Black and minority ethnic group involvement in health and social care research: A systematic review

Abstract Background Patient and public involvement (PPI) in research is growing internationally, but little is known about black and minority ethnic (BME) involvement and the factors influencing their involvement in health and social care research. Objectives To characterize and critique the empirical literature on BME‐PPI involvement in health and social care research. Search strategy Systematic searches of six electronic bibliographic databases were undertaken, utilizing both MeSH and free‐text terms to identify international empirical literature published between 1990 and 2016. Inclusion criteria All study designs that report primary data that involved BME groups in health or social care research. Screening was conducted by two reviewers. Data extraction and synthesis Data extraction and quality appraisal were performed independently. Data extraction focused on the level(s) of PPI involvement and where PPI activity occurred in the research cycle. Studies were quality‐assessed using the guidelines for measuring the quality and impact of user involvement in research. Data were analysed using a narrative approach. Main results Forty‐five studies were included with the majority undertaken in the USA focusing on African Americans and indigenous populations. Involvement most commonly occurred during the research design phase and least in data analysis and interpretation. Conclusion This is the first systematic review investigating BME involvement in health and social care research internationally. While there is a widespread support for BME involvement, this is limited to particular phases of the research and particular ethnic subgroups. There is a need to understand factors that influence BME involvement in all parts of the research cycle.


| INTRODUCTION
Patient and public involvement (PPI) in research has been defined as "research being carried out 'with' or 'by' members of the public rather than 'to', 'about' or 'for' them". 1 Involving patients and public in research has become central to the health research policy context, both in the UK and internationally. [2][3][4] This is reflected in the proliferation of official guidance on principles of best practice in PPI in countries such as UK, 1 Canada 3 and Australia. 5 The growth of PPI has been underpinned by suggestions that actively involving service users in all aspects of the research process has the potential to improve research quality, relevance, impact and integrity as users can provide unique and relevant perspectives. 6,7 There is also a moral or human rights aspect to involving those who are the recipients of care and responsible for financing these services, either in a direct (fee for service) or indirect (eg, via taxation funded systems) manner. 8 In the UK, interest in PPI within the National Health Service stems from demands of the public for a greater voice and involvement in making decisions regarding their services and politicians' demand for greater efficiency, service quality and effective use of public funds that reflect the influence of the New Public Management approach to health services. 9 Terminology within this area is varied and lack consistency, as researchers tend to use participation* engagement † and involvement interchangeably. 10 This study will focus on involvement of people from Black and Minority Ethnic (BME) backgrounds, as existing systematic reviews in the area of PPI and health and social services have assessed the impact of PPI on health and social care research 4 and UK healthcare services more widely. 11,12 The reviews offered evidence that involvement can have positive impact on research in terms of enhanced research quality, relevance and ensuring appropriateness 4,11,12 as well as identifying a smaller evidence base of challenges to PPI 4 such as potential tensions between researchers and PPI contributors ‡ due to different perspectives; results may not be perceived as important and concern that PPI contributors may disseminate the results before publishing in academic journals. Moreover, these studies 10,11,13 found that a key limitation to PPI evidence base was poor quality of reporting impact, with a small number of studies defining PPI, little theoretical conceptualization, lack of robust measures of impact and lack of detailed account of descriptive evidence.
Other reviews on the topic have focused on how to identify and engage patients in health services research, as well as benefits and barriers to patient engagement. 14 Despite the widespread promotion and inclusion of PPI in the last ten years, some have suggested that involvement is limited in scope as to whom is involved, with PPI activity not mirroring the diversity of the population. 10,15,16 For example, studies have reported either a difficulty in involving or a lack of involvement of diverse groups (eg, BME groups) in health and social care research. 17,18 In addition, people from BME groups often experience inequalities in care, [19][20][21] may also have specific health and social care needs and/or views on service design/provision, which, if not taken into account during the research or service design phase, could mean that the end product/services(s) might not necessarily reflect their specific needs and therefore might not be used/acceptable and therefore effective. In order to address this, it is not only important but also necessary to involve members of BME communities in a PPI capacity so that health and social care research is more relevant to them and produces appropriate research and health outcomes.
Traditionally, in Western literature, the term BME, Black, Asian and Minority Ethnic (BAME) Groups or "ethnic minority" has been typically used to refer to minority populations characterized by their non-White origin. 22,23 However, the definition of minority ethnic groups varies and is naturally defined and described differently in different countries. The label of BME is therefore somewhat of a blunt instrument, which attempts to provide comprehensive coverage for what are highly nuanced, locally defined populations which may likely have differing health and social care needs. It is difficult therefore to have a holistic definition for ethnic minority groups that can be applied globally and which acknowledges this complexity. However, the BME label is seemingly most prominently used and recognized within the health and social care literature, the topic of this research and therefore will be utilized in this study. Furthermore, this study will utilize from the included studies, the authors' reported definitions of what constitutes a BME population in their individual studies.
In summary, although the lack of BME involvement in health and social care research is clearly identified as an issue from the existing PPI literature in the area, no study to date has provided a systematic review of PPI for people from BME backgrounds in health and social care research. This study intends to fill this gap in our current knowledge.
This study therefore aims to review the literature in order to: 1. Identify, characterize and synthesize the literature on the involvement of people from BME groups in health and social care research 2. Identify any reported factors that may promote or inhibit BME involvement in health and social care research.

