Study design overview
The study, which built on the findings of the realist review described in Chapter 2,44 was conducted in three phases and embedded in a continuous process of participatory and dissemination activities; it is summarised in . First, we undertook a scoping exercise, combining questionnaire data with NHS audit data, to generate a descriptive understanding of existing group-based care and the current state of diabetes care for young adults. Second, we carried out in-depth participatory co-design activities to develop a model of group clinic-based care that was then implemented in two NHS diabetes services (‘group clinic sites’). Third, we evaluated this group clinic model using mixed, qualitative and quantitative methods, and a cost analysis, with a further two NHS diabetes services used as ‘control sites’. Our evaluation assessed both the impact of the group clinics on young adults and the potential impact of group clinics within the NHS. Our quantitative evaluation of group clinics was not designed to provide a definitive estimate of differences in outcomes; rather, it was intended to contextualise the qualitative evaluation and, in combination with the cost analysis, to guide the feasibility of a future at-scale evaluation.
Our study design and methods have been published in a protocol paper.1
Changes to the original protocol
There were three changes from our original (published) protocol, highlighted below.
Revised aims of National Diabetes Audit analyses
We had intended to perform unit-level (i.e. per clinical service) comparisons of the care delivered to all patients at our group clinic and control sites, using the NHS Digital National Diabetes Audit to provide a descriptive background and a comparison with national data. However, after data access had been granted by NHS Digital, it became apparent that complete unit-level data from our research sites were not available because the required data sets were not complete and, at one research site, could not be disaggregated at unit level from within a larger hospital trust provider. We, therefore, undertook an alternative analysis by using data collected at the individual level from our recruited research participants and comparing this with national-level data from the National Diabetes Audit (NDA).45
Inclusion of an additional research site for group clinics
During the course of recruitment to our group clinic care model at Newham University Hospital, it became apparent that we would not meet our recruitment target because of a recent reduction in the total clinic population (owing to recent local service changes), as well as incomplete uptake. We, therefore, made a proactive decision to open a new research site at the Central Middlesex Hospital to run the group clinics. Central Middlesex serves a comparable multiethnic, deprived, urban population to Newham University Hospital. The inclusion of this new site had both advantages and disadvantages to the research process. The advantages were that we could evaluate the delivery of group clinics in an NHS setting with different staff, care processes and organisational structures, and we were able to study the group clinic model at a more advanced stage of its development following its co-design and implementation at Newham. This provided new insights and richness to the evaluation of implementation, and contributes to the generalisability of the research findings. There were fewer disadvantages: the variation in delivery of the care model made the direct comparison of quantitative data between Newham and Central Middlesex challenging, and, therefore, data from both sites have not been aggregated.
Comparative quantitative analyses
We had planned to undertake an intention-to-treat analysis, in which we would compare participants at the intervention sites who had been offered and agreed to participate in the group clinics with those who had not been offered group clinics. However, owing to the challenges of recruitment at the group clinic sites, and the variable attendance at group clinics of the recruited participants, we instead offered the group clinic to all young adults at each clinic site and compared those who did with those who did not attend (DNA). As planned, we then compared characteristics and trajectories of all participants at intervention sites with participants at control sites (intention-to-treat analysis).
Theoretical approach
Our work was underpinned by a set of theoretical ideas that influenced intervention development and implementation, and informed qualitative data analysis and interpretation; a more detailed theoretical analysis will be included in upcoming publications. Our approach to intervention development and implementation was underpinned by complexity theory, primarily drawing on an understanding of complex systems as characterised by uncertainty, unpredictability and emergence.46,47 This means that we treated the introduction of group clinics as a complex change process in which mechanistic replication and standardisation were not sufficient, and attention to the dynamic properties of context and the ongoing tensions raised was necessary.46,48 Following previous work,46,49 the principles driving our implementation effort can be summarised thus: acknowledging unpredictability, harnessing the capacity of implementation teams to self-organise differently in the different settings, facilitating interdependencies between clinical and operational processes, and encouraging sense-making and experimentation with different group clinic delivery formats. In addition, we focused on developing adaptive capability in staff so that they could make good judgements on how best to introduce this new model of care, attending to human relationships that would make this complex change feasible (e.g. through goodwill and reciprocity), and, finally, harnessing conflict productively to contribute to positive solutions.46 Following a complexity-informed approach, our evaluation also produced a nuanced account of how change came about by drawing on different data collection methods and through developing close relationships with the field sites.
