Article Text

Original research
Socioeconomic inequalities and adverse pregnancy outcomes in the UK and Republic of Ireland: a systematic review and meta-analysis
  1. Katie Thomson1,
  2. Malcolm Moffat1,
  3. Oluwatomi Arisa1,
  4. Amrita Jesurasa2,
  5. Catherine Richmond1,
  6. Adefisayo Odeniyi1,
  7. Clare Bambra1,3,
  8. Judith Rankin1,3,
  9. Heather Brown1,3,
  10. Julie Bishop4,
  11. Susan Wing4,
  12. Amy McNaughton4,
  13. Nicola Heslehurst1,3
  1. 1Population Health Sciences Institute, Faculty of Medical Sciences, Newcastle University, Newcastle upon Tyne, UK
  2. 2Primary Care Division, Public Health Wales, Cardiff, UK
  3. 3Fuse, The Centre for Translational Research in Public Health, Newcastle University, Newcastle upon Tyne, Tyne and Wear, UK
  4. 4Health Improvement Division, Public Health Wales, Cardiff, UK
  1. Correspondence to Nicola Heslehurst; nicola.heslehurst{at}


Objective There has been an unprecedented rise in infant mortality associated with deprivation in recent years in the United Kingdom (UK) and Republic of Ireland. A healthy pregnancy can have significant impacts on the life chances of children. The objective of this review was to understand the association between individual-level and household-level measures of socioeconomic status and adverse pregnancy outcomes.

Design Systematic review and meta-analysis.

Data sources Nine databases were searched (Medline, Embase, Scopus, ASSIA, CINAHL, PsycINFO, BNI, MIDRIS and Google Scholar) for articles published between 1999 and August 2019. Grey literature searches were also assessed.

Study selection criteria Studies reporting associations between individual-level or household socioeconomic factors on pregnancy outcomes in the UK or Ireland.

Results Among the 82 353 search results, 53 821 titles were identified and 35 unique studies met the eligibility criteria. Outcomes reported were neonatal, perinatal and maternal mortality, preterm birth, birth weight and mode of delivery. Pooled effect sizes were calculated using random-effects meta-analysis. There were significantly increased odds of women from lower levels of occupation/social classes compared with the highest level having stillbirth (OR 1.40, 95% CI 1.23 to 1.59, I2 98.62%), neonatal mortality (OR 1.39, 95% CI 1.22 to 1.57, I2 97.09%), perinatal mortality (OR 1.39, 95% CI 1.23 to 1.57, I2 98.69%), preterm birth (OR 1.41, 95% CI 1.33 to 1.50, I2 70.97%) and low birth weight (OR 1.40, 95% CI 1.19 to 1.61, I2 99.85%). Limitations relate to available data, unmeasured confounders and the small number of studies for some outcomes.

Conclusions This review identified consistent evidence that lower occupational status, especially manual occupations and unemployment, were significantly associated with increased risk of multiple adverse pregnancy outcomes. Strategies to improve pregnancy outcomes should incorporate approaches that address wider determinants of health to provide women and families with the best chances of having a healthy pregnancy and baby and to decrease pregnancy-related health inequalities in the general population.

PROSPERO registration number PROSPERO CRD42019140893.

  • epidemiology
  • maternal medicine
  • fetal medicine
  • public health

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Strengths and limitations of this study

  • Search strategy broad and wide-ranging resulting in the inclusion of 17 130 631 mothers and babies in the analysis from 35 studies.

  • No restriction by pregnancy outcome; results therefore provide a comprehensive overview of the effects of socioeconomic status.

  • Results are limited to the UK and Ireland only.

  • Between countries and over time, the occupational classifications used have changed.

  • Occupational data dominates the evidence base; limited data were available for other measures of inequality (eg, housing, income and car ownership).


There has been a sustained increase in infant mortality in England, disproportionately affecting the most deprived areas,1 2 and a reduction in life expectancy for the first time in 120 years, with similar trends in the rest of the UK.1 The evidence base that demonstrates the increase in children and families living in poverty, and the association with increasing mortality rates and reduction in life expectancy, is compelling. The recent Marmot review1 identified that 30% of children in England are living in poverty, and those in the most deprived 10% are almost twice as likely to die and more likely to have illness or long-term disability.

Having a healthy pregnancy is vital to improve the life chances of future generations, to prevent perinatal and infant mortality and to give every child the best start in life. For example, babies born small for gestational age (SGA) or preterm are more likely to die during infancy and childhood.3–5 However, studies in high-income countries (HICs) show worse pregnancy outcomes among women living in deprived communities.6 7 Socioeconomic inequalities disproportionately affect women, which impacts on their own health and the life chances of their children. Female life expectancy has declined in England’s most deprived areas over the past 10 years, which has been attributed to austerity, whereas it has slightly increased for men.1 Women, particularly lone mothers, have the highest gap between income and adequate living standards and are the most likely to be living in food insecure households,8 which is another measure of poverty. The evidence relating to the disproportionate impact of the COVID-19 pandemic on the most disadvantaged, and the potential for future inequalities in health consequences due to political and economic pathways,9 will only serve to increase these inequalities.

