Prediction of aggression in psychiatric patients: A review of prospective prediction studies

This review analyzes prospective studies on the prediction of aggression in psychiatric patients. Five aspects are explored: sample characteristics, setting and time perspective, predictor variables, criterion variables, and outcome. The current review reveals that prospective studies are scarce and that they constitute a hetereogeneous group concerning the five aspects explored here. Apart from findings reported in three of the studies, predictive validity in the reviewed studies overall was low. There was a systematic relationship between decreasing base rate values and increasing negative predictive values in the studies reviewed. The expected relationship between increasing base rate values and increasing positive predictive values was also confirmed. Attempts are made to relate predictive accuracy to the nature of predictors and follow-up records of aggression. Suggestions for improvements are presented.