Impact of Community Based Peer Support in Type 2 Diabetes: A Cluster Randomised Controlled Trial of Individual and/or Group Approaches

Background Diabetes peer support, where one person with diabetes helps guide and support others, has been proposed as a way to improve diabetes management. We have tested whether different diabetes peer support strategies can improve metabolic and/or psychological outcomes. Methods People with type 2 diabetes (n = 1,299) were invited to participate as either ‘peer’ or ‘peer support facilitator’ (PSF) in a 2x2 factorial randomised cluster controlled trial across rural communities (130 clusters) in England. Peer support was delivered over 8–12 months by trained PSFs, supported by monthly meetings with a diabetes educator. Primary end point was HbA1c. Secondary outcomes included quality of life, diabetes distress, blood pressure, waist, total cholesterol and weight. Outcome assessors and investigators were masked to arm allocation. Main factors were 1:1 or group intervention. Analysis was by intention-to-treat adjusting for baseline. Results The 4 arms were well matched (Group n = 330, 1:1(individual) n = 325, combined n = 322, control n = 322); 1035 (79•7%) completed the mid-point postal questionnaire and 1064 (81•9%) had a final HbA1c. A limitation was that although 92.6% PSFs and peers were in telephone contact, only 61.4% of intervention participants attended a face to face session. Mean baseline HbA1c was 57 mmol/mol (7•4%), with no significant change across arms. Follow up systolic blood pressure was 2•3mm Hg (0.6 to 4.0) lower among those allocated group peer-support and 3•0mm Hg (1.1 to 5.0) lower if the group support was attended at least once. There was no impact on other outcomes by intention to treat or significant differences between arms in self-reported adherence or medication. Conclusions Group diabetes peer support over 8–12 months was associated with a small improvement in blood pressure but no other significant outcomes. Long term benefits should be investigated. Trial Registration ISRCTN.com ISRCTN6696362166963621


RAPSID Background and Methods
David Simmons Version 1.0 6/22/2014 Page 3 Individual buddy approach: where a patient mentors and supports another patient.

Individual expert patient approach:
A structured group training programme providing "expert" patients with skills to train other patients in self management/problem solving.

Traditional self help groups:
People meeting for mutual support in a relatively unstructured way. Such groups exist all over the world and although considered important, there has been little evaluation of whether they lead to behaviour change or improvements in health.

Structured group education approach: Group education incorporating patients learning
and supporting each other during the course (10,11).
Previous research suggests that peer support interventions are welcomed by participants, but has not provided robust evidence for its utility across all cultures (20). Two randomized trials have suggested improved self efficacy and HbA1c from peer-led self management support within Spanish-speaking communities (19) and increased physical activity from peer support amongst African-American women (21). Other trials are underway in Dublin (evaluating group support in primary care, including an educational component) (22) and Warwick (evaluating telecare by peers) (23). A model for how peer support could lead to sustainable behavioural changes that will improve long-term diabetes outcomes has been proposed by Heisler (17) We wonder whether Heisler's model focuses on individual interactions between peer and person with diabetes, and understates the role of the context and social milieu in which the peer support occurs. This is a key component in Bandura's social cognitive theory (24), under which, people, environment and behaviour are all influencing each other. We have therefore included an additional construct of social context which is consistent with Bandura's Theory, can include 1:1 or group/community settings and which may be vital for some individuals.
Bandura's approach could provide a framework (Appendix 1) to train peers for a support role, and also aid understanding of the ways in which peer support may be helpful (25). Aspects relevant to peer support include promoting self efficacy and motivation, buddying, mentoring and modeling within a social context which promotes and values skill acquisition, learning, skill implementation and coping.
We are unaware of studies which have compared individual and group approaches to peer support, and wonder whether patients vary in their responsiveness to such interventions.
Groups could incorporate "social context", particularly if wider commonalities were present beyond diabetes (eg community links). We believe that comparing such approaches would assist in understanding the mechanisms behind any benefits or, alternatively, what the reasons are for any lack of benefit eg is it the patient mix, the intervention, the implementation of the intervention or even the conceptual framework. A concept analysis by Dennis has contributed to our thinking about the role of peer supporters and the training that they will need (26).
Heisler proposes that peer support can play an important role in identifying ways to overcome barriers to self care. We would agree with this and have created a framework (Appendix 2) to achieve this that has been used in New Zealand, Australia and the US (5,6,27,28). Also shown are examples of how the framework can be used to generate discussion between peer and person with diabetes in either a group or 1:1 situation.

