Table 2.
Mediation (by healthy lifestyles) analysis of socioeconomic status and all-cause mortality
| High SES | Medium SES | Low SES | |
|---|---|---|---|
| No. of participants (n) | 2383 | 8110 | 11,600 |
| Deaths (n) | 1298 | 5514 | 8909 |
| Follow-up (PYs) | 13,608.3 | 42,001.4 | 53,066.1 |
| Mortality rate (95% CI)a | 9.5 (9.0–10.0) | 13.1 (12.8 to 13.5) | 16.8 (16.5 to 17.1) |
| Associationb | |||
| Total effect; HR (95% CI), p | 1 [Reference] | 1.135 (1.067 to 1.204), < 0.001 | 1.161 (1.088 to 1.229), < 0.001 |
| Natural direct effect; HR (95% CI), p | 1.135 (1.067 to 1.205), < 0.001 | 1.175 (1.102 to 1.245), < 0.001 | |
| Natural indirect effect; HR (95% CI), p | 1.000 (0.996 to 1.004), 0.936 | 0.988 (0.983 to 0.992), < 0.001 | |
| Mediation proportion; % (95% CI), p | − 0.1 (− 3.8 to 3.3), 0.936 | − 8.9 (− 16.6 to − 5.1), < 0.001 | |
| Excess relative HR solely due to interactionc; HR (95% CI), p | − 0.0021 (− 0.0960 to 0.1300), 0.932 | − 0.0363 (− 0.1313 to 0.0590), 0.438 | |
All models were adjusted for sex, age, marital status, residence, co-residence, comorbidities, ADL disability, and self-reported health
ADL activities of daily living, CI confidence interval, HR hazard ratio, PYs person-years, SES socioeconomic status
aPer 100 PYs
bNatural direct effect and natural indirect effect estimated the effects of SES on mortality that did not or did act through the mediator (i.e., healthy lifestyles), respectively. Mediation proportion estimated the percent of SES effect, on the log(HR) scale, that acted through the mediator, i.e., healthy lifestyles. The results were calculated without considering exposure-mediator interaction
cThe results were considered with exposure–mediator interaction. Since the interaction was not statistically significant, we only showed the excess relative HR that was solely due to interaction