Using Machine Learning Algorithms to Enhance the Management of Suicide Ideation

Area of Science:

• Psychiatry
• Computer Science
• Public Health

Background:

• Combat veterans with mental health conditions are at high risk for suicidal ideation and behavior.
• Existing methods for identifying suicide risk lack predictive accuracy in treatment-seeking populations.
• Machine learning (ML) offers a novel approach to analyze complex health data for risk prediction.

Purpose of the Study:

• To apply machine learning algorithms to predict suicidal ideation (SI) in a treatment-seeking veteran population.
• To identify key variables associated with SI risk using pattern recognition.
• To explore the potential of ML as a screening tool for clinicians.

Main Methods:

• Utilized questionnaire data from 738 patients (veterans, Canadian Forces, RCMP).
• Employed machine learning (ML) pattern recognition methods to analyze multivariate data.
• Identified patterns associated with suicidal ideation.

Main Results:

• Achieved over 84.4% accuracy in predicting SI using 25 variables.
• Obtained 81% accuracy with as few as 10 variables, primarily from the Patient Health Questionnaire (PHQ).
• Quality of health emerged as a stronger predictor of SI than occupational experiences.

Conclusions:

• ML can accurately predict SI in veterans, with potential for broader application to the general population.
• Patient Health Questionnaire data is crucial for identifying SI risk.
• ML-assisted screening tools could significantly aid clinicians in managing suicide risk.