Table 1.
Variable layers used to impose blacklist in a Bayesian network.
Table 2.
Summary of study participants.
Frequencies (%) are shown for categorical variables and Mean ± SD for continuous variables. p-values represent results from Kruskal–Wallis tests (categorical) and Spearman correlations (continuous) with PCS and MCS.
Fig 1.
Causal Bayesian network for health-related quality of life in Korean adults aged 50–81 years.
The directed acyclic graph represents causal relationships among determinants of physical (PCS) and mental (MCS) health-related quality of life. Nodes represent variables, and directed arcs indicate causal pathways from parent to child nodes. Dashed lines denote variable pairs with directional uncertainty (arc directions 0.51–0.57). Variables are color-coded by domain: demographic (yellow: SEX, AGE), socioeconomic (blue: MARRY, JOB, INCOME), lifestyle behaviors (green: EXER, DRINK, SMOKE), anthropometric/physical activity (pink: BMI, ACT), psychological (gray: INSOMNIA, PSS), medical conditions (orange: nCNCD), and HRQoL outcomes (red: PCS, MCS). Abbreviations: ACT, daily physical activity volume; AGE, age; BMI, body mass index; DRINK, alcohol consumption; EXER, regular exercise; INCOME, monthly household income; INSOMNIA, insomnia severity; JOB, employment status; MARRY, marital status; MCS, Mental Component Summary; nCNCD, number of chronic non-communicable diseases; PCS, Physical Component Summary; PSS, perceived stress; SEX, sex; SMOKE, lifetime consumption of ≥400 cigarettes.
Fig 2.
Estimated interventional probabilities and risk ratios of poor health-related quality of life.
Top row: Physical Component Summary (PCS); bottom row: Mental Component Summary (MCS). Left panels show interventional probabilities of scores below the 25th percentile (poor HRQoL). Right panels show risk ratios relative to the reference category, with the vertical dotted line marking RR = 1. For INSOMNIA and PSS, estimates are weighted averages from BNinsomnia (INSOMNIA→PSS, weight 0.56) and BNpss (PSS→INSOMNIA, weight 0.44). Abbreviations: as in Fig 1.
Fig 3.
Estimated interventional probabilities and risk ratios of good health-related quality of life.
Top row: Physical Component Summary (PCS); bottom row: Mental Component Summary (MCS). Left panels show interventional probabilities of scores above the 75th percentile (good HRQoL). Right panels show risk ratios relative to the reference category, with the vertical dotted line marking RR = 1. For INSOMNIA and PSS, estimates are weighted averages from BNinsomnia (INSOMNIA→PSS, weight 0.56) and BNpss (PSS→INSOMNIA, weight 0.44). Abbreviations: as in Fig 1.
Table 3.
Additive interaction, total effect, and relative interaction for selected risk-factor pairs on poor PCS and poor MCS.
For each pair, interaction and total effect are weighted averages (and ranges) across BN_insomnia (INSOMNIA→PSS) and BN_pss (PSS→INSOMNIA) with weights 0.56 and 0.44, respectively. Relative interaction is the ratio of additive interaction to total effect.
Table 4.
Additive interaction, total effect, and relative interaction for selected risk-factor pairs on good PCS and good MCS.
For each pair, interaction and total effect are weighted averages (and ranges) across BN_insomnia (INSOMNIA→PSS) and BN_pss (PSS→INSOMNIA) with weights 0.56 and 0.44, respectively. Relative interaction is the ratio of additive interaction to total effect.