RAPSID Background and Methods
David Simmons Version 1.0 6/22/2014 Page 5 We propose the following principles for the trial, beyond the pre-set requirements: Discussing and addressing barriers to self care should provide a context for appraisal and informational and emotional support.
The trial should be able to combine evaluating the efficacy of the intervention with an evaluation of the mechanisms and issues behind the intervention. This would mean that even if the trial is negative, new knowledge would be gained for future trials.
The trial should allow a comparison of the efficacy of 1:1 peer support, a group approach, a combined approach and normal care without peer support. We believe this question, of whether an individual or group approach is best is important and may remain unanswered unless a specific trial is undertaken The trial should assess uptake of group peer support and any wider population impact.
We have therefore devised a 2x2 factorial trial with an initial barriers survey and geographical clusters (defined by local government boundaries known as Parish Councils and including one or more villages or small towns) being allocated to neither intervention, a 1:1 peer support intervention, a community based peer support group intervention or both interventions.

Recruitment or approaches to reaching intended audience
Two major recruitment methods into the trial will be used: 1 General practices possess lists of almost all local patients with diabetes and will be approached to send out an invitation to patients to participate in the trial. Practice registers will be screened for diabetes, excluding people known to have Type 1 diabetes, dementia, psychotic illness or if it might be unsafe for a peer supporter to visit them at home. Invitation letters, and one reminder, will be sent to all residents with diabetes in smaller Parish Council areas, and a random sample of those with diabetes from larger areas anticipating that one third will join the trial, aiming for 20 people per cluster. Invitation letters will include the patient information sheet, consent form and a baseline survey relating to barriers to care (6). Participation rates will be maximized by using the Dillman method (52) in relation to format, inclusion of a stamped addressed envelope and follow up of the mail out. PCRN-EoE and Diabetes Research Network staff will support practice staff with recruitment.
2 All of the villages and towns have a means to contact local residents through notice boards, parish magazines, social clubs, Women's Institutes and local Parish Councils. These organizations will be approached to help invite people with diabetes into the trial. Materials distributed will include the same pack as supplied to patients invited by general practitioners.
Participants recruited by community advertisement will also be screened.
Potential participants responding to the invitation will be contacted for an appointment for baseline assessment at a convenient venue (including the local general practice where appropriate). Participants will provide written informed consent at this assessment. The three towns with existing diabetes support groups will be excluded from the trial (approximately 25% of those with diabetes), but residents will be invited to contribute to training and back-up cover for the peer supporters recruited for the trial (Cambridge, Wisbech, Huntingdon).
We believe that there are a number of reasons why general practice teams, potential peer supporters and participants will be keen to participate in this trial.
For general practices, this study has the potential to improve their patients' health by overcoming some of the barriers which frustrate current care. Local practices have responded enthusiastically to other diabetes studies (46,53). Practical, governance and financial aspects of study participation will be aided by PCRN-EoE, which will also be able to take over research nurse work on the study if needed.
Potential peers will benefit from interesting and personally relevant training. Many people with diabetes are keen to share experiences and support others, but a particular advantage of doing so through this trial is that it will be provide a safe and supported framework for the peers. Their expenses will be paid, but they will not be paid for their time. Such a payment could undermine their role as "peers" who share common experiences with the other participants and could be viewed as against the "spirit of peer support" (22)).
Potential participants will be individually invited by their general practitioner as well as through community networks. Peer support will be established on a local community basis and endorsed by local health services. Educational resources will be available to all participants and the trial will be run in a way to ensure that all participants know that their experiences and outcomes are important to the study team.
In view of the behavioural nature of the trial, and the need for a control group, mass media will not be approached, although if recruitment is low, this decision may need to be reviewed. In the trial documentation, participants will be advised that they will receive different forms of educational and/or personal support.
Approaches to implementing peer support programs that address the 3 core components: We have devised a 2x2 factorial study design with geographical clusters allocated to neither intervention, a 1:1 peer support intervention, a community based peer support group intervention or both a 1:1 and group intervention. All three programmes would assist daily management and living with diabetes as well as social and emotional support by promoting self efficacy and providing buddying, mentoring and modeling. The group approach would also be able to provide wider social support. All participants will receive access to educational materials and normal care from their healthcare providers. The three intervention groups will draw their discussion topics from the barriers to diabetes care identified in the mail survey (filtered to protect privacy in the group sessions), and also from a list of core topics for the programme to cover. Within the combined individual and group support arm of the trial, participants will be encouraged to agree which topics should be covered individually, and which should be discussed in the group sessions. The qualitative analysis will explore the impact of the different delivery modes on the content discussed and participants' perspectives on its value to them. We have summarised the way that the different intervention programmes will provide the three core components of peer support.