Compared with other HICs, the UK and Republic of Ireland (ROI) perform poorly in terms of both infant and maternal mortality.2 10 11 For example, differences in infant mortality are linked to the level of deprivation, and there is evidence for widening inequalities in the poorest households since 2013.2 Slowdown in life expectancy has been worse in the UK than almost all other HICs across Europe, whereas wealth inequality has increased faster.1 Health costs associated with inequalities in England are estimated to be more than £12 billion1; therefore, there are potential economic benefits as well as moral and ethical reasons to invest in reducing health inequalities. Despite this, local governments responsible for delivering public health agendas in England have had a 77% budget reduction over 10 years. These cuts have been inequitable, regressive and applied to the most deprived areas of England with the greatest health needs, contributing to widening health inequalities.1

Central to understanding the relatively low performance of the UK and ROI compared with other HICs is documenting the scale of inequalities related to adverse pregnancy outcomes across the social gradient. Important individual factors to explore the ‘causes of the causes1 include employment, education, income and housing. While there have been a number of systematic reviews that examine the impact of deprivation,6 12 13 these are out of date and focus on area-based measures (such as the Index of Multiple Deprivation) as an inadequate proxy for individual measures of socioeconomic status (SES), which can be more difficult to collate. There is some evidence from a meta-analysis published in 201214 that individual-level social class is significantly associated with an increase in low birth weight (LBW) and infant mortality, and the authors concluded that future research should focus on the more proximal individual-level determinants of health. Using individual-level socioeconomic indicators provides a richer way of understanding the relative importance of particular SES factors and their direct and indirect impact on a range of adverse pregnancy outcomes. This is an essential first step to understand and then mitigate their effect on the most vulnerable groups in our communities. This systematic review and meta-analysis aimed to understand the association between individual-level and household-level measures of SES and adverse pregnancy outcomes in the UK and ROI. We hypothesised that there would be an association between low SES measures and increased risk of adverse pregnancy outcomes.


Protocol, search strategy and study selection

The systematic review was registered with PROSPERO (CRD42019140893) and conducted in line with MOOSE guidelines15 (online supplemental appendix S1). Nine electronic databases were searched between 1 January 1999 and 7 August 2019 (online supplemental appendix S2). The start date of 1999 was included as it coincided with the start of the English Health Inequalities Strategy.16 The databases were searched using key words and Medical Subject Headings (MeSH) developed by an information specialist at Newcastle University (CR).

Searching for epidemiological studies using databases alone systematically misses a proportion of relevant studies; therefore, in line with MOOSE guidelines, supplementary searches were carried out. Backwards and forwards citation chaining of included studies were undertaken and relevant systematic reviews were assessed. Finally, literature searches comprising key data sources for the UK and ROI (online supplemental appendix S3) were run, and experts in key organisations in each of the countries were contacted.17

The inclusion criteria for the review were determined a priori in terms of Population, Exposure, Comparison, Outcome and Study design18:

  • Population: the unborn fetus, neonate (child under 28 days of age) or mother in the UK or ROI.

  • Exposure: any individual-level or household-level socioeconomic factor such as employment, education, income, housing, poverty/low wage, social isolation and car ownership. Aggregate area-level measures of SES (such as the Index of Multiple Deprivation) were excluded.

  • Comparison: studies with individual-level or household-level exposures that compare low to high SES.

  • Outcomes: pregnancy outcomes related to the fetus, neonate or mother. We did not exclude outcomes if they were relevant to the fetus, neonate or mother during pregnancy or within 28 days postnatal as we wanted to provide a comprehensive overview of the existing evidence base.

  • Study design: observational studies with cohort, case–control, longitudinal or cross-sectional designs.

The results from the database searches were imported into Endnote,19 and duplicates were removed. Screening the titles and abstracts was managed in Covidence.20 Title, abstract and full-text screening were completed independently in duplicate. Where multiple studies reported data from the same cohort, decisions on which study to include were based on sample size, time period and categories used to define determinants and outcomes. We included any studies that met the inclusion criteria and reported an association between the exposure and outcomes of interest, regardless of whether this was the primary aim of the study. Decisions on inclusion and exclusion of studies were recorded using a Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart.

Data extraction and quality appraisal

Data extractions were carried out using a standardised protocol independently by two reviewers for each included study (KT, OA and MM), and discrepancies were validated by a third reviewer (AO and NH). Data were extracted on the characteristics of the included studies and the results reported on associations between inequality exposure variables and pregnancy outcomes (online supplemental appendix S4). Authors of primary studies were contacted if needed to request additional data or clarify reported results (online supplemental appendix S5). The Newcastle-Ottawa scale was used to assess information bias, selection bias and confounding in case–control and cohort studies (online supplemental appendix S6).21 Stars were awarded for high-quality aspects of study design. A maximum of eight stars was attainable for cohort studies (0–2 stars deemed as low quality, 3–5 moderate quality and 6–8 high quality) and nine stars for case–control studies (0–3 stars for low quality, 4–6 for moderate quality and 7–9 for high quality). Any discrepancies on screening, extraction and quality appraisal were resolved through discussion between the reviewers (KT, OA, MM and AO) and the project lead (NH).

