Enhanced motivational interviewing for reducing weight and increasing physical activity in adults with high cardiovascular risk: the MOVE IT three-arm RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 10/62/03. The contractual start date was in January 2013. The draft report began editorial review in April 2018 and was accepted for publication in September 2018. 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 reviewers 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

Khalida Ismail has received honoraria for educational lectures from Sanofi SA (Paris, France), Novo Nordisk (Bagsværd, Denmark), Janssen Pharmaceutica (Beerse, Belgium) and Eli Lilly and Company (Indianapolis, IN, USA). Kirsty Winkley received consultancy fees from Merck Sharp & Dohme (Kenilworth, NJ, USA).

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Ismail et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) 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: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Epidemiology of cardiovascular disease and its risk factors

Cardiovascular disease (CVD) is the leading cause of mortality, morbidity and disability in the UK and in other developed countries. 1 However, many of the major determinants of CVD are modifiable, including tobacco smoking, a diet high in saturated fat, high low-density lipoprotein (LDL) cholesterol, obesity, a sedentary lifestyle, hypertension and diabetes mellitus. 2–6 Lower educational attainment and lower socioeconomic status are associated with a greater risk of CVD, and this association is strongest for females. 7 The risk of CVD varies markedly between ethnic groups, with a higher rate of ischaemic heart disease in those of South Asian ethnicity and a higher rate of cerebrovascular disease in those of African ethnicity among those living in England and Wales. 8

Although CVD remains the most common cause of death in developed nations, mortality rates have been falling. Between 1981 and 2000, CVD mortality in the UK fell by 62% in men and 45% in women. 9 Cohort studies and prediction models suggested that a fall in the prevalence of tobacco smoking, a decline in population blood pressure levels and changes in cholesterol levels were important contributors. 9,10 Population-wide changes in modifiable risk factors, such as dietary intake, can bring about substantial benefits and further changes in blood lipids, particularly reductions in levels of non-high-density lipoprotein (HDL) cholesterol. 10 However, limited changes in physical activity (PA) and rising levels of obesity have limited the decline in CVD mortality. 10 Further efforts are therefore needed to bring about positive changes, particularly in diet, obesity and PA.

Cardiovascular risk identification

The NHS in England introduced the Health Check programme in 2009 as part of the Department of Health and Social Care (DHSC)’s long-term vision for the future of public health in England. 11 In offering Health Checks to all those aged 40–74 years without a known diagnosis of CVD, the aim is to prevent heart disease, stroke, diabetes mellitus and kidney disease and to reduce health inequalities. The risk assessment includes collection of demographic data, smoking status, cholesterol level, blood pressure and diabetes mellitus status. An individualised risk management plan is given in accordance with the assessment to support lifestyle changes, such as referral to smoking cessation sessions, exercise prescriptions, lifestyle advice and signposting to local resources.

A lower-than-anticipated coverage of NHS Health Checks has been reported,12 with regional variations in attendance ranging from 27% to 52%,12 with greater uptake in patients of older age and in regions of lower deprivation. 13 Reductions in CVD risk have been reported for those attending Health Checks;14 however, a systematic review of the implementation of Health Checks found no reduction in mortality or morbidity. 15 The programme has attracted criticism owing to a lack of an up-to-date evidence base and Clinical Commissioning Groups (CCGs) lacking the resources to implement them. 16

Several risk algorithms have been developed and validated to estimate the risk of developing CVD based on known risk factors; these risk algorithms include the Framingham Risk Score,17 the ASSIGN score,18 QRISK^®19 and QRISK^®220 (both QResearch, Nottingham, UK). The latter is now recommended by the National Institute for Health and Care Excellence (NICE) for the identification of people at risk of CVD up to the age of 84 years in England. 21 The QRISK2 algorithm takes into account self-report ethnicity, deprivation and other relevant clinical conditions, and is updated annually to reflect changes in clinical evidence, data recording and population demographics. Once a high CVD risk is ascertained, the primary prevention of CVD is recommended via lifestyle advice and interventions.

The evidence for increasing physical activity

Physical inactivity increases mortality and the risk of diseases such as CVD and diabetes mellitus. 22 The DHSC advises adults to perform ≥ 30 minutes of at least moderate-intensity PA on ≥ 5 days per week, in ≥ 10-minute bouts, for optimum health benefits. 23 Walking is the most common form of PA in adults and is associated with reductions in CVD and all-cause mortality, and walking pace is a stronger predictor of improved outcomes than walking duration. 24 Walking is promoted as a near-perfect exercise as it has the lowest risk of harm and is now included in UK public health policy. 25,26

However, the proportion of those achieving PA recommendations is low, particularly when objective measures are used to assess PA. In England, 39% of men and 29% of women self-report achieving the recommended PA levels, but objective assessment of PA using accelerometers in a subsample of the Health Survey for England found that only 5% of men and 4% of women aged 35–64 years achieved the recommended levels. 27

Systematic reviews and meta-analyses report moderate positive short-term increases in PA following lifestyle interventions, in either a group or an individual format, but findings are limited because most studies used self-report measures. 28,29 Evidence for brief PA interventions suggests improvement in PA in the short term, but there is limited evidence for the long-term impact, cost-effectiveness and deliverability in a primary care setting. 30,31 A review of 32 trials reported that walking interventions led to improvement in a number of cardiovascular risk factors, including blood pressure and weight, but not in lipids. 32 The majority of the reviewed intervention trials recruit motivated volunteers and report low response rates, which may limit the representativeness of observed findings.

The contents of assessed interventions to promote PA differ significantly, but social support and cognitive–behavioural therapy (CBT) strategies, rather than health education alone, are recommended in older adults. 33 The use of pedometers as a method of self-monitoring can increase PA and improve health in the short term. 34 Compared with usual care (UC), the use of pedometers and a brief walking intervention in primary care led to improvements in 12-month PA in a sample of adults not achieving the recommended activity levels at baseline. 35 The intervention was as effective when delivered by post as when delivered by nurses in primary care, suggesting that PA can be improved in physically inactive patients with minimal resources. Details of the components used in trials of interventions to promote PA, including information on the behaviour change techniques (BCTs) used, are recommended to improve implementation and evidence syntheses. 29,36

The evidence for dietary interventions

Modest beneficial changes to dietary intake, specifically changes to fat, fibre, fruit and vegetable intake, are found following healthy diet interventions in primary care, but there is variability in intervention design and delivery as well as the methodological quality of previous studies. 37 Based on the limited evidence available, estimates of the cost-effectiveness of dietary interventions in primary care suggest that interventions need to be targeted towards the older population at greatest risk of disease to be cost-effective. 38

A systematic review of randomised controlled trials (RCTs) assessing generic dietary advice for reducing CVD risk found modest beneficial effects on mean total and LDL cholesterol levels and small reductions in blood pressure up to 12 months later, suggesting that dietary advice may contribute to an improved CVD risk profile. 39 However, the longer-term effects of dietary interventions are unknown owing to limited follow-up periods of reviewed studies. Compared with UC, dietary interventions produce modest weight losses that diminish over time. 40 Further work is needed to understand how the modest benefits of dietary interventions may be maintained.

