SHORT-TERM CLINICAL-PREDICTION OF ASSAULTIVE BEHAVIOR - ARTIFACTS OF RESEARCH METHODS

Objective; The apparent accuracy of predictions of assaultive behavior in psychiatric inpatients varies substantially, depending on the meth od used to study the prediction. The authors explored the effects of d ifferent measures and sampling strategies on short-term clinical predi ctions of dangerousness. Method: The index subjects were patients who were rated by intake clinicians as potentially highly assaultive on th e ward (N=32) and patients who were involuntarily committed on grounds of danger to others (N=32). The respective comparison groups comprise d patients predicted by clinicians not to be assaultive (N=32) and pat ients committed for reasons other than danger to others (N=40). The te xt of unit meetings and data from chart reviews were used to determine the occurrence and dates of violent acts, seclusions for violent acts or threats, and violent threats. Results: There was a significant dif ference in the rate of inpatient violence between the subjects rated a t admission as potentially assaultive (75.0%) and patients rated as no t potentially assaultive (12.5%), but the difference in the rates of v iolence between the patients who were (56.0%) and were not (42.0%) inv oluntarily committed as dangerous to others was not significant. Most of the violent acts occurred relatively late in the hospitalization, b ut seclusions occurred almost exclusively in the initial stages of hos pitalization. Conclusions: The reported accuracy of clinical predictio ns of assaultive behavior is markedly affected by the choice of sampli ng strategy, comparison group, outcome measures, and follow-up period. Including seclusion and violent threats in the outcome variable appea rs to lead to deceptive findings.