Predicting batterer recidivism five years after community intervention

The purpose of this study was to examine batterer recidivism rates 5 years after community intervention and to determine differences that discriminate between recidivists and nonrecidivists. Of the 100 men included in the sample, 40% were identified as recidivists because they were either convicted of domestic assault, the subject of an order for protection, or a police suspect for domestic assault. A discriminant analysis was conducted using a variety of background and intervention variables. Five variables were selected that significantly discriminated between recidivists and nonrecidivists and correctly classified 60.6% of the cases. Men who had been abusive for a shorter duration prior to the program, court ordered to have a chemical dependency evaluation, in chemical dependency treatment, abused as children, and previously convicted for nonassault crimes were more likely to be recidivists. Variables relating to intervention did not significantly predict recidivism. Implications for community intervention programs are discussed.     

Gender expectations and their impact on black female offenders and victims

The purpose of this study is to investigate the relationship between failure to register (FTR) as a sex offender and subsequent recidivism (N = 2,970). No significant differences were found between the sexual recidivism rates of those who failed to register and compliant registrants (11% vs. 9%, respectively). There was no significant difference in the proportion of sexual recidivists and nonrecidivists with registration violations (12% vs. 10%, respectively). FTR did not predict sexual recidivism, and survival analyses revealed no significant difference in time to recidivism when comparing those who failed to register (2.9 years) with compliant registrants (2.8 years). Results fail to support the supposition that sexual offenders who fail to register are more sexually dangerous than those who comply with registration requirements. The punitive emphasis on registration enforcement may not be justified and might divert limited resources away from strategies that would better facilitate public protection from sexual violence.     

Violent and sexual offences: a validation of the predictive quality of the PCL:SV in Switzerland

Used to diagnose psychopathy, the Psychopathy Checklist: Screening Version (PCL:SV) is still rarely used in the German-speaking part of Europe. The aim of the present study was to conduct a first-time evaluation of the predictive validity of the PCL:SV for sexual and violent recidivism in the German language area. PCL:SV scores for 96 violent and sexual offenders in Switzerland were assessed. These assessments were based on data taken from the forensic psychiatric expert opinions. The scores were then compared to subsequent recidivism as shown in the official criminal records. ROC analysis indicated a cut-off score of 14 for maximized overall predictive accuracy. Consistent with various results from past research in North America and Europe, the PCL:SV yielded a satisfying predictive accuracy (ROC area=.69). This degree of accuracy demonstrates the PCL:SV's usefulness for risk assessment of sexual and violent delinquency in the Swiss cultural and German language area.     

Typologies,risk and recidivism in partner-violent men with the B-SAFER: a pilot study

The research describes the criminal profile of 100 imprisoned partner-violent men (PVM) in Spain, and the follow-up for an average of 15 months of 40 released cases. The ability of the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER) to classify offenders according to typologies and to predict recidivism is analyzed. The results show that PVM have low level of specialization (only 45% limit their criminal activity to intimate partner violence (IPV)) and high level of recidivism (47% previously have been in prison, and 41% have prior arrest for IPV). The B-SAFER shows a high capacity to classify according to batterers typologies (accuracy of 79% with a score ≥13) in two groups: non-pathological and antisocial/pathological offenders. After prison release, 17.5% relapsed (15% in IPV), and 66% have done so within the first year. The B-SAFER had a predictive accuracy of 70% (sensitivity 100%). From antisocial/pathological group, 21% have recidivate compared to 12.5% in non-pathological aggressors, with an over-representation of antisocial/pathological subtype among recidivists (71%). The best predictive variables are the justification of violence, age at first imprisonment, and treatment. There is a 9% of recidivism among treated offenders compared to 50% in the untreated group.     

Improving Risk Assessments for Sex Offenders: A Comparison of Three Actuarial Scales

The study compared the predictive accuracy of three sex offender risk-assessment measures: the RRASOR (Hanson, 1997), Thornton's SACJ-Min (Grubin, 1998), and a new scale, Static-99, created by combining the items from the RRASOR and SACJ-Min. Predictive accuracy was tested using four diverse datasets drawn from Canada and the United Kingdom (total n = 1301). The RRASOR and the SACJ-Min showed roughly equivalent predictive accuracy, and the combination of the two scales was more accurate than either original scale. Static-99 showed moderate predictive accuracy for both sexual recidivism (r = 0.33, ROC area = 0.71) and violent (including sexual) recidivism (r = 0.32, ROC area = 0.69). The variation in the predictive accuracy of Static-99 across the four samples was no more than would be expected by chance.     

