Article Text
PDF
Abstract
Background Socioeconomic position (SEP) is among the most important determinants of variations in health outcomes. This systematic review aimed to summarise the association between socioeconomic disadvantage and the risk of severe maternal morbidity (SMM) and maternal mortality (MM) across high-income countries.
Methods A comprehensive search was conducted in the MEDLINE, EMBASE, CINAHL and PsycInfo databases and Google Scholar from January 2000 to June 2023. Peer-reviewed papers from observational studies conducted in Organisation for Economic Co-operation and Development countries were included. Meta-analyses of comparable studies, a narrative summary and a harvest plot were undertaken.The risk of bias was assessed using a modified Newcastle-Ottawa tool.
Results The final review included 52 papers. In the meta-analyses, compared with the least amount of neighbourhood deprivation, neighbourhood income, neighbourhood poverty and years of education, the ORs for SMM in the highest group were 1.45 (95% CI 1.13 to 1.85), 1.48 (95% CI 1.34 to 1.63), 1.61 (95% CI 0.97 to 2.66) and 1.29 (95% CI 1.22 to 1.37), respectively. Similarly, the ORs for MM among least versus highest amount of neighbourhood deprivation, unemployed versus employed, lower versus higher occupational group and years of education were 2.10 (95% CI 1.57 to 2.81), 1.86 (95% CI 0.95 to 3.66), 1.61 (95% CI 1.03 to 2.51) and 1.90 (95% CI 1.29 to 2.79), respectively.
Discussion In high-income countries across the different measures of SEP, socioeconomic disadvantage is associated with increased risk for SMM and MM. There is a need for interventions across multiple societal levels that will be effective in reducing these inequitable outcomes.
PROSPERO registration number CRD42023399267.
- OBSTETRICS
- PUBLIC HEALTH
- EPIDEMIOLOGY
Data availability statement
Data are available upon reasonable request. Template data collection forms, data extracted from included studies, data used for all analyses and analytic code are available upon request from the corresponding author.
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: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
In high-income countries, research has consistently demonstrated that socioeconomic disadvantage, measured across various indicators of socioeconomic position (SEP), is associated with an increased risk of adverse perinatal outcomes. However, no published systematic review has comprehensively summarised and critically assessed the evidence linking socioeconomic disadvantage to maternal health outcomes across different SEP definitions. Thus, the aim of this review was to review how SEP is assessed in studies of severe maternal morbidity (SMM) and maternal mortality and to synthesise the evidence on the association between socioeconomic disadvantage and the risk of SMM and maternal mortality, considering different SEP measures across various high-income countries, healthcare systems, populations and study designs.
WHAT THIS STUDY ADDS
Our study showed that socioeconomic disadvantage is a risk factor for adverse maternal health outcomes across many high-income countries, regardless of how socioeconomic disadvantage is measured. Individual-level measures did not appear to have a notably larger effect size than area-level measures.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Developing and evaluating interventions targeted at both individual and neighbourhood-level disadvantage may be effective in reducing adverse maternal health outcomes in high-income countries. However, more longitudinal studies are needed with standardised definitions of socioeconomic disadvantage to understand the causal mechanisms in more depth, the relationship between different measures of SEP at different levels and at different times in the life course, and which factors may be amenable to interventions to reduce the inequality gap demonstrated.
Introduction
Socioeconomic position (SEP) is frequently implicated as a contributor to disparate health outcomes. SEP is a multidimensional concept encompassing economic resources, power and status that can influence health throughout life via different pathways and at different levels, such as individual, household and neighbourhood.1
The maternal mortality ratio (MMR), defined as the number of deaths of women during pregnancy or up to 42 days after the end of pregnancy, is 13 per 100 000 live births across high-income countries.2 However, for every woman who dies, many more experience life-threatening pregnancy complications, which are described as ‘severe maternal morbidity’ (SMM). The rate of SMM varies between 0.13% and 2.25% of live births, depending on the country and the definition of SMM used.3 Morbidity studies, therefore, have greater power to draw robust conclusions on healthcare quality and intervention priorities. Furthermore, although individually rare, SMMs collectively are a considerable burden to healthcare systems and to women and have the potential for long-term physical and psychological impacts.