The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) methodology was used to address the quality of the evidence.22 Only those outcomes considered most important were evaluated using GRADE, and due to the available evidence, only inequalities in occupation/social class were considered. Quality of evidence for each pregnancy outcome was assessed based on study design (the baseline rating for observational studies is low), risk of bias, imprecision of estimates, inconsistency of results from different studies, indirectness of study results (ie, lack of applicability) and publication bias.23

Evidence synthesis and statistical analysis

A meta-analysis was used to calculate a pooled odds ratio (OR) and a 95% CI where there were at least two studies reporting the same outcome and comparable individual-level or household-level SES exposures. Meta-analysis was only possible for occupation-derived social class. Subgroups for this variable were included in the meta-analysis where possible to stratify analysis comparing intermediate-level or lowest-level of occupation/social class to highest level. Heterogeneity was explored through meta-regression for factors including duration of data collected, country, region, date and sample size, location, publication date and quality. I2 values of 25%, 50% and 75% were used to indicate low, moderate and high levels of heterogeneity.24 Publication bias was investigated using Egger’s test and funnel plots. Sensitivity analysis was performed for each meta-analysis by excluding one dataset at a time to identify the effect of any individual dataset on the pooled effect size and between-study heterogeneity. All analyses were conducted in Stata V.16.25

Narrative synthesis methods described by Popay et al26 were used to report outcomes with insufficient data to include in a meta-analysis and to integrate the narrative and meta-analysis data. Preliminary synthesis involved grouping all data into themes based on outcomes reported and type of SES measure. The associations between SES determinants and pregnancy outcomes reported by the studies, or calculated using data reported in the studies, were tabulated according to pregnancy outcome themes (eg, birth weight outcomes) and subgrouped by the level of SES determinant (eg, occupation). Where required, data were transformed to be comparable between studies (eg, using reported frequency data to recalculate ORs into more comparable SES categories). Authors were contacted when insufficient data were reported in the published studies to calculate ORs. Relationships between studies are reported by discussion of reciprocal and refutational patterns in the data, including the statistical significance and direction of effect. The narrative synthesis is supplemented by a discussion of the wider evidence base and the strengths and limitations of this review including the gaps in the existing literature.

Patient and public involvement

Patients or the public were not involved in any aspect of the study design, conduct or in the development of the research question or outcome measures.


The database searches identified 53 821 records after duplicates had been removed. A further 9414 references and citations and 529 grey literature data sources were screened (figure 1), of which 1439 were screened at full-text level (reasons for exclusion are detailed in online supplemental appendix S7). Forty-one studies met the inclusion criteria, six of these reported duplicate data (online supplemental appendix S8). Thirty-five studies were included reporting unique data for 17 130 631 mothers and babies, 27 were conducted in the UK and 8 in the ROI (table 1). The studies were published between the years 1999 and 2019 (with data reported from 1990 onwards), 10 were case control and 25 were cohort studies. The majority of studies reported national-level data, and these datasets are assumed to be nationally representative. All studies scored between three and eight on the Newcastle-Ottawa Scale (online supplemental appendix S9). The studies were rated as either high quality (n=4 case–control, n=18 cohort) or moderate quality (n=6 case–control, n=7 cohort), with no low-quality studies. Most studies (n=23) reported outcomes associated with occupation-derived social class using different classifications (online supplemental appendix S10), which were grouped into: (1) highest level (managerial/professional) occupations; (2) intermediate-level occupations; and (3) lowest level (routine/manual/partly skilled) occupations.27 Seventeen studies investigated associations with education, seven for employment and income/finances, four for housing and one for car ownership. Pregnancy outcomes reported were mortality (stillbirth, neonatal, perinatal and maternal), LBW, preterm birth (PTB), mode of delivery, maternal morbidity and congenital anomalies (table 1).

Figure 1

PRISMA flow chart. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Table 1

Summary of the included studies


Six studies reported stillbirth, neonatal mortality, perinatal mortality and maternal mortality (online supplemental appendix S11 tables a-d). Individual and household inequality measures reported were paternal, maternal and primary household occupation/social class.