Prospective Replication of the Violence Risk Appraisal Guide in Predicting Violent Recidivism Among Forensic Patients

An exhaustive survey of a cohort of forensic patients provided an opportunity for a prospective replication of the predictive accuracy of the Violence Risk Appraisal Guide (VRAG). Data collected during the original survey also permitted a test of the predictive accuracy of clinical assessments of risk on the same cohort. The VRAG yielded a large effect size in predicting violent recidivism (ROC area = .80) over a constant 5-year follow-up and performed significantly better than averaged clinical opinions. The superiority of the VRAG was also observed at very short follow-up times and for very serious violence. Moreover, for 16 subsamples, observed rates of violent recidivism did not differ significantly from the expected rates. VRAG score was unrelated, and clinical judgments inversely related to violent recidivism in the small low-risk sample of female forensic patients. The authors conclude that, regardless of length of opportunity or severity of outcome, actuarial methods are more accurate than is clinical judgment.     

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.     

The effect of gender on violent and nonviolent recidivism: A meta-analysis

A large body of literature has been dedicated to understanding re-offending after release from prison - a phenomenon known as recidivism. A meta-analysis was conducted using fifty- seven published studies on actuarial predictors of both violent and nonviolent recidivism in men and women. The current analysis resulted in 127 effect sizes between violent recidivists and non-recidivists, and sixty-eight effect sizes between violent recidivists with nonviolent recidivists. Several variables (drug/alcohol use, age, and marriage) were predictive of recidivism. Gender differences were also observed. In men, increased violent criminal history was associated with increased violent recidivism. This effect was not true for women. In fact, longer sentences were predictive of violent re-offending in women and not men. Despite limited data on women, pursuing the predictors of recidivism in men and women independently demonstrated that gender differences exist, and supported the need for more data concerning the predictors of recidivism in women.     

Identifying Three Types of Violent Offenders and Predicting Violent Recidivism While on Probation: A Classification Tree Analysis

This study employs classification tree analysis (CTA) to address whether 3 groups of violent offenders have similar or different risk factors for violent recidivism while on probation. A sample of 1344 violent offenders on probation was classified as generalized aggressors (N = 302), family only aggressors (N = 321), or nonfamily only aggressors (N = 717). The strongest predictor of violent recidivism while on probation was whether the offender was a generalized aggressor or not, with generalized aggressors more likely to be arrested for new violent crimes. Prior arrests for violent crimes predicted violent recidivism of generalized aggressors, but did not significantly predict violent recidivism of family only and nonfamily only aggressors. For generalized aggressors and family only batterers, treatment noncompliance was an important risk predictor of violent recidivism. CTA compared to logistic regression classified a higher percentage of cases into low-risk and high-risk groups, provided higher improvement in classification accuracy of violent recidivists beyond chance performance, and provided a better balance of false positives and false negatives. The implications for the risk assessment and domestic violence literature are discussed.     

Psychopathy (PCL-R) Predicts Violent Recidivism Among Criminal Offenders with Personality Disorders in Sweden

Psychopathy as conceptualized with Hare's Psychopathy Checklist Revised, PCL-R, has attracted much research during the 1990s. In the Scandinavian countries, few studies that empirically support the validity of North American risk assessment techniques in our regional context have been published. The purpose of this paper is to explore the predictive power of the PCL-R in a population of personality-disordered violent offenders subjected to forensic psychiatric evaluation in Sweden. Following release from prison (n = 172), discharge from forensic psychiatric treatment (n = 129), or probation (n = 51), a total of 352 individuals were followed for up to 8 years (mean = 3.7 years) with reconviction for violent crime as endpoint variable (base rate 34%). As the estimate of predictive power, the area under the curve of a receiver operating characteristic (AUC of ROC) analysis was calculated. For PCL-R scores to predict 2-year violent recidivism, AUC of ROC was .72 (95% CI: .66–.78). In addition, the personality dimension of psychopathy (Factor 1) and the behavioral component (Factor 2) both predicted 2-year recidivism significantly better than random: AUC of ROC .64 (95% CI: .57–.70) and .71 (95% CI: .65–.77), respectively. We conclude that psychopathy is probably as valid a predictor of violent recidivism in Swedish forensic settings as seen in previous North American studies.     

