Machine learning prediction of depression in culturally diverse families: Findings from the Korea Community Health Survey

Background Although South Korea's overall population is declining, the number of culturally diverse families is increasing. Depression in these families is a significant factor contributing to rising social costs and hindering social integration. Methods To predict depression in culturally diverse families in South Korea, we analyzed 131 independent variables from 2,568 culturally diverse families who participated in the 2023 Korea Community Health Survey. Results We identified 15 key predictive variables and evaluated their effects using the XGBoost model, which outperformed 5 other machine learning models. Stress recognition, experience of extreme sadness or despair, subjective health status, age, and frequency of contact with neighbors emerged as significant predictive factors. Conclusions By conducting a comprehensive analysis of multidimensional indices, this study offers a multifaceted perspective on depression in culturally diverse families.