Data-Driven Identification of Brain-Behavioral and Sociodemographic Predictors of Anxiety Severity in Children Using...

Area of Science:

• Neuroscience
• Developmental Psychology
• Machine Learning

Background:

• Childhood anxiety is common and linked to cognitive control issues and sociodemographic risks.
• The interplay between these factors in childhood anxiety remains poorly understood.
• A data-driven approach is needed to identify key predictors.

Purpose of the Study:

• To employ machine learning to pinpoint the most significant neural, behavioral, and sociodemographic predictors of childhood anxiety severity.
• To investigate interactions between these predictors.

Main Methods:

• 181 children (4-10 years) with anxiety data from prior studies were analyzed.
• Electroencephalogram (EEG) and behavioral data from a Go/NoGo task assessed cognitive control.
• Machine learning models (Random Forest, SVR, XGBoost) identified predictors and interactions.

Main Results:

• Random Forest achieved the highest prediction accuracy.
• Error positivity (Pe), single-mother status, frontal theta power, post-error accuracy, and occipital post-error alpha power were key predictors.
• Single-mother status moderated the effect of post-error alpha power on anxiety.
• Low frontal theta power amplified the association between Pe and anxiety.

Conclusions:

• A combination of cognitive control neural markers, behavioral performance, and sociodemographic factors, particularly single-mother status, significantly predicts childhood anxiety.
• Understanding these interactions is crucial for targeted diagnosis and treatment strategies.
• This study offers insights beyond traditional hypothesis-driven research in childhood anxiety.