Table 5.

Results from causal mediation analyses examining the effect of OAD diagnosis as a mediator for chronic rhinosinusitis and gastroesophageal reflux disease for follow-up beginning September 11, 2006.

	High vs. low exposure			Moderate vs. low exposure
	Estimate	95% CI		Estimate	95% CI
Causal mediation models for chronic rhinosinusitis diagnosis
Controlled direct effect	0.542	0.288	0.856	0.352	0.125	0.637
Interaction in absence of mediator	0.001	−0.017	0.013	0.000	−0.007	0.005
Mediated interaction	0.001	−0.018	0.018	0.001	−0.012	0.013
Pure indirect effect	0.029	0.014	0.046	0.015	0.008	0.028
Total effect	0.573	0.333	0.886	0.369	0.148	0.642
Proportion mediated	0.057	0.026	0.108	0.048	0.021	0.101
Results from conventional models for chronic rhinosinusitis diagnosis
Regression direct effect	0.554	0.322	0.788	0.372	0.157	0.577
Total regression effect	0.658	0.419	0.892	0.419	0.215	0.631
Causal mediation models for gastroesophageal reflux disease
Controlled direct effect	0.244	0.069	0.440	0.135	−0.011	0.295
Interaction in absence of mediator	−0.002	−0.016	0.010	0.000	−0.007	0.006
Mediated interaction	−0.002	−0.014	0.012	0.000	−0.011	0.012
Pure indirect effect	0.034	0.018	0.054	0.018	0.008	0.030
Total effect	0.275	0.103	0.464	0.152	0.012	0.305
Proportion mediated	0.136	0.056	0.332	0.289	0.038	0.696
Results from conventional models for gastroesophageal reflux disease diagnosis
Regression direct effect	0.223	0.068	0.385	0.131	−0.010	0.262
Total regression effect	0.336	0.182	0.495	0.191	0.058	0.329

Causal mediation models are defined in Eqs 5 and 6 in the text; conventional models are defined in Eqs 1 and 3. The quantities shown in the tables are defined in Eqs 7 and 4 in the text. Confidence intervals for causal mediation models based on 1000 bootstrap simulations; confidence intervals for conventional models based are large sample Wald-type intervals.