Background: Structured risk assessment schemes can aid violence reduction in mental health and correctional settings. However, the properties and effectiveness of schemes for predicting imminent (within 24-h) violence have not been comprehensively reviewed.
Objectives: To systematically review the properties and predictive performance of structured schemes that have been used to predict imminent aggression.
Design: Systematic review and meta-analysis.
Data sources: Multiple electronic databases were searched, relevant reference lists and journals were hand-searched.
Review methods: Thirty-two structured risk assessment schemes were identified and used in a search for evidence about the prediction of imminent violence, or their psychometric properties when used for that purpose. Study findings were synthesised narratively and, where statistics about predictive validity were available, subjected to meta-analyses.
Results: Five schemes (Brøset Violence Checklist, HCR-20: Assessing Risk for Violence version 2.0 clinical subscale, Dynamic Appraisal of Situational Aggression and its variants for youth and for women) have been subject to relevant research. Fifty papers were included in a narrative review, and 31 comprising 26 samples involving 282,238 assessments for 2723 individuals were included in meta-analyses. Only the Dynamic Appraisal of Situational Aggression and the Brøset Violence Checklist were subject to implementation research in which the tools were largely rated as having utility for, and being acceptable to, user-participants. Many studies have not established interrater reliability of the tools in the field. After removing outliers, pooled effect sizes for prediction of aggression against people and for any aggression were, respectively: Hedges' g = 1.35[1.15-1.56]) and g = 1.24[0.99, 1.50] for the Brøset Violence Checklist and g = 1.04 [0.69, 1.39] and 0.88 [0.62, 1.15] for the Dynamic Appraisal of Situational Aggression. Moderator analyses revealed pooled effect sizes for both schemes were decreased by low study bias risk and increased by patient forensic vs. civil legal status.
Conclusions: Research supports the value of both the Brøset Violence Checklist and the Dynamic Appraisal of Situational Aggression. Both are generally acceptable and feasible to implement, and their predictive value is broadly equivalent producing large pooled effect sizes. There is less evidence about the Dynamic Appraisal of Situational Aggression. variants for women and youth though initial studies are promising.
Keywords: Aggression; Mental health; Meta-analysis; Psychiatry; Risk assessment; Systematic review; Violence.
Copyright © 2020. Published by Elsevier Ltd.