Development and evaluation by a cluster randomised trial of a psychosocial intervention in children and teenagers experiencing diabetes: the DEPICTED study

Article history

The research reported in this issue of the journal was commissioned by the HTA programme as project number 03/46/09. The contractual start date was in June 2005. The draft report began editorial review in June 2010 and was accepted for publication in October 2010. As the funder, by devising a commissioning brief, the HTA programme specified the research question and study design. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2011. This work was produced by Gregory et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This journal is a member of and subscribes to the principles of the Committee on Publication Ethics (COPE) (http://www.publicationethics.org/). This journal may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2011 Queen’s Printer and Controller of HMSO

Diabetes

Diabetes is the third most common chronic disease in childhood, with 1–2 per 1000 children and adolescents in the UK receiving prescriptions of insulin in recent years (1998 and 2005). 1 Since 1989, the incidence has doubled, with a particularly marked increase noted in the preschool age group. 2 In childhood, the vast majority of affected children experience autoimmune-mediated destruction of their insulin-secreting pancreatic β-cells, which leads to insulin deficiency (type 1 diabetes). As a consequence of insulin deficiency, children develop raised blood glucose concentrations (hyperglycaemia), which lead to excess urinary losses (polyuria) and therefore increased thirst (polydipsia). In addition, insulin deficiency leads to uncontrolled breakdown of fat (lipolysis), as demonstrated by marked weight loss over relatively short time periods. Lipolysis in the presence of insulin deficiency results in ketosis, which, if uncontrolled, may lead to potentially life-threatening episodes of acute diabetic ketoacidosis. The presence of vomiting or development of ketoacidosis are common reasons for children with diabetes to require hospital admission. The acute metabolic consequences of insulin deficiency may be reversed or prevented by the administration of an insulin treatment.

Complications of diabetes

In the short term, excess insulin for requirements may cause hypoglycaemia (low blood glucose levels), which, if severe, may lead to loss of consciousness. By contrast, inadequate insulin therapy may cause symptoms similar to those at diagnosis (see above). In the longer term, chronically elevated blood glucose concentrations leads to an increased risk of clinical complications. In childhood, poor glycaemic control causes growth failure and pubertal delay, which may be reversible with improved clinical management including optimisation of insulin therapy. In the longer term, more serious and eventually irreversible microvascular complications arise. These include sight-threatening retinopathy and renal disease. Initially, renal disease is asymptomatic and detected by increased protein (albumin) excretion in the urine but, if untreated, will eventually deteriorate leading to renal failure and the need for dialysis. A further devastating complication is neuropathy, which may produce a range of symptoms such as impaired peripheral sensation and pain or gastrointestinal and genitourinary problems if the autonomic system is affected, resulting in major adverse effects on quality of life (QoL). In addition to the microvascular complications, macrovascular disease is common, with increased risks of myocardial infarction and strokes in later life. Microvascular and macrovascular complications are rarely seen in childhood, but occur with increasing frequency in young adult life. There is clear evidence, however, that the quality of blood glucose control through childhood is a significant risk factor for the development of many of these complications in later adult life. 3

Psychosocial aspects of diabetes

The management of diabetes is complex, requiring significant practical expertise to optimise outcomes and, unsurprisingly, may result in significant psychological difficulties for young people with diabetes and their families. Variations in blood glucose concentrations, particularly overnight, have been shown to affect mood and behaviour. 4 The difficulties of adhering to a practically demanding regimen may result in overdependence of children on their parents5 or adverse effects on behaviour, including an increase in suicidal thoughts. 6 For the family, managing childhood diabetes brings particular pressures, including the grief experienced by parents at diagnosis. 7 In relation to the challenges of the day-to-day management of the diabetes, problems may occur in communication between parent and child, and there is a risk of increased family conflict with the experience of frustration and guilt at failure to achieve optimal outcomes. 8 Existing psychological issues within families involving functioning, coping and interpersonal relationships may be exacerbated. Psychiatric and psychological problems (including eating disorders and effects on body image, etc., exacerbated by the inter-relationship with insulin and other aspects of diabetes management) are therefore unsurprisingly seen more commonly in young people with diabetes than in the non-diabetic population. 9,10

It is well recognised11 that psychosocial and educational influences play a key role in determining management outcomes in children with diabetes. For example, a large audit in Scotland has shown that throughout childhood family structure is associated with glycaemic control. 12 During adolescence, rapid physical change (puberty) leads to relative resistance to the effects of insulin. 13 Concurrent major developmental changes include increasing independence, emerging sexuality and increased stress from peer and academic pressures. These factors together are often associated with deteriorating glycaemic control. Knowledge and skills imparted by the diabetes teams are especially important tools for the child and their family to achieve optimal glycaemic control during this crucial period.

Diabetes management

The management of diabetes by patients and their family requires them to develop an understanding of the complex interaction of the effects of insulin, food and physical activity on blood glucose concentrations. Treatment of diabetes involves the regular administration of insulin, most commonly by two to four subcutaneous injections daily or through the use of an insulin pump, which provides a continuous infusion of insulin through a subcutaneously sited catheter. A healthy lifestyle is recommended, including regular physical activity and a diet that regulates carbohydrate and fat intake. To optimise diabetes management, it is recommended14 that the patient and his/her family develop a sophisticated understanding of the carbohydrate content of food so that the amount of insulin administered can be finely tuned (so-called ‘carbohydrate counting’). The efficacy of management is monitored in the short term by regular self-measurement (ideally four or more times daily) of blood glucose concentrations and in the longer term by monitoring (3–4 monthly) glycosylated haemoglobin (HbA_1c) levels in blood and regular review in paediatric diabetes clinics.

In the UK, clinical care is usually delivered by paediatric diabetes services established in secondary care. Such services require the multidisciplinary input of doctors with expertise in both paediatrics and childhood diabetes, nurse specialists who liaise between the clinic, the child’s home and school, dietitians, child psychologists, podiatrists and social workers. There also needs to be close collaboration between paediatric and adult services to ensure that as children progress through their teens arrangements are made for their care to be handed over from paediatric to adult services. This is a time when particular difficulties may be encountered by clinical services, as teenagers with diabetes take increasing responsibility for their self-management and also encounter the problems caused by increased insulin resistance during puberty. 15

Adherence to diabetes management

The aims of paediatric diabetes services are to support and educate children and their parents in the care of diabetes, to manage diabetes in a manner that optimises clinical outcomes and to prepare teenagers for young adult life by helping them to become increasingly independent in their self-management. Given the complexities of diabetes management described above, it is unsurprising that many children and their families struggle to adhere to optimal treatment strategies, resulting in adverse consequences for diabetes outcomes in both the short and longer term. In the UK, the National Institute for Health and Clinical Excellence (NICE) has recommended that parents and children be informed that the target for optimal HbA_1c concentrations is values < 7.5%. 16 However, an audit of outcomes in 2002 for children treated in the UK demonstrated that, depending on age, only 14–20% of children cared for in clinical services in the UK achieve these outcomes. 17

The landmark Diabetes Control and Complications Trial (DCCT study) has shown that provision to a group of teenagers and young adults of very high levels of support from the multidisciplinary team to facilitate intensification of their diabetes treatment can produce dramatic improvements in blood glucose and HbA_1c concentrations. 18 After a mean of 6.5 years’ follow-up, the group who received intensification of their diabetes management experienced – by comparison with the control group receiving conventional treatment – a reduction in their risk for the development of retinopathy of 76%, microalbuminuria of 39% and clinical neuropathy of 60%, albeit at a cost of a two- to threefold increase in severe hypoglycaemia. Subgroup analysis has shown similar benefits for the younger participants in this study. Interestingly, even after the discontinuation of the DCCT study when both arms experienced similar HbA_1c concentrations, those who had undergone intensified therapy continued to experience a longer-term benefit of a reduced risk of developing diabetes-related complications, including a near 50% reduction in serious adverse cardiovascular disease event. 19–21

The challenge for paediatric diabetes clinical services, therefore, is how to facilitate patients and their families to make changes in their diabetes management that result in similar improvements in HbA_1c level to those achieved in the DCCT study, with subsequent reduced risks of diabetes-related complications.

Behaviour change

Theories of health behaviour change (e.g. reasoned action theory, the health action process approach) and the research associated with them have clarified the need to look beyond a simple approach to adherence and change based upon the delivery of expert information. 22 As Marteau and Lerman23 have put it, ‘Just telling people they are at risk of developing a disease is rarely sufficient to change behaviour’. Two variables run through many of the theoretical models as predictors of health behaviour change: beliefs about the value of change and beliefs about one’s capacity to succeed (self-efficacy). Thus, for example, the efficacy of theory-based interventions such as cognitive behavioural therapy (CBT) has largely been attributed to their capacity to enhance self-efficacy. 24 Using a skills-based approach to counselling has been found to be effective in a number of fields. 24. 25 So, too, brief interventions have been found to be effective in changing a number of risky health behaviours. 26

A second line of research has focused on how the therapeutic relationship either hinders or promotes motivation to change. For example, an early effort to understand the effective ingredients of motivational interviewing (MI)27 identified a correlation between confrontational interviewing and resistance, and between ‘change talk’ and behaviour change. 28 A meta-analysis of MI29 found consistent evidence for effectiveness in some (e.g. alcohol, drug use), but not in all behavioural domains. Interest in the field of diabetes among young people has also emerged. 30–33 One of the challenges in much of this research, however, has been to clarify exactly what elements of a complex method were used by the interventionists. It does appear that some of the principles of MI can be realised in brief health-care consultations, and that helping patients to clarify for themselves why and how they might change their behaviour (MI) can be more effective than brief advice-giving. 34,35 One recent development has been the first effort to integrate this method with CBT. 36 Put simply, this body of work calls attention to both the direction of consultations about change (towards enhancing coping skills) and the way patients are spoken to (eliciting motivation and solutions from them).

Psychoeducational interventions in diabetes

An NHS health technology assessment (HTA) systematic review of the effects of educational and psychosocial interventions for adolescents with diabetes, which led to the commissioning brief for this study, reported that there were no results from randomised controlled trials (RCTs) of psychoeducational interventions in the UK. 37 However, the review did identify an ongoing study evaluating the effects of MI on behaviour change in teenagers. 32 This trial was based on positive findings in a pilot study in children33 and an RCT involving adults with type 2 diabetes. 38,39 The review commented that small to medium-sized beneficial effects on a variety of diabetes management outcomes have been demonstrated mostly in North American studies. 40 It concluded that there is a need for well-designed clinical trials that recognise the inter-relatedness of various aspects of diabetes management and assess outcomes that are specifically targeted for change, at an appropriate time after the intervention. In particular, the review recommended that such research be developed by a consultation process with stakeholders including patients, their families, health-care professionals (HCPs) and health economists. The commissioning brief for this research project further refined these principles in that effort should be directed towards a generic intervention that does not require delivery by trained clinical psychologists, given their relative scarcity in paediatric diabetes services. 41

Overview of the DEPICTED study

The study described in this report was delivered in two phases. The first phase involved six developmental components required to inform the development of the emerging intervention (health-care staff trained to modify their consultation approach to help them discuss behaviour change skilfully in their patients and families), and was followed by a second phase in which the intervention and training programme were trialled. This overview starts with a brief consideration of the issues relevant at the time to modelling and complex intervention development.

Presentation of this report

The next six chapters present the component studies of the development phase of the DEPICTED (Development and Evaluation of a Psychosocial Intervention for Children and Teenagers Experiencing Diabetes) study. Chapter 2 presents the overarching methodological framework adopted for these studies and Chapters 3–6 report the individual studies. Chapter 7 describes how these were integrated within the intervention and training programme, which are also described in detail, and it also serves to summarise the body of work conducted in the developmental phase. Chapters 8–10 describe the introduction, methods and results of the trial, respectively. The DCE, which explores patient and carer preferences, is described in entirety in Chapter 11, whereas the trial process evaluation (PE) is presented in Chapter 12. Finally, the results from the DEPICTED study as a whole are discussed, with conclusions, in Chapter 13.

