2 53 PREDICTIVE PROPERTIES DYNAMIC INSTRUMENTS the categorical moderators. Given the fact that there were only eight unique samples available, we did not conduct a meta-regression analysis for the three continuous moderators. Table 2.2. Effect sizes for research question 2: Incremental validity of dynamic over static risk assessment instruments by recidivism type (van den Berg et al., 2018, p.185) Recidivism Type Fixed-Effect Random-Effect Q I2 (%) N K L Hazard ratio 95% CI Hazard ratio 95% CI Sexual 1.08 [1.06, 1.10] 1.09 [1.06, 1.12] 18.3 34.3 3,747 19 13 Violent 1.06 [1.04, 1.08] 1.06 [1.04, 1.08] 11.8 15.0 9,019 15 11 Any 1.04 [1.03, 1.06] 1.05 [1.02, 1.08] 16.3 57.1 2,511 11 8 Note. K = Number of studies; L = Number of unique samples Values in bold are statistically significant. p < .05 2.4.3 PREDICTIVE VALIDITY OF CHANGE SCORES Only the studies reporting predictive validity of change scores on dynamic risk factors were included in this step of the analyses. The average follow-up interval was 68.6 months (range = 36–140) for studies predicting sexual recidivism, 51.3 months (range = 36–112) for studies predicting violent (including sexual) recidivism and 72.6 months (range = 36–112) for studies predicting any recidivism (sexual, violent, or any). Observed recidivism rates were 9.0% for sexual recidivism (6 samples, n = 2,043), 9.1% for violent (including sexual) recidivism (6 samples, n = 6,234), and 33.7% for any recidivism (4 samples, n = 1,466). Table 2.3 summarizes the effect sizes of the predictive validity of change scores on dynamic risk assessment instruments developed for adult male sex offenders on the three recidivism types. Effect sizes for the individual samples are presented in the Appendix B3 (available in the online supplemental material). 2.4.3.1 Predictive validity of change scores for sexual recidivism In the overall analysis for sexual recidivism (6 unique samples), the fixed-effect weights of the individual studies (the inverse of the variance) varied between 13.16 and 142.86, with a median value of 22.22. The study producing the largest weight involved a sample of 572 men (recidivism rate was 6.2%) and had more than twice the weight of the next largest study (25.64) and more than 10 times the weight of the study with the smallest weight. To reduce the influence of this study, its sample size was artificially decreased to 208, so that its weight was not more than twice the next largest study weight. After adjusting this weight, the fixed-effect weighted d for change scores on dynamic risk assessment instruments in predicting sexual recidivism was 0.26, 95% CI [0.10, 0.42], N = 2,043, based on 9 studies representing 6 unique samples. The Q indicated that the variability across studies was not significant (Q = 5.8, p = .33). The I2 (13.8%) indicated
RkJQdWJsaXNoZXIy MjY0ODMw