Across high-income countries, the rates of adverse maternal health outcomes have precipitously declined. However, studies have shown that differences in outcomes between women of different social circumstances are widening.4 5 Nevertheless, most of the studies examining the association between SEP and adverse maternal health outcomes have considered SEP as a single measure and at a single level. In reality, different socioeconomic factors operate at different levels and through different causal pathways at different times throughout life.6 Given the sensitive nature of the measures of SEP, proxies are often used, such as area-based measures using an individual’s postcode. The differences in the strength of an association of the distinct measures of SEP, such as low income or occupation, on any increased risk of adverse outcomes are unknown, or whether there is a difference in effect size if SEP is measured at the individual or neighbourhood level.
The aim of this systematic review was to assess whether, in high-income settings, there is a consistent association between SEP and SMM and maternal mortality (MM) across different countries, healthcare systems, populations and study designs. Furthermore, the review aimed to identify which aspects of socioeconomic disadvantage have the strongest association with adverse outcomes. This will help identify both the most at-risk women and which aspects of socioeconomic disadvantage can be targeted for interventions to have the greatest impact on reducing the widening inequalities in SMM and MM.
Methods
The systematic review was registered with the International Prospective Register of Systematic Reviews (PROSPERO) CRD42023399267. This is a database where systematic reviews and meta-analyses are registered to ensure transparency and avoid duplication in research efforts. Methods for the review were based on the Cochrane Collaboration’s Handbook for Systematic Reviews of Health Promotion and Public Health Interventions and followed Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and Meta-analyses of Observational Studies in Epidemiology guidelines for systematic review and meta-analysis.
Patient and public involvement (PPI)
For the development of the study, we consulted birth companions who work to improve the lives of pregnant women who experience inequality and disadvantage, such as women in prison, refugees and asylum seekers or those with social services involvement. Birth companions helped recruit a group of women who have been recently pregnant or given birth and have lived experiences of social or economic disadvantage. These women met with the research group to guide the project, assist with the dissemination of the findings to the communities at the heart of this study and provide feedback on the future directions of research based on the results of this review.
Eligibility criteria
The population, intervention/exposure, comparator, outcome, study design for the study is shown in online supplemental table S1. The inclusion criteria were outcomes reported among women who were either pregnant or up to 1 year postpartum in a high-income (Organisation for Economic Co-operation and Development (OECD)) country and socioeconomic disadvantage measured using Shavers’7 definition (online supplemental table S2). The OECD is a forum where most countries have high-income economies, are ranked as ‘very high’ in the Human Development Index and are regarded as developed countries. The study exclusion criteria are shown in online supplemental table S1.
Supplemental material
Peer-reviewed papers from observational (cohort, cross-sectional or case-control) studies were included if they reported an association between socioeconomic disadvantage and SMM or MM, included a comparator group, and followed the PIECOS framework. Reports from National Maternal Mortality Surveillance Programmes and peer-reviewed published posters and conference proceedings were included.
Database search
A comprehensive search was conducted in MEDLINE, EMBASE, CINAHL and PsycInfo databases, alongside OECD country National Maternal Mortality Surveillance Programmes. The start date was 1 January 2000, as this coincided with the introduction of the Millennium Development Goals.8 Studies were limited to humans with no language limits and full-text availability. Reference lists were screened for studies not included in the database search and relevant systematic reviews were assessed. Google Scholar was used to search the grey literature. Search strategies were built using controlled vocabulary and free-text terms, with input from an experienced clinical librarian. See online supplemental table S3 for details of the search strategy.
Data extraction and synthesis
Studies were extracted to Covidence9 and duplicates were removed.
