Current approaches to stratify patients with psychosis into risk groups are limited by inconsistency, variable accuracy, and unscalability.

This paper will present an overview of current approaches based on a systematic review. It will also present a novel scalable approach based on a total national cohort of 75 158 Swedish individuals aged 15–65 with a diagnosis of severe mental illness (schizophrenia, schizophrenic-spectrum, bipolar disorder, and other psychotic illnesses). We developed predictive models for violent offending through linkage of population-based registers and tested them in external validation. We measured discrimination and calibration for prediction of violent offending at 1 year using specified risk cut-offs.

: A 16-item model was developed from pre-specified routinely collected criminal history, socio-demographic and clinical risk factors. In external validation, the model showed good measures of discrimination (c-index 0.89) and calibration. For risk of violent offending at 1 year, using a 5% cut off, sensitivity was 64% and specificity was 94%. Positive and negative predictive values were 11% and 99%, respectively. The model was used to generate a simple web-based risk calculator (OxMIV).

We have developed a prediction score in a national cohort of all patients with psychosis that can be used as an adjunct to decision-making in clinical practice by identifying those who are at low risk of future violent offending and higher risk individuals who may benefit from additional risk management. Further evaluation in other populations and countries is needed.