Zur rückfallprognostischen Bedeutung des Tatverhaltens bei Sexualdelinquenz

The study examined the possibility of predicting sexual recidivism within a period of 5 years in a sample of 612 male sex offenders by means of behavioural crime scene variables. A total of 20 variables were selected based on a significant bivariate correlation with sexual recidivism. Multivariate analysis was conducted to determine combined predictive accuracy. Step-wise reverse logistic regression extracted 12 variables explaining 22% of total variance in sexual recidivism. The unweighted sum score of these variables showed considerable predictive power (AUC?=?0.76; r?=?0.34). Findings were robust not only for different subgroups of sexual offenders, but also when cross-validated on an independent sample. In addition, the unweighted sum score showed significant incremental validity above and beyond common risk measures [Static99, Sexual Violence Risk- (SVR-)20], both in the original and the cross-validation sample. Thus, the potential of crime scene analysis in predicting sexual recidivism has been clearly demonstrated. Future research should provide a more accurate access to psychologically meaningful interpretations of the risk immanent in crime scene actions by adapting risk scales to certain subgroups of perpetrators.