Despite their greater burden of disease, people from a South Asian background and other minority ethnic communities are underrepresented in clinical and applied health research [1–3] and their recruitment to and successful participation in studies remain problematic [4]. The issues of recruitment and participation of minority ethnic groups pose challenges not only for researchers conducting randomised control trials and cohort studies, but also for qualitative researchers in applied health research [5–7].
In this article, we set out some of the underlying reasons for the underrepresentation of minority ethnic groups in clinical and health research, and discuss how methods of recruitment and data collection can inadvertently serve to exclude them, rendering them what has been termed 'hard to reach'. We draw on a study based in Birmingham in the United Kingdom in which we explored South Asian women's experience of food preparation, and living with and cooking for family members with type 2 diabetes. We were particularly concerned with what Crotty [8] has termed the pre-swallowing domain of behaviour, culture, society and experience which she contrasts with the post-swallowing world of biology, physiology, biochemistry and pathology. Our aim was to find out how eating practices are embedded in social and cultural contexts and in the flow of daily life [9]. In South Asian communities it is women who bear the responsibility for cooking and food preparation. Yet their voices have been largely missing from the health research literature, despite their potentially valuable contribution to knowledge about cooking practices, and the intersection of faith, culture, health and food. We describe an approach of doing research with this particular group which we developed in collaboration with a local artist, in order to create an appropriate environment for the research participants and to provide stimuli for discussion in an effort to gain accounts of experience, views and perspectives, memories and meanings. Although this approach was particularly suitable for this group, we are not advocating the use of Rangoli as a technique to gather data from all those who are underrepresented in research. Rather, we set out the need to adopt creative alternatives to eliciting data which are meaningful to a particular group. This may involve activities and settings not usually associated with research in order to generate evidence for addressing cultural difference in the design and provision of services. Much like a 'bricoleur, a term Lévi-Strauss [10] coined to describe how the use of an object, or in this case an activity, is reconceived to explore their possibilities and re-direct their purpose so that they may serve in a new setting, we used an activity known to research participants in a novel way that facilitated the collection of data. Such a novel approach is likely to involve different techniques for different groups. Finally, we discuss the implications of adopting such an approach in relation to participant recruitment, ethics, data collection, analysis, and quality, drawing out both strengths and limitations.
The making of the 'hard-to-reach'
The exclusion of certain groups from clinical research constitutes poor science [1] in terms of the implications for the validity and generalisability of research findings [11]. However, it also contributes to health inequalities insofar as research affects the allocation of, and access to, power and resources [12]. At a time when the unequal distribution of the burden of ill-health continues to be a growing problem in the industrialised world [13–16], the exclusion of groups most affected by poor health becomes an issue not only of science, but also of ethics and social justice. Everyone has a right to participate in research, and researchers have an obligation to treat potential participants fairly and develop inclusive recruitment practices [17, 18]. The participation in health research by all members of society is necessary to generate inclusive and culturally sensitive research evidence to inform the design of effective, suitable and equitable health interventions and services.
Reasons for the underrepresentation of minority groups in UK and US health research have recently been summarised Rugkåsa and Canvin [5]. They describe a range of issues which we broadly divide into three categories: deliberate, unintended, and conceptual exclusion. Failure to include people in health research due to perceived cultural or communication barriers can be described as deliberate exclusion. Such a decision may be motivated by the composition of the research team where there is an absence of members of minority ethnic groups; a lack of language, interpretation and language skills by the research team; insufficient resources in terms of time and money to facilitate and support the recruitment and participation of minority groups; and a lack of understanding of the importance of including minority groups in research studies for both scientific and ethical reasons. Lo and Garan [19] also propose that negative attitudes of researchers could influence their decision to recruit members of minority groups in their studies because they believe that those who do not speak English and might also lack, for instance, housing or transport, are likely to have difficulty in keeping appointments or complying with the study protocol. Such attitudes will clearly limit minority ethnic representation in research.