Updating the systematic review

In 2001, the NHS research and development (R&D) HTA programme published a systematic review of the effects of educational and psychosocial interventions for adolescents with diabetes mellitus. 37 In summary, this review identified 62 studies, of which 25 were RCTs, mostly conducted in the USA (none from the UK). The mean (pooled) effect size was 0.37 for psychosocial outcomes and 0.33 for HbA_1c with outliers (0.08 without outliers), suggesting that these interventions have small to medium beneficial effects on diabetes management outcomes. The authors concluded that future studies should be evaluated by assessing outcomes that the intervention specifically targets for changes, at an appropriate point in time post intervention to reflect the impact and durability of the intervention. The lack of cost-effectiveness analyses of published studies was highlighted.

When our study was initiated, we undertook an update to this systematic review. At the time of analysis of papers identified, similar structured and systematic review updates of the effectiveness of psychoeducational interventions in children with diabetes were published. 62–64 These published reviews identified largely similar manuscripts and drew similar conclusions to those that we were developing at that time and, therefore, we will draw largely upon their findings.

A further 27 papers had been published describing the evaluation of 24 psychoeducational interventions. 64 As before, routine clinical care seems to produce inadequate metabolic outcomes. Education seems most effective when integrated into routine care, where continued parental involvement65,66 and adolescent self-efficacy is encouraged. Although there was evidence of a methodological improvement in published trials, there was no evidence of improved effectiveness of the interventions. Although psychological interventions seemed more effective in children and adolescents than in adults,63 few studies have investigated the effectiveness of interventions in younger children and most trials remained underpowered to demonstrate modest, but clinically significant improvements in HbA_1c level. An estimated sample size of 360 is required to show an HbA_1c concentration difference of 0.5%, and a sample size of 350 is required to detect a small psychological effect size of 0.3 with 80% power. 64 Most psychological interventions were based around CBT and the limited understanding of the potential of MI in childhood diabetes32,33 was highlighted. 63 No specific psychoeducational intervention could be deemed superior to others. Specifically, there were no interventions that seemed effective when targeting those with poor glycaemic control, and concerns have been expressed62 that targeting only those demonstrating ‘readiness to change’ patterns of thinking may overestimate the effectiveness of certain interventions. 33

Hampson and colleagues37 concluded that agreement was required on appropriate, valid outcome measures for trials of psychoeducational interventions, but little progress has been made in this respect. The need to agree measures that include glycaemic control using common reference methodology, age-validated questionnaires for psychosocial variables and service utilisation and cost measures is clear. Unresolved issues include the relative importance of the content of the intervention as opposed to contact time with the interventionist and whether or not interventions should be combined with other efforts to intensify insulin therapy. 67 Given the increasing importance ascribed to education and the wider number of HCPs, including physicians, providing such education to patients and their families, the importance of understanding the role of self-efficacy, the principles of education and its delivery have been highlighted as priorities for training. 64 Future interventions should be theoretically grounded, with clearly described protocols to allow adequate analysis and reproduction62 and greater priority given to patient preferences. 63

Motivational interviewing

Framework for the methodological approach

The approach of the team to the research development of the intervention mirrored many of the principles they felt could underpin the resulting clinical intervention itself. As a group of experts in the field, we felt we had some ideas that might be useful, but needed to explore how the target practitioners and patients would receive these ideas and what they would find useful.

These questions provided the starting point for the intervention development process, in which the systematic study of the views and experiences of clinicians would be brought to a multidisciplinary group of stakeholders that included parents and children. This stakeholder group would work with the research team to design the guiding principles and structure of an intervention that responded to the needs of all involved. Chapters 3–7 provide an account of this unfolding development process.

Introduction

This section is a description of a survey of practitioners, one of the stakeholder consultation activities designed to elicit information about their experiences of meeting psychosocial needs in clinical practice. The aims of the survey were to understand practitioners’ own assessment of challenges in delivering routine care and their existing approaches to encouraging behaviour change. This information would inform the development of the intervention and the training programme for teams.

Method

A random sample of 112 hospital centres stratified by region was selected from an augmented list of 216 UK hospitals (excluding Northern Ireland) providing services to children and young people with diabetes in the UK. 78 No more than one hospital per trust was selected for inclusion. Consultants (or nominated alternatives) responding positively to an initial postal approach were followed up by telephone interview.

A telephone interview schedule was developed by the research team with additional input from Diabetes UK. Survey domains covered included current and past service innovation and educational approaches, routine care provision, psychological support and clinic characteristics (Box 2).

The survey was anchored on patients at least 12 months post diagnosis. The survey instrument was piloted in six interviews by two members of the research team with four local practitioners. The interview was planned for 20 minutes’ duration, included several open-ended items with standardised probes, and was audio-recorded with respondents’ verbal consent. Two interviewers completed the interviews.

Analysis

Responses to quantitative items were analysed and reported using percentages. Responses to key open-ended items were transcribed, analysed and coded according to emergent themes. A priori categories were not used. Coding of the narrative data was agreed by two researchers (HH and KB) who both independently coded three interviews and then the coding was completed by one researcher (HH) supported by the use of a Microsoft Access 2003 database (Microsoft Corporation, Redmond, WA, USA). Illustrative extracts will be used to support the description of emergent themes with the coding of ‘D’ for doctor and ‘N’ for nurse, followed by their identity number.

Results

Seventy (63%) practitioners responded to the initial approach and 51 clinicians completed the interview, of whom 22 (43%) were from teaching hospitals. Forty-four respondents (86%) were doctors and seven (14%) were nurses. Characteristics of responding practitioners and clinics are summarised in Table 1.

The main responses to the four open questions are summarised in Box 3. The key themes to emerge are described in two sections: Challenges of providing care and Managing behaviour change.

Challenges of providing care

In considering the practitioners’ views of the challenges faced by teams in providing psychosocial care, the dominant theme was the issue of engagement and communication, but within this there were two key areas: the integration of diabetes into everyday life and meeting the needs of different ages.

Discussion

Completing the survey engaged stakeholders in thinking about their services, the challenges they face in relation to providing routine care and their experience of facilitating behaviour change. The high rate of response to this survey demonstrated that this is an area that practitioners recognise as a priority. Respondents outlined the complexity of engaging patients and their families and the importance of communication skills in trying to meet the needs of many different ages, developmental stages and cultural backgrounds within a range of family contexts. They described using a combination of advice, education, listening and shared goal-setting to help encourage their patients to change.

From the responses it would seem that the clinicians were using the three core skills of asking, listening and informing in their communication. There was also implicit recognition that behaviour change is at the heart of the interaction: practitioners described trying to engage patients in making those shifts between the competing demands, yet that process of change was very difficult for the majority to conceptualise or articulate. In considering the most appropriate patient–practitioner interactional model for the intervention in DEPICTED, it had to be effective in addressing behaviour change and incorporate the practitioners’ existing key consultation skills. One model of communication with potential for improving practitioners’ confidence and skilfulness in dealing with behaviour change in routine consultations was to incorporate more of a guiding style into their consultations, encouraging patients to explore their own views of the behaviour and making their own decisions – an approach that has been shown to make change more likely. 42

Although many respondents had received communication skills training in various guises, it would seem that training in communication skills and behaviour change may have been too distant, too general, or not tailored to the context in which they work, to be of use in helping practitioners have a conceptual map of the work and tools available to enable them to function confidently.

The survey contributed significant information to help plan the training programme. It was clear that any intervention had to have a broad application that was flexible enough to respond to the differing needs of a very mixed patient population. It had to facilitate the balancing of the often different priorities of patient, family and practitioner in the consultation process. For practitioners to be able to grasp the relevance of any such training programme to behaviour change, the training needed to be conceptually clear and specific to the context of delivering clinical care within a paediatric diabetes service. By ensuring that the training was more context bound, with the focus on everyday scenarios that have meaning for the practitioners and with the communication skills aspects of training woven into the practice, the aim was to increase the relevance and retention of the information. 55 As well as giving guidance in relation to the development of the DEPICTED study, the findings of this survey underline the importance of the style of training at undergraduate and postgraduate levels across disciplines.

Introduction

The attitudes and experience of professionals in the UK in communicating with children and teenagers with diabetes and their families are unknown. For children and teenagers with type 1 diabetes, consultations are complicated by family dynamics and developmental issues. In the UK, notable attempts to train HCPs in communication skills have occurred in specialties such as oncology and general practice and have met with mixed success. 38,43,79,80 However, there is little published evidence regarding the acquisition and proficiency of communication skills of HCPs in paediatric disciplines. Staff involved in the management of child and adolescent obesity in the USA have reported low levels of self-perceived proficiency in the use of behavioural management strategies, delivering guidance in parenting strategies and in addressing family conflict. 81 This suggests that additional training would be beneficial in improving confidence and skills in these areas. This survey aimed to assess communication experiences, attitudes and training opportunities for HCPs to inform the development of the programme.

Blended learning approaches, which provide a mixture of learning opportunities, have been effective in delivering communication skills programmes. 79,82–84 More recent technological advances, such as CD-ROM (compact disc read-only memory) and web-based programmes, provide a flexible method of education delivery and have been used with some success to teach clinical communication skills. 85 However, such technologies have not been evaluated on a large scale in a multidisciplinary clinical environment in the UK and their potential acceptability to paediatric clinical staff is unknown. Despite the obvious advantages of e-learning (such as the potential to reach large numbers of learners), barriers to the use of e-learning in continuing professional development (such as lack of time and confidence) have been reported and may restrict such developments. 86. 87 However, the use of technology by HCPs in everyday practice is rapidly developing. It is possible that such problems may have been resolved and preferences for training may have moved to embrace such approaches. Therefore, this element of the developmental work also aimed to assess the perceived feasibility of and preferences for various methods of learning among staff working in paediatric diabetes services.

Methods

Questionnaire

The survey covered three broad areas (1) previous experience in communication skills training and its delivery; (2) a scenario-based assessment of attitudes towards addressing different topics in routine consultations; and (3) perceived feasibility of different options for training delivery and skill maintenance. The overall content domain and individual items were developed by a research team comprising psychologists, communication skills trainers and clinical practitioners in paediatric diabetes, among others, and piloted with 11 practitioners working in two paediatric diabetes centres in south Wales. Consultants’ responses to a previous telephone survey of psychosocial service provision for children with diabetes also contributed to the content of the questionnaire.

Data analysis

Data are presented as frequencies, means and medians. Differences in responses to scenarios were analysed using t-tests and analysis of variance (ANOVA). Standard deviations were adjusted to account for clustering of responses within services through inflation by the intracluster correlation coefficient (ICC). 88 Responses from services with just one team member in the sample were excluded from analyses of scenario responses to minimise distortion of the ICC (n = 11). Associations between variables were examined by calculating Pearson’s coefficient. For analyses of responses to all questions on previous experience of training in communication skills and the case scenarios, psychologists and other therapists were excluded from the analysis (n = 14). All data were analysed using Spss version 14.0 (SPSS Inc., Chicago, IL, USA).

Results

Importance and confidence ratings for communicating with patients

The internal consistency of aggregate scores was high for ‘confidence’ ratings (medical issues α = 0.91; psychosocial issues α = 0.83) and reasonable for ‘importance’ ratings (medical issues α = 0.81; psychosocial issues α = 0.69). Internal consistency of aggregate scores was optimised by excluding those topics not falling exclusively into a ‘medical’ or ‘psychosocial’ category, such as a girl’s weight.

For the case scenarios presented, respondents rated both ‘medical’ and ‘psychosocial’ issues as either important or very important to address during routine consultations {mean [standard deviation (SD)] ratings 4.0 (0.68) and 4.5 (0.50), respectively}. Psychosocial issues were given higher importance ratings to address within a routine consultation than medical issues (t = 8.93, p < 0.001). Confidence to address medical issues was high [mean rating = 4.3 (0.66)], particularly among doctors and nurses, but confidence to address psychosocial issues was significantly lower across all disciplines [mean 3.5 (0.75), t = 15.85, p < 0.001; Figure 1)]. The biggest discrepancy between importance ratings for a specific topic and confidence ratings related to the topic of a teenage girl’s low mood (see Box 5). Other issues which respondents rated as ‘important’ or ‘very important’, but had less confidence to address included the impact of parental conflict on a young girl and talking about a teenage girl’s views of living with diabetes. Sixty-eight (27.0%) respondents said they would not consider addressing the impact of parental conflict on a young girl themselves (Table 5).

