Recidivism Estimates Research Articles

Assessing dynamic violence risk: Common language risk levels and recidivism rates for the Violence Risk Scale.

The present study features the development of new risk categories and recidivism estimates for the Violence Risk Scale (VRS), a violence risk assessment and treatment planning tool. We employed a combined North American multisite sample (k = 6, N = 1,338) of adult mostly male offenders, many with violent criminal histories, from correctional or forensic mental health settings that had complete VRS scores from archival or field ratings and outcome data from police records (N = 1,100). There were two key objectives: (a) to identify the rates of violent recidivism associated with VRS scores and (b) to generate updated evidence-based VRS violence risk categories with external validation. To achieve the first objective, logistic regression was applied using VRS pretreatment and change scores on treated samples with a minimum 5-year follow-up (k = 5, N = 472) to model 2-, 3-, and 5-year violent and general recidivism estimates, with the resulting logistic regression algorithms retained to generate a VRS recidivism rates calculator. To achieve the second objective, the Council of State Governments' guidelines were applied to generate five risk levels using the common language framework using percentiles, risk ratios (from Cox regression), and absolute violent and general recidivism estimates (from logistic regression). Construct validity of the five risk levels was examined through group comparisons on measures of risk, need, protection, and psychopathy obtained from the constituent samples. VRS applications to enhance risk communication, treatment planning, and violence prevention in light of the updated recidivism estimates and risk categories are discussed. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Estimating the probability of sexual recidivism among men charged or convicted of sexual offences: Evidence-based guidance for applied evaluators

Risk assessment is routinely applied in forensic decision-making. Although relative risk information from risk scales is robust across diverse samples and settings, estimates of the absolute probability of sexual recidivism are not. Nonetheless, absolute recidivism estimates are still necessary in some evaluations. This paper summarizes research and offers guidance on evidence-based practices for assessing the probability of recidivism, organized largely around questions commonly asked in court. Overall, estimating the probability of sexual recidivism is difficult and should be undertaken with humility and circumspection. That being said, research favours empirical-actuarial risk tools for this task, more structured scales, and the use of multiple scales. Professional overrides of risk scale results should not be used under any circumstances. Paradoxically, however, professional judgement is still required in some circumstances. Risk scales do not consider all relevant risk factors, but the added value of external risk factors reaches a point of diminishing returns and may or may not be incremental (or worse, can degrade accuracy). There are reasons actuarial risk scales may both underestimate recidivism (e.g., undetected offending, short follow-ups) and overestimate recidivism (e.g., inclusion of sex offences not of interest in some referral questions, data on declining crime and recidivism rates, newer studies demonstrating overestimation of recidivism). Given all these considerations and the need for humility, in the absence of exceptional circumstances, I would not deviate too far from empirical estimates.

Incremental Contributions of Static and Dynamic Sexual Violence Risk Assessment: Integrating Static-99R and VRS-SO Common Language Risk Levels

We examined the incremental contributions of static and dynamic sexual violence risk assessment in a multisite sample of 1,289 men treated for sexual offending. The study extends validation work that established new risk categories and recidivism estimates for the Violence Risk Scale–Sexual Offense version (VRS-SO), using the risk assessment common language (CL) framework. Different rates of sexual recidivism were observed at different thresholds of static risk (Static-99R) as a function of dynamic risk and treatment change, particularly for men who were actuarially above or well above average risk (Levels IVa and IVb, respectively). A framework integrating CL risk levels for Static-99R and VRS-SO dynamic scores into overall CL risk levels is presented. We discuss implications for dynamic sexual violence risk assessment regarding the language used for risk communication and the importance of dynamic risk instruments in sexual violence evaluations, particularly when credible agents of risk change may be present.

Private vs. public prisons? A dynamic analysis of the long-term tradeoffs between cost-efficiency and recidivism in the US prison system

ABSTRACT Compared with the rest of the world, the US prison system is characterized by high incarceration rates, high recidivism, and substantial incarceration costs. The US has increasingly turned to private prisons, resulting in a debate concerning the efficacy of these private prisons. While the private system may produce cost savings it may have also have negative consequences including increased recidivism. Although numerous studies have investigated the cost savings associated with private prisons and their effect on recidivism, to our knowledge no work has evaluated the joint impact. In this paper we evaluate these impacts jointly within a dynamic system model, utilizing cost, incarceration and recidivism estimates from the literature. Overall, considering the tradeoff between cost efficiency and recidivism rates, we find that public prisons are less costly in the long term than private prisons. Considering a 25-year time horizon, we find that total inflation-adjusted costs are approximately 1.5% higher for private prisons than public prisons. Considering a 40-year time horizon, we find that private prisons are 3% more costly than public prisons. Thus, results suggest that estimated short-term cost efficiency provided by private prisons may not be worth the long-term consequences of potential increases in recidivism.