Stillbirth was reported by five studies28–32 for 12 642 203 births (online supplemental appendix S11, table a), and four29–32 could be pooled into the meta-analysis. Perinatal mortality was reported by three studies29 31 32 for 11 189 810 births (online supplemental appendix S11, table b), and neonatal mortality was reported by three studies30–32 for 11 580 610 births (online supplemental appendix S11, table c), and all could be pooled in meta-analysis. There were significantly increased odds for the pooled lowest/intermediate occupation compared with highest for all three outcomes (stillbirth: OR 1.40 95% CI 1.23 to 1.59 (figure 2); perinatal mortality: OR 1.39 95% CI 1.23 to 1.57 (figure 3); neonatal mortality: OR 1.39 95% CI 1.22 to 1.57 (figure 4)). All subgroup meta-analyses showed highest odds among the lowest level of occupation/social class compared with the highest level (stillbirth: OR 1.61 95% CI 1.37 to 1.88; perinatal mortality: OR 1.57 95% CI 1.41 to 1.74; neonatal mortality: OR 1.58, 95% CI 1.44 to 1.74; figures 2–4). Heterogeneity was high for all meta-analyses and was not explained by any of the factors included in meta-regression (online supplemental appendix S12Aiii, Biii, Ciii). Sensitivity analysis demonstrated that the meta-analysis results were robust to the effect of any individual studies for all stillbirth analyses (online supplemental appendix S12Aiv) and for perinatal and neonatal mortality overall pooled meta-analyses and lowest versus highest occupation subgroups (online supplemental appendix S12Biv, Civ); however, the significance changed with the removal of the Office for National Statistics (ONS) 2002–2017 dataset in the intermediate subgroup for both outcomes. There was evidence of publication bias for stillbirth (online supplemental appendix S12Aii) but not for perinatal or neonatal mortality (online supplemental appendix S12Bii, Cii). One study28 not able to be pooled in the stillbirth meta-analysis reported increased odds of stillbirth among unemployed mothers (OR 1.24, 95% CI 1.00 to 1.54) and fathers (OR 2.26, 95% CI 1.69 to 3.02) compared with those who were employed (online supplemental appendix S11, table a). Maternal mortality was reported by one study33 that showed a significantly increased odds among unemployed women compared with those in employment following adjustment for confounding factors (AOR 1.81, 95% CI 1.08 to 3.04, online supplemental appendix S11 table d). GRADE scores showed very low-quality evidence for stillbirth and low-quality of evidence for neonatal, perinatal and maternal mortality (online supplemental appendix S13).

Figure 2

Forest plot of stillbirth. REML, Restricted Maximum Likelihood

Figure 3

Forest plot of perinatal mortality.

Figure 4

Forest plot of neonatal mortality.

Birth weight

Thirteen studies29 31 34–78 reported associations between low and high birth weight outcomes and paternal and maternal occupation/social class, education, housing, income/household finances and car ownership. A meta-analysis was possible for occupation and very low birth weight (VLBW), LBW and macrosomia.

Two studies29 31 reported associations between occupation/social class and VLBW (<1500 g, n=10 284 045 births) and macrosomia (>4000 g, n=3 430 632 births) (online supplemental appendix S11, tables e and g), and both were included in the meta-analysis. Six studies reported LBW (<2500 g, n=3 430 632 births) with four29 31 79 80 that could be included in the meta-analysis (online supplemental appendix S11, table f). The meta-analysis for the pooled lowest/intermediate occupation compared with highest identified no significant association with VLBW (OR 1.18 95% CI 0.89 to 1.57), a significantly increased odds of LBW (OR 1.40 95% CI 1.19 to 1.61) and decreased odds of macrosomia (OR 0.89 95% CI 0.79 to 0.99) (figures 5–7). The subgroup meta-analysis identified that the highest odds for LBW were in the lowest level occupation category (OR 1.55 95% CI 1.24 to 1.86, figure 6), whereas for macrosomia, only the intermediate-level versus highest level occupation remained significant (figure 7). There was significant heterogeneity in all meta-analyses for birthweight outcomes that were not explained by any of the factors included in meta-regression (online supplemental appendix S12Diii, Eiii, Fiii). There was no evidence of publication bias (online supplemental appendix S12Dii, Eii, Fii). Sensitivity analysis demonstrated that the results for birth weight meta-analysis were not robust. For LBW, results were robust for the overall pooled analysis and for the lowest level subgroup. However, the exclusion of datasets in the LBW intermediate-level subgroup and all macrosomia groupings resulted in the odds no longer being significant, whereas the exclusion of one dataset in the VLBW analysis resulted in the odds becoming significant for the pooled meta-analysis and both subgroups (online supplemental appendix S12Div, Eiv, Fiv). The GRADE score for LBW showed low-quality evidence (online supplemental appendix S13).

Figure 7

Forest plot of macrosomia.

Additional data were reported for occupation and mid-low birth weight (MLBW) <3000 g29 31 and SGA <5th or <10th percentile79 81 82 (online supplemental appendix S11, tables h and i). Predominantly, the data showed a significantly increased odds of SGA (following adjustment for confounding factors) and MLBW when mothers had lowest level occupations compared with highest,29 31 79 82 whereas the evidence was conflicting for intermediate/non-manual occupations.29 31 79 81 No association was found between SGA and whether the mother had a contract post, worked shifts or long hours or had an active or stressful job.81

Further SES exposures reported for birthweight outcomes were education,81–87 income/finance83 84 86–88 and overcrowding, housing tenure and car ownership,89 the majority of which adjusted for confounding factors. There was significantly increased odds of LBW,83 SGA82 and IUGR84 for lower levels of education when studies used degree-level education as the comparison group (online supplemental appendix S11–table j); however, studies using other comparison groups reported no significant associations with high or LBWs.81 85–87 There was significantly increased odds of LBW in lower income groups83 88 but not for VLBW87 or IUGR,84 or between other finance measures and SGA86 (online supplemental appendix S11, table k). One study89 reported a significantly increased odds of LBW when babies were born to mothers living in overcrowded households and in council/housing association properties but no significant associations with the number of rooms or renting (online supplemental appendix S11, table l). One study89 reported a significantly increased association between LBW and households with no cars and a significantly decreased association for households with >1 car (online supplemental appendix S11, table m).