Unintended exclusion occurs when, even though minority ethnic groups were included in the design of the study, no or few people from these groups actually take part; or when despite efforts made to recruit and support their participation recruitment is low. The reasons for this are manifold and range from insufficient numbers of people from minority ethnic groups being available for recruitment to ineffective recruitment strategies and inadequate resources. Pinto et al [20] emphasise the importance of researchers to actively enhance and support the recruitment and retention of those who are underrepresented in research, such as women from minority ethnic communities who have experienced trauma and substances abuse, in order to produce findings with the potential to address health inequalities. Unintended exclusion also occurs as a result of a lack of willingness by targeted groups to take part in research. The reason is likely to lie in scepticism and distrust by minority ethnic groups of research in general, and research led by members of the majority ethnic group in particular. As Lo and Garan [19] suggest in relation to the USA, a powerful deterrent effect on research participation by black and minority ethnic groups has been the knowledge of previous research scandals, such as the Tuskegee Syphilis study which lasted over 30 years, and it was revealed in 1972 that treatment was withheld from black African American men long after it was known that syphilis could be treated with penicillin [21]. However, there have also been recent allegations of research misconduct in studies in which research subjects were largely members of minority groups, further exacerbating mistrust and suspicion of clinical research [22]. Unfavourable previous experience of involvement in research or concern about whether findings will have unwelcome repercussions for participants or their community will also lead to a lack of willingness to engage with research [23].
Conceptual exclusion is associated with unintended exclusion and refers to cultural differences in the understanding of research as well as differences in priorities and values [1, 24]. We use the term conceptual exclusion more specifically to denote the situation where there is a fundamental mismatch between researchers' and potential participants' understanding of the nature of knowledge production. Researchers' underlying assumptions about what constitutes valuable knowledge may not be congruent with potential participants' views on what constitutes a question worthy of investigation and how to answer it. This is often described as a knowledge deficit, rather than as a difference in conceptualisations of knowledge and relevance. Potential participants may also feel marginalised by the very agencies requesting their participation in research and they may not see their concerns reflected in the scope of the research study. Consequently, different conceptualisations of the purpose and nature of research, different social and cultural norms, and a lack of common goals are likely to lead to non-participation by minority and already marginalised groups.
In relation to these types of exclusion, the term 'hard to reach' has become shorthand for describing people for whom the usual strategies of contact and engagement do not work. It is increasingly being used in the research literature to construct the problem of non-participation in research by certain groups in society. It has originally been linked to problems of access to services in health and social care as well as local government, and is frequently used in discourses about health inequalities [25, 26]. With regard to the research literature, the problem of non-participation and absence from research has been largely presented as a social fact [27, 28] and has not been subject to critical analysis. Reasons for being described as 'hard to reach' include having compelling motivations for remaining 'hidden' such as parents using illicit drugs [28, 29], as well as belonging to groups such as sex workers [30], injecting drug users and men who have sex with men [31]. Groups with whom traditional, discursive interview based research is less appropriate have also been described as 'hard to reach', for example young men from travelling communities, or some people with cognitive or learning disabilities [32]. Minority ethnic groups have attracted the label too [33]. This notion that certain people, by virtue of sharing a certain identity are invisible and inaccessible by ordinary means, has become powerful and ubiquitous [34] and 'reaching' them has become somewhat of a holy grail. However, there are dangers associated with the use of such a term because it suggests homogeneity within and across groups that are described this way [26]. Multiple and various identities of different marginalised groups are thus conflated and essentialised. The term is also potentially misleading and stigmatising, and it tends to imply that the blame for being 'hard to reach' belongs with the group itself [25].
The group with whom we have been working are frequently described as 'hard to reach' because they tend not to be visible in public life. They are South Asian women, many of whom are first generation immigrants. For them the goals and methods of science are not congruent with their priorities and realities. They may be 'hard to reach' primarily in terms of research participation because of the very methods being used to recruit and to collect data from them. For example, letters of invitation addressed to a woman may be opened by the male head of household who decides whether or not it is appropriate for her to take part in research. Participation may be problematic for families adhering to strict traditional social norms where research is situated in the public realm and thus unsuitable for women to enter. Requirements for signed consent pose similar hurdles.
