release planning should contribute to reductions in recidivism.     

MDS use with crime scene data replicates poorly: response to Goodwill,Alison, and Humann (this issue) and Davis (this issue)

Abstract

Examinations of treatment attrition form an important – although sometimes neglected – component of evaluating a correctional programme's effectiveness in reducing recidivism. Previous research has identified offender characteristics that predict non-completion. This study investigated non-completion in 138 high-risk, violent male prisoners attending an intensive cognitive–behavioural programme. Almost one-third of men who commenced it did not complete the 7-month programme. Most asked to leave of their own accord, or were removed for ongoing offending. In contrast to previous research, no support was found for the hypothesis that those who terminated treatment prematurely were more in need of intervention than those who completed the programme; non-completers did not differ from completers on static estimates of criminal risk, PCL-R scores, demographic variables or self-report scales measuring dynamic risk factors. It was concluded that successful prediction using variables related to criminal risk and criminogenic need depends both on the characteristics of programme participants, and on contextual factors such as programme policies: when high-risk high needs offenders are a programme's target clientele, variables related to risk and need will have limited predictive utility. From a practice perspective, the programme was successful in retaining through to completion a relatively untreatable group: high-risk offenders with moderate to high PCL-R scores.     

Against Prevention? A Response to Harcourt's Against Prediction on Actuarial and Clinical Predictions and the Faults of Incapacitation

This article reviews Bernard Harcourt's Against Prediction: Profiling, Policing, and Punishing in an Actuarial Age (2007), in which he criticizes the use of actuarial prediction methods in the contexts of policing and sentencing. I focus on the latter context. I argue that Harcourt has identified an important, and not exclusively American, trend and develops a valid critique of it that should be pushed further. From a theory of punishment perspective, I argue that Harcourt's critique is no less applicable to clinical prediction methods than to the use of actuarial ones. Harcourt's arguments, however, beg a more general explanation of the flaws of incapacitation as a justification for punishment. If we base our objection to the use of prediction methods on such larger grounds, questions arise as to the legitimacy of other practices that are not considered punitive but rather "regulatory" or "preventive."     

How well do dynamic needs predict recidivism? Implications for risk assessment and risk reduction

Purpose

The current study assessed the predictive validity of nine dynamic risk factors in two samples of justice-involved individuals (n = 24,972) to identify promising targets for correctional programming. The study also tested the incremental predictive validity of dynamic risk relative to static indicators of recidivism risk (i.e., criminal history, age and sex).

Methods

The study relied on bivariate correlations and stepwise multivariate logistic regression analyses to test the predictive and incremental validity of dynamic risk constructs measured by the Level of Service Inventory-Revised (LSI-R).

Results

Study findings revealed that while several dynamic risk constructs were significantly related to recidivism (antisocial attitudes, antisocial peers, education/employment, and substance abuse) over a 36-month follow-up period, the inclusion of dynamic risk items on the LSI-R did little to improve the overall predictive validity of the instrument. Across both study samples, static criminal history risk emerged as the most robust predictor of recidivism.

Conclusions

The advantages of third and fourth generation risk-need assessment tools are noted for classification purposes, but the study findings imply that risk prediction is better served by static risk factors. Implications for risk assessment and offender case management are discussed.     

Mandatory life sentences for serious sex offenders in England and Wales: tough,but intelligent too?

Abstract

This paper reviews the changes in risk assessment procedures for sexual offenders over the last 15 years from the viewpoint of two active participants in that change. Best practices with this target group have evolved at a dizzying pace, leaving many practitioners and programme managers uncertain about which tests or procedures they should use and, frankly, wondering why things keep changing. We view this ongoing evolution as very positive. Compared to the early 1990s, evaluators now have better knowledge of the static and dynamic factors associated with sexual recidivism, and a number of empirically validated risk assessment tools. We describe the various risk assessment procedures we have introduced (e.g. STATIC-99, SONAR, STABLE-2007/ACUTE-2007), the reasons why practices have changed and the reasons why practices will continue to change.     

Actuarial Assessment of Sex Offender Recidivism Risk: A Cross-Validation of the RRASOR and the Static-99 in Sweden

We cross-validated two actuarial risk assessment tools, the RRASOR (R. K. Hanson, 1997) and the Static-99 (R. K. Hanson & D. Thornton, 1999), in a retrospective follow-up (mean follow-up time = 3.69 years) of all sex offenders released from Swedish prisons during 1993–1997 (N = 1,400, all men, age 18 years). File-based data were collected by a researcher blind to the outcome (registered criminal recidivism), and individual risk factors as well as complete instrument characteristics were explored. Both the RRASOR and the Static-99 showed similar and moderate predictive accuracy for sexual reconvictions whereas the Static-99 exhibited a significantly higher accuracy for the prediction of any violent recidivism as compared to the RRASOR. Although particularly the Static-99 proved moderately robust as an actuarial measure of recidivism risk among sexual offenders in Sweden, both procedures may need further evaluation, for example, with sex offender subpopulations differing ethnically or with respect to offense characteristics. The usefulness of actuarial methods for the assessment of sex offender recidivism risk is discussed in the context of current practice.