By drawing on ecological theories,50,51 we viewed patient self-management and self-care not as activities carried out in isolation, but as activities nested in multiple, proximal and distal contexts (e.g. family, education, employment, social life) that afforded particular opportunities and constraints, especially so in the context of socioeconomic deprivation. This was acknowledged not only in terms of how group clinic interactions attempted to influence self-care, but also in terms of how the evaluation elicited an understanding of how group clinics worked (or not) for young people.
The evaluation delved further into social theory to make sense of the way that group clinics worked and theoretically substantiate an understanding of their change mechanisms. We drew from ideas on practices of solidarity as theorised by Prainsack and Buyx,52 who conceptualise solidarity as enacted, embodied and contextual, based on a view of personhood as relational (i.e. in which people are dependent on and open to their environments). Our analysis draws not only on instances of solidarity at an interpersonal level (i.e. manifestations of willingness to carry costs to assist others with whom a person recognises similarity in at least one relevant respect), but also on instances of solidarity at a group level, through shared commitment to engage together with others in diabetes care, including through a sense of a joint purpose.52
In contrast to individual appointments, group clinics put relational aspects of self-care to the fore, with the focus on interactions between patients rather than just patient–clinician relationships. This introduced new types of ethics relations between patients and projected new ways of enacting patienthood. We explored what it meant for clinical care to be harnessing experiential knowledge directly through active patient participation in service provision. This included an understanding of how a balance was achieved between biomedical knowledge and practical experiential knowledge in group clinics. Following Pols,53 we examined how embodied patient knowledge becomes transferable and useful in the context of group-based care.54 We asked the following question: what kinds of knowledge are shaped in group clinics and what are the conditions for doing so? This was supplemented by burden of treatment theory, which focuses on how self-management work and responsibility become delegated to patients and the demands that these place on them.55
Finally, another theoretical perspective that informed our analysis of group clinic implementation related to articulation work as the hidden, invisible adjustments and alignments necessary to successfully carry out tasks in sociocultural settings.56 We were specifically influenced by the three different types of articulation work proposed by Allen:56 (1) temporal articulation, with health professionals acting proactively to facilitate care processes or reactively to address unexpected developments; (2) material articulation, referring to how tools and other artefacts become mobilised and embedded in care pathways to maintain their stability; and (3) integrative articulation, which covers relational aspects of receiving input and managing co-ordination and coherence in care, including working with contradictions.
National context
Existing use of group clinics in the NHS
To investigate the existing and potential use of group clinics in diabetes management and treatment, a scoping survey was set up for interested health-care professionals in May 2017. An online questionnaire tool was disseminated to clinical networks and professionals via e-mail and social media.
National Diabetes Audit
The NDA is a large audit managed by NHS Digital. It contains individual- and service-level data on diabetes care processes and outcomes from NHS trusts in England, benchmarked against quality standards (e.g. guidance from NICE).57 We planned to use NDA data to provide a national context to our research and make unit-level comparisons with our research sites. We used the standard NDA audit reporting of eight routine care processes [HbA1c levels, blood pressure (BP), cholesterol, renal function, urinary albumin, foot examination, body mass index (BMI) and smoking review, recorded in the previous year] and three treatment targets (HbA1c levels of ≤ 48 mmol/mol, BP of ≤ 140/80 mmHg and total cholesterol of ≤ 4 mmol/l in the previous year).
As planned, an application was submitted to the NHS Digital Data Access Request service on 22 May 2017 for service-level data on patients aged 16–25 years in the NDA. However, there were a number of changes to the process for accessing NDA data over the study period, which caused a long delay in receiving these data. Our data application (DARS-NIC-228637-P6N0L) was not approved until 14 August 2019, and at this point we were given access to 2016/17 and 2017/18 data.
During the course of this application, a number of unexpected issues arose. First, it became apparent that we would not be able to obtain unit-level data disaggregated by our research sites to compare with data from other units or national-level data. This is because the NDA is designed predominantly for analysis by general practices and Clinical Commissioning Groups (CCGs) rather than hospital services. In addition, data from all patients at Newham University Hospital (NUH) (a group clinic site) and Mile End Hospital (MEH) (a control site) were aggregated and reported by the overseeing NHS trust provider [Barts Health NHS Trust (BH)] in 2017/18. Data were not available from Central Middlesex Hospital (CMH), which is part of the larger London North West University Healthcare NHS Trust. Second, the data extraction process changed to a new system in 2017/18 and we were advised that national comparison would not be possible between 2016/17 and 2017/18. Finally, we were given access only to data with small numbers suppressed. For this reason, statistical analysis of unit-level data was often inappropriate owing to the relatively small sample size.
