A computer-based system for identifying suicide attemptors

Abstract

The following is a report of a collaborative study done by members of the University of Wisconsin-Madison Departments of Industrial Engineering and Psychiatry. The purpose of the study was to reduce the number of deaths due to suicide by providing the clinician with a well-defined set of characteristics with which to evaluate an individual thinking of suicide. Specifically, our model provides an estimate of the chance of death if a suicide attempt is made. Two models were developed to make this estimate. One model was Bayesian and the other was a linear regression model. The two models were evaluated by comparing the predictions made by the models and those made by two groups of clinicians who also predicted the outcome of 32 actual suicide attempts. Among other things, the study suggested that on the average both models were better able to predict the lethality of a suicide attempt than were the clinicians.