Predicting the First Onset of Suicidal Thoughts and Behaviors in Adolescents Using Multimodal Risk Factors: A 4-Year...

Area of Science:

• Adolescent psychiatry
• Machine learning in healthcare
• Developmental psychology

Background:

• Suicidal thoughts or behaviors (STB) are a leading cause of death in youth globally.
• Previous studies predicting STB onset had limitations in data modalities and population focus (adults).

Purpose of the Study:

• To prospectively predict the first onset of STB in adolescents over a four-year period.
• To utilize both a pre-existing STB classification model and a novel machine learning model incorporating 195 biopsychosocial features.

Main Methods:

• Utilized data from 7,503 adolescents (aged 9-11 at baseline) from the Adolescent Brain Cognitive Development (ABCD) project.
• Applied an existing STB history classification model and trained a new elastic net logistic regression model with 195 features.
• Validated the top 15 features from the new model across independent sites.

Main Results:

• Both models demonstrated moderate predictive accuracy for first-onset STB (AUCs ranging from 0.63 to 0.73).
• Consistent predictors included female sex, sleep disturbances, and maladaptive home and school environments.
• Novel neurocognitive and brain imaging risk factors were also identified.

Conclusions:

• The developed models offer moderate accuracy in predicting first-onset STB in adolescents.
• The study reinforces known psychological risk factors and introduces new potential neurobiological markers.
• Further validation in diverse, large-scale samples is recommended prior to clinical application.