Delivering the individual 1:1 approach:
In the 1:1 peer approach, peers would be allocated a caseload of up to 10 individuals depending on their time. The peer would be from any of the 1:1 clusters, but resident as near to the participant as possible. Peers would follow their training in relation to the participant.
Participants would be expected to have completed each aspect of the framework within 6 months.

Delivering the group approach:
The groups would be geographically based and therefore held largely within walking distance of the participant residence. Groups are expected to include 20 individuals on average, but this

RAPSID Background and Methods
David Simmons Version 1.0 6/22/2014 Page 12 will vary according to size of community. The size and dynamics of such groups will be allowed to evolve naturalistically, but will be described. Each group would have two leads, preferably from the cluster, although this may not be practicable for some of the smaller villages. A further volunteer would also be appointed and trained. The group would progress through the barriers framework as with the 1:1 approach, but as a group rather than as individuals. Contact between meetings would generally be encouraged, to widen the peer support action. The group would also be encouraged to develop wider social programmes within their village/town and to attract others with diabetes from the cluster who were not in the trial, providing it did not compromise the dynamics of the group. Alternatively, new groups could arise should membership become too large. Temporary group leaders and new leadership training may be needed in such instances.

Delivering the combined group and 1:1 approach
Peers would be trained to deliver both the 1:1 and group interventions. Participants who drop out of the groups would be followed up 1:1 by the trained peers. Individuals who require additional input beyond the groups would also be followed up in the 1:1 setting, comparable to the 1:1 approach..

The Peers
Patients such information given by a person with diabetes would be taken "in a different way", would help encourage people to overcome difficulties and would come from a different perception of management than a health professional.
The peer would need to be "like" the person being supported in relevant ways eg age, socioeconomic group. Geographically based groups would have this as a commonality.
Young people would be likely to want a young peer/peer group Peers should be voluntary and could be expected to give 4-10 hours/week with no payment but costs covered. Peers would require "basic people skills" and would need careful selection.
Support could be a few minutes on the telephone, face to face or any other way -this would be expected to evolve. Frequency of contact would depend on time available, need (i.e. baseline status) and how far down the support programme the participant was. Dependency would need to be avoided.
Peers will be sought from the recruited cohort as well as local diabetes services and primary care by both recommendation and invitation. Peers would need to come from intervention clusters (or outside the study area for 1:1 peers including Cambridge, Wisbech and Huntingdon). Criteria for selection will be developed. Participants who became peers would be excluded from the randomized trial (but part of a study of the impact on peers themselves).
A succession strategy will ensure that if peer leaves there is more than one peer available. We will train reserve peers from outside the study area, or from areas with pre-existing support groups to provide back-up cover.

Peer Training
Training of peers would be undertaken separately for the 3 intervention programmes. This will cover: A) the theoretical basis of peer support and behaviour change interventions, drawing on

Linkage with local clinical services
Peers will be linked with local services through: The training team will include members of the local diabetes and primary care services Supervision will be by a diabetes specialist nurse employed for this purpose and co-located This linkage framework will be created in a way that should the trial show peer support to be successful and cost effective, the service that trains and supervises peers could be commissioned as part of the local diabetes and primary care services.