For all review stages, the first reviewer was DGB and the role of the second reviewer was split between BO, SB, LT, TB and RK. MK and RR acted as the third reviewer who resolved disputes. Titles and abstracts were screened and the full text of potentially eligible studies was retrieved and independently assessed by two reviewers. For articles published in a language other than English, translations were used for abstract screening.
A standardised form was used to extract data for the assessment of study quality and evidence synthesis (online supplemental table S3). Where a study did not report a relative measure of effect (eg, a risk ratio), this was calculated where there was sufficient information provided.
Risk of bias assessment
Risk of bias was assessed using a modified Newcastle-Ottawa scale10 applicable to pregnancy research11 and cross-sectional studies and assessed risk of bias of the result rather than the study quality. For each domain, studies were rated as having either low, some concerns, high or unknown risk of bias based on the selection of the study population, measurement of exposure and outcomes, comparability and missing data (online supplemental tables S5,S6 and S7).
Data analysis
Meta-analysis was performed for studies that had the same measure of socioeconomic disadvantage, comparator and outcome of either SMM or MM if the data were available. If a study reported the result for SMM combined with MM as the outcome, it was included in the SMM analysis only. If there were two studies with overlapping cohorts that used the same exposure and outcome, the study with the largest population was included. A random effects model using the DerSimonian and Laird approach was used due to the high heterogeneity between studies. ORs and their associated 95% CIs are presented for binary outcomes, as this was the most frequently reported measure. Where required, ORs were recalculated using reported frequency data into more comparable socioeconomic status categories. Given the availability of the data, only unadjusted summary ORs were calculated.
Heterogeneity was assessed using I2 statistic and prediction interval. An I2 value greater than 50% was considered indicative of substantial and 75% considerable heterogeneity.12 A prediction interval is well suited to evaluate the variability of the effect of an intervention over different settings because it provides a predicted range for the true treatment effect in a future individual study.13
A harvest plot was used to summarise the findings from all the studies, including those that could not be included in the meta-analysis. The plots were stratified by statistical significance and the height of the bar corresponded to the percentage measure of association. Summaries for each study were presented as a percentage change in the ORs or risk ratios depending on what was most appropriate based on the availability of studies. A narrative summary was used to report the findings from studies that did not have a dichotomous outcome and therefore could not be included in either the harvest plot or the meta-analysis.
Meta-analyses were conducted using Stata V.17.0 (Stata Corp. 2021, Stata Statistical Software: Release 17, College Station, Texas, USA: StataCorp LLC).
Results
After removing duplicates, 18 612 studies were identified from database searches. A further 549 articles from grey literature were screened. 18 811 were excluded after title and abstract screening. Four additional studies were added to the final list from citation searching. In the final review, 52 studies were included, 40 of which were included in the meta-analysis. 10 studies reported data from overlapping cohorts and were included in the harvest plot if there was no perfect overlap between studies. The PRISMA flow diagram is shown in online supplemental figure 1.
Characteristics of included studies
Online supplemental table S8, S9 and S10 summarise the characteristics of the studies included. 34 studies reported SMM as the outcome and 20 reported MM (two studies reported both MM and SMM independently). There were 31 cohort studies, 10 case-control studies, 4 cross-sectional studies and 7 National Maternal Mortality Surveillance Programmes. The studies were conducted over a range of countries, with the highest number (26 studies) from the USA.
The studies used a range of measures of socioeconomic disadvantage, with the most common being neighbourhood deprivation (15 studies), individual educational attainment (13 studies) and neighbourhood income (13 studies). Different definitions were reported for SMM and MM, with the most common being the Centre for Disease Control and Prevention definition (23 studies) and the International Statistical Classification of Diseases, 10th Revision definition (11 studies) for SMM and MM, respectively.