The evidence for motivational interviewing

Motivational interviewing (MI) is a common approach to behaviour change in health care, defined as a collaborative, goal-oriented style of communication with particular emphasis on the language of change. 41 It is designed to strengthen motivation for and commitment to a specific behavioural goal by eliciting and exploring personal reasons for change within an atmosphere of acceptance and compassion. 42 The appeal of MI is that it is brief, can be delivered by a range of health-care providers, has a competency framework and can be applied to a range of health-care settings, with evidence of benefits to health outcomes when compared with other interventions. 43

Systematic reviews and meta-analyses have demonstrated significant, moderate effects of MI on diet and exercise behaviours,44 as well as on health outcomes such as reduced weight, cholesterol and blood pressure, although the number of trials is small. 45,46 However, a systematic review of MI used in lifestyle interventions for people at risk or with a diagnosis of CVD found little evidence of the benefits of MI, noting the considerable variability between interventions and the outcomes measured and that few have included a long-term follow-up to assess whether or not any observed benefits are sustained. 47 A 14-month follow-up of diabetic patients who had received a MI lifestyle intervention found no benefits in lifestyle, biomedical or psychological outcomes compared with UC. 48 A RCT with a 12-month postintervention follow-up period found that the benefits of up to five MI sessions delivered over a period of 6 months on walking and cholesterol levels were maintained in a sample of overweight or obese patients, but other CVD risk factor outcomes were not maintained. 49 Effects were stronger for those found to be at higher risk at baseline, suggesting that interventions should target high-risk patients to achieve the best results.

A taxonomy of behaviour change techniques

The limitations of current models of lifestyle interventions, particularly their short-term effects, have led to a search for more sophisticated and targeted behavioural interventions. 50 A Cochrane Database Systematic Review of multiple risk factor interventions for the primary prevention of coronary heart disease observed that techniques based on instruction and information were associated with small improvements in lipid levels and blood pressure, especially when embedded in a theoretical framework related to behaviour change. 51 A systematic review of behavioural interventions found that the techniques most effective in improving diet and PA were based on self-regulatory behaviours, such as goal-setting, self-monitoring, giving feedback, utilising social support and MI. 52 Interventions based on a psychological theory, such as the theory of planned behaviour,53 were more effective, as were those for high-risk populations. There is less evidence to support the case for any minimum threshold of intensity, mode of delivery, intervention provider and setting;50–52 therefore, further study is required to understand how the benefits of behavioural interventions on lifestyle and health outcomes can be maintained. Evaluating interventions in the context of a taxonomy of BCTs and an intervention map offers a framework that is easier to teach, test, replicate and translate. 52–54

In a pilot RCT, a group intervention developed in line with evidence of the most effective BCTs52 led to reductions in weight, when co-interventions and comorbidities were controlled for, but did not increase PA at the 12-month follow-up in people at high risk of CVD, compared with UC. 55 As this was a pilot RCT, the authors note that the intervention was acceptable to participants and that outcomes could be improved by adapting the intervention accordingly. The effects of the group-based setting were not compared with a one-to-one setting, and there is little evidence in the literature on the differences between group and individual approaches. It has been found that the benefits of group sessions outweigh individual sessions even when it is not preferred by participants,56 and that group sessions are highly valued by participants. 57 However, it is also reported that at least one individual session is critical to the success of interventions and in ameliorating disengagement in some participants. 57,58 The variability in patient samples involved in these studies, the intervention delivered and the methodology of each study limit the generalisability of the findings.

By enhancing a MI approach with effective BCTs, the modest effect sizes of each may lead to improvement in outcomes. In people with type 1 diabetes mellitus, four sessions of MI alone was not associated with improved glycaemic control, but four sessions of MI followed by eight sessions of CBT was associated with improved glycaemic control, compared with UC. 59 However, the effects were not maintained after 12 months,60 and participants stated that MI helped them to become more ready to change their behaviour but that further support was needed to implement the change. 61 The evidence for enhancing MI with CBT is not consistent, as the landmark COMBINE (Combined Pharmacotherapies and Behavioural Interventions for Alcohol Dependence) study did not demonstrate increased abstinence from alcohol in those receiving the psychological intervention. 62

The role of health trainers

The DHSC recommended the deployment of health trainers into the most deprived areas of the UK in a White Paper published in 2004. 63 A health trainer is employed by the NHS from the local community in which they serve to provide lifestyle advice to those at risk of disease, with the overall aim of the programme being to address health inequalities, an important issue within multi-ethnic and variably deprived settings. 64 The role involves identifying clients from hard-to-reach and disadvantaged groups, providing one-to-one support in identifying potential problems in lifestyles, setting goals, supporting behaviour change and reviewing client progress. 65 The health trainer programme has been found to increase uptake of NHS Health Checks, particularly in men, younger age groups and those from less affluent areas. 66

A comprehensive review of interventions delivered by lifestyle advisors found little evidence of the benefits of interventions promoting healthy diet and/or increased PA in North American trials; there was no effect on weight and little evidence of improvement in PA levels, but the intervention contents, delivery and goals varied considerably. 67 A review of the available evidence on the NHS health trainer programme thus far indicates positive outcomes and acceptability, but critics argue that the models of service provision are varied, evaluations have included no comparator group and there is a lack of evidence for the maintenance of behaviour change. 68 A health trainer programme for CVD risk reduction in patients with at least one CVD risk factor found significant reductions in CVD risk after 12 months for only those who were identified as being at high risk of CVD (a Framingham Risk Score for CVD of ≥ 20.0%) at baseline. 69 The service also led to high levels of behavioural goal achievement and small, but significant, increases in quality of life. However, as this was a service evaluation there was no comparator group and achievement of behavioural goals was self-reported and unrelated to changes in CVD risk. Further work is needed to assess the potential for health trainers to deliver complex behavioural interventions, including the employment of objective outcome measures, the assessment of maintenance of behavioural change and clinical benefits, and comparison of outcomes with a control group.

The case for an enhanced motivational interviewing intervention

At the population level, the potential benefits of reducing weight and increasing PA in those who are at risk of CVD are considerable. Modest, short-term beneficial effects of various behaviour change interventions for the primary prevention of CVD emphasise the need for more complex, targeted interventions and RCTs that assess long-term benefits. MI provides broad appeal for its collaborative patient-centred style, evidence base and deliverability, and by enhancing a MI approach with the specific BCTs that have the strongest evidence, and incorporating techniques designed to enhance maintenance, observed outcomes may be improved. The relative effectiveness of a group intervention versus an individual intervention remains uncertain, but the former offers social support and may be more cost-effective, if it is acceptable to participants. Health trainers may improve the acceptability of interventions to harder-to-reach populations and positive outcomes have been reported in health trainer programmes thus far. The investigation of the potential for a health trainer to deliver more sophisticated interventions, and for the effects to be compared with UC in a RCT setting, is yet to be undertaken.

