Development of a Short-Form Hwa-Byung Symptom Scale Using Machine Learning Approaches

Area of Science:

• Psychiatry and Behavioral Sciences
• Computational Psychology
• Korean Medicine

Background:

• Hwa-byung (HB) is a culture-bound syndrome primarily observed among Koreans, often termed "anger syndrome" or "fire illness".
• Existing assessment tools for HB may lack efficiency for widespread clinical use.

Purpose of the Study:

• To develop a short-form Hwa-byung symptom scale using machine learning techniques.
• To create an efficient and accurate assessment tool for Hwa-byung.

Main Methods:

• Exploratory Factor Analysis (EFA) was used to identify core symptom dimensions.
• Machine learning models including XGBoost, Logistic Regression, Random Forest, SVM, Decision Tree, and Multi-Layer Perceptron were employed.
• A survey of 500 Korean adults using the original 15-item HB scale provided the dataset.

Main Results:

• EFA identified two factors: psychological symptoms and somatic manifestations of HB.
• A two-item scale (Q3 and Q10) demonstrated high predictive power for HB presence.
• Machine learning models achieved high accuracy (around 90%) and discriminative ability (AUC 0.9436-0.9579), with Multi-Layer Perceptron performing best.

Conclusions:

• Integrating EFA and AI/machine learning is effective for developing practical assessment tools.
• The study offers a validated, efficient two-item scale for Hwa-byung assessment.
• This methodological approach can inform the development of efficient assessments in Korean medicine and beyond.