| METHODS
A systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 24

| Search strategy
A comprehensive search strategy included combinations of four main blocks of terms, including and relating to public involvement (patient, public, consumer, citizen, carer, lay, service users, stakeholder, family, relative, survivor), the type of involvement (participation, collaboration, engagement, partnership, consultation, user-led, consumer or *By participation, we mean taking part in research either by completing questionnaires, taking part in interviews or focus groups. † By engagement, we mean sharing information about research with members of the public or disseminating study findings to research participants or members of the public. ‡ For the purposes of this review, people involved in health and social care research will be referred to as PPI contributors. patient panel, advisory board/group/panel), health and social care research (health services, health care, public health, social care, etc.) and ethnicity § (ethnic*, race, cultur*, minorit*) using a combination of MeSH terms wherever relevant and possible (see supplementary file S1 for example search strategy). Six main electronic bibliographic databases were searched for potential studies from January 1990 to November 2015 (and then updated in April 2016 to include any recent and relevant studies as the initial searches were >6 months old): MEDLINE, EMBASE, PsycINFO, Evidence Based Medicine (EBM) reviews (Cochrane library), Cumulative Index to Nursing and Allied Health Literature (CINAHL) and Healthcare Management Information Consortium (HMIC) were searched. Additionally, citation searches and reference lists of included studies and systematic reviews supplemented the database searches (see Figure 2).
A lack of medical subject headings (MeSH) for PPI has been previously reported, and MeSH terms on PPI need to be developed in order to identify more sensitive ways of searching. 10 In order to compensate for this, and identify studies that were not mapped to MeSH terms, free-text terms were used resulting in long and complex search string similar to Brett et al. 10 This approach was necessary because there is no consistency in the way databases index studies relating to PPI, definitions, conceptualizations and poses challenges for developing search strategies. 10 Moreover, several scoping exercises in different electronic databases were applied to maximize the sensitivity and specificity of the developed search strategy. For example, the search term "critical friend" ¶ did not yield any results; the term "PPI" was excluded as it minimized the retrieval of relevant studies as PPI has different meanings.
While the purpose of the review was to identify relevant articles focusing on BME-PPI, studies utilizing CBPR and related participatory approaches identified by the searches and which met the inclusion criteria were included in this review (despite it not being used as a search term or being a focus) due to the fundamental shared characteristic of a philosophy of partnership and collaboration in both CBPR and PPI, between those conducting the research and those for whom the research is focused on or about. Furthermore, the use of INVOLVE's definition in this review, that is to look for research "carried out 'with' or 'by' members of the public" also meant there was scope to include any studies retrieved by the search which appeared relevant to the overall aims of the study. The overlap in terminologies used to describe "involvement" is a complex issue, and a detailed discussion comparing CBPR and PPI is outside the scope of this review. However, a table has been presented (see Appendix 1) to briefly offer an overview of similarities and differences between these two approaches.