The overall aim was to compare the effectiveness of MOVE IT (enhanced MOtiVational intErviewing InTervention), which integrates MI with CBT BCTs, in reducing weight and increasing PA in those at high risk of CVD over 24 months across three arms: (1) enhanced MI in a group format versus (2) in an individual format versus (3) UC. The primary interest was whether or not MOVE IT in a group format was more effective and cost-effective than the individual format or UC because of its potential for social support.

Primary objective

To assess whether or not MOVE IT delivered in either a (1) group or (2) individual format is more effective than UC in reducing weight (kg) and increasing PA (average number of steps per day assessed via accelerometry) 24 months later.

Secondary objectives

• To assess whether or not group MOVE IT is more effective than the individual format in reducing weight and increasing PA 24 months later.

• To assess whether or not MOVE IT, delivered in either (1) a group format or (2) an individual format, is more effective than UC in reducing LDL cholesterol and in reducing CVD risk score 24 months later.

• To compare the number of fatal or non-fatal cardiovascular events, and other recorded adverse events (AEs), per treatment allocation.

• Cost-effectiveness: to assess whether or not MOVE IT delivered in either (1) a group format or (2) an individual format is more cost-effective than UC, in terms of quality-adjusted life-years (QALYs) gained over the 24-month follow-up period.

• Process evaluation: using mixed methods, we conducted the following –

  • Mediational analysis: to examine whether or not changes in behavioural and psychological factors, such as dietary intake, health beliefs, depressive symptoms and self-efficacy, mediate the association between the intervention and outcomes.

  • Participation bias: to assess whether or not the RCT recruited those it intended, we assessed the participation bias by comparing the sociodemographic characteristics and QRISK2 scores of those who responded to the invitation to participate and those who did not.

  • Fidelity analysis: to assess whether or not MOVE IT was delivered in accordance with the manual and to compare whether or not the levels of competencies among the HLFs were associated with variations in outcomes using thematic contents analysis of sessions.

  • Participant and therapist experience: using qualitative methods, we describe the perceived expectations, benefits, strengths and limitations of the intervention from the patient and intervention provider perspective.

Trial design

This was a three-arm parallel RCT for individuals at high risk of CVD. The three arms were (1) UC and enhanced MI in a group format, (2) UC and enhanced MI in an individual format and (3) UC. As participants in the group intervention arm, but not in the other two arms, were clustered within groups, this trial has a partially clustered (or nested) design. Interventions were delivered in 10 sessions across a period of 12 months. Participants were followed up at 12 and 24 months from baseline.

Recruitment

Outcome measures

At the 12- and 24-month follow-up assessments, participants were asked if there had been any change to their sociodemographic status, such as accommodation, relationship status and educational attainment. AE and participant feedback questionnaires were administered. All other baseline measures were repeated at the 12- and 24-month follow-ups (see Table 1).

Trial arms and intervention details

Arm 1: usual care and enhanced motivational interviewing in a group format

Theoretical framework

The intervention is based on the theory of planned behaviour for initiation of behaviour change: to change behaviour, people need to form an intention. 53 Intention formation is influenced by three constructs: (1) expected value or positive attitude (people see the value in making the change), (2) subjective norm (significant others and peers also value the change) and (3) self-efficacy (people believe that they are capable of making the change).

Our intervention taps into all three constructs using principles and techniques from MI,41 CBT91 and social cognitive theory92 (Figure 1). MI is used to support participants in forming healthy intentions. MI is a collaborative conversation style for strengthening a person’s own motivation for, belief in and commitment to change. Hobbis and Sutton91 highlight the gap between translating intention into action and illustrate how CBT can be applied to bridge this gap. For this intervention, techniques from CBT are used to support the transition from intention to action, and action to maintenance. 93 Identifying and challenging unhelpful thoughts or thinking styles can promote more positive emotions and behaviours, for example ‘When I get breathless after some exercise (bodily sensation), this means that I am going to damage my heart (incorrect cognition)’ or ‘I have eaten one doughnut (behaviour), I might as well eat the whole bag (all or nothing cognition)’.

FIGURE 1.

Theoretical framework and intervention map. White background indicates psychological process; green background indicates behaviour change technique.

Social cognitive theory emphasises the importance of significant others in shaping people’s behaviours. The theory of planned behaviour also highlights this aspect through the ‘subjective norm’ construct. In our intervention, social networks from the participants’ own lives and/or group members (in the group intervention arm) were actively utilised to provide practical and emotional support and opportunities for modelling health behaviours during all phases of the intervention.

Clinical effectiveness

Cost-effectiveness

The main perspective for the economic evaluation is that of the health-care system. The EuroQol-5 Dimensions, three-level version (EQ-5D-3L) is used to generate QALYs for use in economic analyses. 98 Intervention costs were calculated taking into account staff time for delivering the sessions and the unit costs include elements for overheads and oncosts and account for the ratio of direct-to-indirect contact time. We assumed that the clinician would be grade 3 and that the unit cost per hour was £32.40. For the group intervention, the costs are apportioned over the number of attendees, which averaged 5.5. Other service use was measured at baseline, 12 months and 24 months using the CSRI. EQ-5D instruments were developed in the 1980s and have been subsequently adapted for use in around 500 studies. Service costs were calculated by combining service use data with information on unit costs. 99,100 We had originally planned to explore the cost of lost employment. However, these data were not adequately recorded and so this was omitted.

Process evaluation

The overall aim of the qualitative analysis is to identify and describe factors and processes that affect the delivery, receipt and outcome of the study to aid interpretation and translation of the observed findings. Process data are analysed before outcome data wherever possible to reduce bias in interpretation. The main themes are described in the following sections.

Statistical analysis

Analysis and reporting is aligned with Consolidated Standards of Reporting Trials (CONSORT) guidelines,101 with primary analyses being on an intention-to-treat (ITT) basis. A description of the sample will be presented using mean (SD) or count (percentage). The baseline characteristics of those who withdrew and those who did not complete the intervention are compared with those of participants who completed follow-up. Descriptive data also include maximum values, minimum values, medians and interquartile ranges (IQRs) when appropriate. Significance is reported to a two-tailed alpha level of 0.05 as appropriate for comparisons.

