Managing with Learning Disability and Diabetes: OK-Diabetes - a case-finding study and feasibility randomised controlled trial

Informed consent

In developing our informed consent procedures and materials we drew on the NHS Health Research Authority’s guidance on supporting informed participation in research, as well as the toolkit supporting informed consent with those lacking capacity [https://connect.le.ac.uk/alctoolkit (accessed 2 January 2018)], the Department of Health guidance on assessing capacity and obtaining consent both for clinical care and research37 and relevant literature. 44 We approached four other research teams identified from the National Institute for Health Research portfolio as working in the area of learning disability, and they kindly shared the materials developed for their own projects (see Acknowledgements). We also worked with our third-sector partners and learning disability specialists to develop appropriate materials.

All researchers were trained in assessing mental capacity by the learning disability consultant on the team (AS). In line with guidelines in the Mental Capacity Act 2005,36 the presumption was of capacity and a judgement of lack of capacity was made only after all practicable steps to help the participant make a decision had been taken by the researcher in collaboration with the person’s supporter. We were mindful that capacity is not fixed and context free, and so capacity was constantly reassessed during the information-giving and data-collection processes. Our aim was to develop a pragmatic approach to consent that balanced ‘self-determination, respect and safety’ with strategies to promote participation in the research. 44

Participants recruited into the study were provided with information about a system of research advocacy that we established with a third-sector organisation, People in Action [http://peopleinaction.org.uk (accessed 2 January 2018)]. They were provided with a confidential number that they or a supporter could call if they wanted to ask questions about the project, express concerns or withdraw if they did not feel they had the confidence to say so directly to a member of the research team.

We were aware that in some cases the supporter of a person with a learning disability may have a learning disability or other mental capacity issues themselves and so we did not make assumptions about their ability to support self-management. A supporter was defined as the main adult who was self-nominated by themselves, or nominated by the participant or by a professional who knows that person, as providing practical help and support in day-to-day living relevant to their diabetes management.

There was a further ethical challenge in considering how to explain the study to those people who had not previously been identified as having a learning disability. This was addressed by developing information materials in collaboration with third-sector organisations and their customers with a learning disability to ensure that the materials were fit for purpose and would be acceptable to our participants. We did not make reference to ‘learning disability’ or ‘learning difficulty’ anywhere in our participant or supporter materials. After suggesting a number of options to service users via People in Action, on its advice we used the abbreviation ‘OK-Diabetes’ to offer a simple memorable project name for participants, again with the aim in mind of not making reference to learning disability in our materials. In each interview, the interviewer discussed with the participant why they had been referred and that the study included people with learning disabilities. It was discussed whether or not a participant minded being in the study and if they felt they had a learning disability.

Safety

We developed a safeguarding policy for the project in consultation with the Safeguarding Adult Partnership Board in Leeds. This set out when researchers should break the confidentiality of the participant and who they should report concerns to depending on the level of perceived risk. The researcher’s obligation to report safeguarding concerns was discussed with the participant. Any non-urgent concerns that did not need an immediate response were reported back to the chief investigator and lead co-investigator for discussion and advice. We also developed a safety protocol for researcher lone working in which researchers were required to call a nominated member of the team prior to going into the interview, and once again straight afterwards.

Informing general practitioners and recruiting by direct referral

We wrote to all general practices at the three study sites using a covering letter from their Clinical Commissioning Group (CCG) lead for either diabetes or learning disability. The covering letters were sent to practice managers and/or diabetes lead GPs, gave CCG endorsement to the study and urged practices to take part. Letters were followed up by telephone calls from a researcher. Practices were supplied with project information, examples of the easy-read information that would be sent to participants, a step-by-step guide to recruiting, a recruitment log to record methods of identification, a referral form with Freepost envelope for successful referrals and an easy-read template letter to send to participants who were unable to be reached by telephone. In our original proposal we intended to develop a checklist for health professionals to help them identify potential participants, particularly individuals who may have milder levels of intellectual disability and not be on the learning disability register. We developed a step-by-step recruitment guide for health-care providers (see Appendices 1 and 2), which, in addition to giving information on cross-checking learning disability registers with diabetes registers, informed practitioners how to run the Read Code search against electronic patient records. Our referral forms included a checklist based on one that was developed by the Royal College of Nursing to identify individuals with a learning disability. However, initial returns of the log, despite numerous missing data, suggested that practices were only running the computer-based searches or were finding them the most productive way of identifying people, suggesting that either this checklist was unhelpful or the majority of GPs were unwilling to try and identify a person as having a learning disability who was not already on the register. For this reason the study did not further develop the ‘simple checklist’ as expected, and instead it focused on register and Read Code searches, as these proved to be the most popular method of identification with clinicians.

Practices were offered the opportunity to meet the researcher to discuss the project, as many had concerns about referring a ‘vulnerable’ patient group. The researcher would attend the practice, explain the project and demonstrate the participant information materials to ensure that practice staff knew the study had taken communication issues into consideration. Over the course of the project, practices that did not respond were written to three times and received at least two follow-up telephone calls. Practices that took part in the study were paid service support costs by the Primary Care Research Network.

A researcher attended CCG meetings, giving talks on the project to practice managers and distributing project information. Time for Audit, Research, Governance, Education and Training (TARGET) training was also attended by a researcher who spoke with practice staff and distributed project information. ‘Research-ready’ practices (through what was then the Primary Care Research Network) were recruited through its regular research network meetings. The Commissioning Support Unit supported the project, circulating project information at its events, in newsletters and to its contacts. The researcher attended Continuing Professional Development training relevant to the project and presented to the clinicians there.

Case finding via searches of general practice databases

We developed Read Code searches to help general practices identify potential participants. Read Codes are clinical codes applied to patient records that can create searchable tags of information about a patient but are not limited to diagnosis codes. They can be about a patient’s identity (ethnicity for example), their social circumstances, living and care arrangements or about their symptoms. There are many Read Codes, and codes are not used routinely or consistently across practices. Some codes will automatically include a person on a list or register, such as the diabetes QOF register, or trigger a required response, for example depression symptom codes can trigger an automated reminder to screen for depression; however, not all codes are linked to registers or actions.

Different computer systems use different codes. Bradford and Wakefield use SystmOne (The Phoenix Partnership, TPP, Leeds, UK), whereas the CCGs in Leeds use a combination of SystmOne, EMIS Web (EMIS Health, Leeds, UK) and EMIS LV (EMIS Health, Leeds, UK). The research team met with system specialists and learning disability experts to design a search that could be run on any practice system to create a list of patients whose Read Codes indicated they might have a learning disability or had accessed learning disability services. It would then be up to a member of their direct care team to review these results. The search included diagnostic codes for learning disability and chromosomal or other physical abnormalities, service access codes (e.g. referred to community learning disability team) and functional ability codes (e.g. cannot read). Other codes were discussed but rejected for generating too many results, for example adding the code for ‘unable to perform personal care activity’ added over 16,000 people to the results for all of Bradford.

Two practices agreed to test the search and review the results. Both practices knew their patients with a learning disability very well. They agreed that the search had identified some of the patients known to the practice as having a learning disability but having no formal diagnosis. However, it was agreed there were too many false positives and several codes were removed that had not generated any new eligible patients. Once refined, the Read Code search of codes indicating potential learning disability was then linked to the records of patients who had type 2 diabetes but were not on insulin to further reduce the number of results and ensure eligibility. This search was linked to two other searches and uploaded to practice systems. Practices were able to search their:

1. learning disability register crossed with their QOF diabetes register (excluding type 1 diabetes and people on insulin)

2. whole list for patients with a Read Code that indicated learning disability and type 2 diabetes (excluding insulin)

3. whole list for patients with a Read Code that indicated learning disability.

A member of clinical staff was asked to review the results provided by searches 1 and 2 and recruit all potentially eligible patients. Search 3 was provided to allow practices to review their learning disability registers if they wanted to do so. Searches were embedded in the clinical systems of each practice so they could be directly accessed through their clinical reporting files. In consultation with GPs, it was agreed that if the patient lists generated by the Read code search were too long, clinical staff would not be expected to review them. In practice, the simplicity of this arrangement proved popular as it took only four ‘clicks’ to run all of our searches. Members of the research team gave support on the use of the searches if required.

Secondary care services

Project coapplicants (DN, RA and AS) were located in secondary care settings. They referred patients to the study and recruited colleagues to refer. One site assigned a clinical studies nurse for recruitment. We approached all secondary specialist learning disability services and diabetes services in the three cities directly by letter, indicating our inclusion and exclusion criteria. We asked them to review their case load for files flagged with a learning disability and to keep the study in mind as they saw new patients. A researcher attended the Comprehensive Local Research Networks Diabetes Research Network meetings and encouraged network members to refer to the study. A researcher visited all learning disability community teams and at these meetings a key person was identified who would be the liaison between the research team and the community teams. The learning disability leads for the hospitals were involved in the project recruitment as a key figure to support recruitment in secondary care. The research team also contacted retinal screening services and specialist dentistry services to identify potentially eligible patients.