Preterm birth

Eight studies79 84 86 90–94 reported a combination of adjusted and unadjusted associations between PTB and occupation/social class, employment status, education, housing and income (online supplemental appendix S11, tables n-q). Meta-analysis was possible for the association between occupation/social class and preterm birth, which was reported by two studies79 90 for 515 752 births (online supplemental appendix S11, table n). The pooled meta-analysis for lowest/intermediate occupation compared with highest identified a significantly increased odds (OR 1.41, 95% CI 1.33 to 1.50; figure 8). The subgroup meta-analysis identified that the highest odds for PTB were in the lowest level occupation category (OR 1.51, 95% CI 1.42 to 1.60). Heterogeneity was moderate for the lowest level occupation subgroup analysis but substantial for the overall meta-analysis and the intermediate occupation subgroup (figure 8), and these were not explained by the factors included in meta-regression (online supplemental appendix S12Giii). Sensitivity analysis demonstrated that the meta-analysis results were robust to the effect of any individual studies (online supplemental appendix S12Giv), and there was no evidence of publication bias (online supplemental appendix S12Gii). The GRADE score for preterm birth showed very low-quality evidence (online supplemental appendix S13).

Figure 8

Forest plot for preterm birth (<37 weeks gestation).

Additional data reported for PTB included employment, education, housing and income or other finance measures. Two studies93 94 reported significantly increased odds of PTB (overall, spontaneous and elective) among unemployed women compared with employed (online supplemental appendix S11, table n). Five studies84 90–93 reported an overall lack of association between education and PTB (online supplemental appendix S11, table o). Two studies84 93 reported the association between income and PTB, and another86 reported the association between financial difficulties and PTB (online supplemental appendix S11, table p). For income, there was a significantly increased odds among women in the lowest income group84 in one study but not in another.93 There were also significantly increased odds among women ‘not managing financially’ but not for ‘being behind with bills’.86 A significantly increased risk of PTB was reported for babies born to women living in crowded homes compared with non-crowded homes,91 and there was conflicting evidence91 93 for rented housing (online supplemental appendix S11, table q).

Mode of delivery

Five studies reported the associations between occupation/social class and caesarean delivery (emergency, elective or any, a combination of unadjusted and adjusted data),95–98 instrumental delivery (all adjusted for confounders)96 and induction (all unadjusted).99 Meta-analysis was possible for caesarean delivery. Any caesarean section (elective or emergency combined) was reported by three studies95 96 98 for 50 345 births, whereas emergency and elective caesarean were reported separately by four studies95–98 for >47 496 births (sample size not reported for one study,97 online supplemental appendix S11, table r), and all were included in the meta-analysis. The meta-analysis identified a significantly reduced odds of emergency caesarean (OR 0.85, 95% CI 0.77 to 0.94; figure 9), elective caesarean (OR 0.83, 95% CI 0.76 to 0.91; figure 10) and any caesarean (OR 0.80, 95% CI 0.74 to 0.86; figure 11) for the pooled lowest/intermediate occupation/social class compared with the highest level. The subgroup meta-analysis identified that the lowest occupation/social class had the lowest odds of all caesarean categories (figures 9–11). There was moderate heterogeneity for any caesarean delivery, and this was significantly reduced in the subgroup analysis (figure 11). Heterogeneity was moderate for elective caesarean, which was reduced by subgrouping (figure 10) and further partly explained by the date of data collection (online supplemental appendix S12Iiii). There was high heterogeneity for emergency caesarean pooled data, which was partly explained by subgrouping (figure 9) and further explained to a very small degree by sample size but not by the other factors included in meta-regression (online supplemental appendix 12Hii). Sensitivity analysis demonstrated that the meta-analysis results were robust to the effect of individual studies (online supplemental appendix S12Hiv, Iiv, Jiv), and there was no evidence of publication bias for any category of caesarean section (online supplemental appendix S12Hii, Iii, Jii). The GRADE score for caesarean section showed low-quality evidence (online supplemental appendix S13).

Figure 9

Forest plot of emergency caesarean section.

Figure 10

Forest plot of elective caesarean section.

Figure 11

Forest plot of any caesarean section.