The differences in treatment effect between the three arms at 12 and 24 months of this partially nested design are analysed using mixed-effects models, with pre-randomisation values as a covariate. 102 In the linear mixed model, ‘treatment arm’, ‘time’ (as a categorical variable with two levels: 12 and 24 months), the ‘interaction between treatment group and time’, ‘borough’, ‘ethnicity’ and the ‘baseline values’ of the outcome variable are the fixed part of the model. The random parts of the models are ‘general practice’ (participants are nested in practices) and ‘therapy group’. The design of the study is complex because we have a partially clustered cross-classified design. In the previous version of the analysis plan, we planned to account for the partially nested design of ‘therapy group’ in an approach that matches the non-parallel data structure. 103 However, preliminary analyses with blinded data revealed that this complex model of a partially nested, cross-classified design did not converge. Thus, in agreement with the Trial Steering Committee (TSC) and Data Monitoring and Ethics Committee (DMEC) (on 6 July 2017), we decided to drop the random effect for therapist from the primary analyses. We are aware that we may underestimate standard errors of treatment effects; however, health intervention therapist effects in similar studies [e.g. the National Institute for Health Research (NIHR)-funded IMPACT104 and Diabetes-6105 RCTs] ranged between negligible and small (ICC range 0 to 0.02).

The dependency of the repeated observations of the same participants at 12 and 24 months and the covariance between the residuals within the lowest-level group ‘patients’ are correlated by using an unstructured covariance pattern model. For the final model, the group difference estimates and associated confidence intervals (CIs) are reported for 12 (for secondary analyses) and 24 months after randomisation.

In a sensitivity analysis, the random effect of borough was replaced by a partially nested random effect for therapist group, as described previously, and results compared.

The described analysis approach provides valid inferences under the assumption that the missing data mechanism can be ignored (missing at random). Sensitivity of results to missing data was further assessed by including covariates predictive of missingness in the analyses model.

The significance level was 2.5% (two-sided) for the two main comparisons of (1) group format versus UC and (2) individual format versus UC. The (secondary) comparison (group format vs. individual format) was assessed on a 5% significance level as our research hypotheses is a null hypothesis of no difference.

The large sample size should have ensured that all possible confounding variables are equally distributed between treatment arms. However, in the sensitivity analysis, we extended the model of the primary analysis by including baseline variables with substantial imbalance, thought to be important in determining outcome. The potential baseline variables were age, sex, IMD, education, marital status and smoking status.

The following further sensitivity analyses were conducted and, for all analyses, changes in predicted treatment outcome differences are presented in the analysis plan – available on the project webpage [see www.journalslibrary.nihr.ac.uk/programmes/hta/106203/#/ (accessed 29 April 2019)].

Ethics issues

The trial was reviewed by the Dulwich REC and has been approved (reference 12/LO/0917). The TSC provided overall trial supervision, supported by the DMEC. The main ethics consideration was to ensure that the risk of harm to participants was minimised and that they were fully informed of any risks. We considered literacy and cultural sensitivities in obtaining informed consent. Other ethics considerations were in ensuring that recruitment and informed consent were handled in such a way that potential participants were not put under pressure to take part and that confidentiality was preserved. All participant data are stored using a unique study identifier and electronic data are password protected.

In general, regular PA is associated with improved health outcomes and this outweighs the risk of sedentary lifestyles. However, sudden increases in vigorous PA for otherwise sedentary individuals is associated with a higher risk of myocardial infarction and of musculoskeletal injuries, which may be pertinent as we were intervening in a group that is at high risk of CVD. However, one of the components of BCTs is to deliver the message that PA should be increased in a graded manner rather than suddenly. The intervention discouraged excessive and/or sudden changes to lifestyles. Weight loss could worsen frailty by accelerating the usual age-related loss of muscle that leads to sarcopenia, but combining weight loss with increased PA can actually ameliorate frailty. 106 Importantly, our intervention is based on healthier diets and gradual and sustainable weight loss as opposed to commercial weight-loss programmes. We considered risks to be small and minimal owing to the exclusion of patients with existing CVD and the use of GP advice when CVD events occurred during participation.

Patient and public involvement

Before the trial began, we held a focus group with 10 patients from a general practice in Peckham, London. All patients were of African, Caribbean or South Asian ethnicity, first generation migrants and at high risk of CVD, and were distributed equally between sexes. The group understood the rationale of the MOVE IT trial. They recognised that daily stressors affect lifestyle choices and welcomed the chance to talk about this and think differently. The preference was for interventions in an individual format, but they could see the potential benefits of learning from others in a group. Flexible appointment times were stressed as being important, so that the intervention could fit around other activities. The patients felt that it was necessary to provide more information about individual health status rather than just telling patients that they are at high risk of heart disease. They also gave tips on how to recruit; they thought that by getting some patients on board, these patients could network in the local community. These comments were incorporated into the intervention.

We invited this group of patients to form a patient and public involvement (PPI) group to help with the development of study documentation, including the patient information sheet and consent form. A PPI group of five patients was formed, and two members remained involved until the end of the trial. Their role during recruitment involved advertising the trial to both general practices and potential participants, advising on the development of poster and leaflet campaigns to increase public awareness and actively recruiting general practice sites across four of the CCGs. As part of the process evaluation, we sought PPI in the development of topic guides to be used in focus groups to gain feedback from participants who attended the intervention.

Throughout the trial, the trial team updated PPI members with progress and welcomed feedback. The PPI members attended all TSC meetings throughout the trial to gain more detailed feedback and discuss points of interest with the trial manager, PI and co-investigators.

Low uptake of NHS Health Checks

In parallel to setting up MOVE IT, there were increasing numbers of reports and concerns that there was a lower-than-anticipated uptake of NHS Health Checks,12 with regional variations in attendance ranging from 27% to 52% and greater uptake among older patients and in regions of lower deprivation. 13 We also observed this when we started recruiting and that this was delaying the project.

It was necessary to amend the protocol to detail that the researcher would repeat all CVD risk algorithm measures at screening. General practice searches for eligible participants could therefore not rely on recent Health Check data but instead on estimates of CVD risk using outdated data or inaccurate substitutions for missing data.

This change to the protocol led to the exclusion of a large number of patients whose medical record data suggested that they were at a ≥ 20.0% QRISK2 score but on screening were found to have a score of < 20.0% (see Chapter 5). Consequently, more resources were required to complete recruitment to target (substantial amendment; approved in July 2013).

Change to recruitment and study time frame

As research assistants were screening a larger proportion of ineligible patients who could not be randomised, the recruitment period had to be extended from 12 months to 21 months to allow the target sample size to be reached. This required an increase in funding to support a greater number of research assistants and for a longer period of time to support the recruitment procedures at a greater number of research sites. As a result, the end-of-study follow-ups were also extended by a further 9 months (minor amendment; approved on 16 March 2015).

Research sites

We initially invited general practice sites with list sizes of > 8000 patients. We extended invitations to smaller practices with > 5000 patients, as we required an increase in the number of sites owing to the high number of responses from ineligible patients (substantial amendment; approved in July 2013).

Randomisation

Randomisation was not stratified by general practice and ethnicity as planned, as the numbers of patients in many practices are not large enough to stratify by both factors. Simple randomisation was agreed to be the best method, with surgery included as a random factor in the model, and emphasis being on more practices and fewer patients per practice (substantial amendment; approved in July 2013).