Within these limitations, we had to change our intended aim of using NDA data to make unit-level comparisons between our group clinic sites and national data, and instead we used the data to make more general descriptive comparisons, as follows:
describe sociodemographic and clinical characteristics of young people (aged 16–25 years) under the care of adult diabetes services in England
study performance on the three treatment targets and eight care processes for young people (aged 16–25 years) under the care of adult diabetes services in England
compare descriptive characteristics of young people (aged 16–25 years) at our research sites with young people and older people (> 26 years) nationally.
We studied NDA data from England cross-sectionally from 2017/18 audit submissions. Data were aggregated by type of diabetes; however, owing to small numbers, rare types of diabetes defined by NDA as ‘other’ (e.g. cystic fibrosis-related diabetes and monogenic diabetes) were included with the T2D data. The aggregation of rare types of diabetes with T2D is unlikely to distort the findings because they are a small fraction of the whole, especially outside tertiary specialist centres (which were not included in our research sites). We obtained national data (from all England sites submitting to the NDA) as well as data from BH (which included our group clinic site NUH and our control site MEH, in addition to two other clinical services). NDA data were not available from CMH. We made descriptive comparisons between young adults (aged 16–25 years) at our research sites and national data on the same age group, as well as people aged ≥ 26 years.
Setting
We included four research sites based at clinical services delivering young adult diabetes care (). These included two sites (NUH and CMH) that would deliver group clinics, and two sites [MEH and Whittington Hospital (WH)] that would act as comparator (control) sites.
Research sites: summary of setting and clinic characteristics
Research participants
We recruited young adults (aged 16–25 years) living with diabetes (of any type) and receiving care at the four research sites, with the aim of recruiting 80–100 participants across all sites. There were no exclusion criteria. Young adults recruited at the ‘group clinic sites’ were invited to join group clinics co-designed and implemented during the course of this research. Young adults recruited at ‘control sites’ were involved in data collection only.
When the study was designed, we did not anticipate that we would invite all young adults (if eligible) under each service to join the study. However, after observing the high patient turnover in the clinics and low attendance rates, we revised this plan and instead attempted to recruit all young adults under each service.
Potential participants were identified by usual-care teams, and were approached by the research team (which included usual-care team members), and were given invitation letters and patient information sheets. Informed consent was then sought from interested individuals. An example of all patient-facing documents (i.e. invitation letter, patient information sheet and consent form) is given in Report Supplementary Material 1. The research was given a branding and logo and named ‘Together Study’, which was used across all patient-facing documents.
Co-design
Methodological approach
Our research drew on key theoretical and practical approaches to participatory research that allow researchers, practitioners and service users to learn together for the benefit of service redesign.58,59 The group clinic interventions were co-designed using a participatory approach to ensure cultural, developmental and practical relevance; enhance recruitment and retention; and attempt to instigate system change and support sustainability.59 Co-design focuses on improving patient and staff experiences, with equal importance being given to each perspective. Participants in co-design, whether patients, staff or other stakeholders,60 are seen as being ‘expert through experience’, and this experience is utilised in co-design to identify opportunities for improvement and adaptations to service design, focused on the functionality (usability) for patients and staff. In co-design approaches, patients and staff work alongside each other to identify problems that can be practically overcome and to develop a jointly negotiated outcome. Co-design with young people may require special facilitation skills and adaptation to be developmentally and age appropriate. The role of the facilitator is key in building trust,61 but there are also considerations around the special challenges of engaging young people to take part and in ensuring informed consent.
We used The King’s Fund Experience-based Co-design Toolkit.31,61–63 This approach was chosen because research64 suggests that co-designed services can lead to service improvement. The EBCD process provides a template for co-designing service development, drawing on the expertise of patients and staff, with regular review and iteration. EBCD gives a detailed insight into the experiences of all participants, including both patients and staff. When EBCD is conducted, patients and staff work separately at first and are then brought together for joint work. This careful approach supports participation from patients and staff (including wider stakeholders), in prioritising aspects of existing care to build on, as well as areas for improvement and change.
We planned a co-design process that would evolve and adapt as the project developed, and would fit around availability of participants, practical constraints and hospital procedures. Previous work65 indicates that there are particular challenges of engaging young people in co-production work, especially if they are also coping with a long-term condition as well as the day-to-day challenges of education and/or work.