Discrimination and Calibration Properties of the Level of Service Inventory–Ontario Revision in a Correctional Mental Health Sample

We examined the predictive properties of the Level of Service Inventory–Ontario Revision (LSI-OR) in a sample of 604 provincially incarcerated men with mental illness from a correctional mental health facility followed up nearly 2 years after release. Recidivism base rates and LSI-OR scores were relatively consistent across major mental disorder categories, but higher among individuals with personality disorder, substance use disorder, or dual diagnosis. LSI-OR scores predicted general and violent recidivism in the overall sample and among specific diagnostic groups. Calibration analyses were conducted to model 1-year recidivism estimates for the overall sample and among individual diagnostic groups associated with individual LSI-OR scores. Good correspondence was observed among the different diagnostic groups, with some difference in recidivism trajectories given the differences in base rate. The results support the predictive properties of the LSI-OR with correctional mental health samples and inform the recidivism estimates associated with LSI-OR scores in this population.

Using the Violence Risk Scale-Sexual Offense version in sexual violence risk assessments: Updated risk categories and recidivism estimates from a multisite sample of treated sexual offenders.

The present study sought to develop updated risk categories and recidivism estimates for the Violence Risk Scale-Sexual Offense version (VRS-SO; Wong, Olver, Nicholaichuk, & Gordon, 2003-2017), a sexual offender risk assessment and treatment planning tool. The overarching purpose was to increase the clarity and accuracy of communicating risk assessment information that includes a systematic incorporation of new information (i.e., change) to modify risk estimates. Four treated samples of sexual offenders with VRS-SO pretreatment, posttreatment, and Static-99R ratings were combined with a minimum follow-up period of 10-years postrelease (N = 913). Logistic regression was used to model 5- and 10-year sexual and violent (including sexual) recidivism estimates across 6 different regression models employing specific risk and change score information from the VRS-SO and/or Static-99R. A rationale is presented for clinical applications of select models and the necessity of controlling for baseline risk when utilizing change information across repeated assessments. Information concerning relative risk (percentiles) and absolute risk (recidivism estimates) is integrated with common risk assessment language guidelines to generate new risk categories for the VRS-SO. Guidelines for model selection and forensic clinical application of the risk estimates are discussed. (PsycINFO Database Record

Supplemental Material for Using the Violence Risk Scale-Sexual Offense Version in Sexual Violence Risk Assessments: Updated Risk Categories and Recidivism Estimates From a Multisite Sample of Treated Sexual Offenders

Supplemental Material for Using the Violence Risk Scale-Sexual Offense Version in Sexual Violence Risk Assessments: Updated Risk Categories and Recidivism Estimates From a Multisite Sample of Treated Sexual Offenders

Incremental prediction of intimate partner violence: An examination of three risk measures.

Improvements in the risk prediction of domestic violence against intimate partners have the potential to inform policing practices in the prevention of further victimization. The present study examined the incremental predictive validity of 3 measures of risk for intimate partner violence (IPV)-Spousal Assault Risk Assessment (SARA), Ontario Domestic Assault Risk Assessment (ODARA), and the Family Violence Investigative Report (FVIR)-for IPV, general violence, and general recidivism outcomes. The sample featured 289 men and women who were reported to police for IPV and followed up approximately 3 years post release. Archival ratings of the 3 measures demonstrated that SARA scores showed incremental validity for IPV recidivism, ODARA scores incrementally predicted general violence, and both tools incrementally predicted general recidivism. The FVIR did not incrementally predict any outcomes. Fine grained analyses demonstrated that the Psychosocial Adjustment domain of the SARA contributed most uniquely to the prediction of IPV. Survival analysis supported the use of the SARA and ODARA in tandem for appraising risk for IPV or general criminal recidivism. Calibration analyses using logistic regression modeling also demonstrated 3-year recidivism estimates for SARA and ODARA scores. Implications for the use of multiple tools in clinical practice are discussed, particularly for combining the SARA and ODARA measures to augment IPV risk assessment and management. (PsycINFO Database Record

