The NIHR under-served programme board agreed that ethical approval was not required for this research, as the survey was deemed service improvement. Participants’ consent was presumed by the active choice of clicking the link to the survey and completing the questions. Responses were anonymous. The survey data were stored on password protected devices and only accessed by the research team. Free text comments were reviewed prior to analysis and any identifying information removed or coded.

An online survey was designed and delivered in smartsurvey.co.uk ³³ . It was produced following a focussed review of the literature. The review was completed between July and September 2021 using the terms ‘under-served’ and ‘underrepresented’. PubMed database was utilised, and the search was limited to articles produced between 2018 and 2021, where full-text or abstracts were available. All article types were included and the bibliographies of each reviewed. A total of 62 articles were identified and reviewed for relevance. After the initial assessment, 47 articles formed the focussed review. That focussed review produced themes which were used to determine the groups identified as under-served and factors that impacted on being under-served. These were peer reviewed within the study team and formed part of the piloting process to formulate the survey responses.

The subsequent survey was piloted and peer reviewed by six members of the NIHR under-served programme board. Feedback from peers was discussed and acted upon and there was one further round of revisions. A copy of the survey is available ³⁴ .

There was no patient or public involvement in this project as it was aimed at professionals working in research.

The survey was advertised through an NIHR communications package. It was circulated to all Local Clinical Research Network communications leads and partner organisations with a request to share. It was promoted via regional NIHR newsletters and the Clinical Research Network (CRN) national newsletter. Further promotion appeared on Twitter and LinkedIn. As an example, in the North East and North Cumbria CRN, the request to share was actioned by the communication lead. This led to the direct emailing of research and development managers, speciality groups leads and core team delivery staff which totalled 89 direct emails. The recruitment aim was to reach those working in the public sector of the UK who were involved in research delivery. There was no defined target sample and there were no specific exclusions.

The survey commenced with an introductory statement and link to the NIHR INCLUDE project. It was split into four sections totalling 19 questions. The first section of nine questions collected demographic and role information. This was followed by one question regarding specialism. The next seven questions were titled ‘Barriers and enablers’ and gathered information on who are the under-served, factors impacting on being under-served and strategies to improve research activity to better engage the under-served. The final two questions allowed respondents to indicate if they wanted to make direct contact with the NIHR under-served programme board and leave their contact details. Question types included closed, open, ranking and four free text response questions. Only complete data for each section was analysed.

Statistical analysis was performed using R software ³⁵ . Descriptive statistics were utilised with associations compared using chi-square test. To test relationships between categorical variables, statistical significance was deemed as p. value of <0.05. Qualitative analysis of the free text responses was completed using a thematic approach ³⁶ . The analysis followed three steps. Initially the free-text comments were read to familiarise and anonymise. The comments were then coded and descriptive themes generated. The themes were guided by the findings from the focussed literature review. Finally, the descriptive themes were reviewed to generate an overall viewpoint ^{36, 37} . Analysis was aided by Microsoft Excel, topic models ³⁸ and tidytext ³⁹ packages.

The process was verified for trustworthiness by the core team. Reflexivity was demonstrated by acknowledging the prior views and knowledge of the research team. Themes were guided by the previous literature, however differing views and information were explicitly sought out. Data were coded by different individuals and data analysis clinics held to reflect on codes and discuss initial theorising ³⁶ .

The authors have experience of survey research (DC, CW, JR), statistical (TB) and qualitative analysis (DC). CW was responsible for the original idea, CW, JR and DC formulated the questions and JR built the survey. MSB has expertise in equality and diversity and aided the qualitative analysis with DC. The study is reported in line with CROSS ⁴⁰ (Checklist for Reporting Survey Studies), STROBE (STrengthening the Reporting of OBservational studies: cross -sectional studies) ⁴¹ and SRQR (Standards for Reporting Qualitative Research) ⁴² guidelines.

Data was collected from 945 fully completed and submitted survey responses. All respondents confirmed they were happy to complete the survey ³⁴ . Responses were received from all geographical areas of the UK. The largest number came from the North East and North Cumbria (n=111, 12.4%). The age, sex and ethnicity profiles reflect that of the UK’s National Health Service (NHS) with 72% (n=660) identifying as female, 82% (n= 748) white and over 30% (n=275) in the 45–54 years age group ⁴³ . Respondents self-identified their sex, the options being male, female, prefer not to say and other (please specify). Figure 1 illustrates the demographic profile of the respondents.

