Proefschrift

2 43 PREDICTIVE PROPERTIES DYNAMIC INSTRUMENTS (e.g., dynamic risk factors, names of dynamic risk assessment instruments). In addition, papers were selected and retrieved using the reference sections of empirical studies and previous reviews, and authors of published dynamic risk assessment instruments were contacted and asked about possible other relevant studies and databases, whether published, submitted for publication, or, at that time, unpublished. 2.3.2 ELIGIBILITY All relevant studies published in English, German, French, or Dutch were included. Study design was not used as a criterion for inclusion or exclusion, although we included this information to test the impact of study design and publication bias by post-hoc statistical tests (moderator analyses). To be included in the analyses, samples needed to consist of adult men with a history of at least one sexual offense, and the risk assessment instrument used should contain only dynamic risk factors or at least include a subscale specifically addressing dynamic risk factors. To be included, studies also needed to include the assessment of the predictive validity of the total score of dynamic risk factors, which could involve the prediction of new charges, convictions, or the revocation of conditional release because of new offending behaviors. Following the recommendations of Hanson and Morton-Bourgon (2009), recidivism was operationalized in three ways: 1) sexual recidivism (vs. no recidivism or nonsexual recidivism) – this category includes possession of child pornography and other noncontact sexual offenses; 2) violent (including sexual) recidivism (vs. no recidivism or nonviolent recidivism); and 3) any recidivism (vs. no recidivism). 2.3.3 FINAL SELECTION OF STUDIES Out of an initial set of 148 potentially relevant studies, 58 met our inclusion criteria. A total of six studies were excluded because effect sizes were not presented in the original publication and could not be obtained from the authors. The data of the final set of 52 studies were used to address the three main questions of our meta-analysis (see Figure 2.1 for a flowchart of study selection). Some studies used partly or fully overlapping samples (e.g., because they repeated analyses after extending the follow-up period). For these cases, the data involving the longest follow-up period were included, but only if the sample size did not decrease by more than 30%. If the decrease in sample size was more than 30%, the data for the next and shorter follow-up period were used. 2.3.3.1 Included risk assessment instruments The studies that were part of this meta-analysis included a total of 14 dynamic and nine static risk assessment instruments (or subscales). The static instruments were used to assess incremental predictive validity of the scores on the dynamic instruments and/or to evaluate the predictive validity of change scores controlled for both static and initially dynamic scores.

RkJQdWJsaXNoZXIy MjY0ODMw