TABLE 5 - Attitudes to addressing topics in scenarios of routine consultations

SDs are adjusted to take account of clustering within services.

Contributes to summary scores for ‘medical’ issues (Cronbach’s alpha: confidence = 0.91, importance = 0.81).

Contributes to summary scores for ‘psychosocial’ issues’ (Cronbach’s alpha: confidence = 0.83; importance = 0.69).

Topic	Percentage rating as either ‘important’ or ‘very important’				Mean (SDa) important rating (all)	Percentage rating as either ‘confident’ or ‘very confident’				Mean (SDa) confident rating (all)	Percentage stating they would not attempt to address topic themselves
	All	Doctor	Nurse	Dietitian		All	Doctor	Nurse	Dietitian		All	Doctor	Nurse	Dietitian
A teenage girl’s …
HbA1c resultb	75.2	78.0	72.5	74.0	4.1 (0.98)	89.2	93.6	89.0	80.0	4.3 (0.77)	1.2	0.9	0	4.0
Insulin regimenb	65.0	64.2	69.2	59.1	3.9 (1.03)	80.0	93.6	86.8	80.0	4.1 (0.98)	9.3	1.9	1.1	56.2
Low moodc	95.6	98.1	97.8	85.7	4.7 (0.56)	42.8	47.7	46.1	26.0	3.3(1.10)	14.9	18.5	8.9	18.0
Views on life with diabetesc	98.4	98.1	98.9	98.0	4.8 (0.46)	66.8	69.7	76.9	42.0	3.7 (0.93)	5.6	2.7	5.5	12.0
A teenage boy’s …
HbA1c resultb	70.1	72.6	72.5	62.0	4.0 (0.97)	89.6	94.4	93.4	72.0	4.4 (0.80)	1.6	0.9	0	6.0
Insulin regimen (encouraging him to talk about)b	81.9	85.0	85.7	70.0	4.3 (0.79)	87.6	92.6	94.5	64.0	4.3 (0.78)	2.4	0.9	1.1	8.0
Life at schoolc	91.5	90.7	93.4	89.6	4.5 (0.69)	77.5	72.2	89.0	68.0	4.0 (0.77)	3.2	3.7	3.3	2.0
Drinking behaviourc	90.0	87.0	94.5	88.0	4.4 (0.69)	69.9	63.8	81.3	62.0	3.9 (0.82)	6.0	9.3	3.3	4.0
A young girl’s …
HbA1c resultb	76.5	80.5	79.8	62.0	4.0 (0.81)	90.3	92.2	93.5	80.0	4.4 (0.80)	0.8	0.9	0	2.0

FIGURE 1.

Importance (a) and confidence (b) ratings for medical and psychosocial issues. a, t-statistics are for the whole sample.

There were no interprofessional group differences in importance given to addressing psychosocial and medical topics within the consultation, but there were interprofessional differences in confidence, with dietitians expressing less confidence across all issues (F_2,229 = 4.12, p = 0.018; Figure 1). Confidence ratings for addressing both medical and psychosocial issues were correlated with years of experience working in diabetes (r = 0.30 and r = 0.36, respectively, both p < 0.001). A weak correlation was found between importance ratings for addressing psychosocial issues and years’ experience working in diabetes (r = 0.15, p = 0.026). The correlation between importance ratings given to medical issues and years’ experience in diabetes was not significant (r = 0.13, p = 0.059). Those who had received specialist communication skills training, such as MI and CBT, reported slightly higher mean confidence ratings for psychosocial issues than those who had not received specialist training (mean = 3.8 and 3.6, respectively), although this difference was not statistically significant (t = 1.6, p = 0.103). No other differences were found between attitudes towards addressing psychosocial issues and previous communication skills training. There was some clustering of importance and confidence ratings within services, particularly for confidence ratings to address psychosocial issues (ICC = 0.08), indicating a tendency for members of the same team to self-rate in similar fashion. The same was true for importance ratings given to medical issues (ICC = 0.04).

Motivating factors for participating in a communication skills learning programme included helping patients talk about their needs more easily, helping to change patient behaviour and learning skills that can be used in life beyond diabetes care (80.0%, 79.2% and 72.8% of respondents agreed with these statements). A total of 19.6% of respondents expressed finding talking with patients sometimes quite difficult as a reason for participation. Agreement with this statement was correlated with fewer years’ experience working in diabetes (r = 0.15, p = 0.016).

Discussion

Although some professionals had undertaken specialist training in communicating with patients, almost one-quarter of nurses had received no such training and half of all professionals surveyed had received no training since graduating. Confidence among doctors and nurses to address medical issues in consultations involving young people with diabetes was high, but confidence to address psychosocial issues was significantly lower. Given that addressing psychosocial needs is perceived as important by both patients89 and professionals, these low confidence ratings reflect a training need and motivation to learn new skills among professionals working in paediatric diabetes and a gap in current care provision.

It is clear that some practitioners feel unprepared to address psychosocial challenges that are commonly found in practice, and it may be that some feel that it is outside their remit to do so. Referral to psychology services may be an answer for some patients with particularly pressing concerns, but access to such support is limited. 90 In many services, practitioners have little option but to address complex psychological and emotional topics themselves. The clustering of confidence and importance ratings found within individual services may be a reflection of differing ‘cultures’ and variations in the availability of specialist support. Establishing models of care to meet families’ psychological and emotional needs that are applied across services is therefore a priority.

High importance ratings given to addressing psychosocial issues, coupled with low confidence ratings, suggest a role for communication skills education to support routine care. In addition, reasons given by practitioners for participating in a communication skills learning programme demonstrate the clinical challenge of behavioural self-management in diabetes and emphasise the perceived importance of effective communication between family and professional.

Among the strengths of this study was the development of the survey instrument by a team of clinical and research professionals with a particular interest in training, enhancing both the face and content validity of the final survey instrument. Furthermore, the scenario-based assessments were developed on the basis of clinical experience and conceptually driven. This was borne out by the high level of internal consistency for three of the four resulting aggregate scales, with scores exceeding the benchmark Cronbach’s alpha of 0.7. 91 Similarly, associations found with other related variables, such as the positive relationships between confidence ratings and years of experience in diabetes, support the construct validity of these measures. Potential weaknesses of the study include the response rate and coverage of the survey. The sample of respondents may be biased towards professionals who are interested in, or seeking training in, communication skills. Therefore, caution should be taken before suggesting that these findings represent the attitudes and training needs of all staff working in paediatric diabetes. However, given that clinicians from approximately one-quarter of all services in the UK took part in the survey, the sample includes a significant proportion of professionals working with children and teenagers with diabetes in the UK.

What messages are there for training provision in this field from this part of the developmental work? There was support for multiple methods of delivery of a learning programme with monthly team-based learning activities rated as most feasible and support given for face-to-face learning, case reflection, colleague and coach observation, and feedback. Training within teams at regular intervals may prove a valuable method of learning, particularly in context. Given the lack of previous experience of online learning in communication skills, there was considerable support for the use of web-based or CD-ROM materials, although potential barriers – such as lack of time – continue to be reported. 85 Unlike previous findings,85,86 lack of skills was barely reported as a potential barrier to accessing web materials and is a likely reflection of the rapid increase in skills in and use of information technology by health professionals in everyday practice. Given potential barriers such as lack of time, electronically delivered learning programmes must ensure flexible delivery, minimal technical demands of the user, and timely support.

Introduction

This section describes part of the preparation for the development of a clinical intervention to improve communication in clinic consultations, deliverable within the context of routine care by the diabetes team. As part of the development of this intervention, the DEPICTED team used focus group methodology to gather contextual information from children and young people with diabetes (and their parents or guardians) about the way diabetes affected their lives and how they felt the doctor–patient relationship worked for them in clinical encounters.

Methods

Focus group methods were adapted for paediatric settings, using previously published guidelines. 92,93 The discussion framework used in the focus group discussions is described in Box 6. These discussions aimed to enable participants to describe the issues that took prominence in their lives, their hopes and aspirations, how these were identified and dealt with by HCPs in the clinic setting, and what patients and families wanted from clinic consultations.

Six audio-recorded focus groups were comoderated by two non-clinical researchers. Potential participants (parents, children and young people), who were identified by a clinician working in a paediatric diabetes service, were sent information sheets and forms for consent to researcher contact. Same-gender and related-age-range discussion groups were arranged, as recommended by earlier research on conducting focus groups with children. 92 Participants were selected to achieve a range of treatment regimens (two, three or four injections per day – insulin pump), family structures (single- or two-parent families, siblings or parents with diabetes) and coping/treatment adherence (e.g. ‘doing well’ or ‘struggling’ from a clinician perspective). Children (aged 7–11 years) and young people (aged 12–16 years) with type 1 diabetes and their parents were invited to participate.

All potential participants expressing an interest in the study were telephoned a few days before the focus groups for the researchers to introduce themselves, reiterate the purpose and format of the groups, re-confirm their decision to participate and to respond to any questions. A specialist nurse from the paediatric diabetes service, familiar to the participants, greeted them on arrival, and was available after the discussions to answer any medical concerns that may have arisen. The specialist nurse was not present during the focus group discussions themselves. Written informed consent was taken before the focus groups started. All participants received refreshments on arrival and a £10 gift voucher as token appreciation. All parents, those participating and those accompanying their children/young people to the venue, completed an information sheet documenting their own age, occupation, family size, child’s age, duration of diabetes, other family members with diabetes and, for accompanying parents, how well they believed diabetes management was going at that time for their child.

To encourage greater openness, children and young people took part in the groups without their parents present in the room. Basic ground rules (e.g. everyone getting an opportunity to talk, one person speaking at a time) were introduced at the start of the focus groups, followed by an introductory session for the parents’ and young persons’ groups, and ice-breaker games for the children’s groups. In addition, the children and young persons groups used a ‘pick a postcard’ strategy to start discussion about what living with diabetes was like for them, based on their choice of picture. During the focus groups, one researcher assumed the role of the main facilitator, whereas the other provided interim and final summaries, logged comments, wrote bullet points on the flip chart, and handled recording equipment and refreshments.

Group discussions lasted between 66 and 98 minutes, including a short comfort break. The discussions were audio-recorded and notes made by the support facilitator. They were transcribed verbatim and loaded into NVivo 2 (QSR International, Doncaster, VIC, Australia), a qualitative software package. Transcripts were coded and emerging themes were identified by one researcher. Two others then read the transcripts independently and agreed the coding and themes. Themes were further developed and linked through discussion between researchers.

Results

Forty-eight patients and parents were invited to participate initially, of whom 39 returned ‘consent to contact’ slips, and 32 eventually participated in one of the six focus groups held (Table 8). The range of duration of diabetes in participating patients was 18 months to 7 years. The ice-breaker activities generally worked well, but were more difficult to control in the younger age groups, particularly the boys’ group, which was less inclined to focus on the task.

Discussion

A broad set of issues associated with their diabetes will occupy the minds of patients and their parents, which may only be partially addressed by a narrow clinical focus on glycaemic management. These can sometimes cause conflicting expectations of the function of clinic visits by children, parents and health professionals, who all have their own personal agendas, however implicit or apparently modest.

The data from the focus group discussions clearly described children’s experiences of poor communication and marginalisation, anxiety experienced waiting in clinic for blood tests and to see ‘the doctor’. Many children expressed their dislike of being measured or examined by people they hardly knew and the ‘adult’ acceptance of medical jurisdiction was not a paradigm they found easy to accept. Most children (especially the younger ones) relied on their parents taking over. Additional barriers to effective communication include duration and frequency of contact, gender, perceived attitudes to children and adolescents, and the presence of other people in the consulting room. This results in children and adolescents taking a passive role in consultations, reluctant to raise personal or sensitive issues or to ask questions that might reveal poor adherence. 89

There were clear differences in expressed preferences for the typical ‘nurse-style’ consultation over the ‘doctor-style’ consultation, as nurses were seen as more approachable and more realistic in their expectations of patients, and often had a longstanding relationship with patients.

