Background Approximately 8% of births in the UK are preterm, and rates have been increasing for the past decade. There are significant inequalities in the prevalence of preterm birth, with some estimates showing 50% higher prevalence in mothers with low socioeconomic status (SES). Understanding and explaining how low SES increases the risk of preterm birth, will be important to reduce these inequalities.

Methods Systematic review of quantitative observational studies of mediation of the relationship between maternal SES and preterm birth. Searches were conducted from five databases covering January 2000 to June 2020, and all were identified for dual screening based on title and abstract review. Inclusion criteria was comparison across SES, preterm birth or gestational age as outcome, and study of mediation of non-genetic variables. All included studies were citation searched. All studies were narratively synthesised and quality assessed.

Results Initial searches identified 58 studies for full-text screening, of which 13 met inclusion criteria, and citation searching identified four more studies. A final sample of 17 studies from 11 countries (five from North America, 11 from Europe, and one each from Africa, the Middle East and South America; one study covered three European countries) were identified for analysis, covering study periods between 1980 and 2013. The majority (10/17) used the difference-of-coefficients approach and estimated proportion mediated. Other methods included multiplication of coefficients and counterfactual approaches. Forty-eight mediators (or groups of mediators) were examined. SES was measured most frequently by maternal education (7/17). All studies used a categorisation of preterm birth. The most frequently examined mediator was smoking (8/17), with the proportion estimated ranging from 2% to 43%. A residual direct effect of SES on preterm birth not via mediating exposures studied remained in over half of these studies. Four studies examined mediation via maternal health (range 6% to 39%), BMI (range 2% to 17%), and depression and anxiety (range from no effect to 44%).

Discussion Our review suggests that inequalities in preterm birth are in part explained by unequal exposure to risk factors across socioeconomic strata. However, limitations in the methods to assess mediation, reporting of mediation results, and measures of SES used mean it is difficult to meaningfully combine findings. A major limitation is the lack of consideration for the theoretical basis for mediation in the data, such as the underlying assumptions. Despite these limitations, these findings highlight potential intervention points for reducing preterm birth inequalities.