How patient and community involvement in diabetes research influences health outcomes: A realist review

Abstract Background Patient and public involvement in diabetes research is an international requirement, but little is known about the relationship between the process of involvement and health outcomes. Objective This realist review identifies who benefits from different types of involvement across different contexts and circumstances. Search strategies Medline, CINAHL and EMBASE were searched to identify interventions using targeted, embedded or collaborative involvement to reduce risk and promote self‐management of diabetes. People at risk/with diabetes, providers and community organizations with an interest in addressing diabetes were included. There were no limitations on date, language or study type. Data extraction and synthesis Data were extracted from 29 projects using elements from involvement frameworks. A conceptual analysis of involvement types was used to complete the synthesis. Main results Projects used targeted (4), embedded (8) and collaborative (17) involvement. Productive interaction facilitated over a sufficient period of time enabled people to set priorities for research. Partnerships that committed to collaboration increased awareness of diabetes risk and mobilized people to co‐design and co‐deliver diabetes interventions. Cultural adaptation increased relevance and acceptance of the intervention because they trusted local delivery approaches. Local implementation produced high levels of recruitment and retention, which project teams associated with achieving diabetes health outcomes. Discussion and Conclusions Achieving understanding of community context, developing trusting relationships across sectors and developing productive partnerships were prerequisites for designing research that was feasible and locally relevant. The proportion of diabetes studies incorporating these elements is surprisingly low. Barriers to resourcing partnerships need to be systematically addressed.


| INTRODUC TI ON
Research on the social determinants of health shows that the living conditions of people with diabetes, including cultural background, economic circumstances and built environment, can interfere with the potential effectiveness of interventions. 1 Culturally adapting interventions can increase their relevance, acceptability and uptake of physical activity and healthy eating. 2 Adapting an intervention refers to 'the process of altering a program to reduce mismatches between its characteristics and those of the new context in which it is to be implemented or used' 3 (p.25).
If mismatches between intervention and target group are reduced, the reach and provision of services may become more equitable for people in minority or vulnerable groups. 4 A recent meta-review found 199 reviews published since 2010 on patient and community involvement in adapting interventions. 4 However, there are no published studies concerning the effectiveness of involving patients and the public in adapting diabetes interventions to reduce diabetes risk or achieve better diabetes self-management. 5

| The theory of involvement in research
People can be involved in many different types of research. The focus of our review was involvement in co-designing and implementing interventions to help people reduce diabetes risk and selfmanage diabetes. The underlying theory for involvement posits that people who have a health condition, their carers and their communities have an important contribution to make in terms of identifying and prioritizing research topics, as well as the ways the research should be conducted.
Involvement enhances research in a number of ways as follows: ensuring the relevance and appropriateness of the research design; developing more effective recruitment strategies; designing research tools that are more appropriate and user-friendly; conducting interviews and surveys with more relevant and acceptable lines of enquiry; analysing data that includes lay perspectives; and better dissemination and implementation of research findings. 6 The lived experience of people with chronic health conditions and their carers is just as important as the professional knowledge of practitioners and the skills of health researchers, 7,8 so contributions at any of these steps could influence the relative success of an intervention. Research can be done to a community, for a community, by a community or with a community. Although these different stances reflect different perspectives on involvement, the relationships between stance and effectiveness are rarely reviewed. 9 Although involvement is stipulated by an increasing number of funders and organizations, we need to know more about how patient involvement can actually contribute and in what circumstances it is useful. 10 The International Diabetes Federation (IDF) Guidelines recommend that people with diabetes work with organizations to provide expert support and recommend that lay health workers actively engage people in communities, responding flexibly to dimensions of culture, ethnicity, psychosocial situations and disability. 11 Such engagement, which is conceptualized as a 'meeting of minds coming together', 12 is based on the principle that the experiences of people who have diabetes and those at risk of diabetes are essential in the collaborative design and delivery of services that work for everyone.
Despite the IDF Guidelines, diabetes research has been challenged by issues such as low uptake of screening and problems with recruitment to education sessions. 13,14 Diabetes research has also been criticized for not considering context, erring on the side of focusing on interventions to change individual behaviour, when effectiveness is actually influenced by factors outside an individual's control, 15 including physical environment (access to healthy affordable food, opportunities for activity), psychosocial context (exposure to stressors, mental health and coping strategies) and biological characteristics. 16 A recent review suggests that culturally tailored diabetes interventions may increase participation by minority and migrant groups. 17 Despite the evidence that co-produced health research can ultimately improve service delivery, patient experience and outcomes, 18,19 we did not identify any reviews exploring the relationships between various types of involvement and diabetes outcomes.
We conducted a review to identify how different approaches to involvement are being used to design and adapt diabetes interventions, and whether involvement contributes to reduction of diabetes risk and improved self-management.
The questions for the review were as follows:

