This study sought to: a) ascertain the effect on rates of violence by varyi
ng its operational definition and b) compare characteristics of violent and
nonviolent patients. Aggressive behavior was recorded daily for every pati
ent (N = 78) during a 2-year period. Standardized rating scales were used t
o rate psychopathology and functioning. Almost two thirds of patients were
aggressive to others, and 26% violently assaulted another person. Official
incident reports Underestimated rates of violence to others, self-harm, and
property damage. Multivariate predictive models that greatly improved accu
racy over base rates showed that violent patients tended to be female, schi
zophrenic (nonparanoid type), and abusive of alcohol before admission. Viol
ence is more common in treatment resistant psychotic inpatients than sugges
ted by incident reports. Standardized definitions of violence are urged in
order to accurately study its prevalence and correlates. Models combining b
oth historical/demographic and clinical data may enhance prediction of viol
ence.