Research measures

Three blood pressure measures were taken. The third, rather than the mean, systolic blood pressure is used in the QRISK2 algorithm and will be reported in this final report (substantial amendment; approved on 9 June 2015).

Intervention details

The proposed health trainer job title was changed to HLF to better incorporate the MI aspect of the intervention whereby the participant was ‘facilitated’ rather than ‘trained’ to make healthy lifestyle changes.

We revised the definition and description of MI to that used in the third edition of the MI textbook. 42 The intervention was updated to include a session 0, which was an introductory session via telephone. The length of group intervention sessions was increased to 120 minutes, rather than 90 minutes, to better match for attention with the individual intervention sessions (substantial amendment; approved in July 2013).

Further to the above changes, we changed the protocol to state that participants would be seen for intervention sessions within 6 months from randomisation, as this time period was not initially stated and the start of intervention sessions was presumed to be immediate. The time period varied greatly owing to (1) large numbers of participants randomised and waiting to be seen, as a result of the accelerated recruitment rate to meet the target, and (2) HLF staff turnover and the subsequent recruitment and training of new staff members, which created delays. When this time period was exceeded, we made every effort to still engage participants in the intervention and continued to follow-up for the ITT analysis (substantial amendment; approved on 9 June 2015).

Additional consent procedures

Participants were given the opportunity to consent to the following additional items post randomisation:

• Once participants had completed all research measures, they were asked if we could link their data to other collaborating research studies in which they also participated (substantial amendment; approved on 21 September 2015).

• We requested access to Hospital Episodes Statistics data,107 information on all hospital appointments for up to a 6-year period beginning 12 months before participating in the study. However, as this was requested after participants completed the study, we did not gain consent from all participants for this and thus did not seek the Hospital Episodes Statistics data and used only CSRI information on hospital services usage (substantial amendment; approved on 21 September 2015). We requested that participants provided consent to these points remotely, via a letter to and a telephone call with the participant, to avoid unnecessary appointment scheduling (minor amendment; approved on 13 January 2016).

• A small number of participants who received the intervention (group or individual) were randomly selected to be invited to a focus group to provide feedback. They were asked to provide written informed consent to take part in an audiotaped focus group (substantial amendment; approved on 7 March 2016).

• A small number of participants were asked to consent to a video recording or photograph of an intervention session to be used to promote the study for educational and training purposes only (substantial amendment; approved on 7 March 2016).

Elaboration of process evaluation: participation bias

We gained ethics approval to collect anonymised demographic data on all patients who were identified as potentially eligible and invited to participate when screening the GP medical records. These data enabled an estimation of participation bias through examining differences in age, sex, ethnicity, deprivation and QRISK2 score between responders and non-responders to the study invitation (substantial amendment; approved on 8 January 2016).

King’s College London network outage and impact on the fidelity analysis

On 18 October 2016, a university-wide information technology (IT) network outage occurred as a result of an infrastructure fault, restricting us from accessing the shared network drives in which various trial data were stored (as per university protocol) and resulting in some data being lost. Although King’s College London IT services made an extensive effort to forensically retrieve data from backup sources, all audiotaped intervention session data were lost from the university-managed shared network drive. We estimated this to be almost 2000 hours of intervention delivery. In addition, we lost data from approximately 200 dietary recall diaries (which could be re-inputted from their paper copies) and 95 accelerometry data files from the 24-month follow-up. For the accelerometry data, we asked the 95 participants to rewear the accelerometer; of those, eight declined and consequently had missing data for the PA outcome. We conducted a sensitivity analysis adjusting for participant rewear of the accelerometer.

The loss of the intervention session data was reported to the TSC. At the subsequent TSC meeting (on 6 July 2017), its members recommended that we attempt to recover the session recordings from other sources and conduct the fidelity analysis in accordance with the original plan as closely as possible. Of the data lost, we were able to recover approximately 400 hours from local backups and physical audiotapes, with the co-operation of university IT services and previous HLF staff. This precluded us from conducting the a priori fidelity analysis and randomly selecting coding data from a larger sample of tapes in the database. Instead, we adapted our analysis plan to code and analyse all of the limited, non-random data that were available. Furthermore, we had planned to use the data from the fidelity assessment to determine whether or not the levels of competencies between the HLFs were associated with variations in patient outcomes or mediated the treatment effect. However, the quality of the data recovered precluded us from conducting this analysis as well. See Chapter 8 for further details on our adapted fidelity analysis.

Practice recruitment

Of the 310 general practices with a list size of > 5000 in the 12 South London CCGs that were invited, 126 (40.6%) agreed to participate. A further nine (out of 145) general practice sites with a list size of < 5000 were recruited. Therefore, a total of 135 (29.7%) of the 455 general practices within the 12 South London CCGs participated. 108 The list of participating general practices is provided in Appendix 2. The distribution of practices participating in each CCG and the differences in IMD 2010 scores of participating and non-participating practices are given in Table 3. There were no differences in general practices’ participation by IMD 2010 score.

There was greater initial interest from practices in outer-London CCGs, such as Bromley, Bexley and Merton, and feedback from inner-London CCGs, such as Lambeth and Wandsworth, was that practices are overexposed to research studies and may not have the resources to participate. No formal process of collecting reasons for non-participation of general practices was conducted.

Participant recruitment and flow through the trial

An electronic search for eligible patients was conducted by each general practice using inclusion and exclusion criteria as Read codes109 and including a calculated estimate of QRISK2 value based on values for risk factors already recorded. Participants were recruited between June 2013 and February 2015; recruitment per CCG is given in Table 4. Figure 2 is the CONSORT flow diagram showing participant progression through the trial. Screening eligible participants at general practices led to invitations being sent to 17,775 patients. An expression of interest to take part was received from 3515 patients (a response rate of 19.8%). Participation biases (the differences between those who responded to the invitation and those who did not) are explored in Chapter 7 as part of the process evaluation. Following a brief telephone screening questionnaire, 3183 patients attended for a first appointment with a researcher and provided consent.

FIGURE 2.

The CONSORT flow diagram. COPD, chronic obstructive pulmonary disease. Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

In the first appointment, eligibility was assessed using all measures required for QRISK2 score calculation. This led to the exclusion of 1332 participants as they did not meet the eligibility criteria. The main reason for this (1105 participants; 83.0% of all exclusions) was that the QRISK2 score was < 20.0% on recalculation at face-to-face screening. For 71 participants, blood test results indicated a raised HbA_1c level (> 47 mmol/mol) and they were directed to their GP for further advice and informed that they were not eligible to participate. Eighty-eight participants did not complete all screening measures and, therefore, QRISK2 could not be calculated (see Figure 2 for further details of exclusions).

Of the 1851 participants eligible to be randomised, 1742 were randomised and the remaining 109 declined to participate further. Of the randomised sample, 697 were allocated to the group intervention arm, 523 to the individual intervention arm and 522 to the UC arm.