Health professionals are expected to work in partnership with young people94 and their parents, but often lack the communication techniques with which to engage them81 in the discussion and management of their illnesses. This is particularly important when applied to the care of chronic conditions such as type 1 diabetes, where day-to-day self-management involving children and their families is crucial in maintaining optimal control and good clinical outcomes. A recent survey showed that 16% of doctors, nurses and dietitians working in UK paediatric diabetes services had received no training in communication skills, and 47% had had no training since graduating. 95 The balance between achieving biomedical outcomes while acknowledging and listening to children’s daily experiences (referred to graphically by Barry and colleagues96 as their ‘lifeworld’) is often unequal, resulting in suppression of their contextualised accounts and a reduction in their autonomy and engagement. This is especially the case if communication is held mainly with the attending parent(s). This approach ignores and contrasts with the considerable role of children in their own diabetes management. 97 In contrast, skill-based approaches that more actively engage patients in their consultation (e.g. MI) can facilitate behavioural changes and improve glycaemic control in teenagers with diabetes, but it takes time to do and requires specialist input. 32

Clarification and understanding of agreed agendas in the clinic setting at the start of a consultation is a clearly important conclusion from the data presented here. So is the communication style adopted by the HCP. A direct consequence of the insights gained from the focus groups has been the development of an interactive agenda-setting tool within the learning programme of the DEPICTED trial. The aim of this tool has been to ensure that everyone in the consultation can raise issues that are important to them.

Many of the themes developed from the data are familiar, but we have developed them to produce guidance in the form of a ‘conceptual approach’ for professionals working in a paediatric context in the UK (although they might need to be adapted for other cultural settings) (Box 11). This approach emphasises that how the patient is involved in the consultation is as important as what is communicated during it.

Introduction

Major UK funding bodies and Research Ethics Committees (RECs) actively support public involvement in research. 50 The National Institute for Health Research HTA programme has developed an evidence-based approach to involving service users in research and development agenda-setting, which includes approaches to reducing barriers to meaningful participation. 98 Similarly, the James Lind Alliance was established in 2004 to encourage patients, carers and clinicians to work collaboratively to identify research questions. 99 The commissioning brief for the DEPICTED study emphasised the need for key stakeholders, including children and teenagers with diabetes and parents/carers, to be actively involved in the development of the research intervention, not surprisingly, given that the intervention had to be ultimately deliverable within the context of routine care without the need for additional clinical support, and acceptable to all stakeholders. This chapter concerns stakeholder involvement, particularly lay stakeholder involvement, at key stages of the research as members of a SAG. Their involvement is described, and the issues and challenges arising are identified and explored within the literature on service-user involvement in health-care research.

The Stakeholder Action Group

The DEPICTED research team worked with a specially constituted SAG that was to advise on the developing research intervention and on the formal trial evaluating the intervention. The SAG was responsible for reviewing relevant evidence provided by the researchers, considering and advising on developing ideas for the intervention, and guiding the research team about plans for evaluation in the subsequent trial. The group was so named to confirm the active role it had in working with the research team.

Lay representatives included teenagers and young adults with type 1 diabetes, and parents of affected individuals approached through independent support groups outside the catchment area of the clinical researchers. This safeguarded confidentiality and professional relations by avoiding the inclusion of patients and parents known to professional attendees. Parents were invited to attend meetings with their children or on their own. In addition, the research team included the mother of a teenager with diabetes in her capacity as a user representative, and a representative from Diabetes UK was invited.

The professional group was selected to represent key professionals with an interest in children and teenagers with diabetes. Professions represented included paediatrics, specialist diabetes nursing, general practice, child psychiatry, paediatric dietetics, clinical psychology, school nursing and social work.

Setting and context

The SAG met on three full days over 10 months. An independent facilitator – a consultant clinical psychologist not involved in diabetes care – was employed to run the meetings, and met with the research team before and after each SAG meeting to clarify roles and responsibilities and review events, respectively. Various research team members facilitated small-group sessions and hosted other plenary sessions. Each SAG meeting was constructed around plenary and parallel small-group sessions. Plenary sessions included presentations by the research team about study activities and developmental concepts of the research intervention. The composition of groups in the small-group sessions varied, for example mixed stakeholder groups addressed ‘What makes for a successful consultation?’, whereas lay and professional stakeholders were separated to consider ‘Choices about service delivery’. Lay stakeholders were further divided into parent and teenager groups to discuss the development of separate patient and parent/carer questionnaires for use in the subsequent trial. Additional materials were made available to view during breaks and over lunch (e.g. scientific and ‘meet the team’ posters). In all meetings, presentations included examples of the developing intervention and made use of both audio and video materials encouraging further input of ideas and comments.

All lay stakeholders provided written informed consent at the start of each SAG meeting. All stakeholders received reimbursement of travel expenses, and lay stakeholders also received £30 vouchers for each meeting they attended. Newsletters were used between meetings to update stakeholders about study progress. At the final meeting, stakeholders were asked whether or not they wanted to continue being informed about the study as it progressed into its trial phase.

Evaluation of stakeholder action group meetings

Stakeholders’ views were audio-recorded, and the data transcribed and analysed following each SAG meeting. Summary written notes taken by researchers observing the group sessions were made available to stakeholders during the day and were subsequently formally transcribed as additional data. Each SAG meeting was evaluated using an anonymous stakeholder feedback form that nevertheless identified whether or not the respondent was a patient, parent or professional. This form addressed expectations about the day, whether or not these had been met, what they did/did not enjoy, their views on information provision, meeting format and practical arrangements. Suggestions for how future SAG meetings could be modified were sought. The evaluation form for the third meeting asked how stakeholders would like to continue being involved in the study.

Outcomes

Between 13 and 17 lay stakeholders (teenagers and adults with diabetes, and parents) and 10 or 11 professional stakeholders attended each meeting. In addition, there were between 13 and 15 research team members present.

How the stakeholder action group influenced the design of the research intervention

Three key outputs arose from the SAG meetings, all of which were implemented in the trial phase of the DEPICTED study:100

1. The Talking Diabetes course. 57

2. The shared agenda-setting tool (3T: TimeToTalk). The concept of a shared agenda-setting tool arising from the focus groups with lay stakeholders was reinforced during SAG meetings, where lay stakeholders helped the research team with choices related to the tool design (Table 9). Advice from both lay and professional stakeholders was used to develop guidelines for subsequent use of the tool in trial centres. Finally, at the third meeting, a naming competition was held for the tool in which all stakeholders and researchers had the opportunity to suggest and select names for the new tool (3T).

3. A DCE questionnaire (see Chapter 11).

Discussion

The requirement by the National Institute for Health Research HTA programme to actively involve key stakeholders, including children and teenagers with diabetes and their parents/carers, reflects the increasing focus on patients and the public being involved at all stages of the research process. 101,102 Kirby103 proposes that service user involvement can occur at three different levels: consultation, collaboration and user-controlled research. Service-user involvement in DEPICTED was primarily at the second level – collaboration – and was viewed as a reciprocal partnership between stakeholders, including service users and researchers. In DEPICTED, stakeholder involvement resulted in major contributions to the design of the research intervention, including reassurance to the research team of the intervention’s utility by ensuring that practical aspects proved acceptable to practitioners, patients and parents. Furthermore, stakeholders’ identification of the need for, and contribution to the design of, an agenda-setting tool (3T), and their specific contribution to the patient preference questionnaire (DCE), have been shown to be valuable given their successful piloting and the subsequent central nature of 3T to the intervention. Involving stakeholders was not simply a philosophical driver in this process; it also directly addressed the validity of the research being delivered. 104,105

These findings confirm that research proposals may benefit from service-user involvement at an early stage in their development rather than simple identification of the research topic and dissemination of research findings, which are the levels of service-user involvement in research most commonly reported106 and are in contrast to the conclusions of a recent consensus study. 107

A survey in 2007 concluded that only a small proportion of NHS researchers were actively involving service users when evaluated against eight consensus-derived indicators of successful service-user involvement. 107,108 As recognised by Barber and colleagues,108 this is partly due to the evaluated projects being undertaken at an early stage in the development of policies on service-user involvement. Nevertheless, although principles for successfully involving service users in health research have been clarified by Telford and Faulkner,107 there is less reported evidence of actual user involvement in NHS research,107,108 although this deficiency has been partly addressed in a formal evaluation of patient and public involvement in the UK Clinical Research Collaboration. 102 The lack of reported evidence may be due to the fact that service-user involvement in research is patchy and inconsistent. 109,110 It is still in its relative infancy, with many practical, ethical, moral, methodological and philosophical questions unanswered. 110

There is lack of clarity concerning the extent to which service users can influence and benefit the research process and low levels of consensus about what it means to involve service users successfully in research. 108,111 The work by Telford and Faulkner107 in identifying clear and valid principles to guide good practice, and the subsequent survey by Barber and colleagues,108 are important developments in an under-researched area, but raise questions regarding how ‘successful’ involvement can or should be measured. Furthermore, the eight indicators relate only to process and do not attempt to measure the impact of successful user involvement upon research outcomes. This concern is raised by other researchers working in the field of involvement, who purport that the lack of an evidence base concerning the impact of involving service users can mean that this endeavour is seen as relatively low status and labelled as an ‘add-on’. 112 They further suggest that the existence of a strong evidence base would significantly contribute to the ‘business case’ for involvement, encourage more general recognition and help protect continued funding.

The experience of the DEPICTED research team demonstrates that stakeholders can make a significant contribution to the design of a complex research intervention, even at relatively early developmental stages. This finding provides evidence that involving lay and professional stakeholders can produce relevant and valid interventions that benefit substantially from their combined experiences. This is consistent with outcomes reported by others investigating the benefits of involving people with diabetes in research. 113 Furthermore, the experiences of the DEPICTED research team, and of the lay stakeholders involved in the study, seem to conflict with the belief that one of the fundamental barriers to the empowerment of service users is researchers’ fear of losing their power and status as ‘experts’. 109 Of course, it is important that equality of power is considered at the planning stage of research. For example, in DEPICTED, an adequate number of service users in the SAG balanced the influence that professionals might exert on the process,114 particularly when a question (e.g. what are the characteristics of a ‘successful’ consultation?) potentially posed a tension between lay and professional viewpoints. The nature of stakeholder involvement reported here strongly suggests that lay stakeholders and researchers can work together as experts, but in complementary dimensions of the research project: expertise by experience and expertise by profession. 115

An indicator of successful service-user involvement in DEPICTED is that despite travel and time commitments no collaborating young person or parent withdrew from the study. This could reflect their inherent interest in the topic under investigation or perhaps the value they placed on being able to influence service improvements, a significant motivating factor for people involved in research. 107,116 It could be argued, therefore, that the lay stakeholders felt empowered through their involvement at this level in the research process. They had views about their experiences of clinic consultations and identified strategies to improve existing services. These views were listened to and directly informed the development of the 3T and DCE questionnaire, as well as strongly influencing the design of the research intervention and the subsequent trial design, all of which confirmed for lay stakeholders the value and purpose of their involvement.

There were specific issues related to working with families. The decision to include parents of participating teenagers was made partly for logistical reasons (parents needed to transport/accompany their youngsters to the meetings). In most cases, teenagers were accompanied by two parents, resulting in excess numbers of parents contributing to the discussion groups. Furthermore, it became clear that teenagers should be in separate discussion groups from their parents to promote their active engagement, to allow them to step outside their usual generational roles and to disclose information that they might not wish to share with their parents. This arrangement proved successful, allowing teenagers to become more vocal within the groups, and is an important lesson for research teams proposing to work with young people and their families in the future.

Although problems were not encountered, it is important to anticipate tensions when research is steered by service users (collaboration) as opposed to seeking their opinion (consultation), as occurred, for example, in the naming of the agenda-setting tool. Although consensus was reached with little difficulty, this process could have proved problematic if there had been disparity between the research team’s choice of name and that of other stakeholders. Greater transparency in decision-making, including clarity about individual roles, should ensure that all can make a valid and recognised contribution. Establishing a consensus-based ‘terms of reference’ for the group at the outset would be a useful future strategy.