Third-sector organisations

A significant amount of support for people with a learning disability is provided by non-statutory organisations, for example in relation to advocacy, housing and family issues, health, leisure and employment. In association with partner organisations, People in Action and Tenfold [www.tenfold.org.uk (accessed 2 January 2018)], we identified third-sector services that may be able to support the identification of potential participants in their client base. We worked closely with the learning disability partnership boards in each area to endorse our project to these services and local authority services. We recruited through a variety of services, including housing organisations, care/support organisations, Citizens Advice Bureaux, day centres, advocacy services, religious groups and community volunteering projects. A variety of recruitment methods was used including becoming familiar faces at learning disability events (e.g. in Learning Disability Week), day centres and learning disability organisation networks.

Baseline researcher interview

The interview (see Appendix 5) used a standardised format to (1) check understanding of the study using an easy-read booklet; (2) assess mental capacity to be involved in the study; (3) obtain written (or verbal when necessary) consent; (4) obtain consent to review medical records for routine clinical measures; (5) establish diabetes management, including diet, physical activity, treatment, self-care awareness and engagement with health services; (6) record mood, feelings about diet, activity levels and having diabetes; (7) identify the role of supporters in the participant’s diabetes management; (8) elicit preferences for further assistance with diabetes management and consent to re-contact for further research; and (9) record a nominated supporter to be involved in further contacts, as applicable.

The interview guide was developed with, and reviewed by, staff from People in Action, a third-sector service provider for people with a learning disability. Measures that required long-term recall (defined as > 1 week) were found to be impractical. The full interview guide was piloted by two employees of CHANGE who had a learning disability and diabetes and was altered based on their feedback.

With one exception, we did not identify standardised existing research measures for mood, diet, attitudes to diabetes or knowledge about diabetes that were appropriate for people with a learning disability in this research setting. 45 Most of the standardised measures that we identified did not have sufficiently robust psychometric properties to overcome doubts about their utility. Even self-report questionnaires with 12–15 items were considered too taxing by our expert advisers and when they depended upon making ranking judgements, such as on a 4-point severity scale, they were beyond the intellectual ability of our participants. Instead, we used such standardised measures as the basis for deriving simple and important questions to ask in the interview. Respondent burden was also a major consideration in our choice of measures and influenced by the cognitive abilities of our sample. A standardised measure that might have been feasible but that takes substantial effort to complete could not be justified unless it met the core needs of the project. For example, although there are standardised depression measures used in learning disabled populations,46 a questionnaire such as the Glasgow Depression Scale,47 even if it could be completed, takes 10–15 minutes with a supporter and would produce a result only indirectly related to our core objectives.

However, because of their particular importance in this study, we did attempt to capture data on diet using the Rapid Eating Assessment for Participants (REAP) – Shortened Version. 45 Although this measure is not designed for people with a learning disability, we supplemented the questions with visual aids, for example, pictures of common foods and a retrospective food diary for the previous week to prompt recall.

The interview employed mainly closed questions (see Appendix 5). When possible, questions were phrased in an explorative conversational form to promote understanding. The interview process was supported by visual aids in the form of laminated picture cards that illustrated the topics the interview would cover and images related to questions about, for example, family, employment, diabetes care, diet, activity and mood.

Evaluating the research process

At the end of each interview the researcher asked the participant how they found the questions on a scale of easy to hard. They were also asked to assess the acceptability of the length of the interview. The interviews were conducted with all registered participants and transcribed.

After each interview, the researcher audio-recorded their own observations using a predefined topic guide. This covered reflections on the consent process and assessing mental capacity for involvement, how well participants understood or were able to respond to questions independently or with support, whether or not the study materials worked well and the researcher’s ability to complete the EuroQol-5 Dimensions (EQ-5D), mood questions and health economics service use questions with the participant and observations about home life and relationship with supporter if present. Journals were initially transcribed to support discussions about eligibility associated with mental capacity; however, they were also used in the analysis phase to supplement the qualitative analysis.

Medical information

A medical information form sent to GPs after the baseline interview included questions to confirm eligibility, obtain recent diabetes-related test results [including glycated haemoglobin (HbA_1c) level], comorbidities (as defined by presence on QOF registers), current diabetes medication and health service usage. The information relating to the last two items was collected on forms designed for the project as part of assessing the feasibility of an economic evaluation in a definitive RCT (see Chapter 3).

Sample size

Across the three sites we expected an approximate total of ≈1400 people in the target population – assuming a 80% population of adults, 6% diabetes prevalence (in whom 20% use insulin) and 2% learning disability prevalence. We therefore aimed to achieve 50% GP involvement across Leeds, Bradford and Wakefield, equating to approximately 117 practices, and we assumed an eligibility rate of 50% to identify approximately 350 people meeting our eligibility criteria. A sample size of 350 people would allow a conservative estimation of the proportion eligible for the RCT of the target population to a reasonable degree of accuracy [to within 5.2% with a two-sided 95% confidence interval (CI)]. It was this original target that is cited in our funding application and in the original study protocol.

Following a Health Technology Assessment (HTA) programme review of recruitment in June 2014, our original target of 350 was lowered to a minimum of 120 following acknowledgement of a lower than expected recruitment rate. The lower recruitment rate was a result of underestimating the use of insulin in type 2 diabetes, less than expected young-onset type 2 diabetes and a stronger than expected bias of learning disability registers to more severe cases. A higher than expected potential eligibility and acceptance rate for the feasibility RCT meant that we were confident with the new recruitment target. Case-finding recruitment was also extended in order to continue in parallel with RCT recruitment.

Patient populations

Analysis methods

As this was a case-finding study for a feasibility study, outcomes were not subjected to formal statistical testing and so no hypotheses have been proposed.

For case-finding and eligibility outcomes, percentages were calculated using the total number of participants or forms expected in the relevant population as the denominator (including all participants with missing data for that variable), whereas for outcomes characterising the population, percentages were calculated using the total number of participants with available data as the denominator. Therefore, participants with missing data for each variable were excluded from the denominator and the number of participants excluded is presented alongside each summary. Analyses were carried out using statistical analysis software SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

To provide a robust estimate of eligibility

Analysis presents the flow of participants through the study within a Consolidated Standards of Reporting Trials (CONSORT) framework, including reasons for exclusion at each stage (referral, interview and consent to be approached for further research). A 95% CI for the proportion of eligible patients out of all referrals was calculated.

To evaluate a simple case-finding method to identify participants

Analysis generated descriptive statistics for patients agreeing to researcher contact, patients for whom a researcher visit was conducted, registered patients and the eligible population, when appropriate.

To characterise the population

Analysis generated descriptive statistics for participant demographics, participant-reported measures and measures of diabetes control including resource use for the eligible population, with the exception of demographics also summarised for the referred population. Measures of diabetes control were also presented categorically according to abnormal ranges on standard criteria. 48–51

Exploratory post hoc descriptive subgroup comparisons of diabetes control were made for BMI and HbA_1c level by the patient’s response to a number of questions, including feelings around mood, diabetes, eating, weight and exercise, whether or not there was a supporter and whether or not they were taking medications.

Development of the treatment as usual materials

Given the variability in care pathways across the three cities, we decided to ensure a basic level of diabetes care and knowledge in all participants. It was agreed this would be done by supplying a standard leaflet to all trial participants. The research team decided to review existing information resources to evaluate if any could be adopted as their standard leaflet.

Members of the third-sector working group, project team members and easy on the i reviewed 23 existing resources developed for use in the UK and aimed at people with a learning disability who have type 2 diabetes (see Appendix 6). We aimed to evaluate these resources in terms of their accessibility and the degree to which they covered the main elements of diabetes self-care, as identified in current guidelines:52

• understanding what diabetes is

• understanding the need for healthy diet and regular meal times

• the importance of doing exercise

• the importance of following treatment/medication regime

• how to access services and important check-ups, such as retinal screening and foot care

• what to do when not feeling well, for example in terms of following treatment

• managing acute complications, that is, hypoglycaemia and hyperglycaemia.

Each resource was evaluated for comprehensive of coverage of these themes, the quality and appropriateness of images, clarity of text, ease of understanding (e.g. inclusion of medical terms) and relevance to our study population. No stand-alone resource met all these criteria to a high level. However, a leaflet developed over 10 years ago by Diabetes UK and a Leeds-based organisation, CHANGE, was identified as the best single resource in terms of the above criteria. This leaflet was no longer available in electronic format, was outdated in terms of its presentation (black and white line drawings) and some content no longer reflected current health advice guidelines. The team therefore approached Diabetes UK and CHANGE to collaborate on the design of a new resource.

CHANGE, working under the guidance of people with a learning disability, provided the graphic design expertise for the resource and reviews by people with a learning disability at each iteration. The new guide was informed by current best practice in developing accessible information provided by our expert advisers and from a review of the research and national guidelines in this area. Ramzi Ajjan, consultant endocrinologist with special interest in diabetes, reviewed the medical content. Christine Harris-Moores, specialist learning disability dietitian, reviewed the dietary advice. Diabetes UK ensured that the content met its current advice guidelines for type 2 diabetes.

