Table 4.

Causal mediation analysis: Natural direct and indirect effects of income at age 3 on SDQ-IH at age 5 and SDQ-IH at age 8 mediated by asthma at age 3 (Eden cohort, No. = 1432)

	SDQ-IH at age 5			SDQ-IH at age 8
	β coefficient	95% CI	p value	β coefficient	95% CI	p value
NDE	−0.37	[−0.50, −0.24]	<0.001	−0.39	[−0.53, −0.24]	<0.001
NIE	−0.04	[−0.08, −0.01]	0.026	−0.04	[−0.08, 0.01]	0.084
Interaction (NDE*NIE)	0.01	[0.001, 0.02]	0.030	0.01	[−0.003, 0.02]	0.117
Total effect	−0.40	[−0.54, −0.26]	<0.001	−0.41	[−0.58, −0.25]	<0.001
Proportion mediated	0.11			0.09

Outcome = SDQ-IH at age 5 and 8 (continuous); Exposure = income at age 3 (continuous); Mediator = asthma at age 3 (binary). CI are calculated using robust standard errors based on the sandwich estimator. CI calculated using bootstrap method produced almost identical results. Fully adjusted for sex, maternal age at childbirth, child birthweight, cigarettes per day during pregnancy and medical centre. The counterfactual values were derived using imputation-based approach in Medflex R-package. Total effect = (pure) NDE + (pure) NIE + interaction term. Estimated using 50 imputed datasets.