| Inclusion and exclusion criteria
Studies were included in this review if they met the following criteria: Population: Adults (18 years or above) classified as being from a BME group(s) explicitly by the authors of the study within the paper itself. Members of any BME groups as defined by the authors of the studies themselves and from any country were included. Studies focusing on migrants including refugees, asylum seekers of different nationalities identified by authors as minority ethnic groups, are included even if detailed descriptors of their ethnicities were not available. In these cases, the population was defined on the basis of "countries of origin." While the populations identified in this review as BME may be different (eg, indigenous peoples) due to characteristics such as language, ethnicity, culture, migration, all of these groups share similar key characteristics in that they are all likely to experience health inequalities, discrimination, racism and stigmatisation that can marginalize these populations and therefore are included in this review. If studies exclusively focused on majority groups or a combination of minority ethnic and majority groups where the data from minority ethnic groups was not clearly identifiable, then they were excluded.
Types of studies: All study designs reporting empirical, primary health or social care research regarding PPI of the population of interest as outlined above were eligible for inclusion.

| Study selection
An EndNote library was utilized to combine and export the results of the searches from different databases. Duplicates (n=1804) were removed prior to the selection of studies (see Figure 2). Study selection was completed in two stages. First, two reviewers (SD and SC) independently screened titles and abstracts in order to identify eligible and relevant studies. Inter-rater reliability was high (kappa coef-ficient=. 89). Subsequently, full text of relevant studies was screened and reviewed in full by the first author for eligibility and a random subset sample of 50% were screened independently by a second reviewer (SC) (kappa coefficient=.73). Any disagreements were resolved through discussions.

| Data extraction and critical assessment
A data extraction form was devised in Microsoft Excel and piloted on four randomly selected studies. The following descriptive data for included studies were extracted: • Study characteristics-authors, year, country, type of study/study design, health topic focus • Participant characteristics-types of people involved in health research (including ethnicity) • Main outcomes-reported methods of recruitment** and communication, † † reported levels of PPI (ie, consultation, collaboration or user control), description of PPI activities within the research study (eg, identifying and prioritizing research topics, research design and management, data collection, data analysis or dissemination) that is where they were involved in the research cycle 27,28 (see Figure 1), and, any reported factors to facilitate or inhibit BME involvement.
The first author extracted the data, and the second reviewer (SC) extracted a random sample of 20% (n=10) independently. No substantial disagreements were observed.

| PPI involvement
Five PPI contributors were involved in reviewing the systematic review protocol, with three of the five PPI contributors coming from BME backgrounds. Three of the five PPI contributors had considerable experience of PPI in research. Two contributors who lacked prior research experience were provided with an informal educational session on systematic reviews, using a lay summary adapted from previous work. 29 This education covered what a systematic review is, the processes involved and advantages and disadvantages of a systematic review. The purpose of this session was to provide them with a basic understanding of the systematic review process, to help them when reviewing the protocol. Despite this educational session, the PPI contributors felt that they did not have the relevant training and skills and as they were relatively new to research. Due to the availability of limited resources and time, it was agreed that the PPI contributors would not be involved further in the systematic review process.
PPI feedback from the review protocol was therefore more focused on clarifying and understanding the review process. However, the remaining three PPI contributors provided feedback on the review protocol, resulting in the addition of two new search terms (steering group and patient advocate). The PPI contributors also **The term "recruiting" here refers to the recruitment of PPI contributors to either carryout research (user-controlled or as peer interviewers) or work in partnership with researchers rather than being participants/subjects in a research study. Methods of recruiting could include advertisements in GP practices, online, etc. † † Methods of communication here include face-to-face, telephone, online via e-mail or forum, postal. 27,28 felt that there was a need to consider how minority ethnic groups were defined, due to the variety of possible definitions internationally. It was agreed that author definition would be an acceptable and consistent definition.

