There is strong evidence that changing people’s health-related behaviour can impact the leading causes of mortality and morbidity [1]. Behaviour change is central in the treatment of chronic illness, and targeting behaviours to prevent and manage chronic illnesses is imperative to deal effectively with increasing numbers of patients and escalating costs [2]. Recent examples of successful interventions have targeted both the behaviour of people with diabetes (diet and activity behaviours) and healthcare professional behaviour (early intervention for diabetes foot ulcers) [3, 4]. Changing behaviour can improve outcomes, with increasing evidence that interventions targeting behaviour change in diabetes can be effective [5, 6].

Despite the potential for behaviour change to improve diabetes outcomes, developing effective interventions is challenging. Diabetes management is complex, encompassing many different behaviours, and patients often struggle to make and maintain the behavioural changes required to manage their condition [7]. As this programme of behavioural research within diabetes continues to grow, how should we decide which behaviours should be prioritised for research? It has been suggested that much health-related research does not address topics which are of importance to patients and clinicians [8]. Seeking the views of patients and healthcare providers should be an essential part of determining the behavioural research agenda, especially to ensure impact from publicly-funded research [9].

Changing diabetes care to implement evidence from research into routine practice is a major challenge within the constraints of the healthcare system [10–12]. Within diabetes, attempts have been made to reduce the research-evidence gap with initiatives such as the Bringing Research in Diabetes to Global Environments and Systems (BRIDGES) project supporting the development of interventions that can be adopted and disseminated in real world settings [13]. Qualitative work with BRIDGES project researchers pinpointed lack of stakeholder and diabetes community links as key barriers in the implementation of diabetes research [14]. Despite the identified need to increase stakeholder engagement, there are few published examples of methods to involve stakeholders in the research process in diabetes.

More recently, efforts have been made to involve stakeholders in the research process by seeking input from patients and healthcare professionals in the prioritisation of research. One approach to collaborative research prioritisation has been driven by the James Lind Alliance which provides guidance on the development of Priority Setting Partnerships [15]. The James Lind Alliance Priority Setting Partnerships aim to bring patient and clinicians together to prioritise treatment uncertainties for research. Treatment uncertainties have been identified and ranked for a range of conditions [16] including Type 1 diabetes [17].

The first step in James Lind Alliance Priority Setting Partnerships is to identify potential research questions of interest to patients and providers. However, involving service users solely in the prioritisation of research questions can be limiting, with evidence that patient suggestions frequently fail to meet the criteria of a researchable question [18]. In addition, focusing solely on treatment uncertainties can limit the scope of research prioritisation. Research exploring the translation of evidence-based behaviour change interventions into practice, for example, does not fall within the treatment uncertainty remit.

The aim of our research prioritisation was to move beyond a more narrow discussion of treatment uncertainties to identify and achieve consensus on shared priority areas for research in diabetes in Ireland. Both the Delphi and nominal group technique processes have been used for the development of consensus in health services research. We also wanted to increase engagement in our programme of research and to build links with relevant stakeholders in Ireland, and so the anonymous approach associated with the Delphi technique was not appropriate to our aims.

The nominal group technique is a controlled group process for the generation and ranking of ideas and for consensus development [19]. The nominal group technique generates a high number of ideas and includes social interaction and discussion while limiting normative pressure for conformity through the use of private individual ranking [19]. Each participant in a nominal group technique process has an equal private ranking vote which provides a democratic process to navigate mismatches between researcher, patient, provider and policy maker priorities [15]. The nominal group technique also avoids limiting patients’ contribution to priority setting by not requiring priorities to be articulated in the form of a researchable question [18].

The aim of the present study was to engage in a nominal group technique processwith key stakeholders to identify, and achieve consensus on, the most important target behaviours for research in diabetes in Ireland. By identifying behaviour change targets that address the needs of the diabetes community, we hoped to enhance the relevance of our programme of research to the Irish healthcare context and to increase the likelihood of future translation of the findings into practice. By focusing on target behaviours rather than tightly defined research questions or treatment uncertainties, we hoped to maximise the potential for patients and healthcare professionals to impact on the development of the research agenda. Finally, by clearly outlining a systematic approach to engaging with key stakeholders we hoped to provide a useful resource for other researchers seeking to engage patients, professionals and policy makers, in the design of research.

The study forms part of a programme of research to develop and evaluate two behaviour change interventions in diabetes in Ireland: one focusing on a behaviour relevant to people with diabetes, and one on a behaviour relevant to healthcare professionals. In identifying high priority behaviours for research, we focused on three key areas: managing Type 1 diabetes mellitus (Type 1 DM), managing Type 2 diabetes mellitus (Type 2 DM) and preventing Type 2 DM. Within each of these areas, we aimed to develop a prioritised list of the most highly ranked target behaviours relevant to people with diabetes and healthcare professionals. We aimed to get the views of people with diabetes, healthcare professionals with clinical experience of diabetes and policy makers working in the area of diabetes.