We made adaptations to the full EBCD process, which ordinarily includes the production of a series of films that help to share perspectives between the groups. We omitted this step because it was decided that this aspect would be too time-consuming and might be a disincentive to young adults and staff engaging with the process. Instead of filming, we relied on audio-recording and verbal feedback at meetings. These changes are consistent with reports from other users and researchers of co-design who have also streamlined these EBCD processes.64
Co-design processes
We undertook co-design in two phases.
Phase 1
In year 1, co-design was used to develop and build the new group clinic-based model of care and involved the following:
Group and individual sessions – group co-design sessions were formed for young people and staff/stakeholders separately, and were complemented by additional one-to-one interviews.
Analysis of the main themes arising from the interviews.
A joint patient and staff/stakeholder event to bring perspectives together.
Further analysis of emerging themes.
Follow-up interviews with additional participants as necessary.
Feedback to participants.
Phase 2
In year 2, the second phase of co-design supported iterative development of the group clinic model, refining its design and delivery after the clinics had already been implemented for 1 year. This involved the following:
Continuous co-design processes
Ongoing co-design was built in alongside the implementation of the group clinics, using discussion and feedback after each session and facilitated by the clinician and researcher in residence. This supported wider participation from young adults beyond the discrete co-design sessions outlined above, allowing iterative adaptation.
Consent/information procedures
Patient participants for co-design were recruited from the clinical settings in which we planned to deliver group clinics. In addition, staff and stakeholder participants were also selected as representatives from organisations (e.g. CCGs) with responsibility for the care of young adults with diabetes. Formal informed consent procedures were used for all co-design participants.
Conducting the co-design sessions
Co-design workshops took place in community and clinical facilities linked to NUH and CMH. Co-design was led by the Association for Young People’s Health (AYPH) (London, UK), an external and third-sector organisation with particular expertise in participation work with young people and health professionals. The clinical teams delivering diabetes care to young adults at NUH and CMH also contributed to the delivery of the co-design process, as did the research team.
The facilitation of co-design sessions was led by experienced staff from the AYPH. Sessions were audiotaped and transcribed. The following principles and approaches were used:
Young people living with diabetes. Each individual was encouraged to tell their own story, recalling their own voice and experiences and communicating their own personal ‘truths’. The facilitator had prepared a workshop of questions and workshop activities to defuse any anxiety or embarrassment and provide a way to encourage sharing. However, the issues raised mainly came from the participants, and discussions followed the line that they wanted to take. Sessions lasted no longer than 2 hours and opened and ended with a ‘check-in’ to raise any emotional anxieties and to ensure that young people left the process in a safe state of mind, feeling supported and listened to. Young people were also given an outline of the next steps in the work and details about how to find further information.
Health-care professionals. It was important to allow staff the same freedom as the patients to share their perspective. Participants included a wide range of professionals and stakeholders, including dietitians, specialist nursing staff, a
CCG commissioner, representatives from primary care, representatives from the voluntary sector (Diabetes UK, London, UK), consultant diabetologists and reception staff.
Joint discussions. To focus and support discussions, all those attending the joint sessions (patients and staff/stakeholders) were asked to prepare in advance three issues to present to the group, instead of collating and sharing issues during the group session. This supported the effective use of the time available and provided a valuable structure for discussions.
Discussion topics introduced in the co-design
We drew on the findings from our realist review to identify the following areas of discussion during co-design:
Group composition and continuity. At the time of our study, to our knowledge, there was little certainty in the literature around the ‘ideal’ composition of group clinics, and we considered the possible importance of age-related developmental stages, sex and disease type in the co-design of our group clinic model. We also discussed the possible importance of factors such as independence (e.g. living at home, away at university), time from diagnosis, family circumstances and general life experiences. Our realist review highlighted the importance of continuity within the group (participants and/or staff) to support relationship building, cohesion and the sharing of stories.
Role of parents. The realist review identified parents as possible active participants in a group clinic model for young adults, and that their involvement could have both positive and negative effects.
Individual versus collective experiences. We identified the need to manage group discussions in such a way that they would support both individual and shared experiences.
Content and approach specific to young people. Our review suggested that many young people prioritised fitting in with their peers as more appealing than closely following diabetes self-management advice, and that this challenge was a potentially fruitful area of focus in the group clinics.
Logistical considerations. We identified a number of logistical considerations important to the design of group clinics, including the time of day they should be scheduled and the location they should be held in.
Clinical aspects. Previous literature
44 highlighted a range of strategic considerations relating to the clinical care delivered in group clinics, including whether group clinics should replace existing care appointments or run in parallel to them (i.e. offer additional care), how the group should be led and facilitated and by whom, how frequently they should be held and how the safety of participants could be ensured through setting ground rules.
