We will use linked administrative crime and education data, including prison discharge and population datasets, focusing on individuals released from prison between January 2008 and December 2021. This linkage enables external validation of a novel violent recidivism risk assessment tool, which could be applied to most of the prison population in England. Violent reoffending will be defined using Home Office offence codes, consistent with the original Swedish risk assessment development study. Key variables include: date of release, offence history and type, age at offence, custodial sentence length, highest education level, gender, alcohol and drug use, employment prior to imprisonment (yes/no), relationship status, previous income (stable/low), and history of mental disorder. The model will be validated by applying regression coefficients from the original Oxford Risk of Recidivism (OxRec) model to predict 1- and 2-year violent reoffending risk in this new dataset, following the recommended three-step incremental approach. This Administrative data UK funded research fellowship began in October 2024, with data access granted in January 2025. In the talk, I will present findings on OxRec’s predictive performance using calibration and discrimination measures at predefined risk thresholds, and assess whether education-related factors, such as school exclusion enhance predictive accuracy. These findings aim to improve violence risk assessment tools and support informed professional judgment and decision-making within the Criminal Justice System.