Additional data for caesarean delivery identified a significantly decreased odds among women in the lowest income quintile84 (online supplemental appendix S11, table s), living in council housing, housing association or privately rented accommodation (online supplemental appendix S11, table t) and among women who bent or stood a lot and physically exerted themselves at work98 but an increased odds among women who mostly sat, undertook demanding tasks, worked during pregnancy and worked more than 50 hours (online supplemental appendix S11, table v).98 There was an overall lack of association with maternal education84 96 (online supplemental appendix S11, table u). There was limited data for instrumental delivery96 and induction99 (online supplemental appendix S11, tables w-x) with some conflicting results depending on parity and previous mode of delivery.

Other outcomes reported

Six studies reported no significant associations between occupation and placental problems,100 maternal morbidity,101 102 myocardial infarction,103 antenatal stroke104 or amniotic fluid embolism105 (online supplemental appendix S11, table y). One study reported associations between education level and congenital anomalies106 with significantly increased odds of congenital anomalies for educational attainment levels below a diploma, degree or higher degree qualification (online supplemental appendix S11, table z). All data were unadjusted for confounding factors.


Meta-analysis from this systematic review of 17 130 631 births in the UK and ROI identified that lower level occupation/social class is associated with a significantly increased risk of stillbirth, neonatal mortality, perinatal mortality, preterm birth and LBW by approximately 40%. Additional data narratively summarised suggest there is also an increased risk of stillbirth, maternal mortality and preterm birth among unemployed parents. Further individual level SES factors with some, although limited, evidence of increased risk of adverse pregnancy outcomes include lowest level of education (compared with women who have a degree), housing and income.

Overall, the data on individual and household measures of inequality suggest women (and families) experiencing lower SES face generally worse pregnancy outcomes compared with women who have a higher social status. Education, housing, income, taxation and social protection policies have undermined social mobility, leading to widening and persistent inequalities.1 Austerity has been associated with widening inequalities in a variety of health outcomes, including infant mortality rate.2 107 In response to economic hardship, women are less likely to adopt positive health-related behaviours due to the need to focus on coping in the short term rather than planning for the future.1 108 As such, poverty experienced by those in lower SES might also be a pathway leading to worse pregnancy outcomes alongside poorer nutrition, housing insecurity, financial stress2 and the wider social determinants of health. These associations may be mediated by health behaviours and other risk factors, whereby mothers facing economic hardship are, for example, more likely to smoke during pregnancy. Our analysis has demonstrated a relationship between low SES and PTB (which increases the risk of subsequent infant mortality), and this is also related to low maternal weight and smoking.82 Furthermore, deprivation is related to poor diet,109 low levels of physical activity,110 increased obesity,111 late access to (and low engagement with) antenatal care,112 increased drug113 and alcohol use,114 and exposure to domestic violence.115 Tackling the fundamental drivers of poor pregnancy outcomes is imperative as they are associated with lower educational attainment and disadvantage throughout the lifecourse,116 117 therefore, contributing to the intergenerational cycle of inequality. This has implications considering the post-COVID-19 economic recovery, and the need for governments to not follow austerity again and instead focus on policies that will not further exacerbate inequalities, particularly for women and children.

This review has many strengths as well as limitations. Our search strategy was broad and wide ranging, which included multiple databases supplemented by searching citations and reference lists and contacting experts in each of the four UK regions and the ROI. The grey literature sources resulted in several large datasets being included from national records in all nations of the UK and the ROI. Using national datasets, supplemented by additional academic publications, has resulted in large sample sizes for some outcome/SES combinations, particularly for occupation and for stillbirths, neonatal mortality, PTB and birth weight. We did not restrict by language, and all screening, data extraction and quality assessments were carried out in duplicate to minimise human error. We also did not restrict by pregnancy outcome; consequently, we have wide ranging outcomes highlighting the effects of SES on the health of the mother and baby during pregnancy and up to 28 days postnatal. However, we did have to exclude studies that reported perinatal mental health or breastfeeding outcomes due to the lack of clarity in reporting the time periods of measurement for these outcomes and therefore these require further exploration capturing a longer postnatal period than 28 days. Using a validated quality appraisal tool, all included studies were rated moderate to high quality. However, our findings refer to UK and ROI only, and therefore, further research should be undertaken to enable an understanding of how pregnancy outcomes relate to socioeconomic determinants across other HICs. Furthermore, rich data sources from the ONS and National Records of Scotland were used extensively in this review to document the association between occupation-derived social class and birth weight, stillbirth, neonatal and perinatal mortality. Over the 28 years for which we had data, the occupational classifications evolved in line with changing labour market conditions as did the reporting of birth registrations; therefore, the categories with which we analysed data are not directly comparable. Broad groupings of lowest, intermediate and higher level occupations were required to be able to pool data, which may not reflect any differences within these groups (eg, any differences between unemployed vs manual employment that would have been grouped as lowest level). Additionally, the use of national datasets such as ONS required the calculation of effect sizes using aggregate frequency data provided in the reports; therefore, we were unable to account for potential confounding variables. There was inconsistency in the reporting of adjusted data throughout, where occupation-derived social class included mostly adjusted data for SGA and instrumental delivery, some adjusted data for LBW, preterm and caesarean delivery, and a lack of adjusted data for other outcomes such as offspring mortality, VLBW and macrosomia. However, most education, income/finances and housing data were adjusted, although reported to a lesser extent overall. We had intended to explore adjustment for confounding variables in the meta-regression as an a priori factor; however, this was not possible due to the limited availability of adjusted data for occupation-derived social class used in the meta-analyses. There were some outcomes where both unadjusted and adjusted data were available (eg, LBW online supplemental appendix S11f). Comparing these, the effect sizes and significance were similar in the univariable and multivariable models that included multiple sociodemographic and behavioural factors. There are some additional limitations with the use of routine data sources such as issues of accuracy of data recorded in routine health records and the comparability and granularity of variables between datasets. However, routine datasets tend to be nationally representative and adequately powered to explore patterns in population health, in a timely manner. These strengths are especially important for research where outcomes are relatively rare, such as stillbirth and congenital anomalies, which would require extensive funding and time to collect prospectively. In addition, for research into health inequalities, the use of routine data are a timely method of identifying potentially vulnerable groups who require additional support. Alternative methods that incorporate lengthy data collection periods could be considered unethical when alternative existing data sources are available.