Evaluating the Impact of “Old” Criminal Conviction Decision Guidelines on Subsequent Employment and Arrest Outcomes

Objectives: Redemption scholars estimate that after an average of 7-10 years pass without a new arrest or conviction, a person’s criminal record essentially loses its predictive value. This article provides the first labor market and recidivism estimates of implementing a criminal background check decision guideline based on this redemption research. Methods: The sample consists of provisionally hired job applicants in New York State’s healthcare industry with at least one prior conviction. A “10 years since last conviction” guideline situated within a highly formalized criminal background check process plausibly creates conditional random variation in clearance decisions, which allows for a regression model to estimate causal effects. Results: Individuals cleared to work because of the 10-year guideline experience meaningful improvements in employment and earnings, but not recidivism on average. However, men do experience reductions in subsequent arrests, which appears to be driven by more complex factors beyond simply time since last arrest. Conclusions: For some individuals, receiving clearance to work even a decade after their last conviction can have not only labor market benefits, but also important recidivism implications. Future research should explore the employment opportunity/recidivism trade-off in adjusting guideline threshold values and consider alternative redemption strategies.

Developing Nonarbitrary Metrics for Risk Communication

Nominal risk categories for actuarial risk assessment information should be grounded in nonarbitrary, evidence-based criteria. The current study presents numeric indicators for interpreting one such tool, the Risk Matrix 2000, which is widely used to assess the recidivism risk of sexual offenders. Percentiles, risk ratios, and 5-year recidivism rates are presented based on an aggregated sample ( N = 3,144) from four settings: England and Wales, Scotland, Germany, and Canada. The Risk Matrix 2000 Sex, Violence, and Combined scales showed moderate accuracy in assessing the risk of sexual, non-sexual violent, and violent recidivism, respectively. Although there were some differences across samples in the distributions of risk categories, relative increases in recidivism for ascending risk categories were remarkably consistent. Options for presenting percentiles, risk ratios, and absolute recidivism estimates in applied evaluations are offered, with discussion of the advantages, disadvantages, and limitations of these risk communication metrics.

On the reinforcing nature of crime and punishment: An exploration of inmates’ self-reported likelihood of reoffending

ABSTRACTCountering deterrence theory, recent scholarship suggests that incarceration may be criminogenic. This study theorizes that two processes influence recidivism expectations: If offending alleviates status frustration, or if criminal justice sanctions normalize deviance and instigate an identity transformation supportive of procrime values, then inmates will be more likely to report an increased likelihood of reoffending upon release. Based on self-report data from a sample of prison inmates, results indicate that a lack of life achievements and greater exposure to criminal justice sanctions discourages a commitment to desistance. Further, the psychosocial reinforcements provided by offending are positively associated with self-reported recidivism estimates.

The MATS-1 Risk Assessment Scale: Summary of Methodological Concerns and an Empirical Validation.

Recently, Wollert and colleagues proposed that sex offender recidivism estimates should be stratified by age and they developed an age-stratified scale called the MATS-1 (Multisample Age-Stratified Table of Sexual Recidivism). The purpose of this article is to highlight concerns with the development of the MATS-1 and to validate the scale using 3,510 sex offenders from 14 unique samples. Concerns with the scale's development fall into three categories: approximations leading to considerable loss of precision, absence of appropriate statistical tests, and the use of inappropriate statistical techniques. The predictive accuracy of the MATS-1 (Area Under the Curve [AUC] = .663) was significantly lower than Static-99R (AUC = .708). The MATS-1 also significantly underestimated recidivism for some offenders. Both the relative and absolute predictive properties of the MATS-1 were not stable across samples. We conclude that the MATS-1 is not appropriate to use for applied risk assessment. Proposals are made for alternate ways to develop risk scales using the age-stratification method.

