Search strategy refinement and implementation

The comprehensive search strategy generated high rates of retrieval of records relevant to the research question of this project.

The search strategy used key terms used in consistently formulated text-based queries and search statements. These terms were based on subject headings, thesaurus terms, or related indexing and categorisation terms, appropriate for each literature database. An example of a detailed final search strategy is given in Appendix 1.

The strategies were also adapted to searching other data sources, such as existing systematic reviews, theses and clinical studies, to ensure the maximum possible retrieval of relevant records. Less-structured search queries were used to search various types of grey literature.

Refinements to the search strategy

In developing the search strategy across a number of the different databases, it became evident that terminologies used for maternal health-care services and immigrant status differ significantly, not only across the original journal articles but also in the different literature resources, in their standardisations of terms for subject headings, thesauri and other indexing and categorisation systems.

For example, across three major databases (MEDLINE, EMBASE and PsycINFO), parallel but differing terminologies were found. Moreover, medical subject headings (MeSH) (originally developed for MEDLINE) are searchable across all three databases, but they do not retrieve comparable answer sets and may need to be replaced or enhanced to do so (i.e. by using Emtree in EMBASE and enhanced MeSH terms in PsycINFO) (see Report Supplementary Material 1).

These concerns required iterative modifications to the original search strategies in each database to ensure that the observed variations in terminology were adequately reflected in both the index terms and the text-based queries that were developed for all of the databases.

Search strategy and study selection

We included all empirically based studies that used a variety of methodologies (see Appendix 2) and focused on immigrant women, maternity care experiences, experiences and interventions, published between January 1990 and January 2018. Studies were published in English and included study locations in the UK (Table 1). We found several studies arising from Scotland and one from Wales; the remaining studies were based in England (see Appendix 3 for the table of excluded studies).

The NS approach relies primarily on the use of words and text to summarise and explain the findings of the synthesis, which are informed by a synthesis of the narrative findings of included papers.86 Gina Higginbottom and Myfanway Morgan have successfully employed this review genre previously and have vast expertise in its usage.87 Narrative synthesis is suitable for both quantitative and qualitative studies, as the emphasis is on an interpretive synthesis of the narrative findings of research, rather than on a meta-data analysis.86 After applying rigorous systematic procedures for searching, screening and selecting literature for inclusion in the review, the articles were appraised for quality and the NS of the findings progressed.

We included studies that focused on immigrant women and adopted the following definition of an immigrant woman for the purposes of our review. A woman is an immigrant if she:

• is born outside the UK
• has lived in the UK for > 12 months or had the intention to live in the UK for ≥ 12 months when she first entered the UK.

Therefore, we included studies on immigrant women in which the population studied fulfilled these characteristics. Based on this, studies that focused on the following population groups were also eligible for inclusion: foreign students, asylum seekers, illegal immigrants and recent legal refugees and immigrants. In many cases the study populations/sample was not accurately and fully described. We therefore used linguistic ability (e.g. the need for an interpreter as a proxy for immigrant status). Our focus was on first-generation immigrant women regardless of their phenotype. Meaning that women of all ethnic groups are including women of white ethnicities, although we encountered few studies that focused on the latter and our review was constrained by the lack of various studies in the scientific literature. For example, we did not identify any studies meeting our criteria that focused on solely white immigrant women, such as Eastern European women.

The search and selection strategies drew on established systematic review methods88 and incorporated recent guidelines for the selection and appraisal of grey literature.89 Two search and selection phases were conducted. The first consisted of searching electronic databases and websites of relevant journals to identify empirical papers published in peer-reviewed journals. We considered empirical papers to be primary research publications that investigated working hypotheses or research questions and tested them by means of observations or experimentation, qualitative investigation or mixed-methods designs. The second phase targeted grey literature and included searches of selected databases, internet-based searches (see Report Supplementary Material 1), reviews of reference lists, and e-mail or telephone contacts with researchers and other stakeholders that had expertise or interest in the target topic. The processes we used to conduct the search and selection process are summarised in Figure 3.

FIGURE 3 Summary of the search and selection methodology. Reproduced from Higginbottom et al.24 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

Our focus was on access and interventions designed to promote or enhance maternity care for immigrant women, as articulated in the original review question. We chose the definition of intervention to be an integration of program facets or strategies designed to create behaviour changes or enhance health status amongst individuals or an entire population’.90 In fact, in practice we identified few interventions that have been rigorously and scientifically appraised (that is not to suggest that these interventions do not exist, merely that they have not been evaluated).