The nominal group technique (also known as an expert panel) was used to identify, and achieve consensus on, the most important target behaviours for research [20]. The nominal group technique was chosen as a systematic process that facilitates both idea generation and consensus development [19].

Procedure

The nominal group technique begins with eliciting participant views on a topic. Similar suggestions are grouped together and a facilitated group discussion during a structured meeting allows for the clarification and evaluation of items. Each participant then privately ranks each item, the overall rankings are calculated, presented, and discussed and the items are privately re-ranked. The process of achieving consensus through an initial ranking, group discussion, and a second re-ranking was decided in advance and follows published nominal group technique guidance [19]. For the current study, a diabetes research prioritisation (DRP) meeting was organised in Galway, Ireland in October 2014. The study process is shown in Fig. 1 and each of the nominal group stages is outlined further below.

Pre-meeting

Stage 1: Pre-meeting generation and collation of health professional and patient behaviours

In advance of the DRP meeting, participants completed an online survey to generate lists of behaviours to target for research. Participants were asked to generate three health professional behaviours and three patient behaviours in each of the key diabetes areas (managing Type 1 DM, managing Type 2 DM, and preventing Type 2 DM). The importance of participants’ own views was emphasised and specific examples were provided for each category of behaviours.

The survey was administered using the Survey Monkey online tool and sent to participants one month in advance of the DRP meeting; a reminder was sent to all participants a week before the meeting. The research team collated all submitted responses by combining similar behaviours to avoid duplication, creating unique behaviours where original submissions included multiple behaviours and creating total lists of behaviours for each key diabetes area.

DRP meeting

Stage 2: Further development of lists of health professional and patient behaviours

Participants attended in person at a three hour DRP meeting and joined small group tables of four at random as they arrived. At the start of the meeting, the research team gave a short presentation outlining the format of the meeting and defined key terms. Participants were provided with the pre-generated lists of target behaviours and engaged in small group discussions to identify any additional behaviours. Small group tables included a mix of patients, healthcare professionals and policy makers. Each small group had an opportunity to feedback additional behaviours to the larger group, and these were added to the total lists. Group discussions were chaired by an experienced facilitator who aimed to ensure everyone had an opportunity to speak. This process was done six times during the meeting, for each of the three key diabetes areas (managing Type 1 DM, managing Type 2 DM, and preventing Type 2 DM), first for patient behaviours and then for health professional behaviours (see Fig. 2).

Stage 3: First ranking of target health professional and patient behaviours

Total behaviour lists were presented back to participants on a screen at the front of the room. Participants privately ranked their top six health professional and top six patient behaviours on paper sheets which were collected by members of the research team.

Stage 4: Calculation of first group ranking, feedback and discussion

The results of the first ranking were manually entered into an excel spread sheet by members of the research team as the meeting progressed. Data entry was checked for accuracy after the meeting by a second researcher; minimal discrepancies were identified. Top ranked priority behaviours were assigned a score of 6, second ranked behaviours were assigned a score of 5 and so on. The total scores for each behaviour were calculated and the results were presented back to the group. The six most highly ranked health professional and patient behaviours in each key diabetes area were highlighted. In a group discussion, the facilitator asked participants to comment on the results, particularly focusing on behaviours whose rankings they found surprising or interesting.

Stage 5: Second ranking of target health professional and patient behaviours

Stage 5 followed a similar procedure as Stage 3 and participants were asked to privately re-rank top six health professional and top six patient behaviours on paper sheets in each of the three key areas.

Post meeting

Stage 6: Calculation of second group ranking

As before, top ranked priority behaviours were assigned a score of 6, second ranked behaviours were assigned a score of 5 and so on and the total scores for each behaviour were calculated. The number of times each behaviour was ranked in participants’ top six and the percentage of participants who ranked each behaviour within their top three priorities were also calculated.

Stage 7: Post-meeting feedback

A summary of the findings was sent to all participants three weeks after the DRP meeting. Participants were sent a link to an online questionnaire and asked to provide feedback on how interesting, enjoyable and useful they found the meeting and to give suggestions as to how the meeting could have been improved.

This study demonstrates that it is possible to work with a group of diverse stakeholders - people with diabetes, healthcare professionals and policy makers - in a consensus process to co-produce a prioritised list of behaviours to target for intervention research. Key stakeholders are more likely to drive forward the implementation of research findings for research topics important to them and over which they feel ownership. By engaging with key stakeholders from the outset, we have begun developing strategic alliances with health partners and service users which will be maintained throughout the programme of research.