Intervention Timeline
The intervention will have 3 phases: Pilot stage to include formative evaluation of a short pilot in a non-study area (Cambridge city).
First 6 months of the trial, implementing a structured programme on a monthly basis Second 6 months of the trial as a maintenance phase, where contact is maintained, but with a structure and frequency as agreed by the participants and peers.
The latter will link into a 6 month translation stage, to allow transition of the optimal trial approach to peer support to a wider community wide approach while some of the trial team remain. This is not funded within the current application and we will seek funding for this purpose from local sources.

Approaches to refining procedures and peer support intervention in months 1 -8
Appendix 4 outlines the key activities in Months 1-8 including trial administration, intervention and evaluation. Of particular importance is the piloting of the peer intervention after the developmental work. The pilot study will be undertaken in months 6-8 in Cambridge, as it is not included in the study itself. Volunteers will be sought through local practices and the diabetes services to pilot the individual and group peer support interventions. After the 3 month trial period, participants may want the intervention to continue and this would provide an opportunity to develop strategies ahead of each stage of the trial.

Approaches to evaluation
Evaluation will include quantitative, qualitative and economic components collected through a range of methods to minimize measurement burden on participants and evaluation costs. Initial barriers mail survey during invitation into trial: This will be undertaken at the time of invitation into the trial and based upon prior studies (5,6). We expect to get some data on those who subsequently are not part of the trial. This will help us to assess the generalisability and reach of the intervention.

RAPSID Background and Methods
Baseline data collection: This will be at a local venue (e.g. general practice, community venue) and include consent to access participant health data (including metabolic and health utilization data). Key metabolic outcomes are HbA1c, lipids and blood pressure, weight and hypoglycaemia. HbA1c and lipids will be largely available from general practice/diabetes services as this is comprehensively (approximately 98%) collected through the Quality and Outcomes Framework for Diabetes. Where this is not collected, blood will be drawn. Blood pressure and weight will be measured in a standardized manner in view of the potential variability in quality from clinically colleted data. Information on hypoglycaemia will be collected by questionnaire. Health service utilization will be assessed through clinical records including ambulance use. Relevant demographic and health data will be collected. Specific questionnaires relating to self management will include those agreed across the Peers for Progress groups and likely to include questions relating to self-efficacy (54), diabetes self-care activities (55), medication adherence (56), family and friends subscale of the Chronic Illness Resources Survey (57). Quality of life will be assessed using the European Quality of Life-5 Dimensions (EQ 5D) which is well validated in the UK (58) and does not overburden participants.
6 months (mid point) data collection: This will include questionnaires completed by telephone and collection of existing health data 12 month (final) data collection: This will include comparable data to the baseline data collection.
Data collection: It is anticipated that some baseline and follow-up data will be collected by practice nurses at the participants' practices (working in separate measurement sessions), or by

Qualitative analysis (Simon Cohn)
A key dimension of this proposal is the suggestion that the value of peer support, both one-toone and in groups, is likely to be complex and multidimensional, and as a consequence difficult to assess solely using pre-determined and restrictive measures. Crucial to this study, therefore, will be the parallel collection of observation data from the peer training, debriefing and intervention sessions, the interview data from a sample of the participants, peers and project staff, in order to provide a more rich and nuanced record of both types of interventions. The aims of the qualitative analysis will be to explore the meaning of peer support as understood by peers and participants, their perspectives on the advantages and disadvantages of individual versus group support and ways that it might be improved. By analysing the data to establish common themes and concerns, we hope to be able to formulate a more sensitive understanding of the central mechanisms at play, and the significant differences when delivered to an individual compared with a group. The qualitative component will primarily be managed by S.
Cohn, who has had extensive experience in similar research using qualified assistants. The data will be collected throughout the intervention period, providing a unique longitudinal record of the project as the interventions are delivered, and patients increasingly participate. Transcripts and notes will all be entered on NVivo data software to provide a common dataset that can be integrated with the quantitative and descriptive variables.

Economic analysis (Amanda Adler)
We will perform an economic evaluation to answer the following questions from the perspective of the National Health Service: Among individuals with diabetes in Cambridgeshire -(1) What are the costs associated with each method of peer support? (2) What is the change in effectiveness (HbA1c) and utility (quality of life) associated with each method of peer support? (3) What are the costs associated with this change? We will measure utility using the EQ5D. We will model cost utility to estimate incremental cost effective ratios comparing 1:1 peer support to the absence of 1:1 peer support and group support to the absence of group support. We will use the UKPDS Outcomes model to assess potential years-of-life gained given changes in HbA1c.