Development of the supported self-management intervention

Supported self-management, which was specified in the commissioning brief as the intervention to be tested, is an existing approach to chronic disease management in which the underlying theory and the basic principles of implementation are already established. Development work in the current project did not, therefore, need to start from the early phases of intervention development as recommended by, for example, the UK’s Medical Research Council53 – elaborating theory and undertaking early-phase proof of concept and efficacy studies. Instead, we planned our development work in five phases:

1. Clarifying the principles of SSM – as summarised in review articles and didactic pieces, and as identifiable from the protocols and final reports of individual studies of self-management in diabetes.

2. Identifying the reported barriers to effective self-management of type 2 diabetes in adults with a learning disability – from published literature.

3. Reviewing existing materials that aim to support self-management of diabetes for people with a learning disability for examples of good practice – in consultation with service users.

4. Synthesising the outputs from the first three phases – to decide on those elements of SSM that are most relevant to the needs of our target population (will respond to likely barriers), and most likely to be acceptable and useful (match the identified good practices). This phase involved a series of problem-structuring and consensus meetings in the research team, checking interim outputs against guidelines on reasonable adjustments and by consultation with experts and service users.

5. Implementation and field testing of early versions – modifying the intervention materials in the light of feedback from research nurses.

At each stage there were regular consultation meetings involving members of the research team with people with a learning disability and their representatives in easy on the i, People in Action and Tenfold, as well as clinical experts in learning disability.

Phases I and II

We realised early in the project that there was no relevant RCT in the area of diabetes self-management in learning disability and, therefore, we did not undertake a formal systematic review of effectiveness studies. Instead, we aimed to use the published literature to identify (1) the principles of self-management that we would hope to embody in our own intervention and (2) those influences on self-management potential in our population that should inform the form or content of a SSM programme.

To help identify published information relevant to our intervention, an information scientist (JW) undertook initial searches in four main areas:

1. SSM in chronic disease

2. self-care in diabetes, including barriers to effective self-care

3. diabetes and a learning disability – a broad search to identify factors that might be specific to the target population

4. descriptions of specific interventions aimed at improving diabetes control in adults with a learning disability.

The search strategies are provided in Appendix 7.

We also reviewed two relevant National Institute for Health and Care Excellence (NICE) guidelines,26,48 two Cochrane Reviews,54,55 a guideline on SSM published by Diabetes UK52 and outlines of national standards in diabetes management from the USA56 and the UK. 57

On the basis of these searches we identified and reviewed 707 titles and abstracts on the topics of self-management of chronic disease, including diabetes, and 350 titles and abstracts on the topic of diabetes in adults with a learning disability.

Titles and abstracts were reviewed by two of the applicants (AH and GL) and full versions of relevant papers were obtained. We categorised retrieved papers as follows:

• reviews of self-management and diabetes (n = 10)24,25,58–65

• individual self-management programmes or interventions (n = 22)66–87

• research protocols describing individual self-management interventions (n = 8)88–95

• observational studies reporting influences on self-management – barriers and enablers (n = 31). 29–33,96–122

Initially, we extracted data from the papers that described the form or content of self-management. We developed an initial framework derived from reviewing the two NICE guidelines,26,48 two Cochrane Reviews54,55 and the guideline on SSM published by Diabetes UK. 52 We then developed this framework by reference to six individual studies in the reports67,74,81,123–125 for which we found reasonably comprehensive descriptions of the intervention. The remaining studies were reviewed against this framework to identify any missing themes, using a modified approach based upon best-fit framework analysis. 126

Next, we reviewed papers describing barriers to effective self-management, influences on interventions or outcomes that were specific to adults with a learning disability. In a series of review meetings we organised the identified influences into a descriptive framework.

Finally, we combined the two frameworks in a narrative synthesis of these literatures to identify the general principles to be adhered to in implementing a pragmatic and sustainable programme of SSM, and the form and content of the specific intervention for this project.

Phase III

For existing self-care resources, a scoping exercise was conducted in which examples of good practice in interventions around health in people with a learning disability were sought from services in the UK. Sources included charities, for example Diabetes UK, local NHS trusts and patient groups, NHS Choices [www.nhs.uk/pages/home.aspx (accessed 2 January 2018)] and Easyhealth [www.easyhealth.org.uk (accessed 2 January 2018)]. The process here was different from that used to develop the TAU intervention described in Development of the treatment as usual materials. In that case we were seeking to identify the best-available resource to adopt in toto; in the present exercise we were seeking specific instances of desirable practice – individual images, approaches to explaining the principles of self-management, or user-friendly resources such as diary sheets.

We identified 18 examples of resources developed in the UK to support self-management of diabetes for people with a learning disability.

Resources were reviewed by a panel consisting of members of the research team and staff and service users from easy on the i. They were asked what they liked and disliked about each resource, and to note implications for the intervention in development. Results were recorded on a structured proforma.

Phase IV

Using the general approach of problem structuring and priority setting,127 preliminary versions of the SSM package – including not only format and content but also tailoring (for easy reading, visibility for those with poor acuity and so on) – were discussed initially in the research team. Finally, we considered guidance on reasonable adjustments to health care designed to ensure access for people with a disability, to check that we were meeting obligations in that respect.

When necessary, a further round of more focused (purposive) reviewing of literature was used to clarify which were the key principles for this population. We generated two lists (Box 1 and Table 2) that we used to help frame final decisions.

Principles for priority setting were that:

• the intervention should respond to known barriers to self-management reported by people with a (learning) disability, including practical problems such as transport, likely attrition from dropout when multiple attendances are expected and ability to accommodate the presence of a supporter

• the format of the intervention should be likely to encourage self-maintained change beyond an early supported element; in our target population this particularly meant that the intervention should involve supporters who were involved with any aspect of lifestyle (shopping, food choice, physical activity, medication monitoring and so on) relevant to diabetes

• the intervention should be designed to be readily integrated into usual health-care provision in the NHS to ensure sustainability.

Based on all the advice we received, we wanted to give particular salience to:

• practical aspects of self-care, such as buying and preparing food, diet change to aim for weight loss and increasing physical activity

• use of simple (accessible) written materials and charts

• supportive contact both with a professional and with an informal supporter, if one could be identified

• use of practical goal-setting, planning to meet goals and self-monitoring.

By contrast, we decided it would be less helpful to focus on:

• a substantial component of healthy living interventions aimed at alcohol and smoking, because they were not major problems in our group

• education beyond the basic information in the TAU booklet

• information technology (IT)-based interventions, such as web, digital versatile disc (DVD), mobile phone, etc., as these were not readily accessible by our participants

• group-based interventions, as attendance is typically poor and individualisation is harder.

Phase V

We decided that professional support would be provided by diabetes nurses with experience in primary care rather than learning disability nurses: very few of the target population would be in contact with (or taken on by) specialist learning disability services. In routine NHS practice, where 20–25% of patients would be using insulin, diabetes nurses would have more relevant experience; for our population (not all of whom would have been told they had, or would self-describe as having, a learning disability) the diabetes background would be more acceptable. We developed a training plan for the research nurses delivering the intervention covering the underlying principles of mental capacity and of self-management, the individualised elements specific to learning disability, trouble-shooting and dealing with problems (such as behavioural disturbance), and the details of the programme and the materials provided with it. Explicit links were made between the practicalities of the intervention and its rationale in self-management principles (Table 3).

The training programme was delivered by two of the researchers (AH and GL) over three sessions of face-to-face contact with the nurses. An additional session on mental capacity assessment was delivered by Alison Stansfield to the nurses and all research interviewers.

Supervised use of the intervention with three initial cases in the RCT was also arranged. In each case the whole intervention was delivered over a maximum of four visits and the nurses met together with Allan House and Gary Latchford after each visit, to discuss any challenges with implementation. On the basis of this experience, early versions of the intervention were modified in format to make them easier for use by the nurses. In particular, the forms for keeping notes on each contact that were prepared for early versions were found to be overstructured by the nurses and intrusive for use in the field. Once they had familiarised themselves with the principles of self-management and the nature of each contact, the nurses preferred to make free-form notes during contact and then check afterwards that they had recorded all the necessary information and completed the essential standardised forms (case report forms) for the trial.

Both nurses and participants involved in the field-testing reported finding the materials easy to use and the nature of the intervention easy to understand.

The final intervention had four standardised components with associated materials. How they were delivered depended on participant and supporter characteristics and preferences.

1. Establishing the participant’s daily routines and lifestyle: this included current diet and activity routines, participation in daytime social activities or work, shopping and food preparation, current self-reported health and self-management. The main aim of this component was to identify the real social and personal influences in the life of the person with diabetes that would limit their ability to self-manage or that might be mobilised as a resource in supporting self-management.

2. Identifying all supporters and helpers and their roles: a key supporter and other helpers were identified when possible at case finding. Key supporters and other helpers were given written information about the project and if they agreed to support a goal set by the participant, then they were given a written reminder of their role. The main aim was to identify people who might be a useful resource in supporting self-management and to ensure that any changes were embedded in the social network for longer-term maintenance of change.

3. Setting realistic goals for change: the main aim was to avoid prescribing change in the way of good dietary practice or other lifestyle change instead but to support goals suggested by the person with diabetes that were specific, simple and achievable, given the person’s current routines and social support, and consonant with their willingness to make change. The main aim was to encourage engagement in a population usually thought of as having little agency and to introduce the idea of selectable elements in a repertoire of self-management options.