Intervention delivery

Delays to intervention commencement

There were unexpected delays to intervention commencement, due to factors mentioned in Chapter 4. The date of intervention commencement was missing for 35 participants (2.9%). We endeavoured to start the intervention within 6 months of randomisation, and this was possible for 986 out of 1185 participants (83.2%): 561 out of 682 (82.3%) in the group intervention arm and 425 out of 503 (84.5%) in the individual intervention arm. The time between randomisation and the start of the intervention for participants randomised to each intervention arm is illustrated in Figure 3.

FIGURE 3.

Time between randomisation and intervention start date (session 0), by intervention arm. Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Participants who were randomised to receive the intervention (in either the group intervention arm or the individual intervention arm) waited a mean of 3.12 ± 2.70 months (minimum 0.16, maximum 15.21, median 2.07, IQR 0.92–4.99 months) from randomisation to their scheduled session 0. The wait time did not differ between the group (3.17 ± 2.78 months; n = 682) and individual (3.04 ± 2.59 months; n = 503) arms [t(1120.5) = 0.83; p = 0.41].

Participant adherence to the intervention

After the completion of each intervention session, the HLF was responsible for recording (1) if the participant set a target for that session (yes/no) and (2) if that target was achieved (coded as no/partially/fully). Session 0 was not included as participants were not asked to set goals for that session.

Of the 1220 participants randomised to the intervention arms, 774 (63.44%) had available adherence data. Of those, 407 (58.39% of those randomised) were in the group intervention arm and 367 (70.17% of those randomised) were in the individual intervention arm.

The 774 participants attended a total of 5932 sessions. Table 8 shows the rate of targets set overall and by intervention arm and Figure 4 shows the number of targets set at each session. The overall rate of targets set differed significantly between arms [χ²(1) = 6.12; p = 0.01] but the distribution of targets set at each session did not differ significantly between arms [χ²(9) = 16.18; p = 0.06].

TABLE 8 - Summary of intervention adherence: target set by intervention arm

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Target set?	Participants randomised to intervention arms with available data, n (%)
	Group (N = 407)	Individual (N = 367)	Total (N = 774)
Target set at session	2582 (90.66)	2852 (92.48)	5434 (91.60)
No target set	266 (9.34)	232 (7.52)	498 (8.40)
Total	2848	3084	5932

FIGURE 4.

Number of targets set, by each session and intervention arm.

For each target set, the HLF also recorded if the participant achieved it; achievement was rated as ‘not at all’, ‘partially’ or ‘fully’. Table 9 shows the number of targets achieved by intervention arm and in total; the distribution of targets achieved (excluding missing data) differed significantly between arms [χ²(2) = 7.25; p = 0.03]. Figure 5 shows the distribution of target achievement at each session number. The rate of target achievement differed by session number [χ²(18) = 116.99; p < 0.0001]. When looking within each session number, the rate of target achievement differed significantly between the intervention arms at sessions 1 [χ²(2) = 6.40; p = 0.04] and 6 [χ²(2) = 6.75; p = 0.03] only.

TABLE 9 - Summary of targets achieved for intervention adherence

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Target achieved?	Targets set by trial arm, n (%)
	Group intervention	Individual intervention	Total
Not at all	143 (5.54)	158 (5.54)	301 (5.54)
Partially	664 (25.72)	679 (23.81)	1343 (24.71)
Fully	1376 (53.29)	1676 (58.77)	3052 (56.16)
Missing rating	399 (15.45)	339 (11.89)	738 (13.58)
Total	2582	2852	5434

FIGURE 5.

Distribution of target achievement, by intervention arm and session number.

Loss to follow-up

Participants were not seen at follow-up for one of three reasons: (1) they had withdrawn from the study, (2) they had died or (3) they were non-contactable at the follow-up due date (attempts were made to contact the participant up to 6 months after the due date). Participants who were randomised to either of the intervention arms but who did not take part in the intervention did not necessarily withdraw from the study, and their data are included at follow-up to carry out the ITT analysis. Table 10 shows loss to follow-up at both the 12-month follow-up and the 24-month follow-up, with the reasons for withdrawal given. Those who were permanently lost to follow-up at 12 months (withdrawn/died) were not contacted again at 24 months, whereas those who were non-contactable at 12 months were contacted again at 24 months.

We exceeded the target follow-up of 83% of participants required by the original sample size calculation. The total loss to follow-up across all arms of the study at 12 months was 11.8%; it was 15.2% in the group intervention arm, 11.9% in the individual intervention arm and 7.3% in the UC arm. The total loss to follow-up across all arms of the study at 24 months was 16.8%; it was 18.4% in the group intervention arm, 19.7% in the individual intervention arm and 11.9% in the UC arm. Data were collected for 91.6% of all participants for at least one of the 12- and 24-month follow-ups, and for 79.7% of participants at both 12 and 24 months. The differences in loss to follow-up between the treatment arms were significant at the 12-month follow-up [χ²(2) = 17.99; p < 0.001] and the 24-month follow-up [χ²(2) = 13.39; p = 0.001]. Withdrawal rates differed significantly between treatment arms at both the 12-month follow-up [χ²(2) = 27.08; p < 0.001] and the 24-month follow-up [χ²(2) = 15.98; p < 0.001].

Accelerometer data completeness

The remaining participants did attend follow-up appointments but did not necessarily complete accelerometer wear for the collection of PA outcomes. Table 11 provides details of the number of accelerometer data collected by trial arm at each time point. The required PA data at baseline was ≥ 5 days of ≥ 540 minutes of accelerometer wear. A number of participants did not wear the accelerometer for the sufficient amount of time at baseline and, therefore, were randomised in error. In the original analysis plan, we stated that we included those with ≥ 4 days of data in the primary analysis to limit exclusions; however, after discussions with the statistician of the TSC (James Carpenter) we included all participants to minimise loss of power and to avoid inducing a potential bias. In agreement with the TSC, for all time points we used the wearable data points if participants wore the accelerometer for at least 1 full day (≥ 540 minutes). Otherwise, the data were considered missing.

The total percentage of missing activity data across all arms of the study at baseline was 0.5% (n = 8 participants) and at 12 months it was 18.6%: 20.1% in the group intervention arm, 20.1% in the individual intervention arm and 15.1% in the UC arm [χ²(2) = 5.9; p = 0.052]. The total percentage of missing activity data across all arms of the study at 24 months was 20.6%: 22.1% in the group intervention arm, 23.3% in the individual intervention arm and 15.9% in the UC arm [χ²(2) = 10.4; p = 0.006]. At least one activity follow-up measurement was available for 88.1% of the study participants: 86.4% in the group intervention arm, 86.6% in the individual intervention arm and 91.8% in the UC arm [χ²(2) = 9.7; p = 0.008].

Baseline characteristics

Tables 12 and 13 present the baseline characteristics and secondary outcome measures, respectively, of participants by treatment arm for baseline comparability. [For baseline values of the primary outcomes, see Tables 16 (PA) and 18 (weight).] A comparison of baseline characteristics between arms did not reveal any major imbalance, including all prespecified variables that may affect primary outcome (age, sex, ethnicity, IMD 2015 score, education status, marital status and smoking status).