The effects of base rate and cutoff point choice on commonly used measures of association and accuracy in recidivism research

Several measures of association and accuracy commonly used in the recidivism literature are examined for their sensitivity to variations in the proportion observed to recidivate (the base rate) and in the proportion selected for classification as recidivists (selection ratio). Logistic regression models are employed on a sample of 11,749 convicted offenders to generate predicted probabilities of four recidivism criteria varying in base rates from 0.06 to 0.48. Cutoff point selections from 0.1 to 0.9 show the effects of cutoff point changes on the following commonly used measures of association and accuracy: RIOC (relative improvement over chance), MCR (mean cost rating), Φ, γ, PRE (proportion reduction in error), and percentage correct. While all these statistics vary across base rate and cutoff points, some vary more than others: RIOC varies across cutoff points more than MCR, MCR more than Φ, and Φ more than γ. Researchers comparing such statistics across studies need be wary of the dangers of ignoring such variation.     

Discriminators of types of recidivism among boot camp graduates in a five-year follow-up study

This was a five-year follow-up study of 572 male graduates of a boot camp in the South. The purpose was to determine what elements of life-course theory discriminated between four outcomes: (1) nonrecidivists, (2) recidivists with felonies, (3) technical parole violators, and (4) parole violators due to drugs. Findings indicated a relatively accurate classification of the first three outcomes.The discriminators between nonrecidivists and various types of recidivists in the study were indicators of informal social controls such as marriage, employment, children, and religiosity. Discriminators of felony recidivists were early sexual and physical abuse, gang membership, peer association, carrying a weapon, and early onset of crime. Discriminators associated with technical parole violation were attachment to female caregivers, self-esteem and self-efficacy, and perception that boot camp stimulated thinking about choices in life. The conceptual and policy implications of these findings are discussed in detail.     

A LONGITUDINAL PERSPECTIVE ON PATTERNS OF CRIME

This article addresses the possibility of prospectively identfying chronic criminal offenders using identifiers available to the criminal justice system. Data were collected as part of a longitudinal study of the lives of men who were in a delinquency prevention program in the early 1940s. For the half who were randomly assigned to the treatment group, a discriminant function correctly identified 67% of the men as recidivists or nonrecidivists. When tested on the control group, it correctly sorted 76% of the men. The author argues that although prospective identification of recidivists may be possible, legal action based on predictions would be wrong.     

Analyzing the Performance of Risk Assessment Instruments: A Response to Vrieze and Grove (2007)

In a recent article, Vrieze and Grove (Law Hum Behav, doi: 10.1007/s10979-007-9092-x , 2007) argue that, because of low recidivism base rates and limited predictive accuracy, an actuarial risk assessment instrument (ARAI) may produce decisions about sex offenders that are worse than simply predicting that no one will commit another sex offense. This article examines: (1) the construction and potential overfitting of ARAIs; (2) the meaning, value, and limitations of ROC areas; and (3) the relationship between the operating point that maximizes an ARAI's correct classifications and the legal criterion-"likely to reoffend"-used for sex offender designations. Contrary to what Vrieze and Grove suggest, ARAIs of modest accuracy yield probabilistic information that is more relevant to legal decision-making than just "betting the base rate."     

Recidivism in subgroups of severe male juvenile offenders

Abstract

Prior research is lacking on the incremental contribution of juvenile offender classification systems in predicting recidivism. To address this gap, the present study examined a five-group classification system of severe adolescent male offenders based on the personality and clinical scales of the Millon Adolescent Clinical Inventory (MACI). Group membership was used to predict adult recidivism while controlling for criminal history. Male juvenile offenders classified as anxious/impulsive were less frequent recidivists than most other groups and had fewer charges after release than some other groups. Offenders classified into the psychopathy group were notable for their high rate of recidivism (nearly 50%). These results highlight the potential utility of offender classification systems for informing risk assessments among severe male juvenile offenders.     

