Suicide risk prediction: a new approach.

Abstract

A computer program that collects suicide risk factors by computer interview from persons with thoughts of suicide and processes the data to provide risk predictions has been written and pilot tested. Patient acceptance of the interviewing technique was good; more than half of the patients interviewed preferred the computer to a doctor as an interviewer. Bayes Theorem is used to process the data collected against a subjective data base. In a retrospective study comparing risk predictions made by the computer with predictions by experienced clinicians, the computer was more accurate in predicting suicide attempters (p .01) and slightly less accurate in predicting nonattempters. The program is economical, can be used wherever a telephone and a computer terminal are available, and is readily and uniformly modified to include new data.