Although guidelines for service user involvement exist in terms of ethical and practical issues, there is a dearth of research evaluating this phenomenon, and different challenges inevitably arise when service users are involved in research as active, rather than passive, participants. 103,117,118 One such issue, particularly when health research focuses on a specific condition such as diabetes, concerns the ability of service users to see beyond their own experiences to view the ‘bigger picture’. Although service-user involvement could be used as a platform for airing particular grievances about service provision, such experiences are also part of the value of public engagement. 119 In DEPICTED, some lay stakeholders attended poorly resourced diabetes clinics and their initial contributions were understandably focused by their own experiences of the adverse effects of such limited resources. However, with a greater understanding over time of the purpose of their involvement, assisted by careful facilitation of discussion groups whose leaders were aware of this issue, the lay stakeholders were able to consider the implications of the research and its effects on patients in a wider context other than their own clinic, which significantly increased the value of their contribution to this part of the research project.

Stakeholder involvement is more broadly represented in this study by inclusion in the research management team of a mother of a teenage patient who participated in a number of roles throughout the study including cofacilitator in SAG meetings. Despite her enthusiasm and commitment, it is important to recognise that such an unfunded contribution requires her to prioritise competing commitments with consequent risks of guilt at being unable to respond to overwhelming numbers of requests from the research team. In retrospect, although her involvement was not tokenistic, it would have been better to have more than one service user at this level of involvement. Consideration should also have been given to including adequate funding of her time commitment in the grant proposal, which others suggest is important in maximising the potential of user involvement in research. 119

Reflections on the process

The research team have learnt important lessons for the future (Table 12). Such engagement has to be well planned and resourced (note that up to 15 researchers worked on each day). Meeting formats should vary to accommodate a range of stakeholder interests and preferences, and to suit the task requirements. We found an external facilitator especially helpful as he provided a challenge to the potentially insular perspective of the research team, helped the research team clarify their objectives for each SAG meeting and had expertise in managing mixed group dynamics. Support for the stakeholders (e.g. exploring and providing for their requirements, providing plenty of information before and during meetings) also seemed to be successful and acknowledged. What we would do differently includes initial eliciting of role expectations from both stakeholders and the research team, which could be formalised in a jointly developed terms of reference. This would include clarity about responsibility for decision-making and safeguarding all stakeholders’ valid contributions. Integration of all stakeholder groups should be addressed, perhaps by an explicit initial exercise, and supported thereafter by adequate opportunities to mix. Finally, working with families requires even further consideration to enable optimal involvement and experience for both children and their parents.

Summary

Service-user involvement in health-care research is increasingly being required by funding bodies. However, the evidence base concerning the impact of involvement, particularly on research outcomes, is minimal. This is an important issue that needs addressing if involvement in health-care research is to achieve higher status and attract continuing funding. This chapter describes the active involvement of lay and professional stakeholders in the developmental stage of DEPICTED. Challenges encountered during the course of the research are identified and reflected upon. Importantly, it is also clearly demonstrated how stakeholder involvement positively affected the study outcomes: that a research intervention was successfully developed only because stakeholders, including teenagers with diabetes, were actively involved.

Introduction

Framework for DEPICTED study intervention development

The framework for developing the DEPICTED study intervention is outlined in Figure 2 and summarises the empirical work planned in the first phase of the study. Existing theoretical approaches and the relevant evidence base have already been described, and the figure therefore focuses upon the modelling process, which is the third substantive component within the ‘Development’ element in the MRC guidance framework for complex intervention development. 120

FIGURE 2.

Intervention development model.

The model describes the three phases that proceeded from an initial consultation process with relevant stakeholders, through to a detailed development stage during which time the intervention starts to take definitive form and, finally, a stage of training delivery and piloting. Research activity within and across each phase is described in the model. The model further identifies which stakeholder group (practitioners, the formally constituted SAG, service users) and research team contributed to each phase and activity. A time line is indicated, although it was also expected that the overall process may not actually occur in such a well-prescribed linear fashion. 120

The research team’s role in addition to co-ordinating and undertaking the component activities was to work with a SAG to define the intervention and develop the training programme that would deliver it in the trial. This is indicated in the detailed development phase in Figure 2 and proceeds into a third phase at which point the intervention was confirmed and training programme finalised. The work of the SAG is more fully described in Chapter 6. Essentially their collaborative role was to provide input and guidance to the research team at key development stages of the intervention. Meeting at three full-day workshops over 10 months, preliminary ideas and outline strategies were presented and discussed in a mixture of small- and large-group formats. Guidance was provided from an overarching ‘in principle’ level through to more detailed input, for example about the design of intervention materials. Although the workshops and intervening contact were organised by the study management team, an independent facilitator was responsible for on-site co-ordination.

Brief description of developmental activities

Description of the Talking Diabetes intervention model

The intervention model is described in Figure 3 and consists of the 3T agenda-setting tool, a menu of strategies to support behaviour change and a conceptual framework that identifies the importance of flexible shifting between consultation styles, with an emphasis upon guiding. Agenda-setting helps patients/families prioritise their concerns prior to the consultation and all parties to agree what to cover in the consultation. It aims to promote parity in deciding what to talk about, to more fully engage patients in their consultation and make them more active partners in the management of their diabetes. Practitioners can select from a menu of strategies that can be used when behaviour change is a possibility. ‘Pros and cons’ may be most useful when a patient appears reluctant to consider a particular change. ‘Importance and confidence’ can help explore a patient’s apparent ambivalence about the ‘why’ and ‘how’ of changing behaviour. Both strategies can help indicate whether or not it will be productive to continue discussing behaviour change within the consultation or move on to other topics. Where patients are ready to plan for change, shared goal-setting through brainstorming encourages them to be active partners in planning and reviewing progress. The value of each of the three consultation styles for different consultation demands provides an overarching framework. All may be appropriate and skilful practice may be defined by flexible shifting between styles. Guiding may be most effective for behaviour change and the use of open questions and effective summarising can support such a consultation focus.

FIGURE 3.

Talking Diabetes intervention model.

Description of the Talking Diabetes intervention and learning programme

The shared agenda-setting tool: 3T

The tool consists of an A4-sized gummed pad of 28 agenda sheets with images of children and young people, and encircled discussion topics that vary by sheet in a sequence of four sheets (Figure 4). The colourful photographic images portray individuals of different gender, ethnicity and apparent mood. Entitled ‘I think I’d like to talk about …’, the sheet offers plenty of blank space and blank topic circles for patients to add their thoughts. The pad’s inner sleeve provides a rationale and instructions for use and a diagrammatic example of a completed sheet. The pad is presented in a rigid plastic folder (Figure 5) of matching design, which incorporates insert pockets (e.g. for storing papers and pen holders).

FIGURE 4.

Sample agenda sheet from 3T.

FIGURE 5.

Front cover of the 3T folder.

Piloting the learning programme

Piloting took two forms. First, a group review of online training materials was conducted with a selection of practitioners and researchers. Secondly, the two face-to-face workshops were piloted in full with members of a single paediatric diabetes team from south Wales. All five trainers were involved in the pilot and, within each workshop, rotated between acting as facilitator and observer. Piloting was a learning opportunity for the trainers; it enabled evaluation and refinement of the teaching strategies (e.g. use of role play) and materials, contributed to developing a trainer’s manual and allowed review of learners’ experience of the online learning. The trainers met in planning and review meetings before, in between and after the two workshops. Feedback from piloting indicated the need for more theory about behaviour change to enhance programme credibility, the need to emphasise the background development work to the intervention and the need for more theory within the first workshop in particular. The role play had worked well. Problems with the online software were reported, with some learners being unable to download the player software that was required to run video components. It was therefore planned that learners would be provided with minimum running requirements and encouraged to access the programme sufficiently in advance of the workshops to allow remedial support. Two further pilot workshops were run following such modifications with a second professionally heterogeneous group of practitioners to confirm the validity of such changes and to provide further experience for the trainers. The balance within each trainer pair (clinician and psychologist) was intended to ensure clinical and theoretical credibility and the value of this appeared to be borne out in practice.

Discussion

A process of complex intervention development has delivered a theoretically driven intervention and associated learning programme for professionals working in paediatric diabetes teams. Development has included modelling intervention and programme components and assumptions within a planned iterative process that has involved a broad multidisciplinary research team and a collaborative group of lay and professional stakeholders. The intervention seeks to improve outcomes for patients by changes in their behaviour facilitated by a modified consultation approach that emphasises engagement and a guiding style supported by specific consultation strategies.

Emergent messages from across the developmental phase studies included the challenges for practitioners in engaging with families – their awareness of the importance of psychosocial issues in helping families, but a lack of confidence in addressing this. Current training opportunities for practitioners have not adequately provided skills that may be useful in negotiating self-management with families, but practitioners were also concerned about not adding to their existing full schedule of work. A lack of engagement by young patients in their own clinical care was borne out by observational findings and by the reports of children and families themselves, who felt marginalised by a clinic agenda that does not fully recognise the concerns of children. Children reported that demands on them were not realistic and not sufficiently positive.

How have phase I studies contributed to intervention and learning programme design?

The development of the Talking Diabetes intervention has been guided by the existing empirical evidence for educational and psychosocial intervention in young people with diabetes. This has resulted in an intervention that can be applied in routine consultations by all practitioners engaging with families and which, beyond the Talking Diabetes training, does not require a specialist (psychological) background or extended consultation time. The intervention would provide the practitioner with a menu of strategies and skills to support patient self-efficacy and engage patients in their own care. A flexible approach to implementation is core to the design and is essential for the intended application across diverse settings, patients with differing requirements and practitioners with varying experience and confidence.

The conceptual model clarified that the intervention is attempting to address not the more severe end of problems that practitioners encounter in clinic but the common clinical challenges found in practice. The learning programme was designed with that focus and uses everyday practice as its framework. Material used in the e-learning was drawn from real-life observation and, in particular, feedback from patients, carers and practitioners. It was felt important that learning could not be all remote, hence the face-to-face workshops would encourage teams to attend rather than individual practitioners. We recognise that this then makes a substantial demand upon practitioners and would be an important component of PE in the trial.

Developing a shared consultation agenda is a good example of how evidence has been integrated in our approach. Previous experience by the research team in the value of this broad approach was influential, but it was also clear from the qualitative work with families that engagement and the dominance of a clinical agenda was problematic. Along with a commercial design company, the stakeholder group of both professional and lay members helped the research team develop their design and implementation of ideas. Piloting showed the intuitive appeal of the approach to both patients and practitioners. The blended learning programme foregrounded the approach in both its e-learning and practical components.

How have phase I studies contributed to trial design?

The design of the intervention and of the professional training programme had clear implications for the design of the trial. Although the primary trial outcome would be a measure of glycaemic control, we would expect the intervention to be effective by a process of initial behaviour change by practitioners leading to an attitudinal shift in patient and families, and a subsequent change in behavioural self-management. It would therefore be important to assess the extent to which these impacts are observed in practice.

First, practitioner engagement with the training itself would be assessed in terms of both contact and response. Key questions would be to what extent do practitioners attend and engage with the training offered. Furthermore, to what extent could we observe the intended behaviours by practitioners in actual practice – would the guiding style be evident in practice and would the strategies such as agenda-setting be implemented as intended? More broadly, what would be the costs and operational requirements of this intervention and training – especially important if this is to be broadly rolled out into routine NHS practice? These issues would be addressed as part of the PE.

Second, what evidence from patients and carers would support the hypothesised pathway of effect? Agenda-setting seeks to better engage families, and the strategies are intended to enhance confidence and self-efficacy among patients who are considering changes in self-management. The detailed design of the trial would need to include measures of process and outcome that can shed some light on the black box of change.

Summary

The Talking Diabetes intervention places patients at the heart of their own consultation, aims to engage them in their own health care and supports all members of the health-care team with behaviour change strategies that can be flexibly deployed in routine clinical encounters. The detailed evaluation of the intervention in a formal trial of effectiveness and cost-effectiveness is presented in the following chapters.

Intervention development

The current trial

Consideration of appropriate experimental design for evaluation within a given context is a key feature of complex intervention development. 120 When evaluating population-level interventions, individually randomised trials are often inappropriate owing to likely contamination of the control group and biased estimates of effect size. 121,123 A cluster, randomised design minimises this source of contamination and is therefore the most appropriate method of evaluation for the current, clinic-level intervention.