| Critical assessment of the studies
There are difficulties when assessing the quality of PPI and related activities, as PPI activities vary from study to study and PPI activities within studies can be designed differently to the actual study itself. 10 Therefore, this review assessed the quality of PPI-related activities rather than appraising the quality of studies themselves, as it allowed a focus on the capture of PPI activity, which is the main outcome of interest for this review.
A critical appraisal checklist designed by Wright et al. 30 was utilized to evaluate the quality of PPI reporting within published research. The first author extracted data and then a second reviewer (SC) checked a random sample of 20% independently. At this stage, disagreements were resolved through discussion with two other authors (SG and RM).

| DATA SYNTHESIS
Data were reported in a narrative fashion due to the heterogeneous nature of the included literature. 31 This entailed familiarization with the papers and identification of data pertaining to themes related to study aims/outcomes. Data were extracted in accordance with the relevant outcomes and have been summarized in descriptive form (See Tables 1-3) in order to draw conclusions about the available evidence.

| Search results
In total, the number of titles and abstracts identified from searching the electronic databases were 5693 papers, and they were screened for eligibility. After reviewing titles and abstracts, this was reduced to 312 papers which were initially eligible for inclusion in the review.
After reading full-text articles, a total of n=41 individual studies (reported across 65 papers) were included. From citation searches and searching the reference lists, a further four studies were identified and included in this review (see Figure 2). Table 1 presents an overview of the included studies. Tables 1 and 2 offer an initial description of the included studies and give an overview of who got involved, how they are involved and reported PPI activities during different stages of the research cycle were extracted.

| Characteristics of studies and populations
This allowed exploring, identifying and describing any patterns both across and within studies by tabulating the extracted data into different clusters as described in Tables 1 and 2.

| Study types
Six studies employed a mixed-methods design with both quantitative and qualitative elements, 39

| Study topics
A range of health conditions were identified: one study on alcohol abuse and sobriety, 68 brain injury, 47

| Involving PPI contributors
Studies generally did not report on how they identified the PPI contributors except for three studies wherein they employed snowballing interviews, 32 through connections and networking with key community informants and community agencies. 33

| Involvement in different stages of the research process
Forty-three studies 32

| Reported inhibitors and facilitators to involvement
Studies often failed to report any inhibitors or facilitators to involving PPI contributors in their research with only 11 studies 32,34,36,41,49,50,52,55,57,63,65 reporting some of the factors that facilitated or inhibited involvement (see Table 3). Inhibitors included cultural challenges, 50

| Summary of the findings
This is the first systematic review characterizing involvement, as opposed to participation, of BME groups in international health and social care research. The results presented in this review highlight where BME-PPI activity occurs, who is involved and how they are involved as well as factors which inhibit or promote that involvement. the best practice of BME involvement or how BME involvement could be improved or performed differently. Some barriers (n=7) and facilitators (n=5) were reported, but this information was largely absent from the identified literature.

| How the findings relate to the wider literature
The present review identified that there was minimal BME involvement during the latter stages of the research process, which reflects the findings of previous reviews exploring the impact of PPI 10,77,78 and another which states that involvement occurred at the early stages of question identification or prioritization. 77 With the exception of 11 studies, 32,34,36,41,49,50,52,55,57,63,65 authors failed to report specific barriers and facilitators to BME involvement. Generally, studies did not offer any rationale for researchers' decision to involve BME-PPI contributors at certain stages of the research process over others, thus highlighting the gap between intended involvement versus actual involvement in practice, echoing the findings reported in previous reviews. 10 The reported barriers, namely time constraints, 36 lack of previous experience in research 41,52 or compensation can be generalized as issues experienced by all those who are usually involved in research. 17 The review did not identify any studies offering insights into personal, social or cultural factors that were specifically relevant to people from BME groups that had an impact on their involvement.
Reviews exploring the impact of PPI 10,11 reflect on the intrinsic problem of assessing impact that is poor reporting and a lack of consistency, as there is a lack of clear structure when reporting PPI in peerreviewed journals. This makes it difficult to understand what works, for whom, under what circumstances and why. 10,27 Similarly, this review identified that evaluation of the quality of PPI activities by BME contributors within studies was challenging due to the variability in the way PPI activities were undertaken and lack of consistency in reporting PPI activities within studies. Studies also did not necessarily provide a comprehensive description of how they identified the PPI contributors and the extent to which they are involved in that particular context.