Evaluation of the group clinic model
Qualitative methods
Embedded research
A ‘researcher in residence’ model was adopted as a practical manifestation of a participatory approach to research and evaluation. The researcher was an integral member of the front-line implementation team, contributed theoretical and practical insights based on research findings, and helped to navigate different bodies of expertise within the study. The embedded researcher worked towards bridging the qualitative and quantitative evaluations, helping to include practitioner and patient views into the design and feeding back early findings to stakeholders. We followed previous experience with how the researcher-in-residence model was applied in a number of different settings.66
Ethnographic observation
The researcher in residence (CP) worked closely with the clinical teams throughout the project and carried out ethnographic observation in the two hospitals. This primarily included three types of observations, involving different degrees of participation depending on the encounters observed:
Group clinics. The researcher was involved in different aspects of the group clinics programme, including planning and setting up, co-ordination between the clinical teams, and de-briefing and discussing ongoing adjustments to the model of care. She carried out ethnographic observation in most group clinics organised in the two hospitals, had informal discussions with patients ahead of the clinic, participated in icebreakers and generally helped with delivery where needed. To support the ongoing co-design of this new model of care, at the end of each clinic the researcher held brief feedback discussions with patients to understand what had worked well and what they thought should change next time (when the researcher was absent this was led by the diabetes specialist nurse or the youth worker). The researcher introduced herself to the group as a researcher from the university interested in finding out how group clinics work for young people, and she worked towards building rapport. Especially with some of the frequent attenders, she managed to build a good relationship and understand people’s experiences not just as a one-off interaction but longitudinally over the course of the group clinics programme. A significant amount of field notes was collected from ethnographic engagement. These notes included information on clinic characteristics, such as session content, context, group dynamics and facilitation style. The majority of group clinics were also audio-recorded with participant consent and transcribed for analysis.
Individual appointments. To gain a wider understanding of standard diabetes care and to be able to draw comparisons with group clinics, the researcher also conducted ethnographic observations in 15 individual appointments in young adult clinics. These were sampled to achieve maximum variation between different consultants and nurses in the two hospitals. Some of these appointments were with patients taking part in group clinics, which allowed a broader understanding of their engagement with care as part of different interactions. It was striking how some patients presented themselves differently in the one-to-one appointments and in the group clinics.
Other interactions. The qualitative researcher also collected field notes from ethnographic observations in co-ordination meetings; facilitation trainings; and other informal interactions with clinical teams in the context of setting up, managing and delivering group-based care alongside standard clinical practice.
Qualitative interviews
Between February 2018 and October 2019, the researcher in residence carried out 31 semistructured interviews with patients, group facilitators and other clinical and non-clinical staff (see Appendix 1 for further details). Interviews lasted 30–110 minutes and followed a semistructured format (see Appendix 2 for interview guides with indicative questions). One staff participant was interviewed twice to reflect on the development of group-based care over time. Another interview involved two siblings who were interviewed jointly about their experiences living with diabetes and attending group care. Most patient interviews took place in hospital settings or other mutually convenient locations either before or after the group clinics, although five patients preferred to be interviewed on the telephone. Staff interviews were conducted in offices or other hospital settings; two took place on the telephone. Interviews formed only part of the encounters with participants in the context of a broader relationship developed with the qualitative researcher during the project (e.g. through ethnographic observations in clinics or informal discussions during clinic set-up). The purpose of the interviews was, therefore, to continue conversations that had already been taking place over the course of the project and to consolidate some of the learning from patient and staff perspectives. The researcher took contemporaneous field notes to contextualize interactions and encounters with research participants, bringing together data from different formal and informal discussions throughout the 2 years of fieldwork. Most interviews were audio-recorded with consent and professionally transcribed; in two of the telephone interviews it was more practical to keep field notes.
Patient interviewees (
n = 19). We recruited 19 young people in interviews, nine female and 10 male, who were between 18 and 25 years of age and from a variety of ethnic backgrounds. Four patients had attended 7–10 clinics, nine had attended three to six clinics and six had attended zero to two clinics (one of whom withdrew from the research study after one clinic and another consented but never attended). Most interviewees were living with
T1D (of whom two were also using an insulin pump) and two had
T2D. This was representative of the broader composition of participants in group clinics, which predominantly included patients with
T1D in both hospitals. Interviews addressed experiences of being diagnosed and living with diabetes as a young person; experiences of receiving diabetes care; and experiences of participating in group clinics, including encounters with health professionals and other young people. Guided by clinicians’ knowledge of patients and the relationships with the researcher as part of the ethnographic engagement, we recruited patients with varied clinical background, time since diagnosis and experiences of diabetes care. Despite our efforts, we were only able to recruit one young person who had not attended any of the clinics; most of those who did not engage with group-based care also declined participation in interview.