4. Monitoring progress against agreed upon goals: we devised a simple system that did not depend on high levels of functional literacy, using tear-off calendar sheets on which participants noted goal attainment in a yes/no format. The main aim was to encourage active participation in an activity that is a core feature of self-management.

We prepared materials to accompany these activities (see Appendices 8–12).

• For the nurses: templates for a weekly timetable, a chart to record friends and family and other helpers, charts to be completed in collaboration with the person with diabetes – my life, my likes and do not likes, looking after my diabetes.

• For the person with diabetes: an OK-Diabetes board to place in a prominent position at home with a visible record of goals including pictorial prompts, for example snack swaps, a written action plan in multiple formats and tear-off slips to record daily actions.

• For supporters and helpers: an information sheet explaining the study and a card summarising what their role was in helping to support the person with diabetes in meeting their chosen goals.

The research nurse worked through the elements of SSM with the participant, explaining how to use materials and suggesting initial actions and activities. Further contact was negotiated with the person with diabetes. We expected that a total of three or four meetings of 30–60 minutes over 6–8 weeks would be provided, followed by telephone support and advice.

We took steps to ensure consistency in use of the SSM: (1) training and supervision sessions with research nurses, (2) annotation of the intervention materials by research nurses and (3) ensuring nurses had other experience and training in diabetes or learning disability care before the RCT.

Search strategy

We ran literature searches on the standard bibliographic databases in July 2013, and updated the searches in July 2015. Full search strategies for each database searched can be found in Appendix 7.

Papers were included if they:

1. described primary research studies

2. involved participants who were adults with a learning disability

3. described a standardised self-management or self-care intervention

4. described an intervention aimed at weight loss or improving self-management of diabetes.

We defined self-management as involving at least (1) some definition of actions relevant to improving the participants’ condition (overweight or diabetes), (2) setting and recording of specific goals or targets related to those actions and (3) monitoring of progress in achieving those goals. The person with a learning disability had to be an active agent in this process – albeit with the help of a supporter at times – so that they were not just a passive recipient of a programme designed and delivered by a third party.

Developing an organising framework

In a series of research team meetings we developed an initial framework for organising the data, based upon our reading of the wider adherence literature as well as the literature identified in the present search. We distinguished steps taken to:

• ensure the intervention was delivered correctly in form, content and quality

• measure for research that these approaches to ensuring quality of delivery were employed

• measure actual provider adherence

• measure participant adherence.

The first two steps fit with existing literature describing ‘fidelity’ – the degree to which provider delivery is in line with the intended form and content. The last two steps fit with the existing definition of ‘adherence’.

To ensure a measure of adherence was applied to all elements of the intervention and not just a selected few, our initial framework had four categories to describe an intervention:

1. The content of the intervention: topics or components covered, such as (for diabetes self-management) shopping for and preparing food, planning physical activity, taking tablets and avoiding unhealthy behaviours such as smoking or drinking too much alcohol.

2. The techniques employed in the intervention: how it is delivered, for example through education, training in goal-setting, use of self-monitoring and feedback techniques.

3. The platform or format by which it is delivered: for example, written materials, group sessions, self-completion charts, web-based resources or text messaging.

4. The degree of individualisation of the intervention: this could mean use of inclusion and exclusion criteria to define the sample from the target population to whom the intervention is delivered or modification of elements of the intervention to suit the needs of individuals within that sample, for example those with visual impairments.

We used this framework for organising information about adherence, and from each paper we also included detail relating to the type of study participants, type of intervention(s), content of the intervention, how the intervention was delivered, how both provider and participant adherence were measured, collected and scored, and how quality and competence were ensured.

Data synthesis

Our search identified 28 relevant studies, which we used to check and refine our framework. 128–155 As there were so few data available regarding treatment adherence in this population and because they were so heterogeneous, there was no possibility of data pooling. We thus undertook a narrative synthesis, organised according to the framework described in Developing an organising framework, to explore the approaches to measuring adherence in the included studies. A first step was to develop a checklist against which to judge the comprehensiveness of the adherence measure, as in Table 4.

Developing an adherence measure

For each component of the intervention (e.g. a list of supporters, identified goals, actions taken to meet goals) a record could be generated either by the nurse or by the person with diabetes. We identified which of these records we could collect routinely for each participant, using the framework as a check to ensure that all aspects of adherence were being recorded. Each of these records was to be captured on a case report form during the trial.

To measure adherence we also developed a scoring system based upon completion rates, which would allow us to estimate the proportion of completed tasks during the SSM intervention, including an overall judgement about whether or not the whole process had been completed at least once. This final score was captured on a summary form to enable collation of these adherence items and recording the adherence score.

The relevant case report forms are in Appendix 13.

Service usage information

One of the aims of the case-finding study was to identify health and social care use (and thus cost drivers) for inclusion in a questionnaire in which the data collected would be used in an economic evaluation of a later intervention. A second aim was to identify suitable outcome measures for use in such an economic evaluation.

A literature review informed the development of a questionnaire delivered as part of the request for health and social care usage from participants’ GP surgeries (see Appendix 14). The literature included health and social care resources used:

• within self-management interventions by people with type 2 diabetes and a learning disability, and the methods by which those data were collected

• by people with type 2 diabetes within self-management interventions

• by people with a learning disability and the methods by which those data were collected.

The literature review also aimed to identify the outcome measures used in any cost-effectiveness analyses. Search strategies used four concepts: learning disability, self-care interventions, type 2 diabetes and cost studies. Three separate searches were run in EMBASE, MEDLINE and The Cochrane Library using combinations of three of the four concepts in each. The searches were not limited by language or date of publication. A ‘cost studies’ search filter was used to limit studies to those including some cost data.

Inclusion criteria:

• cost-effectiveness studies for healthy lifestyle interventions

• cost-effectiveness studies for diabetes self-management

• studies with costs or outcomes for people with learning disabilities.

Exclusion criteria:

• studies reported in abstract form only

• studies relating to self-monitoring

• diabetes drug studies.

Searches undertaken between May 2013 and February 2014 identified 1189 unique references. No papers were identified that included cost-effectiveness analysis of self-management interventions in people with diabetes and learning disabilities. Titles and abstracts were reviewed by two people and data extracted using a bespoke data extraction template. Given the nature of the data extracted, no quality assurance assessment of the papers identified was undertaken.

Eight studies156–163 that detailed cost-effectiveness analysis of self-management interventions in people with diabetes met the inclusion criteria. The selected studies were lifestyle modification programmes and telephone interventions (both automated and non-automated). Intervention cost data were collected using a variety of methods, including interviews with health workers, standardised templates for staff and participant questionnaires. Staff costs were the greatest drivers of cost in relation to intervention implementation and were calculated using official wage statistics or personnel records. Administrative costs, such as telephone charges, printing charges and translation costs, were added to the expense of the interventions. Participant out-of-pocket costs were collected in two158,162 of the included studies, using staff-completed templates and from participant-completed questionnaires. Outcomes measured clinical marker data such as HbA_1c levels, BMI and risk of cardiovascular disease. Five studies156,160–163 also used quality-adjusted life-years (QALYs); two162,163 calculated QALYs using participant-reported EQ-5D.

Six studies164–169 that detailed costs or costs and outcomes of interventions for people with learning disabilities met the inclusion criteria. Resource-use data were collected in all studies: four of these164,165,168,169 used the Client Service Repository Inventory (CSRI), which is a research instrument that provides retrospective information on service use, and three studies165,168,169 used a proxy informant who was familiar with the person with a learning disability. Resource use (for which GP visits, nurse visits, hospital inpatient and outpatient visits were the most frequently utilised, with chiropodist, psychiatrist and dentist also having high utilisation rates), staff (salaries and training) and accommodation were the main drivers of cost in relation to the cost-of-illness studies;165,167 medications, transportation and carer time were also included. Health outcomes were measured in four of the studies using different methods: QALYs derived from EQ-5D,164 social interaction activity patterns166,168 and an aggression scale plus a Quality Of Life Questionnaire. 169

Overall, the diabetes studies showed a variety of data collection methods, with staff time for intervention delivery as the greatest driver of costs. In the studies including people with learning disabilities, hospital-based care, GP, nurse care and accommodation were the main drivers of costs.

Based on the information retrieved from the literature review and with input from experts on the study team, resource-use questionnaires were developed for completion by GP surgeries, using a modified version of the CSRI (for form 10 see Appendix 4, and for form 2 see Appendix 5). The forms were field-tested with people with a learning disability working with our third-sector collaborators, to ensure that they included all of the main items of relevance. In self-management interventions, patients’ costs are likely to be more important than when considered in more conventional interventions; therefore, the forms identified patient resource use from the perspective of the health and social care sector and from a wider societal perspective. The service usage form was field tested and assessed for content, clarity and acceptability (assessed by proportion of questionnaires returned and missing items).

The measure of outcome within the studies varied. However, the EQ-5D and the Health Utilities Index have been shown to be superior compared with other preference-based measures of health for this population. 170 None of the studies identified in review had used the Health Utilities Index and, thus, based on the frequent use of EQ-5D in the included diabetes literature, the EQ-5D was chosen to be assessed for feasibility as an outcome measure for any subsequent RCT.