We adopted the population, intervention, control and outcome of interest (PICO) approach to implement the search strategy as follows:

• population – immigrant women
• interest – maternity care
• control – non-immigrant women (implicit comparator emerging in the results)
• outcome of interest – experience of care.

Therefore, our search strategy development was based on:

• search concept 1 – pregnancy, childbirth (implicitly females requiring maternity care), explicit terms covering women/females requiring all types of maternity care (antenatal, perinatal, postnatal, etc.)
• search concept 2 – immigrant populations (which would not fully distinguish between ‘new’ and ‘second-generation’ immigrants, this would be done at the selection stage)
• search concept 3 – terms used to identify access to, use of, deficiencies in, etc., service provision (to help identify groups with poorer health outcomes or vulnerabilities)
• our final answer set of citations included concepts 1, 2 and 3.

Data search strategy and implementation

Development of literature database search strategies and search implementation

Team member Jeanette Eldridge is an experienced information scientist with whom we developed the scoping search strategy (see Box 2). An exemplar of final search strategies (see Appendix 1) incorporated revisions, as per the knowledge base of the academic team members (GH, MM, KB and CE) and included internet and literature database searches. Jeanette Eldridge constructed the detailed search strategies for each literature database and conducted the searches after review of these strategies by the entire team, including the PAG. The search focus was on studies that described access and compared interventions to improve maternity care experiences for immigrant women in the UK (Box 3).

Search implementation included independent double-screening and team review of ambiguous studies. Our search strategy was adapted for each of these databases so that relevant controlled vocabulary, searching techniques and keywords were used consistently. The reference lists of the included studies were reviewed for relevant citations. Additional hand-searches were undertaken of major relevant journals (e.g. Journal of Immigrant and Minority Health, Journal of Health Services Research & Policy and Canadian Journal of Public Health) and in reviews published by topical research groups (e.g. Reproductive Outcomes and Migration, an international research collaboration).41

We used a screening tool (see Table 1) to select studies congruent with our review questions (Box 4) and to ensure robustness of the NS.86 A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)91 flow chart illustrates the selection process that we implemented (Figure 4).

What interventions exist that are focused specifically on improving maternity care for immigrant women in the UK?

• How do these interventions address inequality?
• How do accessibility and acceptability manifest as important dimensions of access to maternity care services, as perceived and experienced by immigrant women?

FIGURE 4 The PRISMA flow chart of the final selection process. Reproduced from Higginbottom et al.24 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

We included studies that focused on immigrant women; therefore, we included studies on immigrant women in which the population studied fulfilled the characteristics previously mentioned (see Search strategy and study selection).

A relevancy appraisal of each record was undertaken by first reviewing the title and abstract (GH, CE, MM, JE, KB and BH). Records were shared with team members via the EndNote, Version 7, collaborative function (Clarivate Analytics, Philadelphia, PA, USA). We retrieved potentially relevant articles for further assessment against our inclusion criteria, with the entire team making the final decision. We used a PRISMA flow chart (see Figure 4) to document the steps used in the search and selection process.91

Management of non-empirical reports and grey literature

Grey literature is a field in library and information science that deals with the production, distribution and access to multiple document types produced at all levels of government, academia, industry and other organisations. It can be in electronic and print formats, and it may not necessarily be controlled by commercial publishing (i.e. in cases in which publishing is not the primary activity of the producing body).89,92 It is not published commercially or indexed by major databases, although can have an impact on research, teaching and learning.89,92 Some examples include theses and dissertations, conference proceedings and abstracts, newsletters, research reports (completed and uncompleted), technical specifications, standards and annual reports.

We utilised the approach of McGrath et al.93 to systematically review high-quality international grey literature (and non-empirical reports identified in the database searches). We also followed the principles expounded by the US National Library of Medicine.89 The process consisted of (1) identifying the grey literature (including constructing a list of websites); (2) selecting documents via screening, preliminary selection, final selection and data extraction; (3) organising and tabulating data into categories; and (4) undertaking a narrative review of data. Our previous database searches had identified some grey literature, but further identification of grey literature required searches of the following databases: Web of Knowledge Science Citation Index, Web of Knowledge ISI Proceedings, ProQuest Dissertations & Theses Database and the Cochrane Methodology Register. We also searched using Google and Google Scholar (Google Inc., Mountain View, CA, USA). In addition, we hand-searched the reference lists of all synthesised materials and approached relevant organisations, especially through the contacts of the PAG. We also corresponded with information specialists and other experts within this field. The PAG assisted with the identification and final selection of grey literature and the interpretation of findings.

