Version 0.1 – March 2019
OxRec is a web-based risk calculator for assessing risk of recidivism in prisoners. This tool was developed by the Forensic Psychiatry and Psychology group at the University of Oxford and is available on the OxRisk.com website.
How was it developed?
OxRec was derived and externally validated in Sweden, using national population register data for all 47 326 prisoners released between 2001 and 2009. Potential predictive variables for inclusion in OxRec were drawn from carefully reviewing the evidence on criminal history, sociodemographic and clinical factors, and considering only those that were both reliable risk factors, and relatively easy to score.
First, the OxRec model was derived using the data from 37 100 prisoners, and 14 mostly routinely collected criminal history, sociodemographic and clinical factors were included in the final model. Then, the model was externally validated in a distinct group of 10 226 prisoners to establish its performance in a new sample. The model generates a simple risk calculator (OxRec) that generates two outputs – categories (low, medium, high) and probability scores (%) based on 1 and 2 year violent and any reoffending.
The research behind OxRec has been peer-reviewed and published in the Lancet Psychiatry journal (Open Access). OxRec has also been externally validated in the Netherlands, with a few predictors adapted to account for local needs. Other validations are planned in Australia, Canada, Finland, Scotland, Sweden, and the US. For a full bibliography related to OxRec, visit the OxRisk.com website.
When should OxRec be used for?
OxRec is designed to predict violent reoffending within 1 and 2 years in two populations – released prisoners and, in the Netherlands, it has been adapted for use with probationers.
Who should use OxRec?
Criminal justice professionals and healthcare staff can use the OxRec tool, as scoring to specific cases requires interpretation of some of the predictor variables. Individuals who use OxRec are encouraged to read other background literature (including the published papers about OxRec, all freely available on OxRisk.com) and any training videos developed locally within individual countries.
How should the outputs be interpreted?
The outputs from the OxRec calculator include probability scores for violent reoffending within 1 year and 2 years. A probability of 10% for 2 years means that, on average, 1 in 10 individuals with the same predictors will be violent over the next 2 years. This is a similar interpretation to the chance of rain in a weather app – so a 50% chance of rain at a certain time will mean that on similar days as today, there is a 1 in 2 chance of rain.
The calculator also provides a risk category, based on pre-specified risk levels identified on the calculator (and which are different for 1 and 2 year risk).
How should each item be scored?
See item descriptions on the scoring sheet. The best and most recent source of information should be used if possible. Some variables can be scored as unknown, and OxRec will display risk ranges to reflect this uncertainty.