I have a question about the mediation effect/ analysis from this data. What is important to look for to find out where is mediation present?
Table 5.
Randomized total, natural direct and natural indirect effects (risk ratios, 95% CIs) of incident obesity at selected ages and of persistent obesity vs never obese during follow-upa on poor physical functioning at 60–64/50 years (mediated by time-varying inactivity)b
1946-NSHD1958-NCDSIncident obesity at 60–64 yearscIncident obesity at 53 yearsIncident obesity at 43 yearsPersistent obesity (from 36 years)Incident obesity at 50 yearscIncident obesity at 42 yearsPersistent obesity (from 33 years)Randomized total effect 1.32 (0.75, 1.88) 1.53 (0.91, 2.15) 2.32 (1.13, 3.51) 2.91 (1.14, 4.69) 1.14 (0.94, 1.34) 1.22 (0.96, 1.48) 1.53 (1.12, 1.93) Randomized natural direct effect (not via physical inactivity) 1.50 (0.89, 2.11) 2.27 (1.13, 3.41) 2.84 (1.16, 4.51) 1.20 (0.94, 1.45) 1.49 (1.09, 1.88) Randomized natural indirect effect (via physical inactivity) 1.02 (0.99, 1.04) 1.02 (0.97, 1.07) 1.03 (0.96, 1.10) 1.02 (1.01, 1.03) 1.03 (1.01, 1.05) a
Follow-up refers to ages 36 to 60–64 years in 1946-NSHD and 33 to 50 years in 1958-NCDS.
b
Adjusted for: (i) baseline confounders: gender, early-life and adult social class, early adult BMI, mental health, smoking, physical activity (1958-NCDS only), highest educational qualification, illnesses: arthritis/rheumatism, diabetes, heart trouble, high blood pressure and asthma; and (ii) time-varying confounders: smoking, depression and self-rated health (1958-NCDS only), see Supplementary Table S1, available as Supplementary data at IJE online, for details.
c
For incident obesity at 60–64 years in 1946-NSHD and at 50 years in 1958-NCDS, the randomized total effect is not mediated by inactivity: we assume inactivity precedes obesity (i.e. there is no measure of inactivity between obesity and physical functioning), see Supplementary Figure S1 and appendix, available as Supplementary data at IJE online, for details.