| ME THODS
Our questions reflect a contingent review design (Table 1) where findings from the first set of included projects inform the retrieval and synthesis of subsequent projects to answer different questions within the review. 20 A two-stage approach to identifying literature was used because we aimed to do a realist synthesis, and our realist synthesis was dependent on identifying projects that describe the process of participation.
Realist synthesis was used to explore how involvement was facilitated or hindered by different contexts (RQ2.1) and whether interactions between researchers, patients and communities influenced the design, delivery and outcomes of the intervention. 21 We chose a realist approach because, although the degree of involvement can be seen and reported, the underlying explanations for involvement or processes that trigger involvement are hidden. 22 Involvement may be influenced by a number of aspects of the surrounding context. For example, researchers may be reluctant to involve people because they cannot see the value of it, they may feel that the time needed cannot justify the expense, and they may feel unequipped with the requisite skills for engagement.
People with diabetes may lack confidence to participate because they do not understand research or recognize the value of their lived experience. These attitudes and feelings about engagement act as mechanisms, which either enable or constrain involvement in research. These mechanisms have been documented in reviews of patient involvement. 6 However, the ways in which contexts act to either create negative attitudes and feelings towards diabetes interventions, or mitigate and promote positive mechanisms for managing diabetes, have not been systematically documented. Realist synthesis enables the interactions between contexts and mechanisms to be mapped and related to outcomes. 23 It is therefore a promising approach to explaining what works for whom and in what circumstances to promote diabetes involvement.