Empirical and consultative work during the intervention development phase helped formulate and operationalise the Talking Diabetes intervention (described in Chapters 2–7 of this report). The development stage provided evidence that the intervention is feasible for teams managing care, and is acceptable to patients and carers. A RCT was therefore needed to test its effectiveness. The aims, objectives and methodology used in the trial are described in the following chapter.

Trial design and objectives

Trial design

The study was a pragmatic, cluster RCT (Figure 6 – an overview of trial design). Twenty-six teams were randomised to receive training at the start (intervention group) or the end (control group) of the 1-year study period. Multicentre approval was granted by Berkshire REC (07/MRE12/9) (see Appendix 1) and site-specific approval was granted by local RECs at all trial sites and all participating acute trusts.

FIGURE 6.

Trial design.

Trial procedures

Primary and secondary outcomes

The primary trial outcome was glycaemic control, assessed at the individual level using HbA_1c value. Secondary trial outcomes included QoL, other clinical (including BMI) and psychosocial outcomes (assessed at participant level) and cost (assessed at clinic level), and are detailed in the following sections.

Statistical methods

Participant flow

Of the 30 diabetes centres formally approached to participate, 26 centres were recruited into the DEPICTED trial (Figure 7). Half were randomised to the intervention arm and the other half to the control arm, balanced by list size. Each centre was asked to recruit a minimum of 30 patients in order to achieve our desired sample size. Of the 1673 eligible patients, 1262 were approached. Control centres recruited a total of 334 patients and intervention centres recruited 359 patients, totalling 693 subjects. Baseline HbA_1c measurements were obtained for 356/359 (99.2%) subjects in the intervention arm and 333/334 (99.7%) in the control arm. At 12 months’ follow-up, HbA_1c measurements were obtained for 342/359 (95.3%) subjects in the intervention arm and 318/334 (95.2%) in the control arm.

FIGURE 7.

DEPICTED centre and participant flow chart.

Response rates for baseline questionnaires were 83% for patients and 82% for carers and dropped to 64% and 65%, respectively, at the 12-month follow-up. Numbers completing individual outcome measures are given later in this chapter (see Tables 24–31). Response rates for the DCE questionnaire were 65% for patients and 63% for carers.

Recruitment

Centres were recruited between 30 August 2007 and 2 April 2008, and patients between 30 January and 25 September 2008.

Baseline data

Table 18 summarises baseline demographic information for control and intervention groups. Variable counts are due to missing item data that were assumed to be missing at random.

Baseline data indicate that the randomisation achieved adequate balance for all demographic variables including the primary outcome, HbA_1c levels. There were slightly more males in the intervention arm, but adjusting for gender in the primary analysis did not influence the result.

Cluster-level balance was examined for patient demographic data. Summary data for HbA_1c levels, age, age at diagnosis and gender of patients are given in Table 19 and indicate adequate balance across centres. There was variation in the number of patients recruited by each centre. More centres in the intervention arm achieved the minimum recruitment targets than in the control arm. As both the number of centres and patients required were originally inflated to account for dropout rates that were not observed, these variations do not affect the overall power of the study.

Numbers analysed

All analyses were carried out according to the intention-to-treat principle and all centres and participants were analysed as randomised. All 26 centres (13 in each arm) were included in all the primary and secondary analyses. For the primary outcome, 657 patients had HbA_1c measurements at both baseline and follow-up and were included in the analysis. Sensitivity analysis was carried out using routine clinic HbA_1c measurement to replace missing central laboratory levels where possible (HbA_1c, if measured, taken within 100 days of baseline or 12-month follow-up). All four missing baseline values and 7 out of 32 missing follow-up HbA_1c levels were included for this analysis. For each of the secondary outcomes, numbers included are given (see Tables 24–31).

Checks for bias

As there were missing HbA_1c data for the primary outcome, baseline summary patient demographic data were also tabulated for the group that had complete HbA_1c data at baseline and follow-up compared with those with missing follow-up HbA_1c data. Table 20 indicates that those patients missing follow-up measurements had slightly higher baseline HbA_1c levels, had slightly lower BMI and were more likely to be female.

In order to reduce allocation knowledge bias, unblinded cluster randomised trials should aim to recruit all subjects before the allocation of the centres is revealed. In practice this was not always possible. In DEPICTED, 213/693 (30.7%) of subjects on centre eligibility lists were approached and consented prior to revealing allocation to the centres. Baseline data for the groups recruited before and after revealing allocation to the centres for training purposes are given in Table 21. There was no evidence of allocation knowledge bias for patient demographic data.

Primary outcome

The distribution of the HbA_1c data was examined and was slightly positively skewed. A natural log transformation was performed for multilevel regression analysis. Summary data have been tabulated in the original scale for ease of interpretation (Table 22). The HbA_1c levels in both arms increased by a similar amount from baseline to follow-up. The intervention effect in the log scale can be interpreted as percentage change and it can be seen that there was a 1% increase in HbA_1c levels in the intervention arm compared with control, which was not statistically significant. The addition of HbA_1c data from routine clinic visits to replace missing central laboratory values did not alter the result (see Table 23). Although gender was significantly associated with follow-up HbA_1c levels, adjusting for age and gender did not alter the results and there were no significant interactions between intervention arm and age or gender. These data are also shown in Table 23.

Secondary patient outcomes

Validation was carried out on all secondary outcome scores. Factor analysis indicated that for all outcome scores individual items contributed to a single construct in each case. A table listing the Cronbach’s alpha statistics is given in Appendix 3. The distributions of all patient and carer secondary outcomes were examined. A degree of negative skew was observed for some of the scores, including Health Care Climate Questionnaire (HCCQ) and Diabetes Continuity of Care Scale (DCCS) and ‘Importance’, whereas the Patient Enablement score was slightly positively skewed. The degree of non-normality was within the limits of the methods used and all scores were left untransformed for multilevel analyses.

Secondary carer outcomes

The scores for the carer data have been calculated as for the patient scores, with the same ranges and direction of effects and are interpreted in the same way.

Exploratory analyses

Attendance data were examined via CRF completion rates and it was found that in the control arm 11/334 (3.3%) patients did not attend at all, whereas in the intervention arm 4/359 (1.1%) did not attend any clinic sessions [difference and 95% CI 2.2 (–0.1 to 4.8)]. The mean (SD) number of clinic visits during the period of the intervention was 3.5 (1.1) for both the intervention arm and the control arm. In order to investigate any possible dose effect, the number of clinic visits was added to the multilevel model for the primary outcome. The number of visits was not statistically significant and there was no significant interaction between number of visits and trial arm.

Direct costs: DEPICTED training

The costs of training 79 trainees across 13 intervention sites are shown in Table 32. The total cost was £170,895. Of this, £46,377 (£3567 per site) was preparation and delivery costs incurred by the DEPICTED team.

A breakdown of costs incurred by intervention sites is shown in Table 33. The mean (SD) cost per site was £9575 (£4831). The number of trainees per site varied between 3 and 12. The mean (SD) cost per trainee was £1614 (£463). The final column shows the cost per site including the cost of preparing and delivering DEPICTED training (£3567 per site). The total mean (SD) cost per site is £13,146 (£11,698). The number of staff per site and the degree of their engagement with training (both entering and completing) were factors that varied according to site. Variations in clinic list size would be reflected in differing number of staff across sites available for training. Practitioners who completed all modules of training would have incurred more cost than those who maybe did not.

Table 34 shows the annuitised training cost per site and per eligible patient. The latter figure, which varied from £14 to £71 (mean £49, SD £15), together with the cost of one 3T agenda-setting tool (£18.04), represents the direct per-patient costs of DEPICTED.

Indirect costs: patients’ use of NHS resources

Data on indirect costs were from CRFs. The number of CRFs completed reflects the number of clinic visits post recruitment during the follow-up period, which varied between 1 and 6. Patients’ use of NHS resources is the sum across all CRFs.

Data were bootstrapped (1000 replications) taking account of clustering effects at the centre level. Table 35 shows the mean (SD) number of contacts by group. There was virtually no difference in number of (post-recruitment) clinic attendances between patients in intervention and control sites. For the remaining variables the low means (apart from contacts with nurses on the diabetes team) are due largely to most patients having zero contacts for that resource item. The intervention group had significantly fewer contacts with community/GP nurses (p = 0.01) and more home attendances by ambulance crews (p = 0.05).

Mean (SD) total costs by study group are shown in Table 36. The first four rows show patients’ total NHS resource over 12 months, including clinic visits. The amortised per-patient costs of DEPICTED training (£35) and the cost of the 3T tool (£18) are added for patients in the intervention group.

Total costs for the intervention group were higher than for controls, but the difference was not statistically significant (p = 0.10; 95% CI –32.22 to 402.14).

With regard to non-NHS costs, a comparison of cost of carer time off work in relation to their child’s health showed no significant differences (intervention mean £100, SD £273 vs control mean £86, SD £221; p = 0.61). Similarly, there was no difference in patient’s time off school for any health reason (intervention mean 3.6 days, SD 5.4 days, vs control mean 3.9 days, SD 7.5 days; p = 0.73).

Sensitivity analyses

Future work

The economic evaluation plan for DEPICTED included a cost-effectiveness analysis assessing costs against the primary outcome (HbA_1c levels). However, the results showed the difference in the primary outcome to be close to zero and not statistically significant. As DEPICTED training costs applied only to the intervention arm, the control arm had lower overall costs. The absence of a statistically significant difference in effect, however, is no longer a justification to adopt a cost-minimisation analysis approach and a cost-effectiveness plane should still be produced. 138,139

In the analysis reported above we used a bootstrap technique to account for any cluster effects on costs. Production of a cost-effectiveness plane when there are cluster effects poses methodological difficulties and a number of different analytical techniques are available. Bachmann and colleagues140 compared these techniques, but their own data had relatively good characteristics (a balanced cluster with 50 observations available for each cluster). Before producing a cost-effectiveness plane, the DEPICTED data will be used to further examine the performance of these techniques, while at the same time dealing with challenges posed by the near-zero effect, which is conventionally the denominator in the cost-effectiveness ratio.

Introduction

A recent methodological development to elicit patient preferences is known as ‘stated choices’. Stated choice studies in health care describe services in terms of collections of attributes. 132 By varying the levels (ranges) of these attributes, different ‘treatment profiles’ are created. Patients are asked either to order (ranking experiments) or to choose between a set of choices (DCEs) to infer the relative importance of different attributes. The researcher can manipulate attributes and levels to study how patients react to actual treatment options or processes of care. However, choices are made using hypothetical scenarios and may not reflect actual behaviour. It is therefore important that the treatment profiles are realistic and rigorously developed to permit valid inferences about behaviour. 141–143

Modifying the clinical encounter to engage patients and families better, to enhance their clinic experience and to support self-management are aims of the Talking Diabetes intervention. In this DCE we aimed to formally identify the key components sought in a routine consultation in paediatric diabetes and determine the relative importance that patients and carers attach to these components.

Furthermore, the DEPICTED study offered an important opportunity to investigate two methodological issues in the use of DCEs. First, it allowed exploration of the feasibility of using a DCE with young respondents and, second, it enabled comparisons of patients’ and parents’ preferences. Only a few studies have included young participants144,145 and fewer have compared health professionals and carers. 146 However, no study was found that compared children’s and carers’ preferences.

Methods

A DCE involves five steps: (1) identifying attributes and their levels; (2) designing the experiment (identifying the choice sets to use); (3) piloting the questionnaire (e.g. to address cognitive burden); (4) administering the questionnaire; and (5) analysing and interpreting data. 147 Steps 1–3 are fully described in Appendix 4, and more briefly summarised in this chapter. Step 4 is reported in Chapter 9 of this report and step 5 is described in full below.

Results

The DCE questionnaire was administered to all the carers (n = 693) participating in the study and to patients aged 12 years and over (n = 355), with 435 and 230 questionnaires being returned, respectively, giving a response rate of 63% and 65% for carers and patients, respectively.