Risk of bias
The results of the modified Newcastle-Ottawa Score were divided into study types (online supplemental tables S11,S12 and S13); the justification for the score is shown in online supplemental tables S14, S15 and S16. None of the studies had a low risk of bias scores across all the items. The greatest risk of bias was seen in the comparability domain, with all but one study scoring high—these studies conducted univariable analysis or adjusted for covariates on the causal pathway. 17 studies did not provide sufficient information to gauge the completeness of the data and therefore were scored as ‘unknown’ for risk of bias in the adequacy of follow-up of cohort/missing data heading. Six studies had a high risk for selection bias due to the poor representativeness of the cohort to the population. 32 studies that used record linkage to determine SMM/MM scored ‘some concerns’ for risk if bias for assessment of the outcome (cohort studies) or the case (case-control studies). Other potential sources of bias (eg, relating to other aspects of selection or the exposure and outcome) were less of a concern due to the data from the majority of studies being routinely collected and the objective nature of the outcome.
Meta-analysis
Neighbourhood deprivation, neighbourhood income and individual education were associated with an increased risk of SMM, with summary ORs of 1.45 (95% CI 1.13 to 1.85), 1.48 (95% CI 1.34 to 1.63) and 1.29 (95% CI 1.22 to 1.37), respectively. However, there was no evidence of an association between neighbourhood poverty and SMM (figure 1). Neighbourhood deprivation, individual occupation and individual education meta-analyses were associated with MM, with summary ORs of 2.10 (95% CI 1.57 to 2.81), 1.61 (95% CI 1.03 to 2.51) and 1.90 (95% CI 1.29 to 2.79), respectively. There was no evidence for an association between individual employment and MM (figure 2).
Forest plots of unadjusted ORs (and their 95% CIs) for severe maternal morbidity among women who (A) had less than or equal to high school education only compared with those who had a higher education, (B) lived in the most deprived versus least deprived neighbourhoods, (C) lived in neighbourhoods with the highest versus least household median income and (D) lived in neighbourhoods with the highest versus least amount of poverty. Data have been collapsed into comparable categories.
Forest plots of unadjusted ORs (and their 95% CIs) for maternal mortality among women (A) who lived in the most deprived versus least deprived neighbourhoods, (B) who had the highest versus lowest level of educational attainment measured at the individual level, (C) whose husband or partner had a routine/manual occupation versus professional/managerial occupation and (D) who (or whose husband) were employed compared with those who were unemployed measured at the individual level. Data have been collapsed into comparable categories.
Heterogeneity
There was considerable heterogeneity in all but two of the meta-analyses, the exception being neighbourhood deprivation and MM and individual occupation and MM, where the heterogeneity was substantial (I2=62.3% p=0.007, prediction interval: 0.87–5.09) and (I2=70.9% p=0.032, prediction interval: 0.01–261.50) (figures 1 and 2).
Harvest plot
The studies that reported dichotomous outcomes are displayed in figures 3–5 using a harvest plot. These studies showed different measures of socioeconomic disadvantage associated with increased risk or odds for SMM and MM—the percentage increase ranged from 4% to 510%. The studies that did not find evidence for an association were more likely to be case-control studies.14–20 No study showed a significantly decreased association between socioeconomic disadvantage and SMM.
A harvest plot showing the association between individual socioeconomic disadvantage and SMM stratified by measure of socioeconomic disadvantage. If a study was included in the meta-analysis but also reported a result of a multivariable model, this result was included in the harvest plot. Numbers are study references as shown in the online supplemental table S8. SMM, severe maternal morbidity.
A harvest plot showing the association between neighbourhood socioeconomic disadvantage and SMM stratified by measure of socioeconomic disadvantage. If a study was included in the meta-analysis but also reported a result of a multivariable model, this result was included in the harvest plot. Numbers are study references as shown in the online supplemental table S8. SMM, severe maternal morbidity.
A harvest plot showing the association between socioeconomic disadvantage and maternal mortality stratified by the measure of socioeconomic disadvantage and whether a neighbourhood or individual measure. If a study was included in the meta-analysis but also reported a result of a multivariable model, this result was included in the harvest plot. Numbers are study references as shown in the online supplemental table S8.