| Scoping and mapping
A sensitive search was conducted (GA, JH) in order to determine how many diabetes research studies actively involved patients and the wider community in diabetes intervention studies. The scoping was used to make decisions on how to organize studies, based on their respective contexts, populations and approaches to involvement.
The scoping search strategy was based on previous systematic reviews of patient and community involvement in research. 6,24 We searched health databases (Medline, CINAHL, EMBASE), and search terms were broad in order to map the extent of involvement. Population was defined as people with diabetes or at risk for diabetes who are patients, potential patients, carers of people with diabetes and people from organizations that represent people who use diabetes services. The wider community was defined as: health and community service providers, community organizations with an interest in addressing diabetes and diabetes risk, researchers and policymakers.
Involvement was defined as engagement, collaboration or partnership working across the population and wider community that aimed to co-design and co-implement interventions that drew upon the expert knowledge of all those who were involved. Involvement in diabetes research was defined as participation in the priority setting, design, implementation and/or evaluation of diabetes initiatives.
All interventions reducing risk for developing diabetes and promoting self-management of diabetes were included. There were no limitations on date or language. All study types were included, as diabetes interventions have used a range of designs from clinical trials to community-based participatory research. 4 As projects were identified, we identified a number of issues concerning completeness of reporting involvement. We attempted to address these 'thin descriptions' by locating all of the papers for each project using a technique called 'cluster searching' (see Figure 1). Project names and members of the author team were used to identify additional papers describing a particular project. 25 Several of the projects noted that they were aiming to pilot or conduct trials measuring diabetes outcomes, so the search was updated in 2016 and again in 2017 to identify the completed trials and also checked for additional projects. We concurrently searched for papers that presented theories or conceptual frameworks for involvement. Several typologies and frameworks were identified that have conceptualized involvement in health research. 26,[29][30][31][32] The scoping search that was conducted in 2015 produced 3241 hits. After removing duplicates and papers that were not about diabetes, 2716 titles and abstracts were screened for relevance by dividing the set across the review team (JH, MG, MK, JHa, TD).
Relevance was defined as the pertinence of the paper to the review question and concordance with the inclusion criteria. 27  We used a recent conceptual analysis of involvement 26 to classify projects by types of involvement (see Box for definitions).
As the primary aim of most of the papers was to report on a diabetes intervention rather than on involvement, papers needed to be assessed for relevance. 28 Descriptions of involvement were scattered across the introduction, background and discussion sections of the paper in many instances. The quality of information within the papers was assessed on ability to contribute: • data and insight about the contribution of involvement to the process of designing and implementing diabetes interventions; • explanations of how involvement may be related to diabetes outcomes.
The final set totalled 92 articles (see Reference section for detailed list of papers by project).
Data extraction was done by members of the team (JH, MG, JHal, MK) with a subset of articles, in order to develop a data extraction template based on previously developed frameworks. [29][30][31][32] The first data extraction template was structured by the stage of the project at which participation occurred: priority setting, proposal writing, intervention design and implementation.
Implementation included involvement in recruitment, delivering the intervention, data collection and analysis. We then used a recently published concept analysis for community involvement to conduct the synthesis. 26 There were four concepts contained in the framework that were reported in projects, which we used to develop the theory of involvement. These were as follows: • Frequency of contact, for example one-off or specific consultation vs regular contact.
• Contact vs collaboration: People who interact for a specific purpose have a different relationship with the researchers than people who actively collaborate as members of the research team.
• Reciprocal relationships are developed when people with relevant lived experience are able to contribute their knowledge and skills as equal members of the team.
• Shared control over the research process leads to identification of relevant topics and production of knowledge that is useful for diabetes management.
Propositional statements describing the relationship between context, mechanisms and outcomes were drafted for each project • Targeted consultation: people are contacted and consulted on specific aspects of the study, for example tasks such as a research proposal, wording of information sheets or surveys. Those involved may not be otherwise involved in the design of the study and may not receive much information regarding subsequent progress, outputs or impact.
• Embedded consultation: People with relevant lived experience are consulted regularly throughout the research cycle from initial ideas and proposals to dissemination of findings. People may be individual representatives on steering or advisory groups; or be representing a user-led organization. The research team retains ownership and control over the research study with regular input from the public.
• Collaboration and co-production: People with relevant lived experience are active members of the research team, contributing to key decisions regarding the research process as well as the findings. Relationships are reciprocal and collaborative, with shared control across researchers, patients and the public, based on specific areas of expertise.
• User-led research: People with lived experience are supported to lead the research, and take a systematic approach to directing the team through each stage from selecting the topic, writing proposals, designing the intervention, collecting and analysing data, and disseminating findings.
refined and patterns (demi-regularities) were identified between the types of involvement used in different contexts and outcomes of involvement.

| RE SULTS
Projects were the unit of analysis for this review, rather than individual articles, so results are presented by citing projects (see Reference section). The projects were organized by types of involvement (see Box 1).
Targeted and embedded involvement took place in a context where researchers organized one-off or limited consultation with patients or members of the public to inform a specific research element (Table 2).
Two projects focusing on priority setting had frequent contact and one-off contact, respectively 46 The different approaches to targeted involvement suggested that frequent involvement establishes/strengthens relationships and iterative discussions can produce consensus on research priorities as well as useful patient information. In contexts where involvement is used to extract information via a one-off encounter (Brown 2006), the utility of the involvement remains unclear.
A second set of projects characterized as embedded involvement. 85, 86 used semi-structured focus groups and interviews at several points to inform design of an intervention and materials. Researchers retained decision-making authority and control.
Repeated contact with the same set of service users was not reported. They concluded that participation by patients and general practices appeared to be inadequate and suggested that there was a 'need to rethink context and the hierarchical relationships between children, young people, parents and professionals with regard to "partnership and participation" in diabetes decision-making, selfcare and self-management' 86  with clearer strategies for supporting patients with pre-diabetes management 45 ; co-developed materials, which increased understanding of diabetes risk and were incorporated into support services, training and manuals that were useful in informing people with diabetes, other community members and professionals [50][51][52][53] .
Although initial involvement may inform project design, researcher control over subsequent stages of projects may undermine initial co-development. Two projects that consulted patients and the wider community to inform research design, Peer Support in Diabetes 118-119 and RAPSID 120-122 maintained control over recruitment and content of the intervention. Patient recruitment was low, and both projects found that peer supporters who were recruited dropped out. The RAPSID project noted the need to recruit peer supporters directly rather than through clinicians and that recruitment to trials requires very careful preparation, management and an understanding of the population involved. We compared these projects with another set that maintained user relationships through the stages of recruitment and delivery of the intervention.