All components of the NIHR infrastructure and Clinical Specialties in the NHS were represented. Further information on the NIHR infrastructure and clinical specialities can be found at: https://www.nihr.ac.uk/explore-nihr/support/research-infrastructure.htm and https://www.nihr.ac.uk/explore-nihr/specialties/. To identify any differences in enablers and barriers across clinical settings, sub-set analysis via specialism was completed. These were: hospital-based specialties (n=323, female 222, male 91), mental health (n=124, female 82, male 38), and primary care (n=118, female 90, male 24), including public health and community. These three sub-sets comprised 60% (n=565) of the total responses.

The commonest area worked in was secondary care (41%, n= 371, female 259, male 102), followed by research higher education institutions (39%, n=355, female 242, male 97). The least cited place of work was the ambulance service (0.3%, n=3). When asked, 75% (n=587, female 290, male 153) of those who responded stated they were involved in identifying participants for research studies, with 69% (n=624, female 442, male 164) saying they had direct contact with research participants and finally 54% (n=491, female 353, male 128) declared they were involved in consenting participants for research studies.

The respondents were asked to identify who they felt were under-served in ‘your area’, ‘your area’ being open to the respondent’s interpretation. Five open text boxes were provided to insert a descriptor. Respondents were not asked to rank the five descriptors. In total 2782 responses were made across the five open text boxes. The responses were collapsed into categories. Table 1 illustrates the top seven categories and the total responses for each. The most cited group was ethnic and cultural minorities, followed by those deemed as having a lower socioeconomic status, then those where a given age was used as an identifier (the old, very old, young adults, adolescents, children and neonates). In the category where age was used as an identifier, ‘adolescents/young adults’ were put forward as a group not previously identified. The ethnic and cultural minority group contained a broader descriptor than found in the literature which focussed on Black and Hispanic groups.

Subset analysis by specialism revealed that the hospital-based group placed ‘condition specific’ (9%, n=76) above ‘mental health’ (6%, n=52), with those specialising in mental health putting ‘mental health’ in second place (18%, n=79) followed by ‘lower socioeconomic status’ (14%, n=62) ‘learning/intellectual difficulty/disability’ (11%, n=48) and age related 8% (n=36). The primary care-based specialisms elevated ‘access’ above ‘condition specific’ (p=0.038).

Respondents ranked 10 factors taken from the literature, in terms of how they impacted on being under-served (the 10 factors can be found in Figure 2). Respondents ranked language and literacy skills as the most impactful, with the least impactful being religious and cultural beliefs. Figure 2 illustrates the ranking of factors.

Sub-set analysis via specialism indicated statistically significant variation in greatest impact. Hospital-based specialisms (32%) and primary care specialists (30%) ranked language and literacy skills as the most impactful when compared to mental health 11% (p<0.0001). Mental health services ranked a lack of focus on the under-served as more significant: mental health 22%, primary care 15% and hospital-based 12% (p=0.04). In addition, significant differences were demonstrated for socioeconomic status, with this being ranked as a more significant factor in primary care research: primary care 22%, mental health 11% and hospital-based 10% (p=0.004).

A total of 824 (610 female, 214 male) responses were received from those who answered question 8 (Are you involved in identifying research participants? Yes/No) and question 15 (How equipped do you feel in dealing with the under-served? Yes/Partially/Not very well/No). Of the 538 (392 female, 146 male) who stated they were involved in identifying research participants, 12.3% (n=66, 44 female, 21 male) suggested they felt equipped to meet the needs of the under-served, with a further 48% stating they were partially equipped (n=258, 179 female, 79 male). Of those not involved directly in identifying research participants (n=286, 218 female, 68 male), 8% (n=22, 18 female, 4 male) said they felt equipped to meet the needs of the under-served (p=0.06). A significant difference was also seen in those who identify research participants who felt partially equipped (n=258, 179 female, 79 male) when compared to those who do not identify research participants (38%, n=108, 81 females and 27 males), p=0.006. As expected, those who identify potential research participants in their role felt more equipped compared to those who do not (p=0.0004), similarly staff receiving consent from research participants felt more equipped than researchers who do not (p=0.0001). Those who receive consent from research participants were more likely to feel partially equipped to address the needs of the under-served compared to those who do not consent (p=0.013), suggesting that face to face contact is associated with increased confidence in engaging the under-served.