Outcomes and management of identified records

The retrieved data sets (Box 5) were downloaded into an EndNote library. Duplicate records were identified and retained in a separate group within the EndNote library. The downloaded records normally included the title and abstract. Keyword terms and other bibliographic information were also included when available.

• Ovid MEDLINE: 1948 to the present.
• MEDLINE In-Process & Other Non-Indexed Citations: to daily update.
• Ovid EMBASE: 1980 to Week 11 2017.
• Ovid PsycINFO: 1972 to March Week 3 2017.
• CINAHL Plus with full text/EBSCOhost: January 1990 to January 2017.
• MIDIRS on Ovid: 1971 to April 2017.
• Thomson Reuters Web of Science:a 1900 to 2017.
• ASSIA on ProQuest: 1987 to the present.
• HMIC on Ovid: 1979 to January 2017.
• POPLINE (via www.popline.org/): 1970 to the present.

When individual full-text documents were required, they were obtained through a number of routes, including the EndNote ‘Find Full Text’ option, subscriptions to electronic journals and access to other physical or electronic resources available through the University of Nottingham, interlibrary loans of articles or books and direct contact with the authors or the originating institutions.

When records could not be definitively selected or excluded, they were assigned to a separate group and further reviewed by two members of the team. The full library was then annotated to show the final decision.

The bibliographic databases that were searched are listed in Box 5.

Data extraction

We conducted the following foundational activities in order to extract data:

• Textual description: a systematic textual narrative was written for each study. We used headings adapted from Popay et al.:86 setting, participants, aim, sampling and recruitment, method, analysis and results (see Appendix 4).
• Tabulation and summarisation of all studies to be included: these tables described the attributes of the studies and the results. Information was extracted from the textual description using the same headings as above and additional headings as necessary. Papers in the portable document format (PDF) were imported into ATLAS.ti qualitative data analysis software (Scientific Software Development GmbH, Berlin, Germany) using the ‘Attributes’ option to allow the tabulation of relevant data.

Element 2: developing a preliminary synthesis

Our analysis process began with the selection of included papers by the research team, with the final selected papers being reviewed by one researcher and the data of each paper extracted into a one-page textual summary (see Appendix 5). A second team member confirmed the study selection; furthermore, the entire team confirmed the selection at the reflective team meeting. We shared advance information with team members prior to meetings so that views could be elicited in an efficient fashion, with careful recording of views and decision-making processes in meeting minutes. Meetings included incisive debate regarding contentious papers. Team members Myfanwy Morgan, Catrin Evans and Gina Higginbottom are experienced systematic reviewers (evidenced in previous publications94), so we harnessed this expertise for the benefit of this review; pertinent practice perspectives by CM and Janette Eldridge provided technical systematic knowledge.

We engaged in careful reading and re-reading of papers to ensure familiarisation with the content. Once the final selection for inclusion was established, we undertook a tabulation of key variables of all included studies and an exemplar can be found in Appendix 5. Tabulation and extraction of key findings was facilitated by the production of the one-page textual summaries, as this format facilitated comparison and contrast of study findings in a systematic and coherent fashion.

The project principal investigator (GH) also undertook a detailed analysis of the narrative findings of each study using ATLAS.ti computer-assisted qualitative data analysis software, engaging in a line-by-line analysis of the narrative findings of each included study. The analysis created over 250 codes. The two perspectives were then merged to create a cohesive interpretation. The aforementioned processes are human resource intensive, iterative and demanded reconsideration several times. However, differences in interpretation between the researchers may still exist, although these differences are likely to be marginal and will not have an impact on the critical conclusions and implications for practice. Inevitably, the perspectives and foci of the research studies included limit our conclusions.

In element 2 (Table 2), we also applied the critical appraisals tools, made an appraisal and included this in the one-page textual summary we created (see Appendix 5).

We proceeded to group and cluster the studies in respect of the data extracted for the textual description. This enabled identification of patterns between and within studies. During this grouping process, the research questions remained salient in our cognition. A particular feature grouped data, for example the method, the country of origin of the sample studied, the maternity care setting, or the main findings. We identified major themes in the narrative findings (see tables in Report Supplementary Material 1).

Following construction of the preliminary themes, we produced code/narrative theme tables to demonstrate how the basic meaning units related to the theme, utilising the codes produced in ATLAS.ti and aligning these to the manually extracted key findings. We defined codes individually in ATLAS.ti, enabling the essential meaning of key phenomena to be captured, facilitating cross-comparison and variations in phenomena to be established and observed. Codes are defined by Miles and Huberman95 as:

. . . tags or labels for assigning meaning to the descriptive of inferential information compiled during a study. Codes usually are attached to ‘chunks’ of varying size words, phrases, sentences or who paragraphs, connected or unconnected to a specific setting. They can take the form of a straightforward category label of a more complex one (e.g. a metaphor).