This review is subject to the usual limitations of observational research whereby we cannot claim that there is a causal relationship between individual and household measures of socioeconomic disadvantage and pregnancy outcomes. As such, the quality of evidence highlighted by GRADE was typically low or very low. Although outcomes were downgraded (eg, for inconsistency as heterogeneity was high for all meta-analyses), due to the use of national datasets showing large effects, upgrading also occurred. We have also only found a limited number of studies for some outcome/SES combinations, and therefore, it is possible that there is publication bias (that negative results are less likely to be published) and selective reporting within studies that may impact the assessment of risk of bias. To negate this, where possible, we calculated associations between inequality exposures and pregnancy outcomes when these had not been published (eg, if frequency data were reported in a paper but no association or by using frequency data published in nationally representative datasets such as ONS). In the meta-analysis, only stillbirth had significant evidence of publication bias. The positive associations with stillbirth may therefore be an artefact of significant associations being more likely to be published as only one reported association included in the meta-analysis was not significant. However, we should also consider the sample sizes that contributed to these significant associations, the largest being provided by the ONS (2002–17) and including >7.8 million women in the comparison of lowest versus highest occupation-derived social class and >6.5 million women in the intermediate versus highest comparison. There are some methodological deviations from the published protocol. First, we limited studies to the UK and ROI only, due to the volume of studies from European HICs that we retrieved in the search. Second, we anticipated only including a meta-analysis if there were three of more studies reporting associations between the same exposure and outcome. However, we made the decision to include meta-analysis of two studies in the final paper due to the limited amount of data available for some combinations of exposure and outcome. Most of these studies reported multiple effect sizes for subgroups of participants, which could contribute to the overall pooled estimate and be valuable to explore under-reported exposure and outcome combinations.

While there was substantial evidence for some SES measures, there were very limited data available for housing, income and car ownership, and a lack of data for other measures of individual or household inequality that may be important including other measures of poverty and social isolation. Furthermore, there was a lack of consistency in definitions used between studies, particularly for education, which made direct comparisons difficult, and it was not possible to pool these data in a meta-analysis. Future studies should explore these other axes of inequalities in addition to gain a broader insight into the relative importance and interplay of socioeconomic variables and pregnancy outcomes. Similarly, although we did identify a range of pregnancy outcomes, there was an absence of evidence available for other outcomes that require further exploration, such as miscarriage, postpartum haemorrhage, gestational diabetes, hypertensive disorders of pregnancy, measures of newborn health (eg, Apgar score), attachment and the need for high-risk healthcare services (eg, neonatal intensive care). This systematic review did not include pregnancy outcomes beyond 28 days postnatal, such as breast feeding, postnatal depression and infant mortality. These additional inequality measures and health outcomes warrant further research to supplement the existing evidence base and inform public health and maternity policy and practice, particularly in light of rising inequalities in infant mortality in the UK.2

The early-life health outcomes and exposure to SES inequalities identified in this review have life-long consequences; for example, the prevalence of severe mental health problems is around three times higher among children in the bottom quintile of family income compared with those in the top quintile.118 Over the last 10 years, austerity measures have impacted those poorest in our society the most. As a result of tax and welfare policies, between 2010/2011 and 2015/2016, the number of people in work and living in poverty has increased to 3.7 million.119 Furthermore, funding for supporting community services such as Sure Start Children’s Centres has been widely cut, despite having been shown to reduce hospitalisations and improve inequalities, including reducing SES inequalities in infant mortality.120 Indeed, research found that SES inequalities in infant mortality decreased during a period of social policy investment in England (the English ealth inequalities strategy 2000–2010).107 Consequently, more children are born into poverty with associated inequalities in health outcomes such as those identified in this review, as well as infant mortality that disproportionally impacts deprived communities.2

