The Accuracy of Predictions of Violence to Others

摘要: Objective. —To assess the accuracy of clinicians in predicting violence mental patients. Specifically, to determine if can predict when variation rates attributable age, race, and sex is controlled. Design. —Two samples psychiatric patients, matched on sex, admission status, were followed up community during a 6-month period. One group included individuals assessed by emergency department as likely be violent another person follow-up period; other was comparison group. Patients provided self-reports incidents, "collateral," ie, an individual with detailed knowledge patient's life, this same information. Official records also reviewed. Setting. —Patients recruited metropolitan hospital. collaterals interviewed their homes or public places community. Other Participants. —A consecutive sample coming into daylight evening shifts obtained. A total 2452 patients approached for consent 1948 consented. final 357 whom included. Main Outcome Measures. —Patients', collaterals', official records' reports incidents which patient laid hands threatened someone weapon. Results. —Violence period reported approximately 45% cases: 36% 53% cases predicted violent. Overall clinical significantly better than chance, but predictions female patients' not chance. Conclusions. —The level using collateral higher has been obtained methods. Clinical judgment adds predictive accuracy, overall modest particularly low ( JAMA . 1993;269:1007-1011)