Missing data

Patterns of missing follow-up data for the primary outcome were investigated by comparing the baseline data of those with missing primary outcome data at follow-up with the data of those who provided follow-up data. Participants missing 12-month follow-up assessment data walked significantly less at baseline (6509.0 vs. 6802.6 steps per day; t = 1.99; p = 0.048) and were significantly more likely to be current smokers [23.1% vs. 14.3%; χ²(2) = 13.6; p = 0.001]. Participants with missing 24-month follow-up assessment data walked significantly less at baseline (6376.9 vs. 6833.1 steps per day; t = –2.78; p = 0.006), were significantly more likely to be current smokers [22.0% vs. 14.4%; χ²(2) = 6.4; p = 0.041], were significantly more likely to have no formal qualifications [23.8% vs. 33.2%; χ²(2) = 12.3; p = 0.002] and were significantly more likely to be more depressed (PHQ-9 score of 1.89 vs. 2.47; t = 244; p = 0.015). There were no differences between numbers of participants missing at either follow-up point compared with participants who were present in age, sex, weight, BMI, QRISK2 score, relationship status, ethnicity, IMD 2015 score111 or alcohol consumption (AUDIT score).

Primary outcomes

Physical activity

Table 16 presents a descriptive summary of the PA (number of steps) outcome by trial arm and time point. These are the imputed data (methods are described in Chapter 3) and are used for the analyses. Report Supplementary Material 1 presents a descriptive summary of the original (non-imputed) data, which are simply the means of all wearable data points per participant. We observed a near-perfect correlation (r’s > 0.98) between the imputed and original data at each time point. Figure 6 shows the mean PA level with 95% CIs for each arm over time.

TABLE 16 - Summary of PA (number of steps) outcome, by trial arm

Time point	Trial arm			Total (N = 1742)
	Group intervention (N = 697)	Individual intervention (N = 523)	UC (N = 522)
Baseline
n (%)	695 (99.71)	519 (99.24)	520 (99.62)	1734 (99.54)
Mean (SD)	6692.78 (2702.08)	6820.87 (2747.26)	6781.18 (2708.37)	6757.63 (2716.55)
12-month follow-up
n (%)	557 (79.91)	418 (79.92)	443 (84.87)	1418 (81.40)
Mean (SD)	6847.70 (2921.44)	6918.56 (2867.54)	6738.29 (2872.85)	6834.41 (2889.32)
24-month follow-up
n (%)	543 (77.91)	401 (76.67)	439 (84.10)	1383 (79.39)
Mean (SD)	6412.51 (2716.39)	6584.61 (2729.06)	6520.21 (2861.38)	6496.60 (2765.79)

FIGURE 6.

Mean PA (number of steps) with 95% CIs for each arm over time. Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Assessment of assumptions

Assessment of standard deviations between arms and over time

Assessments of SDs reveal only small differences in variances between groups or over time for PA (see Table 16), which further justifies the use of a mixed-effects model with continuous outcome data assuming a multivariate normal distribution for the analysis of PA.

The primary analysis (mixed-effects regression) of the treatment effect on PA was conducted on 2801 observations. There were 133 unique general practices, with a mean of 21.1 observations for each (range 1–99 observations). There were 1534 unique participant identifiers, with a mean of 1.8 observations for each (range 1–2 observations). Table 17 presents the pairwise comparisons. The fixed and random-effects outputs can be found in Report Supplementary Material 1.

There were only minor and non-significant differences between treatment arms in PA levels at 12 or 24 months. At 12 months, compared with participants in the UC arm, participants in the individual intervention arm walked, on average, 210 steps (95% CI –19.5 to 439.9 steps) more and those in the group intervention arm walked, on average, 131 steps (95% CI –85.3 to 347.5 steps) more. All differences (including limits of the 95% CIs) were less than the minimum clinically significant difference (MCD) of 675 steps as defined in the study protocol. Similarly, at the 12-month follow-up and using 97.5% CIs, we observed minor and non-significant differences between the individual and UC arms (mean difference 210.22 steps, 97.5% CI –52.44 to 472.89 steps) and the group and UC arms (mean difference 131.10 steps, 97.5% CI –116.35 to 378.55 steps).

Weight

Table 18 presents a descriptive summary of weight by trial arm and at each time point. Figure 7 shows the mean weight with 95% CIs for each arm over time.

FIGURE 7.

Mean weight with 95% CIs for each arm over time. Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Reproduced from Ismail et al. 110 © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

Assessment of assumptions

Assessment of standard deviations between arms and over time

Assessments of SDs reveal little differences in variances between groups or over time for each scale (see Table 18), which further justifies the use of a mixed-effects model with continuous outcome data, assuming a multivariate normal distribution for the analysis of weight.

The primary analysis (mixed-effects regression) of the treatment effect on weight was conducted on 2981 observations. There were 133 unique general practices, with a mean of 22.4 observations for each (range 2–104 observations). There were 1595 unique participant identifiers, with a mean of 1.9 observations for each (range 1–2 observations). Table 17 presents the pairwise comparisons. The fixed and random-effects outputs can be found in Report Supplementary Material 1.

There was a small but significant mean difference between the individual and UC arms of –0.55 kg (95% CI –0.95 to –0.14 kg) and between the group and UC arms of –0.52 kg (95% CI –0.90 to –0.13 kg). However, the differences (including the 95% CI limits) are below the MCD of 1.25 kg. There was no difference between the group and individual intervention arms (mean –0.03 kg, 95% CI –0.43 to 0.37 kg). At 24 months, no significant differences were observed. Similarly, at the 12-month follow-up and using 97.5% CIs, we observed minor differences between the individual and UC arms (mean difference –0.55 kg, 97.5% CI –1.01 to –0.08 kg) and the group and UC arms (mean difference –0.52 kg, 97.5% CI –0.96 to –0.08 kg).

Sensitivity analyses

Secondary outcomes

Table 19 presents the pairwise comparison output for each of the secondary outcomes. We did not observe any treatment effects for any of the secondary outcomes at 12 or 24 months.

Summary

The health economic analyses show that the group intervention is not a cost-effective alternative to UC as it produces fewer QALYs and results in higher health-care costs. The individual intervention is more effective than UC in terms of QALYs produced but the higher costs result in an ICER that is in excess of the £20,000- to £30,000-per-QALY threshold usually used in England to determine cost-effectiveness and so is also not cost-effective. The ICER for the individual intervention arm compared with the group intervention arm is below this threshold, but UC is still the preferred option. Sensitivity analyses did not alter the results in a substantial way.

Background

Lifestyle intervention trials have previously reported low response rates and do not always recruit those most at risk, which will limit the generalisability of observed associations. 112,113 Failure to recruit those with the highest risk of CVD to the MOVE IT trial could have led to limited representativeness and underestimated effect sizes, and could have contributed to increasing rather than reducing health inequalities. It is important, therefore, that RCTs such as MOVE IT report on participation bias.