Staff interviewees (n = 11). We recruited three diabetes consultants, three diabetes specialist nurses, one youth worker, one research nurse, one dietitian, one psychologist and one sexual health advisor who delivered sessions and/or supported the group clinic programme. Discussions covered clinicians’ experiences of providing care for young people with diabetes and their views on how diabetes health services could better meet the needs of this population, as well as their experiences of setting up and delivering group-based care in the two hospitals involved. We focused our sampling on clinical and non-clinical staff who had been involved in the group clinic programme (as group facilitators or in other roles) to understand their views and experiences.
Documents
We collected all documentation produced in co-design sessions, project and steering group meetings, facilitation training and other interactions. Other materials collected as part of our ethnographic fieldwork in group clinics included outputs of group activities using flip charts, icebreaker materials, sick-day rules diagrams and other artefacts used in the group context ().
Photographs illustrating the group clinic materials and room set-up (photographs A and D from NUH; and B, C and E from CMH).
Qualitative data analysis
Owing to the iterative nature of the research, the researcher in residence carried out analysis in parallel to data collection throughout the project, with emerging findings presented and discussed in team meetings and used to adapt the model of care. The analysis drew on different theoretical lenses (see Theoretical approach), such as complexity approaches in health services research47,49 and frameworks on invisible, hidden work,56,67 as well as burden of treatment and patient work theory,55 solidarity practices52 and critical perspectives on patient expertise.53 We moved between an inductive and a deductive approach to our analysis, paying attention to emergent themes and using substantive theory as a sensitising device to drive further interrogation of the data. We also drew on the programme theory developed as part of our realist review in earlier stages of the project (see Chapter 2) and refined its different components based on further analysis of our empirical data. We continuously developed the overall narrative emerging from our analysis and iteratively added to the coding framework. A second researcher (AF), with a background in psychology and complex intervention development, was involved in later stages of the analysis of the group clinic transcripts, independently verifying the coding framework and developing consensus on emergent themes. Interim analysis was also driven by pragmatic requirements to inform implementation, as well as to ensure that no safety issues were raised by participants.
NVivo version 12 (QSR International, Warrington, UK) and Microsoft Word (Microsoft Corporation, Redmond, WA, USA) documents were used to support management and analysis of qualitative data, which included anonymised interview and group clinic transcripts, ethnographic field notes, e-mail communications and other documents. An example NVivo analysis is presented in Appendix 3.
Quantitative methods
Data collection
Members of the research team at each site (both group clinic and control sites) collected sociodemographic and questionnaire data through face-to-face or telephone interviews with recruited study participants. We were unable to collect data on participants who did not wish to be recruited, but collated brief information on their reasons behind that decision. Clinical data were accessed through each site’s clinical record system. Data were then collated on a standard template [see the NIHR project web page; URL: www.journalslibrary.nihr.ac.uk/hsdr/NKCR8246) using the study number of each participant as the only unique identifier. No identifiable information was recorded on this template.
Baseline data were collected at entry to the study. Follow-up data were collected approximately 1 year later (a window of 9–15 months was used to allow co-ordination with clinic visits, university holidays and other practical considerations).
The following data were collected at baseline:
Sociodemographic characteristics – age, sex, Index of Multiple Deprivation (
IMD) quintile, ethnicity, English as first language (yes/no), education and employment status (education/employment, both/neither).
Clinical characteristics (including health-care activity) – type of diabetes (1/2/other), most recent HbA
1C mmol/mol level, reported frequency of blood glucose monitoring per day, age at diagnosis, use of technology within the last year and previous attendance at any group education for diabetes. In addition, we captured the percentages of planned diabetes appointments, emergency department (
ED) attendances (diabetes related), inpatient admissions (diabetes related) and primary care consultations, all within the previous year.
Patient-reported instruments – the Problem Areas In Diabetes Score and Patient Enablement Instrument were used.
68,69 Relevant information about these instruments is presented in
; for example questionnaires see the
NIHR project web page (URL:
www.journalslibrary.nihr.ac.uk/hsdr/NKCR8246).