Four free-text questions were analysed using a thematic approach. They covered positive innovations, barriers to innovations, challenges to engagement and further training requirements. From these responses enablers and barriers were identified and suggestions for training put forward.

Several enablers were recognised with comments suggesting strategies to enhance engagement. However, these were often accompanied by the barriers indicating why they could not be implemented. The most cited enabler was community outreach.

Community engagement was seen as beginning prior to any notion of conducting research. It included genuine participation by the community in setting its own research agenda via collaborative approaches and continued after the completion of any research project.

Linking to this was the use of Patient and Public Involvement and Engagement (PPIE) groups. These groups were highlighted as enabling, however there was a degree of scepticism about how representative they were of their group.

Outreach also included the movement of research from a secondary care setting to the community. It was suggested methods and methodologies should focus on assisting the delivery of research in the community.

Further enablers were identified such as strategies to improve the accessibility of study documentation and the information presented in them. Suggestions were put forward covering multiple different modalities: translation and interpretation services, video, audio, digital platforms, social media, sign language and braille. Again, these suggestions were accompanied by the barriers to implementation.

Additional enablers to engaging a diverse research population were having a workforce that reflected a varied background and (where possible) the demographic profile of the under-served group. Recruitment via General Practice was thought to better reflect the demographic profile of the area and innovative ways to reimburse participants were also thought to enable engagement.

At an organisational level it was suggested that study design and approvals processes should reflect those of the urgent public health studies carried out during the COVID-19 pandemic. Highlighted were changes in ethical approval, community engagement, consent processes, data collection methods and feedback of results.

Sub-set analysis indicated the most cited enabler from both primary care-based and mental health groups was community outreach, with hospital-based specialties citing changes to research design and methodology.

The most cited barrier to engaging under-served groups was time and funding; this was reflected across the whole dataset and the sub-sets of primary care and mental health specialties. The respondents were aware of the strategies to enhance engagement; however, these all required more time and funding, and that time and funding was not planned into project designs. They felt that the current systems and processes around grants/funding applications and ethical approval were obstructive and risk averse. Funding processes did not consider the planning needed to design studies that are co-produced with communities and approvals processes question innovative recruitment strategies and reimbursement of participants. Comments suggest a disconnect between policy and practical application.

Barriers also included issues with communicating research to specific groups who may be under-served. Respondents highlighted problems with translation and interpretation services and how useful they are in the field of research.

The concerns expressed centred around availability when required, knowledge levels and personal views of the translator or interpreter. Cost of translating written materials and the time taken were also thought to be obstructive. These were the most cited barriers in the hospital-based specialism. There were also comments which suggested that while the use of digital platforms may for some enhance engagement, it conversely disenfranchises some others.

Suggestions about additional information and training to aid researchers were diffuse. They included ways to conduct community engagement including methodologies, methods, inclusive recruitment strategies for both participants and PPIE, funding applications and innovative reimbursement methods. Further general training was requested in cultural competence, taking in cultural awareness and sensitivity and general equality and diversity. Some specific under-served groups that were mentioned regarding engagement activities were the LGBTQi community, those with disability and the homeless. Finally, requests were made for training in consenting those where English is not their first language or their comprehension of English may be reduced, and those who may lack capacity. Sub-set analysis indicated all three groups wanted information on general awareness of under-served groups and improved ways of communication with these groups. The mental health and hospital-based groups also requested training on research design methods.

In summary, the findings from the survey generally reflected the literature in this area. Those working in the field of research felt only partially equipped to deal with engaging the under-served and requested additional information, support and training. The responses overall are context specific and focussed around the respondent’s area of speciality. The main enabler to engaging the under-served was thought to be community outreach. However, there was a consistent reference to lack of time, funding, capacity and support to develop site and project level interventions to engage the under-served. Respondents were aware of numerous strategies to improve the diversity of research participants and there was a sense of divide between policy rhetoric and ability to practically apply.