Background Prevalence of preterm birth (PTB) is increasing in England. Given significant inequalities in prevalence by maternal socioeconomic status (SES) and subsequent negative consequences on health and educational attainment for children, this represents a significant risk of widening population health inequalities. The aim is to understand how different exposures to risks mediate the relationship between SES and PTB.

Methods Analysis of routinely collected data for all births at Liverpool Women’s Hospital between April 2009 and March 2020, covering an area of North West England with high levels of deprivation. The exposure was a dichotomised measure of maternal SES at birth (two most deprived Indices of Multiple Deprivation deciles compared to other eight deciles), and the outcome was PTB (gestational age <37 weeks) compared with term. Mediators were maternal smoking status, medical conditions, obstetric conditions, and BMI at booking. Covariates included maternal age at birth, method of labour onset, parity, baby sex, and previous PTB. Mediation was assessed using multivariable logistic regression for modelling the effect of SES on PTB after adjusting for covariates (minimally adjusted) with and without adjustment for the potential mediators and applying difference-of-coefficients method; the difference in odds ratio (OR) for SES between these models. Proportion mediated is (OR_{mediator-adjusted} - OR_{minimally-adjusted})/(OR_{minimally-adjusted}-1). All analysis was done in R.

Results Analysis sample consisted of 81,680 births; 7.4% were preterm, while 64.1% were to low SES mothers. Of mothers, 19.1% were current smokers, 4.3% had medical conditions, 10.1% had obstetric conditions and 48.6% had healthy weight BMI. The OR for the effect of low SES on PTB in the minimally adjusted model was 1.26 (95% CI 1.19–1.34). When mediators were included, OR reduced by 42% for smoking. OR for low SES did not change when maternal medical conditions or obstetric conditions were included, while OR increased when BMI was included (1.32, 95% CI 1.22–1.44).

Discussion Our findings demonstrate that significant inequalities exist in the prevalence of PTB. The initial results from the mediation analysis suggest that smoking is partly explains these inequalities. Over half of this inequality is not explained by mediators included here. Limitations include unavailable data on other mediators (e.g. maternal mental health) and in the analysis approach. The next step is to analyse the data using the counterfactual approach to mediation. Reducing PTB is a priority for Maternity and Children’s services, and the NHS’ Long Term Plan. These results suggest that action on smoking during pregnancy will reduce inequalities.