At the follow-up visit, all of these variables were again collected and recorded in the study template, with the exception of age, sex, IMD quintile, ethnicity, first language being English and type of diabetes, for which duplicate information was not necessary.
Quantitative data analysis
Four main sets of quantitative analyses were performed to compare the following:
baseline sociodemographic and clinical characteristics of all participants, with comparisons made among sites, and among those at group clinic sites who did and did not choose to attend group clinics
trajectories of participants who did and did not choose to attend group clinics (‘difference in difference’ analysis)
differences in trajectories of participants attending group clinics according to the number of group clinics attended (‘dose response’ analysis).
All groups were compared using chi-squared tests for categorical variables and t-tests for continuous variables with normal distribution. Trajectories were analysed by subtracting the follow-up from the baseline values to derive a new set of variables that reflected change over time.
In addition, linear regression models were used to investigate the association between clinic attendance and trajectories among participants at group clinic sites. Initially, the independent variable was entered as a binary variable (attended any vs. attended no group clinics) and each derived trajectory variable was entered in turn as the dependent variable. These models were then repeated, adjusting for the participants’ diabetes type, age, sex, ethnicity, age at diagnosis and deprivation. Unadjusted and adjusted models were then performed in which attendance was entered as a categorical variable (attended zero, one, two or three or more clinics), with zero clinics used as the reference category.
p-values of > 0.05 were not considered to be significant. Values below this level are highlighted but are interpreted with caution. It is recognised that we are making multiple comparisons and some values of < 0.05 owing to chance would be expected.
Quantitative methods: health economics
An NHS perspective was adopted, spanning primary and secondary health-care sectors. Economic evaluation methods followed the Guide to the Methods of Technology Appraisal 2013,70 which provides guidance on how to assess the clinical effectiveness and cost-effectiveness of new and established technologies in the NHS.
Microcosting of the group clinic intervention
Microcosting was used as a means to estimate the economic cost of the group clinic intervention to the health system (the NHS) using Personal Social Services Research Unit (PSSRU) 2018 costs.71,72 Microcosting is particularly applied to the costing of new interventions; including the large variability across providers of the group clinics for young people with diabetes we also included a bottom-up construction of the costs associated with co-design and delivering the intervention.
The cost of co-design included time spent preparing workshop materials, recruiting participants, running and recording the workshops, transcribing records and analysing transcripts.
The clinic running costs included staff costs of running the clinics, preparing the room and materials, arranging appointments, chasing non-attenders, booking the venue, arranging refreshments and making patient notes. Data on resources associated with designing and delivering clinics were collected prospectively using purposely designed questionnaires; staff completing these questionnaires were encouraged to report all staff work and time commitments, including that of ‘hidden’ work. Staff time was costed using the NHS pay scales 2018–19.73
Cost of intervention per clinic and per participant
The estimated cost of the intervention was based on the number of clinics and the number of patients who attended the clinics using PSSRU costs.72 The average cost per participant for each centre was derived by dividing the total cost of running clinics by the number of attenders (per-protocol analysis). Sensitivity analyses were conducted, varying the number of clinics and the number of participants for both CMH and NUH.
Use of health-care resources
Data on the use of primary and secondary health-care services (usual care) for a 12-month pre-intervention period was extracted from clinical records. These data included the number of contacts with a diabetologist, diabetes specialist nurse, dietitian or psychologist (planned and attended); unplanned contacts with a GP, practice nurse or diabetes specialist nurse; accident and emergency (A&E) attendances; and hospital admissions. Individual-level resource use data were combined with unit costs to calculate the total cost of health services use for each participant. Primary care consultations and referrals to community care were costed using the National Schedule of Reference Costs 2017–18.74 The list of unit costs used for costing health care services are included (see Appendix 4). Data analyses were conducted in Microsoft Excel® 2016.
Summary of data sources
A summary of the data sources used in this project is given in .
Project management and governance
The study received ethics approval from the Office for Research Ethics Committees Northern Ireland (ORECNI) on 23 February 2017 (reference 17/NI/0019) and is registered on the UK National Research Register as ISRCTN 27989430.
Standard rules applied for data security, confidentiality and information governance. Informed consent was sought for ethnographic observations during group clinics and interviews, and for accessing routinely collected NHS data on participants. Confidentiality and safety among group clinic participants was a priority, and all participants who attended were asked to agree to a code of conduct and confidentiality to ensure that all clinic discussions were kept within the group.
