Identifying potential offenders on the basis of police records: development and validation of the ProKid risk assessment tool

Purpose - The purpose of this paper is to describe the development and validation of the ProKid risk assessment tool, which was designed to enable Dutch police officers to identify youths with elevated risk of committing violent and/or property crimes. Design/methodology/approach - The ProKid algorithms were based on the results of logistic regression analyses of police data from a sample of 31,769 adolescents in the former police regions "Amsterdam-Amstelland" and "Gelderland-Midden". For the validation, logistic regression analyses were performed on police data of youths in the police region "Amsterdam-Amstelland" who had been in contact with the police in 2011 (n = 39,977). Furthermore, receiver operating characteristic analyses were performed to assess the instrument's accuracy. Findings - Results indicated that higher ProKid risk categories were associated with greater odds of being registered as a suspect of either a violent or property offence. The instrument was found to have good predictive accuracy. Area under the curve values ranged from 0.83 to 0.84 for violent offences and from 0.82 to 0.83 for property offences. Practical implications - The current study demonstrates that ProKid is a valid and accurate tool to be used by police officers to identify youths with elevated risk of future violent and property offending. The automated assessment procedure enables a quick screening of large numbers of both non-offenders and offenders. This study confirms the value of official police records for assessing the risk of future offending for preventive purposes. Originality/value - The present study demonstrates the validity of a risk assessment tool based on Dutch police records for both non-offenders and offenders.