Severity Classification of Anxiety and Depression Using Generalized Anxiety Disorder Scale and Patient Health...

Area of Science:

• Mental Health
• Psychometrics
• Computational Statistics

Background:

• Scalable mental health screening is vital, particularly in low- and middle-income countries (LMICs).
• Existing tools like the Generalized Anxiety Disorder Scale (GAD-7) and Patient Health Questionnaire (PHQ-9) face feasibility challenges in large-scale use.
• Shorter versions (GAD-2, PHQ-2) reduce burden but lack symptom diversity.

Purpose of the Study:

• To optimize anxiety and depression screening using Classification and Regression Trees (CART).
• To identify concise, high-performing decision rules from GAD-7 and PHQ-9 items.
• To validate the reproducibility of these rules across independent datasets.

Main Methods:

• A large-scale cross-sectional study of 20,585 Brazilian adults using digital outreach.
• Anxiety and depression assessed via GAD-7 and PHQ-9.
• CART models trained and tested on bootstrapped samples with 10-fold cross-validation and hyperparameter tuning.

Main Results:

• CART models generated concise rules (2 items for GAD-7, 3 for PHQ-9) with high accuracy for minimal/mild and severe cases (AUC > 0.900).
• Sociodemographic variables were not included in the final classification paths.
• Models demonstrated stable performance and reproducibility across 5 independent datasets.

Conclusions:

• CART models provide simplified, symptom-specific pathways for anxiety and depression severity stratification.
• These rule-based models are efficient alternatives to fixed short forms, preserving symptom diversity.
• Findings support adaptive, cost-effective screening models for resource-limited settings and LMICs.