Résumé
Few studies have examined the patterns of change in risk-relevant propensities among sex offenders. Given that existing evidence suggests that offenders’ risk of reoffending can change across time, the ability to detect change in measures of risk-relevant propensities would be useful for release, management, and treatment decisions. Of the studies that have examined change in such factors among sex offenders, researchers assumed that the observed scores were without error, making it impossible to disentangle the effects of true change from increased reliability. To date, no studies have examined whether this assumption is correct. The current dissertation advances knowledge by examining change in sexual offending risk using statistical models that separate the effects of measurement error from true change. Two studies were conducted, both using repeated assessments of the Acute-2007 scores for sex offenders on community supervision (Hanson et al., 2007). In the first study, the temporal stability of the factor structure of the Acute-2007 was examined across three time-points for 317 sex offenders. The Acute-2007 met the assumption of measurement invariance and, as such, observed changes were attributed to true change on the risk-relevant propensities assessed by the Acute-2007. The factor analyses supported a one-factor model; however, additional analyses found that two factors (i.e., Approach and Collapse factors) were differentially related to other similar risk measures and recidivism. In Study 2, sex offenders were found to decrease on risk-relevant propensities during the course of community supervision. Higher risk offenders were found to change less than lower risk offenders on the Acute-2007 total scores after controlling for initial scores. The rate of offender change also significantly differed across supervision officers on behavioural indicators of intention to reoffend (Approach total scores) but not indicators of coping skills (Collapse total scores). Exploratory analyses found clear incremental effects of initial and reassessed scores for the prediction of any recidivism. The findings support the Acute-2007 as assessing dynamic, changeable risk-relevant propensities for offending. Additional studies would be required to identify the best methods of integrating change in risk assessments of sex offenders.