Health economics analysis

Completeness of the data collected in the questionnaires was recorded using descriptive statistics, detailing the number and percentage of questionnaires returned and the number and percentage of missing items within the returned questionnaires.

Unit costs for health service resources were obtained from national sources such as the Personal Social Services Reference Unit (PSSRU),171 the British National Formulary (BNF)172 and the NHS reference cost database. 173 If national unit costs were unavailable, the finance departments of trusts participating in the study were asked to provide local cost data. The mean of these costs was used as the unit cost estimate in the analysis. This included the costs of health and social care (service provision and use of other health and social care services) and took into account the productivity costs (time away from work) and out-of-pocket expenditures incurred by the patients (e.g. travel expenses, over-the-counter medicines and supplements and additional costs/savings of any dietary changes). Burden on the supporter was also considered (e.g. productivity costs and out-of-pocket expenses).

Nurse visits

Initially, to increase efficiency, we aimed to determine whether or not we could obtain certain outcomes (HbA_1c levels, BMI) from GP records of diabetes clinical QOF reviews. However, as a result of our experience during case finding, it became clear that such an approach was not feasible because we could not obtain information in a timely and comprehensive way, and we therefore obtained research outcomes ourselves unless participants refused to have measures taken by the research nurses but agreed to us reviewing their medical notes.

Participants were, therefore, asked to undergo a blood test with the research nurse to establish physical measures at baseline and at follow-up. All participants were offered anaesthetic cream for the blood test. Only if a participant declined the physical measures or a blood test did the nurse then attempt to obtain the results from GP records. Whenever possible, the nurse undertaking the follow-up visit was blind to treatment allocation and was one of two nurses who had not seen the participant at baseline, been told their participant’s allocation or delivered the SSM.

Researcher baseline and follow-up questionnaire

For participant- and supporter-completed data, we collected information on supporter details, exercise, service use, medications, employment, accommodation, food consumed, EQ-5D and mood. For the supporter, we collected information on supporter details, service use, employment and EQ-5D. Each question of the EQ-5D was also printed on a separate A4 laminated sheet as an interview aid. We made two changes to the measures we used on the basis of the case-finding study. First, we dropped the REAP because our participants were not able to provide the information necessary to complete it. Second, we decided to use a standardised measure of mood [Patient Health Questionnaire-2 (PHQ-2)],177 as our participants clearly recognised and understood the questions we asked about distress and indicated that it was an important topic to them.

We also recorded participant-reported resource use questions for the questionnaire, based on findings from the literature review and expert input from the study team. The breadth of these questionnaires was wider than the GP-completed questionnaires (reported in Review of general practitioner medical notes) and took a societal perspective.

The interviewers recorded their perceptions of how easy they thought the questions were for the participant to answer.

The feasibility of collecting information on out-of-pocket expenses of supporters relating to travel for participants’ health-care appointments was assessed. Given the expected close involvement of supporters in the self-management process and potential impact on their health-related quality of life, the feasibility of collecting EQ-5D data for supporters or informal carers was also assessed (paid or formal carers/supporters were excluded).

Review of general practitioner medical notes

A brief health and social care questionnaire was again sent to general practices to complete using participants’ medical records, giving details of participants‘ service and hospital usage for the period 4 months post randomisation. The content remained consistent with that requested in the case-finding study with the addition of two new questions about service use on the advice of experts in the team, a diabetes educational course [Diabetes Education and Self Management for Ongoing and Newly Diagnosed (DESMOND) programme] (www.desmond-project.org.uk/) and a chronic illness course (X-PERT; X-PERT Health, Hebden Bridge, UK). In fact, although both are referral options in the regions where participants were located,67,178 no participant had attended either course.

Most recent test results for QRISK^®2 score, serum creatinine level, microalbuminuria concentration and diabetes medications were also requested, as were the dates of the participant’s most recent retinal screen and foot examination.

Other costs

We collected information on the resources associated with development and delivery of the intervention. These were based on data such as administrative and researcher records, as well as a detailed description of the development process including number of sessions, staff involved, nurse training and travel, and cost of materials.

Qualitative data collection

Three types of data were collected: participant and support interview data, researcher journals and nurse journals.

Recruitment and retention outcomes

1. The number and proportion of initial referrals found eligible for the case-finding study, screened for eligibility for the RCT, found to be eligible, consented and randomised to the RCT.

2. The demographics of (1) participants referred, registered and found eligible at case finding and (2) participants found eligible, consented to and randomised for the RCT.

3. The proportion of randomised participants who have a supporter.

4. The reasons for non-participation (participant and supporter).

5. The method of identification for randomised participants and supporters.

6. The proportion of randomised RCT participants with all the required baseline and follow-up assessments completed, number of withdrawals from follow-up data collection, reasons for withdrawal and number of losses to follow-up.

7. The length of time taken to conduct trial procedures including time from first contact to randomisation, follow-up completion and receipt of medical records.

Intervention delivery outcomes

1. The proportion of participants randomised to SSM who attended at least one session of SSM with the research nurse and who completed all required sessions, the number of dropouts from SSM and the reasons for dropouts.

2. The agreed method of measuring participant, supporter and research nurse adherence to SSM, including uptake and adherence rates and assessment of the feasibility of using a standardised measure: the measure we developed took the form of a checklist used to record which elements of the self-management intervention had been completed, as identified by an independent researcher from the nurses’ case records and from the follow-up research visit (see Appendix 13).

3. The detailed description of what treatment was delivered and received in each arm, including a comparison of TAU across arms from medical records and an assessment of the feasibility of collecting data on TAU pathways.

4. The assessment of the feasibility of collecting data on adherence to SSM and the standard leaflet.

5. The preliminary assessment of the acceptability of the intervention, including negative outcomes, hospital attendances and relevant unexpected severe adverse events.

Feasibility outcomes

1. The assessment of the feasibility of blinding researchers to treatment allocation.

2. The proportion of participants who refuse physical measures with available and timely QOF data at baseline and (if necessary) at their follow-up assessments.

3. The completion rates for other data collected, including assessment of the feasibility of collecting health economics data (e.g. participant EQ-5D, NHS and supporter costs or medication use).

4. The missing item-level data on self-reported questionnaires.

Statistical outcomes

• The distribution and variability of candidate primary and secondary outcomes at 4–6 months post randomisation (e.g. HbA_1c levels, BP, BMI, waist-to-hip ratio, EQ-5D, vascular/microvascular risk markers, marker of microvascular disease or participant mood), and assessment for the potential for outcomes to show an effect in a definitive trial, that is, their sensitivity to change.

• The proportion of participants classed as abnormal on standard criteria for medical markers.

Qualitative outcomes

1. The positive and negative experiences of implementing self-management.

2. The perceptions of the standard leaflet and experience of being in the TAU arm.

Sample size

To address the feasibility objectives, we planned to recruit 80 participants in total, randomised equally between intervention and control arms, in order to obtain follow-up data on at least 30 participants per arm, as recommended by Lancaster et al. 179 This assumed loss to follow-up would be no greater than 25% at 6 months. As this study was designed to assess the feasibility of conducting a definitive trial, a formal power calculation was not appropriate because effectiveness was not being evaluated. Part of the rationale was to gather data to inform the design of a definitive trial, so estimates of non-adherence and loss to follow-up rates in this patient group were intended to inform power calculations for the definitive trial.

Populations

Participants considered eligible for the feasibility RCT consisted of all randomised consenting participants fulfilling all of the inclusion criteria and none of the exclusion criteria. The intention-to-treat population consisted of all randomised participants, regardless of their adherence to the study protocol or eligibility, and participants were grouped according to the treatment to which they were randomised. The screening population consisted of all participants in the case-finding eligible population who agreed to be contacted in the future.

Analyses

All data analyses and summaries were performed using SAS version 9.4. As this was a feasibility study, data were summarised using descriptive summary statistics and estimation (with 95% CIs) only, with no formal statistical testing carried out on any of the outcomes.

The analysis summarised outcomes, presenting descriptive statistics, according to the intention-to-treat population unless otherwise stated. Percentages were calculated using the total number of participants with available data as the denominator, and the number of participants excluded is presented alongside each summary.

The feasibility, and success, of the recruitment strategy was evaluated by summarising the screening, eligibility, consent and randomisation processes, and included the numbers of participants involved at each stage and reasons for non-participation. A recruitment flow diagram depicted the course of participants throughout the recruitment process from case-finding referral through to randomisation in the feasibility RCT, and a recruitment graph presents monthly and cumulative recruitment (see Figures 2 and 3).

In addition to the outcomes listed, the following were also summarised overall and by trial arm: (1) safety, including related unexpected serious adverse events, unplanned hospital admissions, deaths and negative life events; (2) responses to the researcher structured follow-up interview; and (3) additional participant information and living arrangements.

Statistical outcomes

To inform the sample size calculation for the definitive trial, we report summary statistics (including variability), and corresponding 95% CIs overall and for each treatment arm for candidate primary and secondary outcomes at baseline and follow-up (e.g. HbA_1c levels, BP, BMI, waist-to-hip ratio, EQ-5D, vascular/microvascular risk markers, marker of microvascular disease or participant mood). Histograms and box plots present the distribution of outcomes. Outcomes reported outside the windows around randomisation and follow-up were included in the analysis. Measures of diabetes control were also presented categorically according to abnormal ranges on standard criteria, as in the case-finding study, with additional microvascular markers (kidney function) categorised as follows:

• Microalbuminuria level: for men ≤ 3.5 mg and > 3.5 mg and for women ≤ 2.5 mg and > 2.5 mg, indicating early signs or presence of kidney disease.

