노인의 주관적 건강 관련 요인의 변화: 18년 추적 자료와 머신러닝 분석

Longitudinal Changes in Factors Associated with Self-Rated Health in Older Adults over 18 Years: A Machine Learning Approach

Background: This study investigated key determinants of older adults’ self-rated health (SRH) and how their relative influence has evolved over time. Methods: Temporal changes in the relationships among SRH-related factors were examined using 18 years of longitudinal data. First, a network-tree model revealed gradual, time-dependent shifts in the correlation structure surrounding SRH. Then, random forest models were applied to assess changes in the predictive importance of each factor. Results: Variables reflecting physical function, emotional well-being, and social participation consistently influenced the prediction of SRH. Although physical function indicators, such as the presence of chronic disease, were among the strongest predictors, their influence has diminished in recent years. In contrast, self-efficacy exhibited a steady increase in predictive importance. Conclusion: These findings offer new insights into the evolving determinants of perceived health in older adults and inform potential directions for health-promotion policy and intervention design.