‘Code families’ can be produced in ATLAS.ti and imported into the graphic ‘network builder’ in ATLAS.ti, resulting in topical or hierarchical graphic outputs. We assigned the codes to code families as illustrated in Figure 5, which shows all the codes associated with code family entitled ‘communication’. This code family is a constituent of theme 3 and serves as an exemplar of the extensive analytical work undertaken. ATLAS.ti facilitated a more sophisticated and nuanced analysis than via manual extraction. These data contributed to the production of a descriptive narrative in respect of each theme. Moreover, we created a tabular matrix illustrating how the themes that occurred are distributed over the individual studies (see Appendix 6). This distribution, to some extent, illustrates the weight of evidence in each theme.

FIGURE 5 Graphic representation of communication theme using ATLAS.ti. HCP, health-care professional. Reproduced from Higginbottom et al.24 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

During the analytical processes we interrogated the data identifying using the following concept suggested by Roper and Shapira:96

• setting – the environment or context
• activities – patterns of behaviour that occur often
• events – rare and infrequent activities
• relationships and social structures – kinship, friendship, bonds, enemies, hierarchical
• general perspectives – the group’s shared understandings
• specific perspectives on the research topic(s) – how people understand the phenomena
• strategies – ways of achieving goals
• process – flow of events how things change over time
• meaning – significance and understanding of behaviour
• repeated phases – depictions of thought processes.

We presented these analytical processes and products at our reflective team meetings to ensure the rigour and robustness of our analytical steps. Through reflection and debate, team members were able to challenge initial interpretations, identify outliers and confirm the coherence of emerging interpretations. Team members independently produced interpretations of the possible meanings of these preliminary themes in order to achieve abstraction, prior to formulation of the higher-level final themes.

Our thematic findings are evolved via a systematic, comprehensive and thorough collaborative team process of comparing and contrasting the emergent themes to establish the ways in which these themes revealed and provided insights into the concepts embedded within the review questions. This iterative process, similar to qualitative research, involved deconstructing the narrative findings into meaning units and social processes as they manifested in the maternity care experiences of immigrant women. As mentioned, individual team members engaged in independent theming of tabular and coded data. We subsequently merged these individual perspectives to form the final harmonised themes, representing a ‘meta-inference’ in respect of the narrative findings of the included studies. Meta-inference is a term used in mixed-methods research to describe the merging of findings from the positivistic and the interpretative paradigms, as is the case in this NS. Tashakorri and Teddlie97 describe meta-inference as ‘an overall conclusion, explanation of understanding developed from the integration of inferences obtained from the qualitative and quantitative strands’.

We have constructed the themes in an indicative fashion (i.e. containing implicit indications) to provide tangible guidance for policy and practice that might be developed into transformational policy- and practice-relevant strategies that will benefit immigrant women and the NHS.

Element 3: exploring relationships within and between studies

Patterns emerging from cross-literary comparisons were subjected to further rigorous evaluation to identify factors that may explain differences, including variance in women’s experiences, in the effects of maternity interventions and in the implementation of maternity services (entire team involvement). This evaluation contributed to our understanding of barriers and enablers that shape maternity services for immigrant women. We also evaluated not only the relationships between the study characteristics and the reported findings, but also the ways in which these relationships may correspond with those reported in other types of literature, including non-empirical and grey literature (which was evaluated separately). Careful attention was paid to the heterogeneity of research methods, methodologies and population characteristics encompassed in the literature through the application of narrative methods.98 Such methods are particularly suited to synthesising such findings.86 Narrative methods99 help us to understand and acknowledge the broader influences of theoretical and contextual variables, such as race, gender, socioeconomic status and geographical location. These methods also enable researchers to understand the shaping of differences between reported outcomes for various study designs, in this case those related to childbearing populations, and the development and implementation of maternity services and health interventions across diverse settings.

Data synthesis and establishing relationships

Grouping and clustering

The data extracted in the tabulation allowed papers to be grouped and thus enabled patterns between and within studies to be identified. Groupings were organised by a particular feature (e.g. location, method, ethnic groups, form of analysis, or main findings) (GH, CE, BH and the PAG).

Thematic analysis

Systematically recurrent or salient themes or concepts across studies were identified. ATLAS.ti was used to manage the data and relevant themes were identified on an inductive basis. We generated over 250 codes in ATLAS.ti and these were mapped against the tabular thematic analysis (see Figure 5).