| Collaboration and coproduction across all stages of the research
Projects reporting involvement across all stages collaborated to explore local environments, design and conduct interventions. These This exploration led to an increased awareness that diabetes was a problem that needed to be addressed.
Partnership governance was instrumental in making shared decisions, and negotiating differential skills and expertise. Co-design and training for local researchers increased skills 55,57 .
Partnerships felt that the engagement and empowerment of people in defining and finding solutions to diabetes for their community lead to a sense of responsibility and control over programmes, which should theoretically promote change 58,108,[133][134] .
Co-learning from exploration of local context was used to develop theory-based protocols that were culturally acceptable, producing accessible and efficacious interventions [89][90][123][124] . Using local facilitators for interviews and focus groups meant that people could freely express opinions, which was key to considering acceptable interventions 77,82,131 .
The feasibility of making lifestyle changes was considered in relation to cultural norms and local resources. For example, the availability and cost of food was mapped and used to leverage local resources to provide healthier and more affordable choices 102,129 .
In some projects, one approach to intervention development was tried and evaluated before revising [114][115] or moving on to another strategy 95-100 while others adopted a multifaceted approach which included a number of local activities which ran in tandem 72-74, 87-90, 130 . Responsiveness to local conditions was noted to be key in designing interventions 107 .
Materials and approaches to supporting people with diabetes and the wider community drew upon local values, lifestyles and traditional social structures 60,90,123,130 . Local workers and diabetes patients tailored educational teaching style and messages, incorporating components that culturally resonate 50-53, 56, 62, 70, 73, 79, 81, 83, 84, 93, 95, 103, 117, 124, 131, 133 . Co-created information ensured cultural appropriateness, respect for local practices and feasibility and acceptability of advice. Community presence contributed to the design of effective and efficient recruitment protocols and ensured the acceptability of the methods to potential participants 55,104 . This was associated with high levels of satisfaction with programmes 124 .
Projects that aimed to co-design controlled trials encountered a number of challenges. Funders found it difficult to accept that there were ethical issues to using a control group design 117  Trials used inclusive approaches to establishing control groups.
There was randomization at household level, with additional family members assigned to the same intervention 55,115 , and active control groups where all families received a version of the intervention 99-100 .
Stepped recruitment successfully retained control group participation because people believed that they would eventually receive the intervention 132 . Ensuring sustained participation, however, incurred more costs (time, money) as a result.
High rates of recruitment and retention were attributed to locally led recruitment teams who were sensitive to neighbourhood constraints and possibilities, and who were involved at proposal, recruitment and data collection stages 55-56, 100, 115 , 124, 132 .

What contributes to the success or failure of involvement?
Theories about involvement 5,6,31-33 and project data were synthesized to construct a series of propositions for how involvement works, in what contexts, for whom, at different stages of design and implementation of the intervention ( Table 2).

| Exploring relationships between the process of involvement and health outcomes
Our final review question asked: Does successful involvement in adapting diabetes interventions benefit people with

Stage of research project Relationship between context, mechanisms and outcomes
Priority setting When setting priorities, the involvement of people with experience of diabetes may increase relevance but only if there is a context where: • there is productive and sufficiently long interaction between researchers and patients/communities • there is facilitation and/or training providing opportunities for people to share knowledge and experiences • the people involved are similar or the same as those that will be designing the intervention When these conditions exist, mechanisms are triggered where: • researchers aware of their own stance and are willing to relinquish control • community members feel safe to share concerns and disagree • people commit to working out differences The outcomes of the priority setting process are as follows: • problem framing which allows patient and community concerns to be foregrounded • agreed topics that are taken to funding bodies • raised awareness of health issues and possible solutions that are relevant to patients/communities • continued interest in participating in design and mobilizing to take action

