| Summary: | Purpose: This study aimed to identify predictors of health-related quality of life (HRQoL) among adults with multimorbidity using Andersen's behavioral model and to develop a predictive model with a random forest(RF) algorithm. Methods: We conducted a secondary cross-sectional analysis of the 2019-2020 Korea National Health and Nutrition Examination Survey (KNHANES) including 858 adults with two or more chronic conditions. Predictors were grouped as predisposing, enabling, need, and health-behavior factors. HRQoL was measured with the EuroQol five-dimension index (EQ5D) value based on the Korean tariff. Model performance was evaluated using mean absolute error, root mean squared error, and the coefficient of determination. Variable importance and Shapley additive explanations (SHAP) were used for interpretation. Results: The mean HRQoL was 0.90 (standard deviation 0.14). The model achieved a mean absolute error of 0.08 and a coefficient of determination of 0.15. The strongest predictors were subjective health status, days of walking per week, age, monthly income, and private medical insurance; Shapley analyses confirmed their direction and relative influence. Conclusion: Andersen's model provided a coherent framework to interpret determinants of HRQoL in multimorbidity. The findings support routine assessment of subjective health, promotion of walking, and resource linkage for older adults and individuals with limited socioeconomic resources. Interpretable machine learning may aid early risk stratification and tailored nursing interventions.
|
|---|