Ideas webbing and concept mapping

‘Ideas webbing’ conceptualises and explores connections among the findings reported in the review studies and it often takes the form of a spider diagram. ‘Concept mapping’ links multiple pieces of information from individual studies using diagrams and flow charts to construct a model with relevant key themes (all team members were involved).

Critical reflection

Popay et al.86 recommended that a summary discussion of the synthesis should be provided, which includes the following: (1) methodology of the synthesis, focusing on the limitations and their possible impacts on the results; (2) evidence used in terms of quality, reliability, validity and generalisability (for quantitative papers), and possible sources of bias (for qualitative papers); we applied the principles of Lincoln and Guba100 (confirmability, transferability, credibility and dependability); (3) assumptions made; (4) discrepancies and uncertainties identified, and how discrepancies were handled; (5) areas in which evidence is weak or non-existent as possible areas for future research; and (6) whether the evidence is ‘thick’ or ‘thin’ and the similarities and differences in the evidence (entire team).

Methodological quality of included studies and quality appraisal

Key to ensuring the robustness of the synthesis is the methodological quality of key literature, the quality and quantity of the evidence base and the analytical methods used to develop the NS. We critically appraised research studies and weighted each accordingly (GH, MM, CE, and BH). Studies that lacked technical quality or were of questionable methodological integrity101 were critiqued and/or noted for key flaws in the synthesis. Although we identified a number of studies rated as weak, we did not reject any studies on the basis of quality. Despite debates existing regarding the availability of well-established methods for evaluating the reliability and trustworthiness of qualitative and mixed-methods studies, the methodological analysis and synthesis were conducted in a relational and systematic manner.

Appraisal of included studies was assessed using tools from the Center for Evidence-Based Management (CEBMa).102 We used Good Reporting of A Mixed Methods Study (GRAMMS) for the mixed-methods studies.103

We used high, medium and low as appraisal categories.94 This approach is congruent with recent publications from the Cochrane Qualitative Research Group’s Confidence in the Evidence from Reviews of Qualitative Research (CERQual) publications, as they use this type of evaluation104 and it was previously used by Higginbottom et al. in published studies.87 Studies are classified in three domains, high, medium and low, to enable a ‘macro’ evaluation:

• High was assigned to studies that used a rigorous and robust scientific approach that largely met all CEBMa benchmarks, perhaps equal to or exceeding 7 out of 10 for qualitative studies, 9 out of 12 for cross-sectional surveys, or 5 out of 6 for mixed-methods research.
• Medium was assigned if a study had some flaws but these did not seriously undermine the quality and scientific value of the research conducted, perhaps scoring 5 or 6 out of 10 for qualitative studies, 6–8 out of 12 for cross-sectional surveys, or 4 out of 6 for mixed-methods research.
• Low was assigned to studies that had serious or fatal flaws and poor scientific value, and scored below the numbers of benchmarks listed above for medium-level appraisals in each type of research.

The past decade has witnessed a growth in the development of new approaches to systematic reviews, especially in the domain of qualitative evidence synthesis. Concurrently, innovative approaches to assessing quality have evolved. Popay et al.86 suggest that we do evaluate the ‘richness’ of studies. Furthermore, this research team have utilised this approach in previous funded studies and publications.86 Popay et al.86 defined richness as ‘the extent to which study findings provide in-depth explanatory insights that are transferable to other settings’ (Table 3). We used established criterion in previous studies87 and appraised all the studies in this review using this evaluative tool (Table 4).

Element 4: rigour, reflexivity and the quality of the synthesis

Reflexivity in research demands a self-conscious and explicit acknowledgement of the impact of the researcher on the research processes, interpretations and research products. Reflexivity also requires acknowledgement of inherent power dimensions, hierarchies and prevailing ideologies that might influence interpretations and subsequent knowledge production and research products. Our social identities, gender, sexuality, professional socialisation, ethnocultural orientation and political lenses further coalesce to provide a specific perspective on any given phenomenon.

The review team members come from diverse ethnocultural backgrounds and have researched in the fields of ethnicity and health for > 20 years in several global contexts, conducting many externally funded studies. We are imbued with a strong personal and professional commitment to the eradication of inequalities and allegiance to contemporary equality and diversity agendas. From a reflexive perspective, this is important given that immigration is a global phenomena and given the inherent vulnerability of some immigrant women.

Reflexive analysis alerts us as researchers to emergent themes and informs the formal and systematic process of analysis. Murphy et al.105 define reflexivity as:

. . . we mean sensitivity to the ways in which the researcher’s presence in the research setting has contributed to the data collected and their own a priori assumptions have shaped the data analysis.