Design of the intervention
During the design stage, the same context is important, with the added provisos that: • relevant stakeholders and agencies need to be included • stakeholders need experience in facilitating partnership working • safe and comfortable spaces need to be identified to encourage participation of new stakeholders If these conditions are in place, then members of the project group will feel: • empowered to judge the feasibility and appropriateness of the emerging design • clear on who has the expertise to undertake different study tasks If successful, the end products are more culturally acceptable interventions, more appropriate approaches to recruitment, and more user-friendly information and tools. Excluding people from the design process may lead to project information that is difficult to understand and less culturally acceptable Implementation stage Implementing an intervention needs some key resources, which include the following: • training for local people and community workers that is appropriate for their needs and levels of knowledge • supportive supervision for people who are explaining the project to potential participants, providing the intervention, collecting and analysing data • knowledge of the ways in which different stakeholders can contribute to the initiative When these resources are in place, the people delivering the interventions feel empowered to exercise judgement when recruiting to the study, confident to tailor the intervention based on patient/community needs and concerns, comfortable to raise issues about acceptability and suggesting how components can be modified diabetes, communities and practitioners lead to achievement of health outcomes?
We assessed the impact of involvement on the research process as well as on health outcomes. A participatory research impact framework 34 was used to chart reported benefits (Table 4).
Projects agreed that successful inclusion of people influenced decisions about research priorities and topics. Projects that con- Projects using a collaborative approach related involvement to achieving health outcomes. Cohesive collaborative partnerships were seen to impact local community health (HEED) and culturally relevant programmes which were sensitive to literacy were associated with efficacy in diabetes management (San Francisco) evidenced by improved HbA1c, weight loss and improved insulin resistance (LLDPP). Conversely, projects that were unable to embed involvement or mobilize partnerships were unable to show clinical effect (British Columbia; EPIC; PSPD; RAPSID;).

| D ISCUSS I ON AND CON CLUS I ON S
This review found that projects which promote frequent and regu- Although we were able to follow our original protocol, 35 our theory depended on the extent of explanation and attribution in papers reporting the process of involvement. We compensated for thin reporting by using cluster searching to identify additional reports for each project. Our finding that good rates of recruitment and co-development of recruitment influence high uptake and retention rates is consistent with a review by Horigan et al 36 that found non-attendance occurs when these factors are not addressed. Our finding that delivery by local workers influences retention is consistent with trials noting that community health worker intervention groups have higher completion rates than usual care groups. 37 The importance of reciprocity is supported by a realist review 38 and a validated model for community-based participatory research that demonstrate the importance of relationships when working in academic/community partnerships to achieve individual and system outcomes. 39 To the best of our knowledge, there are no other reviews attempting to link involvement in diabetes interventions to health outcomes, but there is strong evidence across a range of health conditions that community engagement is effective. 40 However, involvement requires investment in new ways of working, which research funding rarely covers. The development of partnership trust takes time, as was acknowledged by a number of the teams included in the current review. Therefore, we need to ask how limited resources can best be used to incorporate patient and community experiences into diabetes research. We need to define what 'good enough' involvement actually means; in some contexts, targeted engagement may be sufficient, while in others, ongoing involvement and active collaboration may be needed.

TA B L E 4 Reported benefits of involvement
It has been argued that research funders currently prioritize research designs that favour controlled studies at the expense of ignoring complexity in health interventions. 41 Further, in the case of diabetes, research projects are generally expected to generate evidence of clinical impact over a relatively short time period of 6-18 months. The successful projects in our review placed health outcomes within a wider and longer term perspective and defined impact as achieving understanding of community issues and context, developing trusting relationships across sectors and developing productive partnerships.
These dimensions of impact are prerequisites for designing research that is feasible and locally relevant as well as robust. 42 Research teams need to incorporate the key aspects of working in partnerships and community development that were identified in our review. In addition, more support needs to be dedicated to funding different types of research that foreground interactions, interconnectedness and understanding of how integral community systems are to reduction of diabetes risk and self-management of diabetes.

ACK N OWLED G EM ENTS
We would like to thank Nina Skille, Norwegian Diabetes Association, Norway, and Berit Rokne, , Western Norway University of Applied Sciences, Bergen, Norway, for their participation and feedback on earlier drafts.

CO N FLI C T S O F I NTE R E S T
The authors confirm that they have no conflicts of interest.

DATA ACC E S S I B I L I T Y
Data sharing is not applicable to this article as no new data were created or analysed in this review.