South Asian women and qualitative interviews
Qualitative researchers have addressed and tackled the challenges of research recruitment and participation of those whose voices are similarly seldom heard [5, 6, 39, 35–37]. Some have highlighted the importance of carrying out detailed formative research in a particular community before starting recruitment and employing a number of different recruitment methods depending on the communities' local ethnic identities and social networks [12, 38]. Others have discussed particular methods to increase participation: for example, the use of data collection techniques based on traditional community social processes [39], the forging of close working relationships with local staff [40], interviews being carried out by 'cultural insiders' [41] the development of transcultural awareness and sensitivity [23] and accessing gatekeepers, advertising, snowball sampling and building reciprocal relationships [42].
However, we suggest that it is not only the methods of recruitment which have the potential to exclude certain groups and communities, but also the methods of qualitative research themselves. Research methods are not culturally neutral even though their assumptions are rarely discussed. One particular method continues to be the mainstay of qualitative health research: the research interview. Lawton [43] points out that interviews invite a certain kind of communication in which interviewees may be more concerned about the way they present themselves in relation to what they believe the interviewer's views are than conveying their beliefs and practices in more direct ways. Furthermore, the question-answer style of semi-structured interviews, or indeed the open-endedness of unstructured interviews, tends to be culturally alien to many first generation South Asian migrants. It may work to exclude South Asian women because the interviewee is unlikely to share the assumptions and expectations of the researcher/interviewer, even if s/he has a similar background. This is because research and research training in the biomedical and social sciences tend to have a particular model of the autonomous, rational, individual self, in control of their own lives and able to make decisions that are in their own best interests. This leads to an expectation that interviewees perform in particular ways and present accounts of themselves in particular ways [44]. Even if recruitment is successful, this lack of congruence is likely to lead to the danger of flat and one-dimensional data, i.e. data that fail to make visible the assumptions and explanations behind attitudes and experiences. Although the challenges of cross-cultural interviewing are raised [45], the effects on and implications for the quality of data are rarely addressed. We do not claim to be able to solve this problem. However, careful consideration and attention to culturally sensitive ways of facilitating expression and communication between the participant and researcher can lead to creative ways of generating useful data.
The limitations of the research interview in qualitative health research in general, not just with regard to groups and communities which have been labelled 'hard to reach', have also been acknowledged [46]. One-off interviews can be collected with relative ease by relatively junior research staff and therefore are often favoured in research grant applications. However, the lack of skills on the part of the interviewers to move beyond responses that would be perceived as socially acceptable to the researcher by the interviewee, taking interview data at face value and analysing it using tight pre-established frameworks often results in a lack of explanatory power of findings [47]. As a consequence, one-off interview based studies rarely move beyond the first, although undoubtedly crucial, step of developing descriptive categories. They are also at risk of generating findings and recommendations which tend to reflect common sense and single perspective views, rather than finely grained insights and explanations necessary to understand peoples' wider social and cultural contexts which shape the way people think about, for example, health and illness, suffering, self-care, gendered social roles and responsibilities, childrearing, and work.