Discussion

A rigorously developed DCE questionnaire with five categorical attributes of three levels each modelled both patients’ and carers’ preferences for the clinic consultation. In terms of consultation style, reportedly experienced patients in intervention sites report that their doctor addresses both them and their parent more frequently than do patients in control sites. Patients in intervention sites also report that their doctor alone sets goals more frequently than do patients in control sites. The former is consistent with the trial results, but the latter is not what would have been expected if trained practitioners are attempting to share decision-making with patients. Neither, however, reached statistical significance at the conventional level.

The ranking exercise showed no significant differences between control and intervention groups. However, the ranking differed between patients and carers, with patients listing ‘who sets the goals’ and ‘continuity of care’ as most and least important, respectively, whereas ‘continuity of care’ and ‘who the doctor talks to’ were listed by carers as the most and least important, respectively.

All attributes and levels were statistically significant, and both patient and carers wanted the other to be part of the consultation as well as part of the goal-setting. Trial allocation was not associated with any difference in either patient or carer preferences.

Strengths and weaknesses of the discrete choice experiment

A strength of this DCE study was that it was conducted alongside an RCT, which enabled assessment of any preference changes produced by the intervention. It also offered an opportunity to investigate the feasibility and acceptability of this type of questionnaire in a young population. The complete data sets supplied by over 90% of responding patients, coupled with the fact that only one respondent constantly chose the constant scenario, support the use of DCE questionnaires with this cohort of participants.

The response rate was good for this type of questionnaire. However, both groups were less representative of the cohort at baseline, which needs to be accounted for when interpreting the results.

Only a small number of DCE surveys have been conducted in the field of diabetes (including three with UK participants). 149–151 One investigated doctors’ preferences for a report card,152 whereas the others focused on patient preferences for alternative routes of insulin administration,150 patients’ willingness to pay for insulin delivery systems149,153,154 and treatment preferences and medication adherence. 151 However, none of these studies related to the clinic consultation and none of them was limited to people with type 1 diabetes or included children.

Future work

The current data set will be used to investigate how other reported psychosocial outcomes relate to DCE choices. It will also explore the feasibility of combining both patient and carer data sets for analysis and the result of using effects coding as opposed to using the dummy coding.

Conclusions

This study presents the first example of using a DCE to explore young patients’ and their carers’ preferences for clinic consultations in the management of diabetes. The approach was shown to be practical, with piloting vital to ensure feasibility and interpretation of the results (see Appendix 4).

The results show that both patients and carers prefer the doctor to address both parties and carers prefer their children to be involved in deciding goals on how to manage their diabetes, whereas the children want the doctor to be involved in the goal-setting process. This is consistent with the aims of the DEPICTED trial, although the DCE did not identify any trial effects. Future work will carry out subgroup analysis, which may produce a clearer picture of respondent preferences and intervention effects.

From a policy perspective, the study identified the key attributes of a routine consultation in paediatric diabetes services. However, the results might not reflect entirely the preferences of patients with higher concentrations of HbA_1c, or of their carers.

Introduction

The intervention was intended to influence how the diabetes team interacts with patients and carers through strategies and skills used by trained practitioners within the consultation. It was not intended to impact directly on structural issues (e.g. number of clinics held, length of consultations, the physical space provided, number of routine home visits, school visits or telephone/text contacts). Nonetheless, it was also important to determine how the intervention was delivered in practice and whether the particular context of the trial may have influenced implementation (either facilitative or inhibitive). A third major element in evaluating the trial process was to assess the impact of training upon practitioner performance. Therefore, the aims of the embedded PE were to explore:

• clinicians’ perceptions about how the intervention was received by clinical teams and by families and, in particular, factors that may have facilitated or hindered implementation and effectiveness

• systematic changes within services during the study period, which may have resource usage implications

• evidence of training impact upon practitioner performance.

The first two aims were addressed through interviews with informants from each clinical team, and the third by the rating of routine consultation recordings.

Local researcher interviews

Implementing the intervention

The results described below focus on the implementation and resulting outcomes of the Talking Diabetes intervention, and only data from intervention centre interviews are presented.

3T agenda-setting tool

Centres adopted a variable approach to implementing 3T, some providing folders to all their patients and others distributing the pads to trial participants only. Some centres utilised many methods to distribute the pads (e.g. posting to patients prior to appointments, giving them to patients as they arrived at the clinic, as well as having pads available in the waiting area). Other centres made the pads available to patients at the clinic only.

The use of the 3T pads (Box 16) by families was varied; some respondents reported that patients utilised the folder a great deal initially but recorded fewer agenda items as time progressed. Respondents felt that this was due in part to many of the issues raised by patients having been addressed at previous clinic visits. In general, most reported that the use of the pads simply declined as the study progressed. Of those who reported a decline, most inferred that the approach of the consultation nevertheless remained very much focused on ‘is there anything you’d like to talk about today?’ Many interviewees also commented that they used the ‘rate your diabetes’ question to initiate the consultation, in the absence of any written items on the 3T pad.

In terms of the age group of participants who utilised the 3T notepads (Box 17), many respondents felt that the younger age group were more likely to record their thoughts, with fewer older children and teenagers using the pads.

A few centres indicated that they would continue to use the agenda-setting pads, although one particular centre [ID 22] had implemented its own version of the 3T pad (post study period), adapted from a combination of the Talking Diabetes pad and an adult goal-setting tool.

Practitioner performance assessment

Key trial findings

Training HCPs to improve their consultation skills, particularly in relation to behaviour change, as developed in DEPICTED, could not be shown to impact beneficially on glycaemic control as measured by HbA_1c levels. Furthermore, patients in clinics where staff had undergone training may have experienced a reduction in confidence in their ability to manage their diabetes, whereas those in the control arm showed, surprisingly, a reduction in barriers and improvement in adherence to their diabetes management. However, patients in intervention centres did report an increased ability to cope with their diabetes as a result of their clinic visit, although this effect was only found in the short term. By contrast, parents of those in the intervention arm experienced greater excitement about clinic visits and an improvement in the continuity of care without the adverse effects seen in their offspring.

DEPICTED has demonstrated that a high-quality, complex, pragmatic trial of health service delivery can be successfully conducted in a currently challenging clinical environment, recruiting teams and patients from clinical services reflecting a wide range of philosophies and research experience within England and Wales. Evidence from this study has demonstrated that paediatric health-care teams can be successfully trained to improve their consultation skills using a combination of workshop and internet-based training, albeit with evidence to suggest that reinforcement of these skills remains an important need. The workshop component of this training represented a significant contribution to the overall costs (both time and financial) of the intervention.

In the following discussion, we initially focus on the findings of the trial phase, highlighting the strengths and weaknesses of the study, including generalisability of the findings. Thereafter, we consider the interpretation of the study results, taking into account the hypothesis and statistical issues relevant to the methodology. 156 Considerations for future delivery of clinical services and research are subsequently addressed in the final chapter.

Strengths and limitations

A cluster design was used to good effect in this study. We provided for the additional number of subjects required and accounted for possible dropout by over-recruiting both centres and patients. All centres recruited into the study completed their participation and the study dropout rate was extremely low, especially for HbA_1c blood samples, providing additional statistical power for both primary and secondary analyses.

Interpretation of results

The failure of the intervention to impact on the primary outcome – HbA_1c levels – demonstrates that the training as encompassed in the DEPICTED trial is not effective in changing glycaemic control over a 1-year follow-up after team training. It is possible that this disappointing outcome has been due to one or more of the following influences: (1) the improvement in practitioner expertise in communication may have remained insufficient to achieve beneficial changes in patient behaviour; (2) the extent of contact between HCPs and patients over the course of the 1-year follow-up (typically three to four brief clinic visits) may have been insufficient for improved communication skills to have impacted on patient behaviour; (3) the effect of training may have been diluted by the impact on patients of practitioners within the team who had not undergone training; and (4) the training, as developed in the DEPICTED trial, did not induce the skills required to facilitate behaviour change in the patient. In retrospect, an extended-length pilot study of the training might have proved useful in highlighting some of these issues. However, on the basis of the results, the training as defined in the DEPICTED trial cannot be recommended to produce clinically significant changes in children with diabetes.

Figure 8 portrays training received by practitioners to enhance their skilful consultation practice. It is more likely that the training will be effective in changing practitioner behaviour if they subsequently value the importance of the guiding approach and feel confident in applying the strategies and skills taught. Skills acquired in training need to be deployed in routine consultations with families. Within the trial there was evidence of high levels of initial exposure of practitioners to the training – good levels of attendance of workshops and, despite some logistical challenges, good engagement with the web-based training. The DEPICTED PE also provided evidence that trained practitioners do adopt a more guiding style when consulting with families about behaviour change, and there is further evidence of the use of some of the strategies in these consultations (particularly agenda-setting).

FIGURE 8.

Model of potential intervention pathway.

The impact of a trained practitioner can be conceived as ranging from proximal to distal. Among the former would be enablers for behaviour change, such as perceived importance of behaviour change and confidence in making such changes (e.g. self-efficacy). In practice, most patients and carers reported the value of change, even at baseline, so this was unlikely to be further modified by the intervention. At 12 months’ follow-up, confidence in managing diabetes was actually reduced for patients in the intervention arm, despite the increased ability to cope with diabetes as a consequence of the clinic visit reported in the short term (another proximal outcome). Physical outcomes for patients included HbA_1c levels – the primary trial outcome – but which is clearly a distal outcome dependent upon several intervening steps. Similarly, the health-related QoL (HRQoL) subscale ‘symptoms’ reports experience of the effects of poorly controlled diabetes and is also a distal outcome. Evidence of the intervening processes leading to these outcomes is provided by the QoL subscales ‘barriers’ and ‘adherence’, and the diabetes ‘care/mismanagement scale’. These record the patients’ appraisal of tasks and behaviours that are associated with self-management. A change in psychosocial factors for patients (e.g. confidence, enablement) may be expected to lead to attempts at behavioural change that these three self-report measures would assess. However, there was no direct attempt to measure behaviour in DEPICTED, for example by recording diet, exercise, medication use or self-monitoring behaviour. Although trial evidence shows that control patients experienced reduced barriers and improved adherence is unexpected, it is consistent with their higher levels of confidence.

Other impacts of the intervention, but ones which may not be directly implicated in behaviour change or physical outcomes, are the patient’s and carer’s appraisal of the clinic team. Perhaps most surprising is the absence of a difference between patients in each study arm in their appraisal of the clinical teams. The HCCQ includes items that are directly relevant to the intervention (e.g. providing choices and options, encouraging question asking). The improvement in continuity of care for carers in the intervention arm may reflect an enhancement in team communication as a result of explicit agenda-setting and greater attempts to engage with and understand the patient and family. Using Figure 8 as a guide, there is evidence of a change in practitioner behaviour following exposure to training, and of an impact upon proximal patient and carer outcomes. There is some evidence of an impact on patients’ appraisal of their own self-management, although no direct evidence of behaviour change. There is no evidence that such changes then led to objective physical outcomes for patients which would be considered as distal outcomes given the training intervention being trialled.

Building on the model (Figure 8) and following the logic of the MRC guidance on complex interventions, it is advisable to interpret the results of this study within two domains. These are firstly, efforts to change the behaviour of practitioners (and the team as a whole) and secondly, their subsequent efforts to change the behaviour of patients (children and parents). The DEPICTED development process paid attention to both domains: the learning programme to the first, and 3T with guiding and useful strategies to the second. Figure 8 illustrates how these domains might be linked together in the treatment of paediatric diabetes.

The economic evaluation showed that, although training costs appear high, when considered as an investment, the cost per patient receiving care from a trained team is modest. Overall, costs were only marginally and not statistically significantly higher for the intervention group, even when taking account of training costs.

Conclusions

Contributions of authors

All of the following named authors contributed to the development of the research question and study design, study implementation (including membership of study management group), analysis and/or interpretation of data and submission of the final report. Contributions to individual study outputs/particular study contributions are denoted below in publication citations.

Professor John Gregory (Professor & Honorary Consultant in Paediatric Endocrinology) Co-principal investigator, lead applicant and study guarantor; led the development of research question and study design, study implementation and report submission; and contributor to stakeholder meetings and learning programme trainer.

