Predictors of incidents during short leave: An analysis of prisoner personal files in the Lower Saxony prison system

Abstract

Abstract This study aimed to investigate the prediction of incidents during short leave (ISLs) for inmates in the German prison system using static risk and protective factors that had been assessed at the start of the incarceration. Data from the personal files of inmates in Lower Saxony were analysed to examine the prevalence of ISLs and to assess the ability of the Offender Group Reconviction Scale – Version 3 (OGRS 3) to predict ISLs. In addition, random forest models were used to identify potential predictors that could further improve the prediction performance. The results showed that the OGRS 3 had significant predictive validity for drug abuse, late return, and any ISL but not for new offences during short leave. However, the OGRS 3 exhibited problems with false positive rates in its prediction of ISLs. The random forest models did not substantially improve the prediction of any ISL, but they did improve the prediction of drug abuse during short leave. This study highlights the importance of considering dynamic factors and using a more comprehensive approach in risk assessment for ISLs.