New or additional insights may be gained by the use of methods which are relatively new to health research and which look to the arts, humanities and social sciences for tools which are transferred into the health care context. They can provide insights into the experience of health and illness, lay and patients' knowledge and perceptions, and views on services and interventions. They include, inter alia, 'lifegrids' to explore the course of illness experiences [48], 'walk-arounds' or 'participatory maps' to gain an in-depth understanding of context, decision making and individual experience [49], photo elicitation to investigate peoples' perceptions of 'being at risk' from developing an illness or doing physical activity, and what recovery from illness means to patients [50, 51]. Although visual and arts-based approaches have been used with children and in therapeutic contexts, the usefulness of their application in health care research has not yet been established through rigorous analysis and evaluation. There is a case to be made to push qualitative approaches in health research further, in a way that is less descriptive and has more explanatory power [47]. This is important at a time when public health policy makers recognise that a focus on personal choice and responsibility in health related behaviour is insufficient to address the wider structural and material conditions in which individual risk of chronic illness is embedded. Indeed, lifestyle oriented interventions aimed at individual behaviour change which do not take account of the circumstances in which people live are unlikely to have a significant impact upon people's ability to manage chronic illness such as type 2 diabetes and are being critiqued for neglecting the wider social, cultural and economic context driving these behaviours [14]. We have been developing a pragmatic approach to methods for eliciting data that has greater potential to bring to the surface contextual, social and cultural factors in order to contribute to a more holistic view of the design of services, interventions and programmes to support people in leading healthy lives or managing chronic illness.
Background to the study
In the UK, people of Indian, Pakistani, Bangladeshi and Sri Lankan descent are referred to as South Asian. They are the largest minority ethnic group and comprise the majority ethnic group in several urban locations [2]. South Asian communities living in the UK are diverse in terms of nationality, religious belief and cultural and social practices. The term South Asian is a blanket term which reduces to it different parameters such as national boundaries, language and religion. In doing so, it gives the illusion of homogeneity and fails to take account of the different features of identity. Within these groups there are no traditional eating practices or uniform perspectives on, for example, lifestyle, adherence to medical advice, and education. However, there is a tendency for South Asian culture to be portrayed as uniform and rigid, and as a negative influence on the self-management of type 2 diabetes and the prevention of complications [52]. South Asian diets in particular are implicated in causing the condition as a result of their high sugar and fat content. Diet is also an important factor in managing this condition, underpinning the need for further evidence about choice of foods and their preparation as well as their social and cultural meanings. A number of research teams [36, 53, 54] have explored the beliefs of South Asian people who have been diagnosed with type 2 diabetes. However, as Lawton et al. [36] suggest, given the importance of the sharing of food, hospitality and food consumption as part of worship as well as the distributed responsibility for food preparation across several female family members, there is a need for research to investigate social practices and beliefs regarding food and health more broadly. Furthermore, recent research [55] has added to our knowledge of the importance of food practices in the construction of ethnic identities in first and second generation South Asian immigrants living in Canada.
Research design
In order to explore the relationship between food practices and health in South Asian communities, it was important for us to consider the research design and methods in order, first, to recruit an adequate number of participants from a range of backgrounds; and, second, to avoid collecting one-dimensional data, composed of mainly routine responses that would be perceived as socially acceptable to a researcher, especially a cultural outsider. Collecting high quality and relevant data was crucial to ensure our analysis was able to produce rich descriptions and work towards a more theoretical understanding of how material, social and cultural factors are connected and influence each other in ways that produce effects on health. We paid particular attention to the environment in which data collection was to take place as we were concerned about its effect on participants' responses, and on the relationship and interaction between participants and researchers. We worked on the assumption that where and with whom food is discussed are important considerations in terms of participant engagement and data quality. This view was informed by our previous observations of multidisciplinary clinics for patients with type 2 diabetes, where women's discomfort when being asked to discuss their diet and food with a dietician in a hospital clinic room was clearly visible. We were mindful of the importance of appropriate spaces that could underpin the production of an atmosphere that was relaxed and conducive to talk about food and health. With regard to the relationship and interaction between participants and researchers, we planned to create a safe, facilitative environment which we could achieve by a women-only group of both participants and researchers. To this end, we accessed a wide range of pre-existing groups serving South Asian women in community centres and voluntary or faith-based organisations, such as gardening, cooking and traditional South Asian arts clubs, a language and employment skills centre for women, and a day care facility for South Asian elders. The research team consisted of women from a range of ethnic backgrounds.
