Mediation analysis
Gestational age features on the causal pathway between chronic hypertension and perinatal mortality as an mediator (Supplemental Figure 2 ); therefore, any adjustment for preterm delivery in the assessment of the perinatal mortality risk in relation to chronic hypertension would lead to collider bias.(32) To resolve this issue, we undertook a causal mediation analysis based on the counterfactual framework(33, 34) to decompose the total effect (TE) of chronic hypertension on perinatal mortality into the natural direct effect (NDE; the effect of the exposure [chronic hypertension] on the outcome [perinatal mortality] if the mediator [preterm delivery <37 weeks] were set to what it would have been, probably contrary to the fact, in the absence of the exposure) and the natural indirect effect (NIE; the effect of the exposure on the outcome when the exposure is present after setting the mediator to what it would have been, probably contrary to the fact, with versus without the exposure).(35, 36) We also estimated the controlled direct effect (CDE; the effect of chronic hypertension on mortality that is not mediated through preterm delivery). The TE, NDE, NIE and CDE were estimated on multiplicative scale with RR and 95% CI as effect measure.
In addition, we estimated the proportion mediated (PM; the proportion of the TE that is mediated through preterm delivery), and the proportion eliminated (PE; the proportion of the chronic hypertension-perinatal mortality association that might be eliminated by blocking the effect of chronic hypertension on preterm delivery). The PE is the proportion of perinatal deaths that could be prevented by designing interventions to reduce preterm delivery associated with chronic hypertension. All 95% CI estimates were based on 1000 bootstrap samples.