• Estimated glomerular filtration rate (eGFR): ≥ 90 ml/minute/1.73 m², normal function; 60–89 ml/minute/1.73 m², stage 2, mildly reduced function; 30–59 ml/minute/1.73 m², stage 3 moderately reduced function (an important cut-off point for clinical care); and < 30 ml/minute/1.73 m², stages 4 and 5 (would not expect this value for people living in the community).

The risk of metabolic complications according to waist circumference and waist-to-hip ratio was categorised as follows:

1. Waist circumference: not at increased risk, men ≤ 94 cm and women ≤ 80 cm; increased risk, men 94–102 cm and women 80–88 cm; and substantially increased risk, men > 102 cm and women > 88 cm. 180

2. Waist-to-hip ratio: not at increased risk, men < 0.90 and women < 0.85; and substantially increased risk, men ≥ 0.90 and women ≥ 0.85.

Sensitivity of change for candidate primary outcomes was assessed based on within-person change in HbA_1c levels and BMI between baseline and follow-up. Summary statistics, effect sizes and histograms depicting the distribution change are presented separately by trial arm.

Sensitivity analyses

Non-parametric bootstrapping was used to determine the level of sampling uncertainty around the ICER. The bootstrap approach is a non-parametric method that considers the original sample as though it was the population and draws multiple random samples from the original sample by generating 10,000 estimates of incremental costs and benefits from the trial results. The results are presented using cost-effectiveness scatterplots to illustrate the uncertainty surrounding the cost-effectiveness estimates. On the cost-effectiveness plane (which plots incremental QALYs against incremental costs), a result is considered cost-effective if it falls on or below the given cost-effectiveness threshold. The CEAC is derived by calculating the proportion of bootstrapped estimates that are cost-effective across a range of willingness-to-pay thresholds, to show the probability that SSM plus TAU is cost-effective across different threshold values. 183,184

In line with NICE-recommended practice, any cost-effectiveness analysis will require QALYs and utility weights for each health state observed in a trial population. A 2001 HTA report,170 which considered general health status measures for cognitively impaired populations, found the EQ-5D to be superior compared with other preference-based measures of health. We therefore explored the feasibility of use of the EQ-5D for participants and for supporters or informal carers (not including ‘paid’ supporters) and of obtaining details of resources used from GPs.

Identification

To evaluate our case-finding methods, the role of referrer and method of contact are presented in Table 8 for all referrals and the referred and eligible populations. The most frequent method of identification was through searching a GP learning disability or diabetes register, leading to 119 (33%) referrals overall, whereas searching the e-health record system using Read Codes led to 78 (21%) referrals. The largest proportion of referrals was made by primary care staff, who provided 202 (55%) referrals. The first 50 practices recruited were asked to complete a log that detailed each stage of identification (see Appendix 1) and to record how many potentially eligible people were found by this method and how many were later found to be ineligible. Only 20 logs were returned, and many were incomplete or had duplicated data in multiple fields. It was agreed that it was not possible to include these data in the analysis because of the high numbers of missing data.

Once identified, the referrer contacted participants to request consent for researcher contact, with 111 out of 365 (30%) of all referrals and 60 out of 147 (41%) eligible participants being contacted in person, which suggests a slightly increased consent rate when patients were approached in person.

Table 9 provides further information on the method of identification.

Table 10 presents the QOF registers (excluding diabetes) that eligible participants were found to be on from their GP medical record check. Of the 103 participants with a GP medical record check and complete data relating to QOF registration, 101 (98%) were on at least one non-diabetes QOF register. Of these 101 participants, the most frequent registers were (1) the learning disability register for 91 (90%) participants (despite most of these not being originally referred following a search of that register), (2) the cardiovascular illnesses register for 37 (37%) participants, (3) the obesity register for 33 (33%) participants and (4) the mental health register for 17 (17%) participants.

Recruitment and resources

The number of contact attempts made by the researcher in order to arrange a face-to-face researcher visit for the 236 participants who agreed to researcher contact are presented in Table 11.

A total of 815 contact attempts were reported overall, with a mean of 3.5 contacts and a median of two contacts per participant and ranging from a single contact to 27 contacts. The most frequent method used by the researchers to contact participants was the telephone, accounting for 713 contact attempts. The median time from first to last contact was 3 days (range 1–437 days) in participants agreeing to the researcher visit and 16 days (range 1–349 days) in participants who did not agree to a researcher visit. Once participants had agreed to a researcher visit, the majority (183/186, 98%) required only a single visit to introduce the study, obtain consent and conduct the researcher interview. Three participants required a second visit to complete this interview. The average duration of each researcher visit was 1 hour; however, visits ranged from 15 minutes to just over 3 hours.

Table 12 summarises the involvement of a supporter throughout the recruitment process: referrer contact; provision of contact details to the researcher by the referrer; researcher contact; presence during researcher visits; and informed consent to the study. Of all 325 referred participants, a supporter was contacted by the referrer to obtain consent for researcher contact for 124 (38%) participants. This was either as the sole contact or alongside contact with the participant. The family members of a further 44 (14%) participants were contacted. For the majority of the 236 participants who agreed to researcher contact, the participant’s details, as opposed to the supporter’s, were provided to the researcher by the referrer (n = 143, 61%). Of the 186 participants receiving a researcher visit, a supporter was present in at least one visit for 128 (69%) participants, and of the 172 consenting registered participants, 99 (58%) also had a consenting supporter.

Establishing eligibility

Uncertainty during eligibility assessments was a common cause of extra work in recruitment. Uncertainty as a result of ambiguity or missing information in the referrals booklet was common, in 46 (13%) of all 365 referrals, with uncertainty on each criterion as follows: if the participant (1) had a mild or moderate learning disability in 22 (6%) referrals; (2) was taking insulin in 18 (5%) referrals; (3) had type 2 diabetes in six (2%) referrals; and (4) was aged ≥ 18 years in two (0.5%) referrals.

Furthermore, the researcher was uncertain whether or not the participant had capacity to take part in the study at the researcher visit for 3 out of 186 (2%) participants. At registration (172 participants), there was uncertainty about whether or not eight participants (5%) had type 2 diabetes and whether or not 11 participants (6%) required insulin.

Table 13 outlines communication problems for referred participants identified by the referrer prior to passing on contact details to the researcher. Of the referred population, 36 (11%) were reported to have communication problems, of which 13 (36%) were as a result of deafness and 12 (33%) were because of a speech difficulty. This information was missing for 70 (22%) of the referred participants when the participant had not agreed to researcher contact.

Of the 236 participants agreeing to contact by the researcher, 106 (45%) stated a preference to be contacted via their home telephone, 50 (21%) via their mobile phone, three (1%) via e-mail and 30 (13%) did not mind. Seventy-nine (54%) eligible participants reported having a mobile phone, with only 16 out of 79 (20%) having a smartphone. Only 33 (22%) eligible participants said that they used the internet, 18 of whom said they accessed the internet with a helper.

Demographics

Participant demographics across the referred and eligible populations were similar (Table 14). The mean age of eligible participants was 54.4 years [standard deviation (SD) 12.8 years], 50% were male and 85% were white.

Participant-reported details during researcher interview

During the researcher interviews, participants were asked a series of questions relating to their living and supporter arrangements, medication, diabetes self-care activities, shopping, cooking, eating habits and diet, use of services for diabetes, comorbidities, satisfaction with lifestyle, desire for change and help looking after their diabetes, difficulties carrying out self-care activities and about their well-being, mood and feelings about having diabetes.

Out of the 147 eligible participants, 103 (71%) reported that they did not live on their own. The most frequent relationship of persons living with the participant were others in a shared house for 41 (43%) participants and immediate family for 37 (39%) participants (Table 15).

The majority of participants (n = 130, 88%) reported having a supporter (Table 16), someone in their life who helps them with their diabetes and/or shopping and cooking, with 108 participants (74%) having someone who helped with both diabetes and shopping and cooking. Of the 130 participants with a supporter, 71 (56%) lived with this person, most commonly with paid staff on site.

We wanted to know about participants’ understanding of their diabetes and its implications and we found that most participants had ‘found out’ about their diagnosis more than 2 years ago (n = 111, 79%). Twenty-five participants (18%) found out between 6 months and 2 years ago, and only five (3%) had been diagnosed in the previous 6 months.

Table 17 provides participants self-report of medication versus reports from medical records.

Self-reported diabetes medication use is presented in Table 18. Most participants were helped to remember to take their medication in some way and did not collect their own prescriptions. Of the 118 participants taking tablets for their diabetes, 35 (30%) reported that they missed their medications at ‘less than once a week’ or more frequently.

When questioned about self-care for diabetes, only a minority were able to report awareness of specific activities (Table 19). For example, only 35 (24%) of eligible participants said that they had been told in the past that they needed to check their blood sugar, but as self-monitoring of blood sugar is no longer recommended for type 2 diabetes, lack of monitoring is not inherently problematic. However, all type 2 diabetes patients will be advised to check their feet, but only 51 out of 145 (35%) or participants said that they knew to check their feet and only 74 out of 154 (51%) knew to have their teeth and gums checked.