The significance of offending behavior for predicting sexual recidivism among sex offenders of various age groups

Based on a sample of 682 male sex offenders, this study investigated variables of the offending behavior for predicting sexual recidivism in different age groups. The sex offenders were allocated into four age groups. For each group, those characteristics of offending behavior showing a significant bivariate correlation with sexual recidivism were extracted. Using logistic regression, we then analyzed their incremental validity above and beyond a previously developed Crime Scene Behavior Risk (CBR)-Score, which measures the risk of sexual recidivism without taking into account the offenders' age (Dahle, Biedermann, Gallasch-Nemitz, & Janka, 2010). Age-specific offending behavior variables with incremental validity were combined with the general CBR-Score into age-specific scores and examined for their predictive accuracy. We also analyzed the extent in which these age-specific scores showed incremental validity above and beyond the Static-99 (Hanson & Thornton, 1999). For three of the four age groups, age-specific Crime Scene Behavior Risk-Scores could be determined which were incrementally valid above and beyond the Static-99. Predictive accuracy varied between AUC=.74 and AUC=.90 (r=.28 to r=.49) depending on age group. The results are discussed within context of recent findings on the latent dimensions of actuarial risk assessment variables.     

Prediction of violent and sexual offences: A replication study of the VRAG in Switzerland

Abstract

The Violence Risk Appraisal Guide (VRAG) is a widely used actuarial risk assessment instrument and has been validated in different countries. However, there is no investigation supporting the predictive accuracy of the VRAG in the German language area. The VRAG scores of 79 violent and sexual offenders in Switzerland were assessed based on data from their psychiatric expert opinions. The VRAG scores were compared to subsequent recidivism as shown in the official criminal records. Consistent with past research in the English language area, the VRAG yielded a satisfying predictive accuracy (ROC area = .73), demonstrating its usefulness for risk assessment of violent delinquency in the Swiss cultural and German language area.     

Youth Psychopathy and Criminal Recidivism: A Meta-Analysis of the Psychopathy Checklist Measures

Although narrative reviews have suggested that "youth psychopathy" is a strong predictor of future crime and violence, to date no quantitative summaries of this literature have been conducted. We meta-analyzed recidivism data for the Psychopathy Checklist measures across 21 non-overlapping samples of male and female juvenile offenders. After removing outliers, psychopathy was significantly associated with general and violent recidivism (r (w)'s of .24 and .25, respectively), but negligibly related to sexual recidivism in the few studies examining this low base rate outcome. Even after eliminating outliers, however, considerable heterogeneity was noted among the effects, with some of this variability being explained by the gender and ethnic composition of the samples. Effect sizes for the small number of female samples available for analysis were mostly small and nonsignificant, and psychopathy was a weaker predictor of violent recidivism among more ethnically heterogeneous samples. In relation to predicting both general and violent recidivism, psychopathy performed comparably to an instrument designed specifically to assess risk, the Youth Level of Service/Case Management Inventory (Hoge & Andrews, 2002).     

Battered women's perceptions of risk versus risk factors and instruments in predicting repeat reassault

This study partially replicates and expands on a previous study that showed women's perceptions of risk to be a strong predictor of reassault among batterers. The current study employed a larger and multisite sample, a longer follow-up period of 15 months, and multiple outcomes including "repeated reassault" (n = 499). According to the multinomial logistic regressions, women's perceptions of risk improved prediction with risk factors (ROC area under the curve improved by.04 and sensitivity of repeated reassault increased 12 percentage points). In comparison to simulated risk instruments, women's perceptions by themselves were better predictors than the K-SID, similar in predictive ability to the SARA, and almost as strong as the DAS. The best prediction of repeated reassault was obtained using risk markers, including women's perceptions (ROC AUC =.83; 70% sensitivity) or by using the DAS and women's perceptions together (ROC AUC =.73; 64% sensitivity).     

Psychopathy (PCL-R) as a Predictor of Violent Recidivism Among Criminal Offenders with Schizophrenia

Hare's Psychopathy Checklist-Revised (PCL-R) was used to test the hypothesis that psychopathy predicts violent recidivism in a cohort subjected to forensic psychiatric investigation and consisting of male violent offenders with schizophrenia (N = 202). Psychopathy was assessed with retrospective file-based ratings. Mean follow-up time after detainment was 51 months. Twenty-two percent of the offenders had a PCL-R score 26 (cutoff), and the base rate for violent recidivism (reconvictions) during follow-up was 21%. Survival analysis revealed that psychopathy was strongly associated to violent recidivism (log-rank = 17.71, df = 1, p < 0.0001). The area under the curve (AUC) of the receiver operating characteristics (ROC) of PCL-R total score to predict violent recidivism varied between different time frames from .64 to .75. Cox regression analyses revealed that other potential risk factors could not equally well or better explain violent recidivism in the cohort than psychopathy as measured by PCL-R.