Narrative summary
Two studies could neither be included in the meta-analysis nor the harvest plot as they did not report dichotomous outcomes. Meeker et al21 reported the measure of association as a per cent increase in SMM per one unit increase in the exposure. There was a positive association for housing violations, neighbourhood lower years of education and neighbourhood median income for SMM—for housing violations, the percentage increase was 3.90% (95% CI 1.40 to 6.21), neighbourhood education 0.29% (95% CI 0.10 to 0.47) and neighbourhood income 0.40% (95% CI 0.01 to 0.79). There was no association between the proportions of owner- or renter-occupied housing and SMM. Frolich et al14 reported a continuous outcome and did not find evidence for an association between neighbourhood median income and MM.
Discussion
Evidence from 52 studies conducted in high-income countries showed that socioeconomic disadvantage was associated with an increased risk of SMM and MM across different populations, social and political systems and healthcare settings. These studies differed in their design, exposure and outcome measures and used a wide range of definitions of socioeconomic disadvantage, SMM and MM.
Furthermore, in this study, the direction and magnitude of association remained relatively consistent across various measures of socioeconomic disadvantage, with no single measure of socioeconomic disadvantage having a notably larger effect size. Different aspects of socioeconomic disadvantage are likely to intersect and act by mechanisms that share complex causal pathways. For example, a woman with a lower level of education may have fewer employment opportunities, leading to reduced income and poorer quality housing, all of which can contribute to poorer prepregnancy health. None of the studies was longitudinal, and therefore, none explored the effects of disadvantage throughout life.
In this systematic review, the association between socioeconomic disadvantage and adverse maternal health outcomes remained even when SEP was measured at the neighbourhood level. It has been argued area measures are a poor proxy for individual SEP as there may be misclassification of the exposure,22 for not everyone living in a deprived area is individually deprived, which may have biased results towards the null. A possible explanation of our study’s findings is that individuals living in the same area are not only likely to share characteristics, but the contextual effect of living in a deprived area is a risk factor for poor outcomes beyond individual-level effect and may describe a unique inequity. The relationship between neighbourhood disadvantage and SMM and MM was consistent across several studies (figures 4 and 5), and although these studies did not show causal relationships, if causation could be proven in future studies, neighbourhood-level interventions could be developed and evaluated in attempts to reduce adverse outcomes, which may be less stigmatising than those based on individual factors.
This review did not consider the complex relationship between SEP and ethnicity, immigration or refugee/asylum seeker status, living in an urban or rural environment or distance to a hospital. Although interrelated with SEP, these determinants are not direct measures of disadvantage, have distinct causal mechanisms on health outcomes, and the effects are likely to vary across countries. One of the most notable of these determinants is ethnicity. Minoritsed ethnicities have a high risk of adverse maternal health outcomes and some are also more likely to experience socioeconomic disadvantage. These ethnic differences in SEP are important contributors to disparities in maternal health outcomes. Studying disparities by ethnicity is complex—ethnicity is a mixture of biological factors linked to ancestral history and geographic origins, alongside the social history of specific groups and factors such as racism. There is evidence that differences in health outcomes between different ethnicities exist even within different SEP strata. Ethnicity and SEP, while interconnected, represent different forms of social stratification, each likely a proxy for specific exposures that can affect health outcomes.1
However, despite consistency in effect direction, many studies did not show an association, possibly due to a lack of power. This was seen mostly in either non-population-based case-control studies or cohort studies conducted in a single or a small number of medical institutions. Furthermore, all but one of the studies had a high risk of bias scores for comparability, as the result was either univariable analysis only or if the authors conducted multivariable analysis, they controlled for multiple and potentially highly correlated psychosocial, behavioural and biomedical variables that might lie on the causal pathway, resulting in over adjustment. Regardless of the measure used, most studies included SEP variables without justifying why a given measure was selected over others, without explaining its meaning, and without fully considering appropriate confounding factors. Additionally, there is still limited understanding of protective factors such as community or social network or aetiological factors that could be amenable to interventions.
