A data science approach to optimize ADHD assessment with the BRIEF-2 questionnaire

Area of Science:

• Neurodevelopmental Disorders
• Clinical Psychology
• Psychometrics

Background:

• Attention deficit hyperactivity disorder (ADHD) is a common neurodevelopmental disorder presenting diagnostic challenges.
• Early and accurate diagnosis is critical for effective intervention and management of ADHD.
• The Behavior Rating Inventory of Executive Function-2 (BRIEF-2) is a widely used tool for assessing executive function deficits in children.

Purpose of the Study:

• To evaluate the efficacy of various feature selection techniques in identifying key BRIEF-2 items for ADHD assessment.
• To determine if a parsimonious set of BRIEF-2 items can accurately predict ADHD-related behaviors.
• To enhance the efficiency of ADHD diagnosis through data-driven item selection.

Main Methods:

• Utilized feature selection methods (Lasso, decision trees, random forest, extreme gradient boosting, forward stepwise regression) on BRIEF-2 data from 59 children undergoing ADHD assessment.
• Collected responses from mothers, fathers, and teachers using the BRIEF-2.
• Performed repeated validation analysis to assess predictive accuracy.

Main Results:

• Lasso emerged as the optimal feature selection technique, balancing accuracy and interpretability.
• Lasso achieved an average positive correlation exceeding 0.5 between informant-reported inattention/hyperactivity scores and predicted scores.
• Effective prediction was accomplished using only approximately 18% of the BRIEF-2 items.

Conclusions:

• Variable selection techniques, particularly Lasso, can significantly improve the efficiency and accuracy of ADHD assessment using the BRIEF-2.
• A reduced set of BRIEF-2 items can reliably characterize a child's condition, valuable in time-constrained clinical settings.
• This approach enhances the comprehension of ADHD and supports earlier diagnostic processes.