Team members hold a deep knowledge of the semantics associated with describing ethnocultural groups and study populations, which varies between global and sociopolitical contexts. The term immigrant is regarded as pejorative in nature in some contexts. We hold deep understandings of ideologies that manifest in research, publications and, indeed, funding and review processes. Two team members are immigrants to the UK and two team members are UK citizens who have been immigrants to other countries. Of note is that the primary researcher and post-doctoral fellow was a male researcher of Pakistani heritage, which may have created some implicit threats/bias to the interpretation of findings and data. We believe our reflective team meetings, consensus decision-making and incisive debates mitigated these tensions.

Our reflective team meetings enabled and facilitated challenges to primary selection of included papers, emerging interpretations and the extent to which individual subjective lenses had influenced these decisions. The development of themes included an exercise undertaken independently by each team (five team members and one PAG member). Our joint deliberations necessitated the constant review and reading and, in some cases, reviewing the theme allocation and evidence. Therefore, we achieved a refined and nuanced approach, gaining insights into the focus of review.

All decisions and processes were reviewed in reflective team meetings by the research team in order to achieve consensus in decision-making. In our proposal (submitted to the NIHR in 2016) we elected to use the newly conceptualised CERQual, owing to the interpretive nature of our review with a focus on narrative findings. This relatively new tool maps out pertinent dimensions demanding consideration of each individual review finding, considering four dimensions: (1) methodological limitations; (2) relevance, meaning the extent to which the relevant study findings interface with the review question; (3) coherence, that is, the extent to which the finding is grounded in the data from the relevant studies; and (4) adequacy of data, an assessment of the richness of the data informing the finding.

We applied these dimensions to our NS findings that arose from qualitative, quantitative and mixed studies, following personal communication with a CERQual author (Dr Andrew Booth, ScHARR, University of Sheffield 2018),104 regarding the utility of CERQual to our NS. Booth agreed that a modified version may be applicable. We developed and applied this modified version. However, we found the application of CERQual to be problematic because the concepts embedded within it are wholly orientated towards qualitative research. There is little guidance on applying this tool, although team members did engage in an online workshop. In our view, this tool is not wholly orientated towards a NS review. Nevertheless, we did engage in this process and, certainly for the qualitative papers, established a moderate to high level of confidence and relevance in respect of our included studies.

Within the published NS reviews we have seen little evidence of attention to publication bias; however, we sought to eradicate any potential bias by undertaking a comprehensive and exhaustive literature review that included grey literature and follow-up e-mails with authors seeking greater clarity and explanation of opaque issues. A number of the included research studies were identified via ProQuest and e-Theses and do not appear as publications in peer-reviewed scientific journals.

At our national research symposium, we presented our preliminary findings to a wide range of health professions: obstetricians, GPs, midwives, two immigrant women with experience of maternity care services, academics, and voluntary and community workers (see Report Supplementary Material 2). The approach can possibly be considered contentious in respect of systematic review, as participants have no knowledge of the original included papers but do hold deep topic knowledge. However, we found broad support for our findings and facilitated group work activities in order to challenge our initial interpretations. These challenges resulted in the construction of theme 5 (discrimination, racism, stereotyping, cultural sensitivity, inaction and cultural clash in maternity care for immigrant women). These focused activities collectively contribute to the confidence in the review findings, providing verification and validation of the themes.

Individual and group work at the symposium focused on four questions (Box 6). The symposium programme consisted of six sessions and is contained in Report Supplementary Material 2. We shared preliminary findings from the review along with commentary from a GP (CM), on the implications of the findings for practice, and a critical race theorist (KS, Director of Centre for Research on Race and Rights, University of Nottingham). Two recent immigrant women (service users) also shared their experiences of using maternity care services in the UK. The participants were afforded the opportunity to reflect on the implications of the findings for policy and practice both individually and in small groups.

1. Please share your perspectives on implications of the findings for maternity care and service delivery (identifying three key points from each group).
2. From your knowledge and/or professional experience, please identify any intervention(s) that have been or are being implemented to address maternity care needs of immigrant women. Please specify the place, the context and contact details from which we can get further information.
3. How can current policy be used to empower immigrant women to receive good-quality maternity care?
4. In your view, who needs to know the implications of findings from this review?

