Related Experiment Videos
Suicide risk prediction by computer interview: a prospective study
H P Erdman1, J H Greist, D H Gustafson
1Department of Psychiatry, University of Wisconsin, Madison.
The Journal of Clinical Psychiatry
|December 1, 1987
Summary
A computer program using Bayesian probability was evaluated for suicide risk assessment. While clinicians excelled at predicting non-attempters, the computer program demonstrated superior accuracy in identifying potential suicide attempters.
Area of Science:
- Psychiatry
- Computer Science
- Risk Assessment
Background:
- Suicide risk assessment is crucial in clinical practice.
- Subjective Bayesian probability models offer a quantitative approach to risk evaluation.
- Comparing computational models with clinical judgment is essential for improving diagnostic tools.
Purpose of the Study:
- To evaluate a computer interview program employing a subjective Bayesian probability model for suicide risk assessment.
- To compare the predictive accuracy of the computer program against that of clinicians.
Main Methods:
- A computer program utilizing a subjective Bayesian probability model was developed.
- Predictions for suicide risk were made by both the computer program and clinicians for 52 patients.
- Statistical analysis, including receiver operating characteristic (ROC) curve analysis, was used for comparison.
Main Results:
- The computer program was significantly more accurate in predicting suicide attempters (p = .001).
- Clinicians were significantly more accurate in predicting non-attempters (p = .01).
- ROC curve analysis indicated better overall discrimination by the computer, though the difference was not statistically significant.
Conclusions:
- The computer program shows promise as a tool for identifying suicide attempters.
- Clinicians remain valuable for identifying individuals at lower suicide risk.
- Further research may refine computational models for comprehensive suicide risk assessment.