Prediction of Suicide in Psychiatric Patients

TLDR

The study aimed to identify patients who would later commit or attempt suicide. The researchers examined 4,800 consecutive Veterans Administration inpatient psychiatric patients using a broad array of instruments previously linked to suicide risk. Although many measures correlated with later suicide, none reliably identified individuals, as all models had low sensitivity, high false‑positive rates, and the low base rate made prediction infeasible.

Abstract

• A prospective research study attempted to identify persons who would subsequently commit or attempt suicide. The sample consisted of 4,800 patients who were consecutively admitted to the inpatient psychiatric service of a Veterans Administration hospital. They were examined and rated on a wide range of instruments and measures, including most of those previously reported as predictive of suicide. Many items were found to have positive and substantial correlations with subsequent suicides and/or suicide attempts. However, all attempts to identify specific subjects were unsuccessful, including use of individual items, factor scores, and a series of discriminant functions. Each trial missed many cases and identified far too many false positive cases to be workable. Identification of particular persons who will commit suicide is not currently feasible, because of the low sensitivity and specificity of available identification procedures and the low base rate of this behavior.