The dynamic prediction of criminal recidivism: a three-wave prospective study.

A three-wave, prospective panel design was used to assess the extent to which static and dynamic risk factors could predict criminal recidivism in a sample of 136 adult male offenders released from Canadian federal prisons. Static measures were assessed only once, prior to release while dynamic measures were assessed on three separate occasions: pre-release, 1 month, and 3 months post-release. Recidivism was coded during an average of 10.2-month follow-up period (SD=19.2). A series of Cox regression survival analyses with time-dependent covariates and Receiver Operator Characteristic (ROC) analyses were conducted to assess predictive validity. Although the combined static and time-dependent dynamic model (AUC=.89, CI=.81-.93) significantly (p<.01) outperformed the pure static model (AUC=.81, CI=.73-.87) the confidence intervals did overlap to some extent. Implications for dynamic risk assessment and management are discussed.