Tables 20 and 21 provide details of participants’ shopping, cooking, and eating habits. Most participants ate food cooked at home rather than out or ‘takeaways’ and 95% said they had access to a kitchen. Most felt well enough to shop for food or cook. One-third of participants said that they read nutritional information on food when shopping but only just over half of these people said they understood the information (16% of the total sample).

Most participants were willing to acknowledge problems with their diet. Table 21 shows that over one-third of participants self-reported that they ‘usually or often’ ate too many fats (butter/oil) or high-fat snacks and too little wholegrain and fruit. Ten participants were unable to respond to the item about ‘wholegrain’ consumption, even when given examples, such as brown bread or high-fibre cereals. This suggests a lack of understanding about the existence of wholegrain foods and/or their contribution to a healthy diet. Excessive alcohol consumption was not commonly reported in this sample.

In relation to the use of services for diabetes (Table 22), most participants reported that they saw primary care staff rather than secondary care staff. The majority of participants reported no difficulty attending appointments for their diabetes and 102 (70%) participants said that they had someone who accompanied them to appointments. A health passport scheme was active in each of the areas but rarely used.

In addition to diabetes, over three-quarters of participants reported conditions that made them ‘poorly’ and for which most were taking medication (Table 23). The most frequent self-reported comorbidities were mental health problems and depression (56/114, 49%) and cardiovascular illness (46/114, 40%).

Participants were further asked about their ‘lifestyle’ – diet, exercise, weight and mobility – how this ‘makes them feel’ and whether or not they ‘wanted help’ in these aspects of their life (Table 24). Most participants said that they did some exercise. Of the 126 participants reporting that they exercised, 38 (31%) said that they exercised every day, 64 (53%) said that they exercised every week and 18 (15%) said that they exercised some weeks. The most common activity was walking (n = 117, 93%), although some people attended exercise classes (n = 13, 10%) or went to the gym (n = 10, 8%). Twelve (10%) said that they went swimming, 8 (6%) said that they cycled and 31 (25%) said that they did other forms of activity. Around 40% of participants said that they had mobility problems. About one-third of respondents said that they did not do enough exercise. Over half of participants said that they were unhappy with their weight and 123 out of 145 participants (85%) said that they would like help looking after their diabetes.

Table 25 presents participants’ feelings about the perceived difficulty of carrying out self-care activities. Those activities most frequently perceived as being ‘hard’ were doing exercise and healthy eating. Taking medication and attending medical appointments were considered ‘easy’ more frequently, often because the participant had help doing these tasks.

Participants were asked about their well-being, their mood and how they felt about having diabetes. It can be seen from Table 26 that 46 out of 147 (31%) participants reported that they felt poorly either sometimes or most of the time, 51 out of 141 (36%) were either unhappy or worried about their diabetes, 83 out of 146 (57%) felt miserable and sad at least some of the time and 16 out of 146 (11%) reported not enjoying the things they used to.

Finally, participants were asked how they found the questions asked during the researcher visit. Of 145 participants who responded, over half of participants, 79 (54%), said that they had found the questions easy, 44 (30%) found them OK and 18 (12%) found them hard.

Physical measures obtained from general practitioner records

It was difficult to obtain medical information from GPs to allow decisions about eligibility. We followed each request with two reminders when we had no reply and further prompts if the practice had agreed to provide information but did not do so.

As we wanted to know if the available results were recent enough for us to use them to check against inclusion criteria in a subsequent trial, we also checked the timing of the last result obtained from the GP in relation to registration. As shown in Table 27, a substantial proportion of available results were dated many weeks or months before the date of registration, insufficiently recent for the purposes of a RCT.

Complete non-response to requests for medical information on eligible participants was also high, for 45 (31%) participants. Clinical characteristics of the eligible participants, as provided by the GPs of participants from their medical records, are shown in Figures 4 and 5 and Table 28. Tables 29 and 30 show other measures of health state and diabetes control.

FIGURE 4.

Distribution of HbA_1c level (mmol/mol).

FIGURE 5.

Distribution of BMI.

Participants had a mean HbA_1c level of 7.1% (54.5 mmol/mol) with a median level of 6.8% (51 mmol/mol). Approximately one-third, 36 (35%), of participants had a HbA_1C level of < 6.5% (48 mmol/mol) and 28 (28%) had HbA_1C values that were ≥ 7.5% (58 mmol/mol). BMI ranged from 16 to 54 kg/m², with a mean of 32 kg/m² and a median of 31.5 kg/m². Almost two-thirds of participants, 66 out of 102 (65%), were obese (BMI of > 30 kg/m²) and 21% (21 participants) had a BMI of ≥ 40 kg/m². There was a positive correlation (Pearson correlation coefficient 0.265; p = 0.0074) between HbA_1c level and BMI based on 101 participants with both a BMI and a HbA_1c result.

Blood pressure was generally within normal limits. However, 14 out of 101 (14%) participants had a systolic BP of ≥ 140 mmHg, and 34 out of 101 (34%) participants had a diastolic BP of ≥ 80 mmHg.

Total cholesterol level ranged from 2.0 to 7.0 mmol/l, with a mean of 4.2 mmol/l. In 57 out of 100 (57%) participants, their total cholesterol level was ≥ 4 mmol/l. Triglyceride concentration ranged from 0.4 to 7.0 mmol/l, with a mean concentration of 2.2 mmol/l and a median of 1.9 mmol/l. In just 3 out of 67 (4%) participants, their triglyceride concentration was ≥ 4 mmol/l.

Subgroup comparisons for diabetes control

We conducted simple post hoc exploratory subgroup summaries to generate hypotheses for relationships between HbA_1c level and BMI and other variables we expected may be associated with adequacy of diabetes management in our population (Table 31). HbA_1c levels were slightly lower in those who named a supporter who was involved with their diabetes management, but BMI was more similar in the two groups.

By contrast, there was a suggestion of an association between BMI, but not HbA_1c level, and participants’ mood and self-reported feelings about a number of aspects of their health and lifestyle, with higher BMIs in participants reporting feeling miserable or sad generally, feeling unhappy or worried about diabetes, feeling unhappy about or wanting to change what I eat and feeling unhappy about or wanting to change weight.

Nearly all participants said that they wanted help looking after their diabetes, with those saying ‘no’ having a slightly lower BMI than those saying ‘yes’. Both HbA_1c level and BMI were higher in those on polytherapy than in those on monotherapy.

For comparison, the local Commissioning Support Unit provided us with aggregate data for the three cities in our study, based on returns in the QOF scheme (Table 32).

These data indicated several points of interest. First, only 0.6% of the total population was on both the learning disability register and the diabetes register. Second, glycaemic control was similar for those who were and were not on the learning disability register. Third, those on insulin and who were on the learning disability register tended to have better glycaemic control than those on insulin who were not on the learning disability register. Finally, those on the learning disability register tended to be rather heavier than those not (one BMI point is equivalent to 2–3 kg in this population).

Resource use

General practitioner and practice nurse visits were the most frequent resources used. Almost 93% of the 121 participants visited the GP at least once in a 12-month period and 90% of participants visited the practice nurse. One in three (35%) participants had been to see a podiatrist and over one-quarter (27%) had been to see an ophthalmologist. Over 21% of participants attended the accident and emergency department (A&E) in the preceding 12-month period and 12% had at least one inpatient stay. Other services were much less frequently used (Table 33).

Sixty-one of the questionnaires included medication information, and the details are given in Table 34.

Case finding and recruitment

The case-finding study resulted in 365 referrals in 325 patients, with some overlap in referrals from different sources, and identified 147 (95% CI 39.8% to 50.6%) eligible and consenting participants, of whom 132 agreed to be contacted in the future about further research in the study.

Following recruitment, medical record information was requested from GPs to confirm eligibility and attain participants’ most recent routine test results and details of QOF registers, service usage and diabetes medications. This proved to be a major barrier to the study because of difficulties in obtaining this information from GPs. To ensure registered participants were not unnecessarily excluded from the eligible population and allow these participants to proceed through to consideration for the RCT, participants who were referred by their GP, nurse or other medical professional (41 eligible participants, 28%) were confirmed as eligible based on information from the researcher visit.

Participants were largely identified through learning disability or diabetes registers and by primary care staff, but third-sector organisations and community learning disability teams also identified patients. Subsequent researcher contact to arrange face-to face visits to obtain consent was resource intensive with numerous contact attempts, in some cases over a prolonged period of time, with a mean number of 3.5 contacts taking place over a range of 1–437 days. Once they were contacted and had agreed to a face-to-face visit, the majority of participants required only a single visit, with a mean duration of 1 hour and ranging from 15 minutes to 3 hours.

Despite the effort put into recruiting from sources other than GPs, and despite the fact that many of our referrals did not come as a consequence of searching GP learning disability registers, the overwhelming majority of eligible participants were included on a learning disability register. In other words, we did not recruit substantial numbers of participants from the hidden majority of unregistered people with a milder learning disability. It is unclear why, but we suspect the answer is that because this group do not have a diagnosis and do not self-identify as having a learning disability; they are simply unrecognised in any formal process of recruitment, even if their difficulties may be recognised informally when direct contact occurs for whatever reason.

