Maternal educational level and preterm birth: Exploring inequalities in a hospital-based cohort study

Abstract

Preterm birth has been related to inequalities in maternal educational level, but the causal mechanism is not entirely known. Some factors associated with preterm birth and low educational level such as chronic medical conditions, pregnancy complications and related-health behaviours could have a mediation role in the pathway. This study aimed to evaluate the association between maternal educational level and preterm birth, analysing the mediation role of these factors. We performed a retrospective cohort study based on hospital electronic records of 10467 deliveries that took place in the Hospital Clínic of Barcelona between 2011 and 2017. Poisson regression was used to obtain crude and adjusted relative risk of preterm birth in women with different educational level and the percentage of change in relative risk was calculated when mediation variables were included in the model. Women with a lower educational level had a higher risk of preterm birth (RR 1.57, 95% CI 1.21, 2.03). The loss of association after the inclusion of body mass index in the model suggests an important mediation role of maternal overweight. Other variables such as smoking, drug use, preeclampsia and genitourinary infections also appear to play a role in the observed inequality between women with different levels of education. Efforts to promote health literacy and to improve preventive interventions, before and during pregnancy, could decrease preterm birth rates and perinatal health inequalities.

Fig 1. Direct Acyclic Graph (DAG) of the association between maternal educational level and preterm birth including confounder and intermediate variables DI: Deprivation Index; EL: Educational Level; BMI: Body Mass Index.

The DAG shows some possible causal pathways between educational level and preterm birth. Age at pregnancy, maternal region of birth and DI of the neighbourhood generate biasing pathways since all three variables are associated with both the exposure and the outcome but they are not in the causal pathway. Adjusting for these variables (represented by framing them) block the biasing pathways and maintain the causal pathways of interest.