Dr Mike Robling [Senior Lecturer, Associate Director, South East Wales Trials Unit (SEWTU)] Co-principal investigator, co-applicant and study guarantor; led the development of research question and study design, study implementation, report submission; chaired study management group; joint lead on learning programme development and process evaluation; and contributor to stakeholder meetings and learning programme trainer.

Dr Kristina Bennert (Research Associate) Qualitative lead responsible for co-ordinating stakeholder involvement, development and field testing of the intervention; and contributor to stakeholder meetings.

Dr Sue Channon (Consultant Clinical Psychologist) Co-applicant; contributed to intervention development; joint lead for process evaluation; and contributor to stakeholder meetings and learning programme trainer.

Professor David Cohen (Professor of Health Economics) Co-applicant; designed and led the economic evaluation; and contributor to stakeholder meetings.

Dr Elizabeth Crowne (Consultant in Paediatric Endocrinology and Diabetes) Co-applicant, contributor to stakeholder meetings and report submission.

Helen Hambly (Research Psychologist) Implemented surveys in the development phase; developed outcome measures for trial questionnaire and for assessing practitioner competencies; and trial data management and contributor to stakeholder meetings.

Dr Kamila Hawthorne (Reader, Primary Care & Public Health) Contributor to stakeholder meetings and trainer for learning programme.

Professor Kerenza Hood (Professor of Medical Statistics, Director of SEWTU) Co-applicant, lead on the quantitative elements and contributor to stakeholder meetings.

Mirella Longo (Research Fellow) Co-applicant, contributed to design of economic evaluation, lead for discrete choice experiment and contributor to stakeholder meetings.

Dr Lesley Lowes (Senior Lecturer/Practitioner (Paediatric Diabetes)  Co-applicant, advised on stakeholder involvement, contributor to stakeholder meetings and learning programme trainer.

Dr Rachel McNamara (Senior Trial Manager) Led trial management, chaired project team meetings and contributor to stakeholder meetings.

Mr Tim Pickles (Junior Statistician) Data management programmer and statistician for performance assessment and other trial data.

Dr Rebecca Playle (Senior Statistician) Responsible for the design of the randomisation protocol and the statistical analysis of the primary and secondary trial outcomes.

Professor Stephen Rollnick (Professor of Healthcare Communication) Co-applicant, lead for intervention development and joint lead for learning programme development; joint lead for process evaluation; and contributor to stakeholder meetings and learning programme trainer.

Dr Emma Thomas-Jones (Trial Manager) Managed 12-month follow-up phase and led the service change assessment (design, conduct and analysis of interviews).

The report authors gratefully acknowledge the substantial contribution of the following in the conduct and delivery of the DEPICTED study:

Administrative, methodological and technical support Mrs Veronica Silva Dunning (trial administrator), Professor Chris Butler (co-applicant), Professor Ian Russell (co-applicant), Ms Helen Stanton, Mr Tim Harris, Mrs Cathy Lisles, Mr Neil Burt, Mr Fasih Alam, Mr Benedict Hodge, Karen Marsh and Nina Gobat (consultation raters), Mrs Carolyn Boston, and members of the administrative team at SEWTU.

Members of the Trial Steering Committee Professor Steve Greene (Chairperson), Dr Julie Edge, Professor Tim Peters and Professor Frank Snoek.

Participating clinical centres and support groups The local principal investigators, members of each clinical team and local UK Clinical Research Network research staff participating in the 26 trial centres, and to all the practitioners and families who contributed to the development and implementation of the Talking Diabetes intervention (particularly from clinics and support groups in south Wales and south-west England). The local principal investigators and participating NHS trusts were:

Mr Paul Langridge (Children’s Community Specialist Nurse) at Countess of Chester Hospital NHS Foundation Trust; Dr Tracy Tinklin (Consultant Paediatrician) at Derby Hospitals NHS Foundation Trust; Ms Francesca Annan (Specialist Paediatric Dietitian) at Royal Liverpool Children’s NHS Trust; Professor Timothy Barratt (Professor of Paediatrics) at Wellcome Trust Clinical Research Facility at Birmingham Children’s Hospital NHS Trust; Miss Caroline Spence (Paediatric Diabetes Specialist Nurse) at Brighton and Sussex University Hospitals NHS Trust; Dr Claire Smith (Consultant Paediatrician) at East Lancashire Hospitals NHS Trust; Dr Fiona Regan (Consultant Paediatrician) at The Hillingdon Hospital NHS Trust; Dr Rakesh Amin (Consultant Paediatric Endocrinologist) at Central Manchester & Manchester Children’s University Hospitals NHS Foundation Trust; Dr Mark Bone (Paediatric Diabetologist) at Central Manchester & Manchester Children’s University Hospitals NHS Trust; Dr Loretta Davis-Reynolds and Dr Sunil Bhimsaria (Consultant Paediatrician) at Barnsley Hospital NHS Foundation Trust; Ms Marie Marshall (Paediatric Diabetes Nurse) at Central Manchester & Manchester Children’s University Hospitals NHS Foundation Trust; Dr Shaun Gorman (Consultant Paediatrician) at Bradford Teaching Hospitals NHS Foundation Trust; Dr Neil Hopper (Consultant Paediatrician) at City Hospitals Sunderland NHS Foundation Trust; Dr Jo Mannion (Consultant Paediatrician) at York Hospitals NHS Trust; Dr Anuja Natarajan (Consultant Paediatrician) at Doncaster & Bassetlaw Hospitals NHS Foundation Trust; Dr Peter Stutchfield (Consultant Paediatrician) at Conway and Denbighshire NHS Trust; Dr Jerry Wales (Senior Lecturer in Paediatric Endocrinology) at Sheffield Children’s NHS Trust; Dr Fran Ackland (Consultant Paediatrician) at Northampton General Hospital NHS Trust; Dr Shailini Bahl (Consultant Paediatrician) at Ashford and St Peter’s Hospitals NHS Trust; Ms Sue Courtman (Diabetes Nurse) at East and North Hertfordshire NHS Trust; Dr Ivor Lewis (Consultant Paediatrician) at Surrey and Sussex Healthcare NHS Trust; Dr Ahmed Massoud (Consultant Paediatrician) at North West London Hospitals NHS Trust; Ms Roma Romano-Morgan (Paediatric Diabetes Nurse Specialist) at The Whittington Hospital NHS Trust; Dr Joanna Stanley (Consultant Paediatrician) at The Lewisham Hospital NHS Trust; Ms Victoria Surrell (Diabetes Specialist Nurse) at Hinchingbrooke Healthcare NHS Trust; Dr Tim Taylor (Consultant Paediatrician) at Royal West Sussex NHS Trust; Dr Nicola Trevelyan (Consultant Paediatrician) at Southampton University Hospitals NHS Trust; Dr Tabitha Randell (Consultant in Paediatric Endocrinology and Diabetes) at Nottingham University Hospitals NHS Trust; Dr Fiona Campbell (Clinical Director) at The Leeds Teaching Hospitals NHS Trust; and Dr Ruby Mathew (Consultant Paediatrician) at Chesterfield Royal Hospital NHS Foundation Trust.

Members of the SAG The young patients, parents, adult patients and diabetes professionals participating in the SAG who critically helped design the trial intervention. To Professor Neil Frude who acted as SAG facilitator.

User representative We particularly acknowledge the dedicated input of our parent and patient representative Mrs Charlotte Crawley (co-applicant) in particular during the development phase of the study.

Other contributors to programme development Students and staff at Whitchurch High School (Drama Club), escape to Design (3T: TimeToTalk), the HealthCare Learning Company Ltd (Talking Diabetes online learning programme, in particular Pat Cannon and Valerie Dougall), Philippa Thomas and Dafydd Gape (learning programme actors).

Institutional support We acknowledge with thanks the trial funders, the UK National Institute for Health Research Health Technology Assessment programme (NIHR HTA), and the financial contribution of Novo Nordisk who provided partial funding for the production of some study materials. SEWTU is funded by the Wales Assembly Government through the National Institute of Social Care & Health Research and the authors gratefully acknowledge SEWTU’s contribution to study implementation.

We thank all those not otherwise mentioned above who have contributed to the DEPICTED study.

Publications

Hambly H, Robling M, Crowne E; Hood, K; Gregory, JW Communication skills of health-care professionals in paediatric diabetic services. Diabetic Med 2009;26: 502–9.

Channon S, Hambly H, Robling M, Bennert K, Gregory JW. Meeting the psychosocial needs of children with diabetes within routine clinical practice. Diabetic Med 2010;10:1209–11.

Lowes L, Robling MR, Bennert K, Crawley C, Hambly H, Hawthorne K, et al. Involving lay and professional stakeholders in the development of a research intervention for the DEPICTED Study. Health Expect 2010. doi: 10.1111/j.1369-7625.2010.00625.x. [Epub ahead of print.]

McNamara R, Robling M, Hood K, Bennert K, Channon S, Cohen D, et al. (DEPICTED): a protocol for a cluster randomised controlled trial of the effectiveness of a communication skills training programme for healthcare professionals working with young people with type 1 diabetes. BMC Health Serv Res 2010;10:36.

Hawthorne K, Bennert K, Lowes L, Channon S, Robling M, Gregory JW; on behalf of the DEPICTED Study team. The experiences of children and their parents in paediatric diabetes services should inform the development of communication skills for healthcare staff (the DEPICTED Study). Diabet Med 2011. doi: 10.1111/j.1464-5491.2011.03292.x. [Epub ahead of print.]

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA programme or the Department of Health.

Contacts with diabetes team

Investigations and appointments

Hospital contacts

Other NHS contacts

Unit costs: time off from work

Hospital and Community Health Services inflation rates

References

Step 1: identification of attributes and levels

The attributes of routine clinic consultations that are of importance to patients and their carers have to be realistic, tradable and relevant to respondents. 141 This was achieved via a series of focus groups. Focus groups can inhibit some quieter participants. Therefore, principles of nominal group technique (whereby contributions from individuals are solicited prior to group discussion) were used to structure and facilitate input from all participants. 114,167 The focus groups involved individual working, group discussions and a formal ranking exercise of key attributes. Separate focus groups were conducted for carers and patients, held as part of the second stakeholder workshop (see Chapter 6).

Each group was run by two facilitators using a semistructured guide to identify key attributes and realistic levels, to ascertain whether or not any attributes dominated and to assess completeness of attributes identification. Discussions were audio-recorded and transcribed. Content analysis was used to identify the attributes and levels and also used data from individual participants’ response sheets. The patient group included 12 individuals (seven boys) aged between 12 and 16 years and the carer group included 11 carers (four with children aged ≤ 11 years).

The final selection of attributes and levels, involved combining some attributes and rejecting those that could not be affected by the DEPICTED intervention (e.g. distance to clinic). A final list of five attributes with three levels each was produced. The same attributes and levels were identified for both groups, with wording adjusted to fit either patients or carers.

Step 2: designing the experiment

The resultant attributes and levels generate 243 (3⁵) profiles (full factorial design). As this would require a large number of respondents, a regular fraction factorial design was used whereby only a subset of the combinations are used to model respondents’ choices (3^5–x). 148 However, the smallest regular fraction is defined by the number of parameters that need to be estimated, including the interactions to be investigated. As the above attributes are all categorical and have three levels each, 11 parameters (10 for the attributes and one for the constant term – explained below) would be needed to estimate the main effect of the model and four additional parameters to estimate each two-way interaction.

The present study determined a regular fraction design of 27 treatment profiles (3^5–2), where three attributes are the generator factors of the other two attributes. Attribute D is derived by the defining equation D = A + B + C and attribute E is derived by the defining equation E = A + 2B.

As 27 treatment profiles is still a considerable number for each individual to answer, it was decided to use a two-level attribute to generate two orthogonal questionnaires (version A and version B). Each questionnaire included the lowest and the highest treatment combinations (00000 and 22222) producing a total of 15 profiles for each version of the questionnaire. In order to construct a pair-wise choice set, it was decided to compare each treatment option with a constant scenario. 132 In this study the middle term of each attribute (11111) was used as a constant scenario.

To facilitate interpretation of the parameters to be estimated by the model, we report the utility model function for the main model only (Figure 1).