Strengths and limitations of the systematic review
This review captured studies from different high-income countries, populations and study designs and demonstrated a relatively consistent direction of effect. The search strategy was broad and wide ranging, and studies were included if information on the prevalence of SMM or MM in each socioeconomic disadvantage category was presented as a frequency only with ORs subsequently calculated. Studies were also included if the association between socioeconomic deprivation and SMM or MM was not the study’s main aim. This allowed more studies to be included potentially reducing publication bias. The inclusion of nationwide maternal mortality surveillance reports and other large cohort studies resulted in large sample sizes for some socioeconomic deprivation/outcome combinations.
An important limitation is that it was not possible to assess potential reasons for the high statistical heterogeneity between studies. All but two of the meta-analyses (neighbourhood median income and risk of SMM and individual educational attainment and risk of SMM) showed that the 95% prediction interval crossed the line of no effect, indicating uncertainty in the estimates. The high heterogeneity could be explained by differences in the populations between countries, measurements of exposure, definitions of the outcome and the low number of studies for each meta-analysis. Furthermore, it was only possible to compute unadjusted summary ORs as many of the studies used different effect measures when presenting the results of the multivariable analysis (ORs, risk ratios and risk differences), and it was not possible to convert results to make them comparable across studies. As there was a limited number of comparable studies in each exposure/outcome category, it was not possible to assess publication bias; therefore, selective reporting cannot be ruled out.
Additionally, the USA is disproportionately represented in this review, with half of the studies included having been conducted there. It has been shown in other studies23 that there may be greater health inequalities associated with SEP in the USA compared with other high-income countries. However, due to the limited number of comparable studies in countries outside of the USA, it was not possible to compare the effect sizes between countries.
Conclusion
This systematic review showed socioeconomic disadvantage to be associated with an increased risk for adverse maternal health outcomes across many high-income countries with different healthcare, political and welfare systems. Although estimates of the magnitude of the relative differences in the adverse outcomes associated with socioeconomic disadvantage are small, the size of the population who experience socioeconomic disadvantage is substantial. Therefore, even a modest difference in the risk can have a large public health impact. Nevertheless, due to the high heterogeneity demonstrated in the meta-analyses, uncertainty exists in the summary estimates. Area-based measures of SEP did not show a weaker association than individual measures; therefore, they may still be useful for research using administrative data, the allocation of resources and targeted interventions. However, more longitudinal research is needed that uses definitions of SEP across the life course, adjusting appropriately for relevant confounding factors, as well as addressing factors amenable to interventions to reduce the inequality gap.
Transparency
DGB affirms that this manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as registered have been explained.
Data availability statement
Data are available upon reasonable request. Template data collection forms, data extracted from included studies, data used for all analyses and analytic code are available upon request from the corresponding author.
Ethics statements
Patient consent for publication
Ethics approval
This study is a systematic review and therefore ethical approval is not required.
Acknowledgments
All relevant personnel involved in the preparation and writing of this paper have been included as contributing authors.
References
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.
Footnotes
RG, MK and RR are joint senior authors.
X @Marianfknight, @drremar
Contributors DGB, MK and RR designed the protocol. DGB wrote the manuscript with contributions from MK, RR, SB and RK. DGB and NWR created the search terms. DGB, BO, TH, SB and RK screen abstracts and full texts and extracted data. Disputes were resolved by MK and RR. DGB conducted the meta-analysis with contributions from RR. All authors read and approved the final manuscript. DGB is the guarantor.
Funding MK is an NIHR Senior Investigator. RG receives salary support from the NIHR Oxford Biomedical Research Centre. The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. The study sponsor and funder played no role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. All authors are independent of all funders.
Competing interests All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare that they have no financial or nonfinancial interests that may be relevant to the submitted work.
Provenance and peer review Not commissioned; externally peer reviewed.
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.