Unveiling Adolescent Suicidality: Holistic Analysis of Protective and Risk Factors Using Multiple Machine Learning...

Area of Science:

• Psychiatry and Mental Health
• Computational Social Science
• Public Health

Background:

• Adolescent suicide attempts are a growing public health crisis.
• Existing machine learning models for suicide risk lack population representativeness and fail to integrate protective factors or established suicide theories.
• Stacked ensemble algorithms, suitable for low-prevalence conditions, have not been evaluated for adolescent suicide risk assessment.

Purpose of the Study:

• To compare the performance of a stacked ensemble algorithm against other machine learning models for adolescent suicide attempt risk assessment.
• To conduct a holistic item analysis to identify both risk and protective factors for adolescent suicide.
• To evaluate the compatibility of identified factors with the Interpersonal Theory of Suicide and the Strain Theory of Suicide.

Main Methods:

• A population-representative dataset of 173,664 Norwegian adolescents (aged 13-18) was analyzed.
• Five machine learning algorithms, including a stacked ensemble model, were trained to predict suicide attempts.
• Exploratory factor analysis was used to identify risk and protective factors and assess theoretical compatibility.

Main Results:

• The stacked ensemble model significantly outperformed other algorithms, achieving 90.1% specificity and 67.5% AUCPR.
• Recent self-harm was the strongest predictor across all models.
• Five additional risk domains were identified: internalizing problems, sleep disturbance, disordered eating, lack of future optimism, and victimization.

Conclusions:

• Stacked ensemble algorithms show promise for improving adolescent suicide risk assessment in low-prevalence conditions.
• The identified risk factors provide stronger support for the Interpersonal Theory of Suicide, suggesting an enhancement to the theory.
• A more comprehensive approach integrating risk and protective factors is crucial for understanding and preventing adolescent suicidal behavior.