| METHODOLOGICAL CRITIQUE
Rigorous search methods were designed to be comprehensive to ensure that all the existing published empirical evidence in this area was identified and the review was reported according to published guidelines. Despite national and international policy initiatives to promote PPI within health services research, to our knowledge, this is the first systematic review, specifically characterizing the international literature specifically on BME involvement in health and social care research. Specifically, who is involved, where involvement occurs and at what level as well as reported factors that appear to inhibit or facilitate BME involvement in such research.
Although, there is a wealth of evidence focusing on CBPR and related participatory approaches, hand-searching of key journals for such studies were not undertaken as this was beyond the focus and scope of the current review. Relevant CPBR studies however were included if they were retrieved by our searches and met the inclusion criteria due to their fundamental shared characteristic of a philosophy of partnership and collaboration between those conducting the research and those for whom the research is focused on or about. The checklist developed by Wright et al. 30 to assess the quality of PPI reporting was useful, but was limited to what was perceived and described as pragmatic involvement, with a focus on PPI-related outcomes on research. Few studies reported barriers to PPI involvement and where they were reported, it was sometimes difficult to judge whether the reported barriers were from the perspectives of PPI contributors or study participants. In such cases, the first author (SD) read and re-read the information and made a judgement to either include it as a PPIrelated barrier or exclude it as it referred to barriers to participation.
Finally, it is also possible that substandard reporting of PPI could be a result of barriers such as limited word count in peer-reviewed journals or because PPI reporting is not a priority making reviews such as this difficult to conduct.
Exclusion of grey literature was a limitation of this review; this is a common problem as factors such as access quality and heterogeneity of grey literature are stumbling blocks to its inclusion in a systematic review. 81 However, as the current review focused solely on published academic peer-reviewed literature in health and social care research and BME involvement within this context, the findings and lessons learnt are likely to be relevant for their intended audience of academics and policymakers. where BME-PPI contributors were involved in data analysis and interpretation. While this may not always be possible, an indication of the consideration of this role for BME-PPI contributors would be appropriate, particularly as interpretation of language and cultural meanings of research findings may otherwise be inaccurate.

| IMPLICATIONS FOR PRACTICE AND FUTURE RESEARCH
Practically, therefore, funders could seek funding applications which account for these difficulties that is those which seek additional finance to allow for language and culturally appropriate methods (eg, use of interpreters, translation of materials) of incorporating BME involvement at all stages of research when appropriate.
This review suggests that understanding specific factors, which may facilitate or inhibit BME involvement in health and social care research in specific BME populations, is also warranted. While this review grouped various ethnicities together in order to learn what may impact on BME-PPI in general, it is important for health and social care studies in the future to target specific BME groups, where appropriate, as diversity exists between and within BME communities and different groups have different needs, priorities and issues. We have however attempted to uncover the common issues experienced by such groups and therefore any common learnings to be had prior to focusing on specific groups. Cultural issues were discussed broadly, and only one study 49 discussed specific cultural issues in relation to language or how input from PPI contributors enabled inclusive participation as researchers had a better understanding of cultural issues. However, there is a paucity of information in relation to how other cultural factors could inhibit or facilitate involvement; for example, women from some BME groups might prefer to be involved as a part of a women's only group or involved individually.
Finally, and perhaps most importantly, we suggest that the manner in which PPI is being enacted by health and social care researchers is narrow, 16 and the reinforcement of this approach by funders and journals only compounds this issue. One possible solution is to learn from the more established field of other participatory approaches. Power differentials between researchers and those researched are perhaps greater in PPI approaches than in other approaches which have co-learning and shared decision-making embedded. Adopting a more co-produced, partnership approach to involvement will more likely enhance the probability of greater inclusivity, thereby helping to address the current imbalance of BME involvement in health and social care research. Beyond, this, agenda setting is also a key. A move towards initiatives such as the James Lind Alliance, for example, is important as they place the power of setting out a research agenda (within a health condition/setting) in terms of priority research questions from the point of view of those directly involved in health itself that is researchers and patients/public (and exclude researchers from this process). One could foresee specific BME populations being targeted and specific research questions being developed, and then translated into partnership working in order to answer those questions.

