2 57 PREDICTIVE PROPERTIES DYNAMIC INSTRUMENTS scores on these instruments. Our findings show that dynamic risk assessment instruments indeed significantly contribute to the prediction of sexual, violent, and any recidivism. In line with Hanson and Morton-Bourgon’s (2009) observations, effect sizes were largest for the type of outcome that the instruments were primarily designed to assess, in this case sexual recidivism. The effect sizes we found for sexual and violent (including sexual) recidivism (d = 0.71; CI [.63, .79] and d = 0.43; CI [.29, .57], respectively) are similar to the mean effect sizes reported by Hanson and Morton-Bourgon (2009) for predominantly static actuarial instruments (d = 0.67; CI [.63, .72] and d = 0.51; CI [.47, .56], respectively). The effect size for any recidivism we found in this study (d = .64; CI [.56, .72]) is higher than the mean effect size for the actuarial overall static risk assessment instruments reported by Hanson and Morton-Bourgon (d = .52, CI [.48, .56]). The relatively small effect size in our meta-analysis for violent (including sexual) recidivism is noteworthy and coincided with great variability in effect sizes among studies predicting violent recidivism and with great variability in the specific (dynamic) instruments used. The latter finding suggests that some dynamic risk assessment instruments may be better at assessing risk for violent sexual offending than others. The current study aimed to investigate the value of dynamic risk assessment in general. For those studies that included more than one dynamic risk assessment instrument, the results of the various instruments were averaged. This makes it impossible to compare and evaluate the predictive properties of individual dynamic risk assessment instruments. Future meta-analyses could take this as their focus. The incremental validity of dynamic over static risk assessment was established for all outcome measures. However, effect sizes tended to be small, which suggests that static and dynamic instruments overlap, at least when it comes to their predictive value. This outcome underpins the question that has been raised by Ward and Beech (2015), whether dynamic risk factors perhaps measure correlates of underlying propensities, like static risk factors do, but in different ways. That is, the possibility should be considered that dynamic risk factors measure clinical features that are associated with the psychological propensities that cause recidivism, instead of assessing these propensities themselves. We also found significant effects for change scores. Statistically correcting for both static and initial dynamic risk scores, change scores significantly predicted all three types of recidivism, indicating that men with a history of offenses who showed larger positive changes (reflecting a reduction in dynamic risk scores) recidivate at lower rates than those who show a smaller change in dynamic risk factors. Much like our findings regarding overall and incremental predictive validity, effects were the largest for sexual recidivism. However, all effect sizes were relatively small. Thus, only a small part of change in recidivism was explained by changes in dynamic risk factors. This raises questions on the nature of the variables currently included in dynamic risk assessments.
RkJQdWJsaXNoZXIy MjY0ODMw