Characterising the sample

The most important finding in relation to living arrangements is that for over half of participants who said they had somebody who supported them, the supporter was a paid member of staff. This matters because, in reality, such support is rarely stable, with a different staff member accompanying participants at health-care and research visits, depending on availability.

Although most participants understood that they had diabetes and knew that they took medication for it, the understanding of dietary and lifestyle management was patchy, with few participants able to interpret nutritional information or knowing what health checks they needed to have. Many respondents had other physical health problems and mental health problems; the latter were especially common. Many respondents said that they felt unhappy, both in general and specifically, because of their diabetes. They reported dissatisfaction with weight and diet and with physical inactivity that they found hard to change. Many agreed that they wanted more help with diabetes and agreed to further research contact.

It proved a challenge to collect routine health data from GP records and, from the absence of recent results in many cases, it is clear that routine data could not be relied upon for outcome measurement in the subsequent feasibility RCT. Data that were available showed reasonable glycaemic control in the sample, and were in line with data for the overall population in the same CCGs provided for us by the Commissioning Support Unit. By comparison, we found high levels of obesity.

The analysis of the results from the GP-collected resource use highlighted the challenge of data collection in primary care. Complete data were obtained for only 33% of eligible participants (48/147). Although 70% of the forms were returned, and the completed forms had few missing items (< 5% missing), there was a high level of service use recorded as unknown. Unsurprisingly, the services provided directly by the GP surgeries had fewer missing or unknown data than those provided by other providers (< 2% and > 12%, respectively). Completion of the resource-use questionnaires was time-consuming and problematic in respect to the intensity of follow-up reminders required to achieve the return rate (including researchers helping with completion in the GP surgeries). Potential reasons for the time-consuming nature of the forms and the missing data were the 12-month time period over which resources were requested and the inability or unwillingness of general practices to complete the time-consuming task.

Implications

What are the implications of these findings for feasibility of a definitive trial? We have shown that it is possible to identify and consent eligible participants from the target population into research. They have important unmet health needs related to type 2 diabetes, especially related to being overweight and inactivity. They have needs for improved self-management in the way of poor knowledge about self-care needs, and many of them express dissatisfaction with their current health and voice a desire to change. We identified two particular challenges. First, despite our best efforts, we identified very few people who were not already known to their GP and other services and were not already on a learning disability register. The hidden majority of people with milder disability remained hidden to us, and yet they may be the population that could benefit most from support in self-management. Second, we were unable to use routinely available primary care data either to confirm eligibility for a RCT or to obtain the comprehensive outcomes necessary for a trial.

Supported self-management

Forty-one (50%) participants were allocated to receive SSM during randomisation. The nurse who was allocated to deliver the intervention saw the participant at baseline, and if no nurse visit had been conducted then the nurse was allocated at random. All participants were seen or contacted to schedule sessions by their allocated nurse.

Thirty-five (85%) participants attended all required sessions, which ranged from two to four sessions, with over three-quarters of all participants (78%) attending at least three sessions (Table 43). Four (10%) of the 41 participants allocated to receive the intervention did not attend any sessions and two (5%) attended only one session and so did not attend all required sessions. Thirty (73%) participants had another person present with them during at least one of their sessions, usually the consenting supporter.

The standard leaflet was not provided to the four participants who attended no sessions, and it was also not provided to a further two participants who attended only their first SSM session.

For the 37 participants attending at least one session, the first session took place a mean of 28 days post randomisation (range 13–40 days), and the intervention delivery period averaged 49 days from the first to last session (range 0–104 days). Table 44 presents the timing of researcher follow-up visit in relation to randomisation and the number of SSM sessions received; no participant was still receiving SSM by the time of the researcher follow-up visit (Figure 10).

FIGURE 10.

Needle plot of SSM session attendance and timing of research follow-up. Timing of researcher follow-up (points) in relation to duration and end of self-management sessions (bars).

A summary of engagement was reported by the nurse who delivered the intervention (Table 45): 23 out of 40 (58%) participants were deemed to be very engaged with the sessions and 12 out of 40 (30%) were deemed to be very engaged with the materials, 15 out of 41 (37%) were reported to have a very engaged supporter (consenting or non-consenting) and 18 out of 41 (44%) had a further or different person (other supporter, partner or family member) who was engaged in the intervention implementation.

A total of 106 sessions were conducted across the 37 participants attending at least one session, with a typical total intervention time of 2 hours (Table 46). Sessions lasted a mean of 45 minutes (range 13–95 minutes), they largely took place in the participant’s home (92%) and the focus included getting started, setting goals, mapping supporter and checking progress.

Finally, Table 47 presents a summary of telephone contact between sessions, excluding contacts for administrative or scheduling reasons.

Adherence and fidelity

Independent review of adherence to, and fidelity of, the intervention took place for all self-management sessions attended by the 37 attending participants.

We assessed which components of the intervention were covered in the first, second, third and fourth sessions, as well as any session overall.

All components of the intervention concerned with assessing day-to-day living arrangements and diabetes management were covered in all participants’ first session, and at least some of these components were revisited during the second session for 10 out of 35 (29%) participants attending at least two sessions.

Goal-setting plans and ‘supporter flashcard(s)’ were covered or provided during first session for just over one-third of participants, were covered for the majority during the second session and in the third session covered for just two participants each.

The most frequent goals identified were to increase physical activity (n = 17, 46%), to eat more fruit (n = 15, 41%), to make snack swaps for food and drink (food, n = 14, 38%; drink, n = 6, 16%) and ‘other dietary improvements’ (e.g. reducing sugar and calorie intake in tea, squash and hot chocolate: n = 10, 27%) (Table 48).

In self-monitoring, 16 (43%) returned no calendar sheets, 5 (14%) returned only unmarked calendar sheets, and 16 (43%) returned sheets in which all or some were marked.

Box 2 lists the SSM materials collected to derive an adherence score for each participant. Table 49 presents the details of adherence scoring for each component contributing to the score, with a possible range of scores from zero to eight depending on the number of completed steps or components (see Appendix 13 for details of the intervention materials used in the scoring). Fidelity scores ranged from three to eight, with the majority of participants scoring six or above (n = 32, 87%). There was evidence that the whole process, all steps, were completed at least once for all but two participants.

Unblinding

The follow-up nurse was unblinded for a total of 34 (42%) trial participants: 20 (48%) in the SSM arm and 14 (34%) in the TAU arm (Table 50). Most unblinding occurred because the nurse who was allocated at randomisation (who told the participant their treatment allocation and went on to deliver the intervention for allocated participants) also undertook the follow-up visit because of staffing availability. This was the reason for unblinding for 12 out of 20 (60%) participants in the SSM arm and 12 out of 14 (86%) participants in the TAU arm. Other reasons for unblinding included seeing the intervention materials in the SSM arm and also being informed by the participant, supporter or researcher. All unblinding in the TAU arm occurred prior to the follow-up visit, whereas in the SSM arm unblinding occurred during the assessment for six (30%) participants and after the assessment for two (10%) of the 20 unblinded participants.

Researchers were unblinded for a total of 16 (20%) participants (Table 51), for which unblinding occurred before the follow-up assessment. Almost all unblinding occurred in the SSM arm with the researcher unblinded for 15 (31%) participants, and only one participant in the TAU arm. For two participants, one in each arm, an alternative blinded researcher was instead able to conduct the follow-up visit. Most researcher unblinding occurred as the researcher was informed by the participant, for 9 (56%) of the 16 participants.

Physical measures

Table 52 summarises the success of attaining the physical measures, height, weight, BMI, waist and hip measurements and BP, and conducting a blood test during the baseline and follow-up nurse visits. For the majority it was possible to collect all measures where a visit went ahead. Most problematic was the blood test, which was not obtained for 5 (6%) of 76 participants at baseline and 4 (5%) of 75 participants at follow-up. However, overall, there was a low participant refusal rate for the blood test (just three refused) and the main problem was poor veins in both arm and hand.

Table 53 presents the number of participants for whom results from GP medical records were required, either because the results could not be obtained during the nurse visit or because there was no nurse visit. GP medical records were required for 14 participants at baseline and 13 at follow-up (the latter had, however, withdrawn from further clinical data collection by follow-up): they were obtained for 11 out of 14 (79%) and 5 out of 13 (39%) participants, respectively. A number of records received did not, however, fall within the ideal 6-week time frame around randomisation or the follow-up researcher visit.

General practices were also contacted to obtain follow-up medical record information for all participants, including a number of physical measures (QRISK^®2 score, serum creatinine level, microalbuminuria concentration, dates of last retinal screening and foot examination), information about patient’s diabetes medications and health-care resource use (Table 54).

The timing of assessments obtained via the nurse only (height, hip and waist measurements and weight) can be seen in Figure 3. Figure 11 presents the timing and source of assessments of the outcome measures: BMI, BP and HbA_1c levels.

FIGURE 11.

Time between randomisation and outcomes. (a) BMI; (b) BP; (c) HbA_1c level; and (d) most recent blood test.