Related Experiment Video
Updated: May 31, 2026
07:31
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
Improving the accuracy of suicide attempter classification.
David Delgado-Gomez1, Hilario Blasco-Fontecilla, AnaLucia A Alegria
1Department of Statistics, Universidad Carlos III, Getafe, Spain. ddelgado@est-econ.uc3m.es
Artificial Intelligence in Medicine
|June 24, 2011
Summary
The International Personality Disorder Evaluation Screening Questionnaire (IPDE-SQ) shows better predictive ability for suicide attempts than the Barratt
Area of Science:
- Psychiatry
- Psychometrics
- Machine Learning
Background:
- Psychometric questionnaires like the Barratt's Impulsiveness Scale version 11 (BIS-11) are used to assess suicidal behavior.
- Traditionally, all items in such scales are considered equally important, but this study questions that assumption.
Purpose of the Study:
- To evaluate the discriminative ability of the BIS-11 and the International Personality Disorder Evaluation Screening Questionnaire (IPDE-SQ) in predicting suicide attempt (SA) status.
- To compare the effectiveness of various classification techniques, including Support Vector Machines (SVM), in analyzing questionnaire data.
- To identify individual items within both scales that possess the strongest discriminative capacity.
Main Methods:
- The study involved 879 participants, including suicide attempters, healthy blood donors, and psychiatric inpatients.
- Data from the BIS-11 and IPDE-SQ were analyzed using multiple classification techniques: Boosting, SVM, linear discriminant analysis, Fisher linear discriminant analysis, and a traditional psychometric approach.
- The classification performance of these techniques was compared to assess their accuracy in predicting SA status.
Main Results:
- The most discriminative item for the BIS-11 was "I am self controlled" (Item 6), and for the IPDE-SQ, it was "I often feel empty inside" (Item 40).
- Support Vector Machines (SVM) achieved a classification accuracy of 76.71% for the BIS-11 and 80.26% for the IPDE-SQ.
- The IPDE-SQ demonstrated superior discriminative ability compared to the BIS-11 in classifying suicide attempter status.
Conclusions:
- The IPDE-SQ items are more effective than BIS-11 items in distinguishing individuals who have attempted suicide.
- The IPDE-SQ, particularly when analyzed with SVM, provides better classification results for suicide attempter and non-attempter status.
- Support Vector Machines (SVM) proved to be the most effective classification technique for both questionnaires in this study.
