Individual and interaction effects of health determinants on health-related quality of life in Korean adults aged 50–81 years: A causal Bayesian network analysis

Health-related quality of life (HRQoL) reflects physical and mental well-being and is increasingly important in aging populations, yet traditional approaches often fail to capture the complex causal pathways among its determinants. We analyzed 2,566 adults aged 50-81 years from the Korean Genome and Epidemiology Study using the Short Form-12 (Physical Component Summary [PCS] and Mental Component Summary [MCS]). A causal Bayesian network was learned using the PC algorithm of Spirtes and Glymour with hierarchical constraints to ensure causal interpretability. We then estimated the causal effects of each variable on tail-defined outcomes-poor (bottom quartile) and good (top quartile) PCS and MCS-and quantified pairwise interaction effects. The network revealed how upstream factors propagate through direct and indirect pathways to shape HRQoL. Notably, PCS and MCS shared common upstream causes but showed no direct causal connection. Quantifying these causal pathways through relative risk (RR) estimates revealed the magnitude of individual factor effects. For poor PCS, severe insomnia (RR = 1.98), high stress (RR = 1.45), low physical activity (RR = 1.39), and multimorbidity (RR = 1.36) were the principal risk factors. For poor MCS, high stress (RR = 3.28) and severe insomnia (RR = 2.72) dominated. Notably, low BMI increased poor MCS risk (RR = 1.20), consistent with frailty pathways. The patterns for good outcomes largely mirrored these findings, with favorable levels showing protective effects. Interaction analyses revealed substantial synergistic effects: severe insomnia with high stress increased poor MCS probability by 6.44 percentage points (pp) beyond additivity, while high stress with physical inactivity added 4.77 pp. For good MCS, low insomnia with low stress (+4.72 pp) and low BMI with exercise (+4.21 pp) showed synergy, whereas stress with inactivity exhibited antagonism (-4.00 pp). These results support integrated interventions that combine sleep improvement, stress reduction, physical activity promotion, and multimorbidity management to improve HRQoL in aging populations.