Our event also included immigrant women with experience of maternity care services who gave presentations regarding their experiences (see Report Supplementary Material 2). We used the event to confirm the credibility of the team’s interpretation and synthesis of the included studies. We gave presentations regarding the theme and then elicited comments. The additional comments elicited at the event were used to inform the discussion; we have not used any of the information elicited in the actual themes, as these are based on the scientific research evidence.

Practical challenges

Use of EndNote for reference management

As the review team was scattered geographically, the sharing features of EndNote proved unhelpful for double-blind screening. References downloaded very slowly from main library.

The EndNote library was set up as a shared library for all of the project team to access. This involved setting up the senior research fellow as the ‘owner’ of the library. The EndNote desktop program installed for each project member could share access to this library and synchronise the content through EndNote online. The library could be backed up regularly using the compressed library feature.

However, issues with synchronisation and with speed of access to the EndNote library prevented full exploitation of the sharing and collaboration features. Although the issues were raised with EndNote support and a number of potential corrections explored, these suggestions did not resolve the problems within a reasonable time, so the project team adopted a different tactic for sharing the EndNote library content for review. The full library content was split into tranches of roughly equal numbers of records and provided to each project member as a separate EndNote library to review individually. The full library content was then annotated to reflect the selections and exclusions from each of the reviewers.

Access to full-text studies

A few small-scale studies that had been conducted back in the 1990s were not available in a digital format. We requested hard copies, but the authors themselves struggled to find them.

Studies published in low-ranked or uncommon journals

The University of Nottingham did not subscribe to some of these journals, so we asked The British Library for access. In some cases, The British Library had to order the articles from another site, which resulted in delayed access.

Lack of full participation from volunteers for patient and public involvement

Some volunteers who signed up initially did not participate later. All were sent invitations to attend the meetings, but a few sent their apologies, others did not respond, and two cited other commitments and resigned.

Conceptual challenges

We observed that authors used a variety of terms to identify participants. Terms used included BME, disadvantaged, immigrants, South Asian origin women, Somali women, Somali refugee women, Palestinian women and Chinese women (including women from Hong Kong, Taiwan and mainland China). However, the term ‘Chinese’ could be used without clearly specifying whether or not the women were born outside the UK. It could refer to immigrants or to UK-born participants who still shared Chinese culture, beliefs and experiences, or to their parents who were born elsewhere.106 The titles and abstracts alone were often insufficient to make a decision about exclusion, so we read all the papers in detail. When there was a lack of clarity about the study sample, we contacted the authors to ask whether or not the participants were immigrants. Maintaining a contemporary contact list of these authors was challenging, as some of them had retired or held new positions. Our first step was to e-mail the authors to confirm immigrant status. In some cases, however, no valid e-mail address was available.

Challenges were accounted identifying specific findings for immigrant women in mixed population studies.

Search results

Studies included in the review, findings and evidence

Overview of the included studies

Our systematic review using NS86 identified 40 research studies in the scientific and grey literature. A broad range of ethnocultural groups and methodological genres are included in this review (see Appendix 6 and Figure 6 for theme development).

FIGURE 6 The total numbers of studies involved in each theme. Reproduced from Higginbottom et al.24 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

Methodological genres

Qualitative studies

Of the 40 studies included in this review (Figure 7; see Appendix 2), we identified 22 as qualitative research studies, employing a range of qualitative methodologies and approaches.4,107–127 However, many of these studies did not specify a qualitative methodological genre, but instead employed a more generic qualitative approach and described only the data collection tools used. For example, some presented multiple longitudinal case studies of participants (asylum seekers and refugees), about their maternity care experiences that included photographs taken by the participants, field notes and observations, in addition to researcher interviews. Another example109 was a case study of an ethnocultural group, immigrant women of Somali origin, that used semistructured interviews and focus groups. Some studies used focus groups and interviews conducted in the language of the population group, for example Bengali, Sylheti, Urdu and Arabic. Others used in-depth interviews, open-ended questions, group story-sharing sessions and individual biographical life-narrative interviews. By contrast, a few studies specified a qualitative interpretive approach, which used hermeneutic phenomenology and focused ethnography.

FIGURE 7 The range of publication dates for the included studies (1990–2016). Reproduced from Higginbottom et al.24 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original figure. (Figure not reproduced here; see the original article.)

Quantitative studies

We identified eight quantitative studies that used a questionnaire as a tool for data collection.14,49,128–133 All of these population-based studies and cohort surveys were cross-sectional; none was longitudinal.