Utilizing a nominal group process approach resulted in the identification of priorities at a broader level than could have been achieved with a more narrow focus on treatment uncertainties. Attendance and engagement with structured education, for example, would not have been identified as a priority if we had followed the guidance provided for James Lind Alliance Priority Setting Partnerships as evidence on the efficacy of structured education programmes is already available. Our approach allowed stakeholders to identify behaviours where regular performance of the behaviour, rather than treatment uncertainty, is the major problem.

The nominal group process findings indicated that in most cases, the most highly ranked behaviour was the same for Ranking 1 and Ranking 2, suggesting that the discussion and re-ranking process did not significantly change participants’ views. However, there were examples where behaviours featured in the top 6 at Ranking 2 but not at Ranking 1, presumably reflecting the impact of suggestions and arguments offered by participants during group discussions. Details of the discussions that might have led to these changes were not recorded, but would make an interesting and useful addition to future nominal group exercises. As expected in a consensus building study, consensus generally increased in relation to endorsement of the most highly ranked behaviours between the first and second rankings.

A number of striking findings emerged in relation to the most highly ranked priorities. For people with Type 1 DM there was general consensus that targeting ‘taking insulin as required’ is the most important behaviour for research. Indeed, the top three highest ranked behaviours were related to effectively monitoring and managing blood glucose, by administering insulin and counting carbohydrates. The skills required for people with Type 1 DM to maintain good glucose control are complex and demanding [21]. A number of studies report that glucose control remains unsatisfactory despite close monitoring and participation in educational programmes by many people with Type 1 DM [22], although there is some evidence that training in flexible, intensive insulin treatment can improve glycaemic control [23]. Future research should focus on developing and delivering effective interventions to promote the skills needed by people with Type 1 DM to achieve good glycaemic control.

For people with Type 2 DM, attending and engaging with structured education programmes was most highly ranked. Research into structured education programmes for people with Type 2 DM in Ireland has been lacking, but recent studies have demonstrated the efficacy of structured education in Ireland [24]. However, a significant proportion of people with Type 2 DM are either not offered or do not attend structured education programmes [25]. Future research should focus on ways to engage people with Type 2 DM in education programmes, as well as ensuring that such programmes are as patient-centred as possible to increase the benefits to participants.

Interestingly, behaviours related to collaborative goal setting and patient-centred care were highly ranked. ‘Engage in collaborative treatment goal setting with patients’ was the top ranked healthcare professional behaviour for the management of both Type 1 DM and Type 2 DM. Among people with diabetes, patients’ perceptions of collaborative care (including collaborative goal setting) have been shown to be associated with patients’ reported self-management [26], however, the relationship between collaborative goal setting and clinical control among remains poorly understood [27]. Future research should focus on interventions to promote the use of collaborative goal setting among healthcare professionals and people with diabetes, and measuring the associated impact on clinical outcomes.

In relation to preventing Type 2 DM, the most highly ranked behaviours were promotion of healthy exercise and diet, with priority given to engaging in healthy behaviours as a family. There is considerable evidence that increased levels of physical activity are related to better clinical outcomes in diabetes, and additional evidence that physical activity interventions are particularly effective when combined with dietary advice [28]. There was much discussion at the meeting around the need to advocate for policy change and attempt to promote environmental change to support healthy behaviour. There is evidence that environmental and policy approaches lead to increases in physical activity [29], and future research should focus on finding ways to maximise the role of behavioural science theory in development and implementation of public health interventions [30].

For healthcare professionals’ behaviours to prevent Type 2 DM, the highest ranked behaviour was ‘attend and engage with behaviour change training’. Recent trials of behaviour change counselling have found limited effects on behaviour change outcomes [31]. Future research should focus on finding more effective ways to train healthcare professionals to promote behaviour change among their patients.

The current study forms the first step in a behaviour change intervention development process. One behaviour relevant to people with diabetes, and one behaviour relevant to healthcare professionals, will be selected as targets for intervention development. As per recommended guidance [32], the selection of target behaviours will be informed by the comparison of the top behaviours identified in each key area under criteria of the likely impact of changing the behaviour, the likelihood of change, the potential positive or negative spillover to other behaviours and the ease of measurement of each behaviour.

To impact, and improve, health and healthcare, behavioural researchers need to engage with stakeholders outside of the research community. In the current study, we have demonstrated that it is possible to engage people with diabetes, healthcare professionals and policy makers working in the area of diabetes to generate and prioritise behavioural research topics. The priorities identified were co-produced by key stakeholders, including patients, healthcare professionals and policy makers, and will inform the development of a programme of behavioural research in diabetes in Ireland. We will continue to engage with these stakeholders by inviting patients, professionals and policy makers to sit on steering committees to move chosen priorities forward. The study also provides a worked example of a research prioritisation process using the nominal group technique which may be a useful for other researchers.