Sex offender risk assessment with the Risk Matrix 2000: validation and guidelines for combining with the STABLE-2007

Comprehensive risk assessment can benefit from combining static and dynamic scales. The purpose of this study was to validate the Risk Matrix 2000, examine incremental validity of the STABLE-2007 and develop structured rules for combining the two scales. In a routine sample of Canadian sex offenders (n = 710), the Risk Matrix 2000/Sex, /Violence and /Combined scales significantly predicted the six recidivism outcomes: sexual, any sexual (including sexually motivated breaches), non-sexual violent, violent, any crime (excluding breaches) and any recidivism (including breaches). Area under the receiver operating characteristic curve (AUC) values were moderate to large. The STABLE-2007 generally added incremental predictive validity to the Risk Matrix 2000. Structured rules and recidivism estimates are provided for combining the two scales. The current study illustrates that predictive accuracy can be improved by combining static and dynamic risk scales in a structured manner.

Preliminary Validation of the ODARA for Female Intimate Partner Violence Offenders

Female perpetrators of assault against dating, cohabiting, or marital partners (intimate partner violence [IPV]) recidivate less than their male counterparts. Risk assessment instruments, though, have been developed almost exclusively on men. In a prospective, masked 9-year follow up of 30 female IPV offenders incarcerated in a correctional treatment institution within one decade, the base rate of IPV recidivism was 23%. The Ontario Domestic Assault Risk Assessment (ODARA) predicted IPV recidivism, receiver operating characteristic (ROC) area = .724, 95% confidence interval (CI) = [0.503, 0.944], but recidivism rates differed significantly from rates based on male samples. Gender-modified items did not improve prediction. We recommend further research with larger samples to seek improved recidivism estimates among female IPV offenders, but in the interim, we suggest the ODARA can be used to apportion intervention resources for female IPV perpetrators.

Absolute Recidivism Rates Predicted By Static-99R and Static-2002R Sex Offender Risk Assessment Tools Vary Across Samples

There has been considerable research on relative predictive accuracy (i.e., discrimination) in offender risk assessment (e.g., Are high-risk offenders more likely to reoffend than low-risk offenders?), but virtually no research on the accuracy or stability of absolute recidivism estimates (i.e., calibration). The current study aimed to fill this gap by examining absolute and relative risk estimates for certain Static sex offender assessment tools. Logistic regression coefficients for Static-99R and Static-2002R were combined through meta-analysis (8,106 sex offenders; 23 samples). The sexual recidivism rates for typical sex offenders are lower than the public generally believes. Static-99R and Static-2002R both demonstrated remarkably consistent relative predictive accuracy across studies. For both scales, however, the predicted recidivism rates within each risk score demonstrated large and significant variability across studies. The authors discuss how the variability in recidivism rates complicates the estimation of recidivism probability in applied assessments.

Improving the Predictive Accuracy of Static-99 and Static-2002 With Older Sex Offenders

Actuarial risk assessment scales and their associated recidivism estimates are generally developed on samples of offenders whose average age is well below 50 years. Criminal behavior of all types declines with age; consequently, actuarial scales tend to overestimate recidivism for older offenders. The current study aimed to develop a revised scoring system for two risk assessment tools (Static-99 and Static-2002) that would more accurately describe older offenders' risk of recidivism. Using data from 8,390 sex offenders derived from 24 separate samples, age was found to add incremental predictive validity to both Static-99 and Static-2002. After creating new age weights, the resulting instruments (Static-99R and Static-2002R) had only slightly higher relative predictive accuracy. The absolute recidivism estimates, however, provided a substantially better fit for older offenders than the recidivism estimates from the original scales. We encourage evaluators to adopt the revised scales with the new age weights.

Recent Research (N = 9,305) Underscores the Importance of Using Age-Stratified Actuarial Tables in Sex Offender Risk Assessments

A useful understanding of the relationship between age, actuarial scores, and sexual recidivism can be obtained by comparing the entries in equivalent cells from "age-stratified" actuarial tables. This article reports the compilation of the first multisample age-stratified table of sexual recidivism rates, referred to as the "multisample age-stratified table of sexual recidivism rates (MATS-1)," from recent research on Static-99 and another actuarial known as the Automated Sexual Recidivism Scale. The MATS-1 validates the "age invariance effect" that the risk of sexual recidivism declines with advancing age and shows that age-restricted tables underestimate risk for younger offenders and overestimate risk for older offenders. Based on data from more than 9,000 sex offenders, our conclusion is that evaluators should report recidivism estimates from age-stratified tables when they are assessing sexual recidivism risk, particularly when evaluating the aging sex offender.