Predictive Properties of the Violence Risk Scale–Sexual Offense Version as a Function of Age

Abstract

The present study examined the discrimination and calibration properties of Violence Risk Scale–Sexual Offense version (VRS-SO) risk and change scores for sexual and violent recidivism as a function of age at release, on a combined sample of 1,287 men who had attended sexual offense-specific treatment services. The key aim was to examine to what extent VRS-SO scores can accurately discriminate recidivists from nonrecidivists among older cohorts, and if the existing age-related adjustments in the instrument adequately correct for increasing age. VRS-SO risk and change scores showed consistent properties of discrimination for sexual recidivism across the age cohorts, via area under the curve and Cox regression survival analysis, as demonstrated through fixed effects meta-analysis. Calibration analyses, employing logistic regression, demonstrated that age at release was consistently incrementally predictive of violent, but not sexual, recidivism after controlling for individual differences on static and dynamic risk factors. E/O index analyses demonstrated that predicted rates of sexual recidivism from VRS-SO scores, particularly when employed with Static-99R, were not significantly different from those observed among age cohorts; however, calibration was weaker for general violence. Implications for use of the VRS-SO in sexual recidivism risk assessment with older offenders are discussed.

Understanding the Latent Structure of Dynamic Risk: Seeking Empirical Constraints on Theory Development Using the VRS-SO and the Theory of Dynamic Risk

“…3We did not impose controls for offense history or age in the change-outcome individual predictor analyses. Olver et al (2020) found that age at release did not moderate VRS-SO change associations with sexual or violent recidivism; the youngest cohort of men (aged <30 years) registered significantly less change than men aged 40–49 years, otherwise there were no significant differences in change among age cohorts. The results of fixed-effects meta-analysis across age cohorts showed that there was no significant variability in effect size magnitudes for change associations with recidivism outcome.…”

Psychometric Properties of a Risk Tool Across Indigenous Māori and European Samples in Aotearoa New Zealand: Measurement Invariance, Discrimination, and Calibration for Predicting Criminal Recidivism

“…Using binary logistic regression in R, we first derived predicted absolute recidivism rates for each score on each subscale by subgroup. Typically (for example, see Gonc xalves et al, 2020;Grego´rio Hertz et al, 2021;Leguı´zamo et al, 2017;Olver et al, 2021), researchers calculate expected/observed (E/O), or predicted/expected indices (Hanson, 2017). These indices are an effect size representing the difference between (a) the number of observed (or predicted, if derived through logistic regression) recidivists and (b) the number of expected (usually derived from tool norms) recidivists.…”

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

“…There are further considerations that may enhance applications of the calculator to account for other factors that may mitigate or exacerbate risk, including (a) advancing age and (b) protective factors. Olver, Beggs Christofferson, et al (2020), in their examination of the predictive properties of the VRS-SO as a function of age, found reasonable agerelated calibration for sexual recidivism projections for older men (i.e., age 60+). Calibration metrics such as the E/O index, however, demonstrated mild overprediction of 5-and 10-year sexual recidivism, compared to the actual observed rates in that subgroup, although use of the Static-99R-VRS-SO model had improved calibration for older men over a VRS-SO only model.…”