In collaboration with a local artist from a South Asian background, we selected a collaborative activity with the aim to mirror food preparation and cooking, thus providing a focus for talk among women. The South Asian craft of Rangoli1 is - like food preparation and cooking - a mainly female activity and shares many of the social and collaborative features of cooking. To enhance this correspondence, the materials used for the Rangoli were dried pulses and grains traditionally used in South Asian cooking. The activity provided a focus for informal conversation and social interaction about food and its role in health, leading to a collaboratively produced piece of artwork that the women could keep. An obvious alternative approach would have been to carry out the data collection in a real cooking session and we would like to consider that in the future. However, there are a number of practical disadvantages including concerns about health and safety and food hygiene which make it a complex activity to plan as a data collection event.
Recruitment
We made contact with several community groups and registered voluntary organisations, serving South Asian women, through their gatekeepers. We sent written information about the project via email or through the post, and then set up a planning meeting which included whoever our contact at the organisation decided needed to be there. In practice, this was often a volunteer or trustee. The recruitment of the women for the Rangoli data collection sessions was carried out by the project lead in the organisation that had agreed to take part. Our aim was to conduct up to six data collection sessions with 15 to 20 participants in each. We ensured that the project leads were clearly briefed about the project and had written information available. With our help, they produced flyers and posters in the relevant language and style. However, word-of-mouth was the predominant method of recruitment.
Ethics and consent
The project proposal was reviewed by the university ethics committee. Given that recruitment and data collection took place in community settings, ethical review through the UK National Health Service was not required. Verbal consent was sought at three levels; at the collective level through the lead person of the community group; second, at individual level when the women who had chosen to attend the Rangoli data collection session were welcomed by the lead researcher who reminded them of the purpose of the session, and finally immediately prior to data collection when the digital recorders were switched on.
Data collection
Each session was facilitated by the artist who led the Rangoli activity while up to six researchers were responsible for the data collection. The room was prepared with two large tables. Each table was set up with one Rangoli board and materials including glue, cereals, pulses, spices, herbs, rice and other dry foodstuffs (Figure 1). Each table accommodated up to twelve women who were seated around it while the artist and researchers moved between the women working at the tables. The artist and researchers introduced themselves as the participants entered the room. Once all expected participants had arrived, the lead researcher opened the session formally with an explanation of the project and our wish to capture in digital audio-recordings the women's views and opinions as they talked to each other and to the researchers. To that end, a digital recorder was placed on each table, and each researcher carried a recorder with her as she moved between the women working at the tables. The role of the artist was to introduce the activity and facilitate the process of constructing the Rangoli (Figure 2 and 3). She demonstrated how to apply the glue on to the board and add the colourful materials to produce the shapes and figures of the Rangoli. Some groups were more skilled than others, as were individuals, and she assisted as necessary to ensure each group of women were able to complete their Rangoli board in around 90 minutes. She responded to requests for information or help, answered questions and also participated in the conversations. The process of the Rangoli construction was flexible and adapted to the level of skill and concentration of the group rather than consisting of a set of reproducible steps. So, for example, the group of women at the South Asian elder care day facility required more assistance from the artist than the group of younger women taking art classes. However, the process of constructing the Rangoli itself was not of prime importance. It merely provided the focus for discussion, the sharing views about food and cooking, the reminiscing and the interactions which were to constitute our data.
All researchers were encouraged to make observational notes, documenting ideas, impressions and insights during the data collection sessions, and any additional information communicated before or after the sessions. These data were shared at the debriefing meetings of approximately two hours, which followed each data collection session. The debriefing meetings were also audio-recorded.
Data analysis
Analysis was based on the constant comparative method [56, 57] in which data collection and data analysis are carried out in parallel. Data were organised using data management software and analysed inductively. The data divided into two broad types - interactive data from discussions among the group, and data obtained through individual interviews in the form of 'off-line' conversations between individual participants and researchers which took place away from the main group. The individual conversations were sometimes initiated by a participant wanting to communicate more information on a particular topic, or on the invitation of a researcher seeking further information on what a participant had contributed in a group discussion. These two types of data, group discussions and individual interviews, were analysed separately, but the findings were later integrated.