Mixed-methods studies

We identified 10 mixed-methods studies using a range of designs (i.e. these studies employed both qualitative and quantitative dimensions).3,20,47,134–140 For example, Duff et al.134 reported a two-stage psychometric study in which focus groups and interviews were used in the first stage to develop a questionnaire for an ethnocultural group (Sylheti). In the second stage, quantitative methods were used to test and evaluate the acceptability, reliability and validity of the questionnaire. Other mixed-methods designs included (1) interviewing a small sample of the participants after collecting data from a large-scale survey; (2) conducting semistructured interviews with a small sample of participants based on quantitative data routinely collected from a large group of participants; and (3) using face-to-face, postal and online questionnaires to collect data. One of the studies used Q methodology, which uses questionnaires with structured and unstructured questions.

Studies focusing on specific ethnocultural groups

The chosen studies included participants from a wide range of ethnocultural groups that originated in diverse countries on different continents, including Asia (e.g. Bangladesh and Pakistan), Africa (e.g. Somalia and Ghana) and Europe (e.g. Poland). In some cases, the sample was drawn from a single ethnocultural group, such as Bangladeshi.136 However, most of the studies were undertaken on mixed samples of immigrant women originating from different countries (e.g. Somalia, Bangladesh and Eastern Europe). Of the studies with mixed populations, we only included those in which the findings where clearly assigned to immigrant women. We encountered a number of relevant studies that did not do so; therefore, we excluded them and entered them in the list of excluded studies (see Appendix 3). Please see Report Supplementary Material 1 for a detailed overview of the characteristics of study participants.

Studies focusing on immigrant women without a clearly specified ethnocultural group

We identified 16 studies that used the term ‘immigrant women’ generically and did not clearly specify an ethnocultural group. In deciding to include these studies, we believed that legitimate proxies for immigrant status could be the specified use of an interpreter or the participants having countries of origin or birth outside the UK. Some studies reported immigrant women arriving from 14 different countries, but did not specify the women’s country of birth; these studies were included.

Studies sorted by immigrant category of the participants

More than half of the included studies (26 in total4,47,49,107,111,112,115,119–121,123–126,128–136,138–140) did not clearly specify the immigrant category of the population they studied (such as economic migrant, asylum seeker or refugee).

Other studies specified the immigrant category as asylum seekers only;109,114,118,122,137 refugees only;108,113,117 or a mix of immigrant categories that included spousal immigrants, economic migrants, asylum seekers and refugees.3,14,20,116,127 Although a range of identifiers were used, we noted a pattern of reducing immigrant women and their complex cultures, ethnicities and lives to simplified contexts or situations (e.g. ‘asylum seekers’) and of using ‘immigrant’ as a generic label for non-British and potentially for non-white women.

Studies sorted by phase of maternity care

Of the included studies, 17 studies3,20,47,108–110,112,118–121,123–125,133,134,136,140 were undertaken across all phases of maternity care (antenatal, intrapartum and postnatal). Nine studies14,49,111,114–117,128,138 reported on antenatal care alone and six studies116,119,126,131,135,137 focused on postnatal care. Two studies122,129 focused on antenatal and postnatal care, but did not include the intrapartum phase. One study107 focused on the intrapartum and postnatal phases of maternity care and one study4 did not specify the maternity care setting.

Studies sorted by year of publication

The dates of publication for the included studies range from 1990 to 2016 (see Figure 7).

Gaps in the evidence

Few studies focus on (1) recent immigrant and refugees (e.g. from Eastern Europe or Syria) or women of white ethnicities; (2) intrapartum care and immigrant women; (3) quantitative work using sophisticated analyses of relationships, predictability or socioeconomic variables relating to immigrant women and maternity care needs; and (4) economic analyses.

Studies excluded from the review

Some studies of mixed populations gave no clear indication of whether or not the findings were specific to immigrant populations (see Appendix 3).

In the UK context, the term ‘BME’ is usually used in research literature141–143 and policy literature6 to describe population groups that have come to the UK from other countries. However, BME is a very broad term that masks the heterogeneity of this population,144 and individuals in this group would have different expectations of health services and diverse engagement experiences.144 For example, BME groups would include both first- and second-generation immigrants. However, the two generations of immigrants would have different experiences when engaging with health-care services. Crucially, the first generation often experiences language barriers, but the second generation is born and socialised in the UK and do not usually experience the same linguistic challenges.145 Nonetheless, both groups are labelled together as BME. Such studies were therefore excluded.

We identified two studies146,147 that reported on mixed samples of non-immigrants and immigrants and provided no separate or clearly identifiable findings for immigrant women. We excluded such studies, as our review focused on only maternity care experiences of immigrant women. Using aggregated findings from mixed samples would have contaminated specific findings for immigrant women in maternity care and would have damaged the integrity of the review.