| CONCLUSION
This review has identified the current state of the international empirical literature on BME involvement in health and social care research.
Overall, the evidence base is considered to be weak as there is limited information on the nature and content of BME-PPI-related activities within health and social care studies and requires further substantive development in terms of understanding factors that influence BME involvement as opposed to involvement more generally, and how PPI can be made more inclusive.

SUPPORTING INFORMATION
Additional Supporting Information may be found online in the supporting information tab for this article.

Definition
Most commonly used definition-INVOLVE define PPI as "research that is carried out 'with' members of the public or 'by' members of the public rather than'to', 'about' or 'for' them" (p.1). 1,77 No consensus on an accepted definition of participatory health research. Viswanathan et al. (2004) 84 define CBPR as "a collaborative research approach that is designed to ensure and establish structures for participation by communities affected by the issue being studied, representatives of organizations, and researchers in all aspects of the research process to improve health and wellbeing through taking action, including social change." (p.22).

Approach and not a methodology
PPI is an approach that can be embedded into the research process at any stage.
CBPR is also an approach that can be incorporated into any stage of the research process.

Origins (top-down/ bottom-up)
Within the UK, especially in the health policy or health service research context, PPI has taken a top-down approach, either as part of official government policy imperative or as a requirement for evidence of PPI by the research funding programmes. There are different rationales for PPI which include moral (right thing to do), instrumental (involvement as a mechanism to achieve better aims) and substantive arguments tend to focus on the publics' contribution towards quality of research 85,86 and typically the researcher controls and leads the projects.
CBPR focuses on reducing health disparities. 87 Here, research is led and controlled by the community with little or no researcher involvement. However, CBPR can also be in the form of collaboration between researchers and community partner. It is a grassroots approach that aims to improve health of the community and eliminate disparities through joint ownership of the research project. Minkler and Wallerstein 87 describe CBPR as a part of continuum with action research on one end to participatory research or participatory action research on the other end (more emancipatory approaches).

Individual vs Collective involvement
PPI tends to focus on individuals or small groups (involvement happens at individual level or collectively). PPI can include collaborative or partnership working, use of a advisory or steering committee as a means of involving patients and public in research. There is emphasis on partnership as the definition of PPI suggests working "with" members of the public.
CBPR focuses on collective identity that is community, and this is a reflection of who, how and where research takes place, emphasizing the importance of equitable partnership throughout the research process, relevant to community and is community based rather than merely community placed. 88 (Continues)

Shared decision making and ownership
PPI can be about shared decision making and sharing ownership. However, the extent to which this happens depends on the a number of factors including needs of the project, the level of PPI utilized in the project (eg, consultation, collaboration/user controlled) and also the stages of involvement (ie, if the researcher chooses how and to what extent the PPI members are involved and whether or not the PPI members have a say in how they want to be involved). Shared decision making and ownership is not the central principle. Viswanathan et al. (2004) 84 emphasize that CBPR is (1) co-learning about issues of concern and, within those, the issues that can be studied with CBPR methods and reciprocal transfer of expertise; (2) sharing of decision-making power; and (3) mutual ownership of the products and processes of research. The end result is incorporating the knowledge gained with taking action or effecting social change to improve the health and well-being of community members. 82 Shared decision making and ownership is a required principle 89 Minority ethnic groups and other groups Advocates of PPI suggest that it is challenging to involve people of BME groups. 10 Evidence seemingly implies that CBPR is the best way to meaningfully involve members of BME groups/vulnerable groups 80,84,87,88 Focus of Impact or outcomes Impact of PPI can occur at project level, for researchers, for PPI members involved. 4,12,26,27 There is shared learning taking place. However, the end outcome is about social change or benefits for community that is capacity building. 84,87,88 Here, both participation and the process are regarded as impact. 89