This study looked specifically for socioeconomic inequalities in pregnancy outcomes in the UK and Ireland alone. These countries have broadly comparable inequalities, which are typically higher compared with other parts of Europe owing to a minimal welfare state.121 Cross-country analysis does, however, suggest there may be important differences in pregnancy outcomes2 10 11 and health more generally.122 Although evidence for occupation/social class differences was greatest in this review, there has in recent times been a move to reframe health inequalities away from individual ‘labels’ to instead focus on power relations and to interrogate the processes that produce them.123 As such, identifying inequalities during pregnancy is a step towards informing where services could adapt provision to effectively support populations that are socioeconomically disadvantaged. However, targeting interventions solely at the most disadvantaged groups will not reduce the steepness of the social gradient; for example, we identified that women in the intermediate-level occupations were also at increased risk of some adverse pregnancy outcomes. Proportionate universalism instead advocates a universal approach with a scale and intensity that is proportionate to the level of disadvantage.124 This requires engagement between health professionals and people from lower socioeconomic groups in shared decision making, which is too often characterised by less information giving, less directions and less socioemotional and partnership building.125 Our review demonstrated that lower SES women are less likely to have a caesarean section, which could be protective against the increased harm associated with an operative birth; however, this could highlight that women from deprived communities are less likely to request a caesarean, or more worryingly, be involved in discussions around birthing options. A greater consciousness of structural constraints, and a willingness by health professionals to systematically break them down, may empower marginalised, lower socioeconomic patients to be more involved in decisions around their care and help reduce inequalities in the long term.

At a policy level, this review highlights that employment and workplace policies need to consider their role in supporting better pregnancy outcomes. The period of employment covers the years when people are starting and raising families and as such is instrumental in reducing the transmission of inequities to the next generation.1 Yet austerity has driven fundamental shifts in the labour market and a push towards employment, whatever its quality or stability, leading to more precarious and part-time opportunities. Since the financial crisis, these changes have given rise to a reduction in real earnings, which for families is exacerbated further by a large reduction in benefits available for working age people and children.126 Indeed, USA research suggests that pushing low-income mothers into work has little health benefit.127 Through shaping policies related to healthcare, public health and social policy, governments can influence the social determinants of health.128 Social protection policies as well as active labour market policies and family policies may help reduce inequalities, although the evidence currently available is inconclusive.126 129

This review has highlighted that systematic collection of routine data in pregnancy allows for associations between upstream socioeconomic factors and pregnancy outcomes to be established as evidenced in the included large national datasets for occupation-derived social class and various pregnancy outcomes. Thus, improvements to the quality and completeness of routinely collected national datasets in the antenatal and perinatal period could have a powerful impact on policy development.

Overall, the findings from this review highlight that pregnancy outcomes are also associated with factors that lie upstream of the health behaviours of women and the care they receive in pregnancy and labour. These findings support the need for ‘whole system’ approaches to explore and address socioeconomic inequalities and pregnancy outcomes. Key areas for action may include: improved surveillance; analysis of the policy landscape impacting pregnant women and their families, with consideration of the role of different sectors and settings (including the National Health Service); and timely and rigorous research and evaluation efforts to further develop understanding of population needs for pregnant women and their families and the ways to address these needs.


The findings from this review provide evidence that there is a worsening of health for both babies and pregnant women who experience higher levels of individual and household socioeconomic deprivation. The strongest current evidence base is for increased risk of mortality, LBW and PTB among lower level occupation-derived social classes. Strategies to improve pregnancy outcomes should consider this evidence base and incorporate approaches that address the wider determinants of health to provide women and their families with the best chance of having a healthy pregnancy and baby and to decrease pregnancy-related health inequalities in the general population.


The project is a collaboration between Public Health Wales and Fuse (UKCRC Centre for Translational Research Public Health). NH, HB, CB and JR are members of Fuse, the Centre for Translational Research in Public Health ( Fuse is a Public Health Research Centre of Excellence funded by the five North East Universities of Durham, Newcastle, Northumbria, Sunderland and Teesside. We would like to thank our colleagues at Public Health Wales, Frank Dixon (National Records of Scotland) and Leslie Marr (The Scottish Health Council) for their help with regards sourcing grey literature. We would also like to thank authors of the studies who provided additional data including Aoife Brick (Economic and Social Research Institute) and Chris Dibben (University of Edinburgh).


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors AM, AJ, JB, SW, CB, HB, NH and JR conceived the study idea and developed the proposal. CR developed the search strategy with contributions from KT, OA, NH, JR, CB, AJ and SW. CR, KT, OA and AO carried out the database searches and supplementary searches. KT, OA, MM, AO and NH completed the screening, data extraction and quality appraisals. MM ran the meta-analysis with support from NH and HB. KT and NH wrote the first draft of the manuscript. All authors edited drafts of this article and approved the final version. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. NH is the guarantor.

  • Funding The review was commissioned by Public Health Wales. Three public health consultants from Public Health Wales are authors on the paper.

  • Competing interests All authors have completed the ICMJE uniform disclosure form at and declare: support from Public Health Wales for the submitted work; AJ, SW, JB and AM are all employees of Public Health Wales.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.