The study was led by Sarah Finer, with co-leadership from Dougal Hargreaves. Owing to Sarah Finer taking maternity leave, Trish Greenhalgh temporarily led the study, with Dougal Hargreaves, from April 2019 to January 2020. The study was delivered and managed by a core working group (SF, DH, CP, SV, MK, AH and GC) and supported by 6-monthly independent steering group meetings, along with ad hoc communication with steering group members when necessary. Monthly research management meetings were held throughout the lifetime of the project, which focused on progress towards short- and long-term milestones, administrative tasks and general project management. These were attended by the core team identified above and often supported by others, such as clinical teams, depending on the focus of the meeting. Action points and minutes circulated after all meetings enabled team members to keep track of study progress and ensured steady progression towards milestones and early identification of potential challenges.
A routine internal audit was carried out by the study sponsor on 10 September 2018. No critical or major findings were identified, and following action of nine minor findings the audit was closed and a certificate was issued on 5 December 2018.
Project steering group
Project steering group meetings were held every 6 months, starting on 6 November 2017 and finishing with a final wrap-up meeting on 20 April 2020. Group members included representatives from Diabetes UK and their young adult panel, along with expert lay people, clinicians, external academics and team members. Support from this group with a wide range of experiences and perspectives was invaluable to the development of the work, and the external, critical viewpoint has been very helpful in raising and highlighting areas that have benefited from further development. In particular, the group has helped the project to maintain a wide perspective so that findings and dissemination could be accessible to a variety of audiences, including commissioners, stakeholders and service users, as well as an academic audience.
Dissemination and patient and public involvement
Patient and public involvement (PPI), and wider dissemination to the academic and practitioner community, were both built into our research from the outset and followed INVOLVE guidance.75 We built on research priorities identified through the James Lind Alliance Priority Setting Partnerships for type 1 and 2 diabetes,76,77 which identified the need to research approaches to the patient-centred management and delivery of social support for people living with diabetes. Our PPI approach was also built on the foundations set by the participatory co-design of the group clinic model.
Key patient and public involvement elements
The main planned elements of PPI in our research were led by our voluntary sector partner, the AYPH, in partnership with the main project delivery team and academic partners.
The overall intention of the PPI and dissemination activities was to share learning about the challenges and achievements of the group clinic model as it was implemented across the sites, so that others could understand and, potentially, replicate the model. As the project unfolded, dissemination was focused on sharing the results of the realist review, the co-designed new care model, its evaluation and messages for generalising findings to a wider context. We also focused on disseminating specific outputs related to co-design, so that others could learn from our successes and failures and apply these to their own service improvement work.
Patient and public involvement processes can bring challenges relating to the relationships of power between patients and public services,62 and we were mindful of this in our PPI and dissemination activities, for example involving a young person in the design and delivery of stakeholder events.
Patient and public involvement methodology
The main methods employed for involving patients and the public, local and national stakeholders and the wider dissemination list included the following:
Co-design. A co-design approach to developing the group clinic model supported involvement from young people, clinicians, commissioners and others from the outset.
Management of the research. An external steering/advisory group was convened and worked alongside the research team throughout the duration of the research, and included a patient representative from Diabetes UK with lived experience of
T1D as a young adult.
Developing participant information resources. Patient representatives and stakeholders were involved in reviewing all patient-facing documents (e.g. information sheets and consent forms).
Dissemination of research findings. We planned to produce a range of events and outputs, including stakeholder engagement events, briefing papers and summaries, and these are discussed in
Chapter 9. Our intended target audiences included young adults with diabetes (‘patients’) and stakeholders with roles in service use, service delivery, policy-making and health service design. In addition to this, we used social media engagement though our website and project-specific Twitter (Twitter, Inc., San Francisco, CA, USA;
www.twitter.com) account.
Dissemination networks
We built wide networks for dissemination, drawing on the contacts of the main project partners. This included the following:
local clinical and academic networks –
institution based: Queen Mary University of London, University College London, University of East London,
BH, Newham
CCG and Newham Borough Council
Local Transforming Services Together Programme
NHS England London Region Children and Young People’s Programme
Improvement Science London
NIHR Clinical Research Networks North Thames.
clinical and patient networks –
strategic and national networks –
national children and young adult working groups chaired by the national leads for children and teenage and young adults
royal colleges, including the Royal Colleges of Physicians, General Practice and Paediatrics and Child Health.
national policy forums –
NHS, Public Health England and the Department of Health and Social Care and their voluntary sector partners via the Health and Wellbeing Alliance (a partnership between sector representatives and the health and care system)
78 and the Child and Maternal Health Intelligence Network.
79
