Acute health and social services for children and adolescents often struggle with youth aggression and violence. Early identification of violence risk during institutional stay can help prevent violent incidents. As such, this study assessed the predictive accuracy of the Violence Risk Assessment Checklist for Youth (V-RISK-Y) aged 12–18 in two different juvenile settings providing 24-hour services for youth. Institutions were included from child and adolescent inpatient psychiatry and residential youth care under child protective services.

A prospective, naturalistic observational study design was employed. V-RISK-Y was administered for youth admitted to four acute inpatient psychiatric units and four acute residential youth care institutions. Incidents of violence and threats during the youth’s stay were registered by institutional staff. In total, 517 youth were included in analyses, 59 of whom were registered with at least one incident of violence or threats during their stay. Area under curve (AUC) and logistic regression analyses were used to assess predictive accuracy and validity of V-RISK-Y.

For the overall sample, V-RISK-Y had good predictive accuracy, and the sum score of V-RISK-Y significantly predicted registered violent incidents. Stratified analyses indicated good predictive accuracy of V-RISK-Y for the inpatient units, but not for the residential youth care institutions.

Findings imply that V-RISK-Y is accurate in identifying violence risk for youth admitted to inpatient psychiatric units but has limited predictive accuracy in residential youth care institutions. Future research should explore approaches to correctly identify violence risk in residential care settings.

Instruments for evaluating the risk of violence towards others have mostly been developed for assessment of risk for recidivism into violent crime in forensic psychiatry. In general psychiatry there is a considerable need for specialised, brief and structured assessment tools to inform risk decisions.

The study aimed to validate a brief structured clinical risk assessment screen of inpatient violence (V-RISK-10), a 10-item structured clinical checklist with a good vignette-based interrater reliability (ICC=0.87). In this study it was used for risk assessment of a one-year sample of patients (N = 1.017) admitted to two acute psychiatric units. Risk assessments at admission were compared to prospective records of aggressive and violent acts during the hospital stay.

Results showed a base rate for aggression of 9%. The predictive validity of the V-RISK-10 was estimated by Receiver Operating Characteristics (ROC). It yielded an area under the curve (AUC) of 0.83, with sensitivity/specificity of 0.81/0.73 and corresponding positive and negative predictive values (PPV/NPV) of 0.24/0.97. The screen was easy-to-use and showed a short completion time.

Despite promising results further validation studies are required before the V-RISK-10 is adopted into routine clinical practise.