Reach to and engagement of intended audience
The GP registers and Quality and Outcomes Framework data provide a denominator for the prevalence of diabetes across the county (3.3%). The local health board (the Primary Care Trust) has offered to assist with any data available. From this and local GP data (anonymised to avoid any privacy issues), we hope to be able to get a denominator for each cluster. This should allow an estimate of participation, both for the trial and uptake of peer support. Characteristics of those participating will be known from the baseline data collection. Some data will also be available on non-trial participants through the mail survey. Further data on individuals initially not engaging may become available in clusters randomly allocated to groups or both peer interventions through the local networking and de novo attendance at the groups. Such individuals will not be seen as part of the randomized controlled trial, but will be invited to be assessed through the trial and any changes assessed as a retrospective pre / post-intervention comparison. Peers will undergo the same measurements as the participants to assess any changes in a pre-post manner. This will allow assessment of the impact of the programme on their health and self-management behaviour.
With routinely collected clinical data (eg HbA1c, lipids, blood pressure) being so comprehensively collected under QoF, we would also hope to be able to assess any change in metabolic control for all of those with diabetes in the cluster. Again, we would find a way by which this data could be provided in an anonymised way. We would also be able to assess some changes in health service utilization within clusters (eg by assessing hospital admission rates by postcode). We will use the RE-AIM (reach, effectiveness, adoption, implementation, and maintenance) framework which takes a broad perspective of issues that are important to policy makers to evaluate the public health benefits of the approaches tested (59).

Intervention fidelity
The extent, completeness and quality of the implementation of the intervention will be assessed during the above qualitative and quantitative evaluation, the peer-diabetes specialist nurse interactions, meeting records and activity logs prepared by the peer supporters. These will include attendance and the content of peer support discussions. We plan to develop this work in further discussions with Wendy Hardeman, a colleague in the Cambridge General Practice Research Unit (60).

Special issues
Reasons for drop out are particularly important and individuals will be carefully followed up and interviewed where possible. Family members and carers would be invited to participate into both 1:1 and group approaches through the person with diabetes. They would be asked to complete a questionnaire adapted from the patient questionnaire and would be included in the qualitative analyses.

Data management and analysis (Toby Prevost-statistician)
The trial design is randomised two-by-two factorial, testing main effects of Peer support versus no peer support (factor 1) and Group versus individual (factor 2). Randomisation to the four arms will be at the level of the village/area ("cluster"), with patients in the same village/area cluster randomised to arm. The justification for cluster randomisation is (1) to enable feasible group sizes in villages/area (2) to minimise contamination arising from contact between participants in the same village/area (3) bearing in mind that, within the two group arms and to a lesser extent in the peer-individual arm, clustering of outcomes would already exist even with an individual randomised design, reducing the marginal effect of cluster randomisation on sample size inflation. A restricted randomisation method that incorporates cluster-level stratifiers will be used, with randomisation performed centrally by a third party independent of

RAPSID Background and Methods
David Simmons Version 1.0 6/22/2014 Page 20 trial coordination, free from access to the study database, and with clusters identified by unique study numbers without access to village/area names. The primary outcome will be HbA1c. A plausible and achieved effect size difference in mean HbA1c in another study (19) is 0.36. We propose a realistic effect size in the range 0.3 to 0.4 in difference in mean HbA1c for each factor as worth detecting. A recent trial in patients recruited from Cambridgeshire and Oxfordshire (53) has demonstrated a standard deviation of HbA1c between patients of 1.25.
The Cambridgeshire ADDITION trial provides a recent and local estimate of the intracluster correlation coefficient for HbA1c of 0.037 (46).
In the proposed study we anticipate being able to recruit a mean of 20 patients per randomised villages/area. Allowing for 10% dropout in access to records of HbA1c at study follow-up, we therefore anticipate a mean cluster size of 18 participants at the analysis stage. With 1520 participants recruited in 76 randomised village/area clusters (20 per cluster), and allowing for four clusters to withdraw from the study without allowing access to patient follow-up data, we anticipate that 1300 participants (90% of 1440) will be followed up with an HbA1c. There will be 90% power to detect a difference of 0.3 in mean HbA1c for each of the two factors at the 5% level of significance, allowing for a design effect due to clustering of 1.63, based on mean cluster size of 18 participants per cluster and ICC of 0.037. There will also be 90% power to detect a difference of 0.4 in mean HbA1c between any two of the four arms, and these comparisons offer protection against any unexpected interaction that might be observed between the two factors. As a sensitivity analysis for this sample size calculation, in the event of a higher observed ICC, for example arising from the interaction between grouped participants, this coefficient would need to be approximately twice as large (0.07) to be able to effect a reduction in power from the desired 90% to a minimally acceptable 80%. For context, the ICC in a previous trial was 0.047 (61) HbA1c follow-up 270 other outcomes follow-up 18 followed-up areas Other measures may not be able to be ascertained from medical records and a reduced followup rate of 75% from questionnaires is assumed. With 1080 followed-up (15 participants per cluster), there will be 90% power to detect sufficiently small effect size differences in each factor of 0.25sd for factorial main effects, and 0.35sd for pairwise comparisons between arms.

RAPSID Background and Methods
Statistical Analysis: All analyses will account for the clustered design through the use of linear mixed effects for modelling continuous outcomes and simple comparisons of proportions, and Generalised Estimating Equations with exchangeable correlation matrix for modeling binary outcomes. Where a baseline of an outcome is available there will be an adjustment for this covariate in order to improve the precision of the estimated intervention effects. Participants with a missing value for baseline will be retained using the method of White et al. The analysis will follow guidelines for the approach for factorial trials (62,63). The methods will be adapted to allow for varying intracluster correlation by arm where this is identified. The influence of missing data for the primary outcomes will be investigated using the multiple imputation method of Rubin outlined in Shafer and the primary intention to treat analysis will be supported by per protocol analysis (64,65). Further before-after analyses will allow assessment of change in those clusters involved in the wait-list design. All analyses will be two-sided and assessed at the 5% level of significance. An interim estimation of variances, cluster sizes, cluster withdrawal, and participant follow-up and questionnaire return rates will be made to ensure adequate study power to detect effect sizes in the stated range. A trial analysis plan will be developed prior to the analysis. This is an experienced research group with wide experience in collaborative research and we are keen to collaborate with other groups.
A The group has been involved in curricula development in DAFNE (a national educational/empowerment programme in the UK), lay health worker training in New Zealand and peer support in breastfeeding and diabetes in the UK and New Zealand.

B
The group has been involved in identifying key common evaluation indicators in metabolic control (eg in diabetes in pregnancy (66), in diabetes overall (67)), self management behaviours (eg readiness to change lifestyle (68), peer support for breastfeeding (35), Barriers

E. IRB APPROVAL AND INFORMED CONSENT
Provide a brief discussion about the ethical implications of this study, how human subjects will be protected, and regulations for assuring such protection that govern the applicants and their institutions.
1. The major ethical issue is around ensuring liaison with the usual health care providers and service purchasers with whom we already work. 2. A further issue relates to the end of the trial and how peer support might be integrated into normal care at the end of the trial if successful. (As peers are to remain unpaid, the only additional expenses would be operating costs, most of which would take place in the fist 6 months of the trial. The area has a Diabetes Care Patient Advisory Committee and discussions will be held depending on the outcome of the trial.) 3. Confidentiality by peer supporters. (Selection of peers should include a general practitioner reference or review of application. Peer training should cover confidentiality) 4. Confidentiality in support groups (Cover in peer training. Include in group rules at start of group) 5. Access to GP records. (Patients will be identified from searches of GP registers, conducted by practice staff, or by PCRN staff with honorary contracts form the practices to cover this. Invitation letters will be sent directly from practices, so recruitment data need not be transferred out of the practices. Individual participants will then provide written consent for research staff to access their records for follow up). 6. We will ensure ongoing support for the peers, which will enable the team to identify and address unexpected ethical issues if they arise. There is a fully functioning Ethics Committee in the area for protection of participants. We will apply for approval before commencing the research and meet all NHS ethics and Research Governance requirements.