The factors found to increase likelihood of participation in walking and lifestyle intervention trials are mid-life, female sex, living in more affluent areas, white ethnicity and university education. 112,114 Ethnic minorities are at higher risk of CVD and related disorders, such as type 2 diabetes mellitus. 115 Regarding the health of participants versus non-participants, some trials report that participants are more likely to be of poorest health,112,116–118 whereas other studies report that participants are healthier or more active. 113,117,119,120

We compared the anonymised sociodemographic data and CVD risk of responders and non-responders to the MOVE IT invitation to participate. We aim to test the hypothesis that people who are older, female, living in more affluent areas, at lower risk of CVD and of white ethnicity are more likely to respond.

Methods

Using a cross-sectional design, we compared anonymous data from medical records of patients who responded to the invitation to participate in MOVE IT and those of patients who did not. We gained REC approval to extract anonymised data for all patients invited to participate unless an informed dissent code was present on medical records.

The measures collected anonymously for all patients invited to participate included age (at time of screening), sex, postcode data to calculate IMD 2010 score,76 QRISK2 score and ethnicity.

QRISK2 score was estimated on medical records via a batch calculator, using age- and sex-based national averages for missing data values.

Ethnicity data from medical records were grouped into white, black African or Caribbean, South Asian, other Asian, other/mixed or missing for this analysis. South Asian and other Asian ethnicities are coded separately because South Asian ethnicity is associated with a higher CVD risk. 72 The category of other/mixed incorporates any ethnicity that is reported and does not fit into the previous categories, as well as including mixed ethnic backgrounds.

Data are summarised as mean (SD) or percentages. The adjusted odds ratios (AORs) for response to invitation by the sociodemographic and CVD risk data collected were calculated using logistic regression in Stata version 14. General practice was included as a random effect in the model to allow for clustering by practice. AORs for age are presented per 5-year increase, for IMD 2010 scores per 10-point increase and for QRISK2 scores per 5-percentage-point increase to provide a better comparison of the strength of the relationships.

Results

Lists of patients invited to participate were not saved by general practices and we returned to practices later to extract anonymous data. Sixty practices provided the original anonymised data. Sociodemographic data and QRISK2 scores were extracted for 8902 patients, representing 50.5% of the 17,618 patients originally invited to participate. Table 27 shows the CCGs from which anonymised data were collected and the deprivation and ethnicity census data of each local authority. 71 General practice deprivation scores (IMD 2010)76 of practices from which anonymised data were extracted did not differ significantly from that of all other practices in South London [t(440) = 0.57; p = 0.57). 108

The sociodemographic and QRISK2 data of responders (n = 1489) and non-responders (n = 7413) are presented in Table 28. The mean age of all patients at the time of invitation was 67.3 ± 5.7 years, and 20.7% were female. The mean IMD 201076 score of patients invited was 21.7, a score that falls within the second-most-deprived quintile, and the mean QRISK2 score of invitees was 25.2%. Of all patients invited, 69.9% had a white ethnic background, 13.9% had ethnic minority backgrounds and 16.2% had no ethnicity data recorded in their medical records.

The logistic regression analyses that were conducted to estimate the odds of response are also presented in Table 28. Response rates were higher with increasing age (AOR 1.19, 95% CI 1.12 to 1.26); the odds of response increased by 19% for each 5-year increase in age. The response rate was greater in male patients (AOR 1.24, 95% CI 1.07 to 1.44) and decreased with greater levels of deprivation (AOR 0.84, 95% CI 0.79 to 0.90); for every 10-point increase in IMD 2010 score, the odds of responding decreased by 16%. Increased CVD risk also lowered the response rate (AOR 0.83, 95% CI 0.77 to 0.88); for every 5-point increase in QRISK2 score, the odds of responding decreased by 17%. The response rate was lower for patients of black African or Caribbean ethnicity (AOR 0.66, 95% CI 0.45 to 0.96) and those with missing ethnicity data (AOR 0.55, 95% CI 0.46 to 0.66) than for patients of white ethnicity. The response rates of Asian and other ethnic backgrounds were not significantly different from those of patients of white ethnicity.

Pairwise comparisons were conducted to explore differences in response between different ethnic minority groups, but there were no significant differences. Response rates of patients with missing ethnicity data were significantly lower than those of patients of South Asian ethnicity (p = 0.002). Owing to small numbers of invitees from non-white ethnic backgrounds, a sensitivity analysis was undertaken to investigate predictors of response to invitation with ethnicity removed from the model. Older age (p < 0.001), male sex (p = 0.017), being more affluent (p < 0.001) and having a lower CVD risk (p < 0.001) remained significant predictors of response.

Summary

In Chapter 5, we demonstrated that the trial successfully recruited general practices that did not differ significantly from all other practices in South London in practice deprivation, and this was also true for the smaller group of practices involved in this analysis. Therefore, the MOVE IT trial successfully recruited from socioeconomically varied areas. For this analysis, we found evidence of participation bias towards patients who were older, male, residing in more affluent areas and at a lower CVD risk and we found that patients of black African or Caribbean ethnicity were less likely to respond than patients of white ethnicity. South Asian patients were as likely to respond as patients of white ethnicity but the QRISK2 score identified a smaller sample size of people of African Caribbean and Asian ethnicities. This could be partly because those CCGs with the highest IMD 2010 scores and greater ethnic diversity were less likely to participate in MOVE IT. It is noteworthy that these CCGs were the slowest to respond to participate although all CCGs were invited at the same time. We also observed high rates of missing ethnicity data on medical records and this group of patients with missing ethnicity data was less likely to respond than patients of white or South Asian ethnicity. These findings suggest that there may have been a participation bias towards those with lower CVD risk compared with those most at risk.

Introduction

As part of the process evaluation of the study, we assessed fidelity, here defined as delivery of the enhanced MI intervention by the HLFs in accordance with the manual. Furthermore, we were interested in determining if the level of HLF competency in delivering the intervention differed between the individual and group intervention arms.

As described in Chapter 3, HLFs received extensive training and supervision before and during the intervention to ensure that they were competent in delivering MOVE IT. HLFs were asked to audiotape each intervention session, and these data were stored on King’s College London’s shared network drives, as per the university protocol. Chapter 4 describes how a university-wide network outage caused the recorded session data to be completely lost, with only a limited sample being subsequently recovered from other sources. In this chapter, we will describe our original aims and methods and how these were adapted following the network outage.

Aims

For the analysis, we were originally interested in answering five questions:

1. What was the degree of adherence by the HLFs to the MI and BCT elements in the group and individual intervention arms?

2. What was the level of HLF competency in delivering the MI aspects?

3. What was the level of HLF competency in delivering the BCTs?

4. Were there any differences in competency in delivering the MI and BCT elements between the group and individual treatment arms?

5. Were the HLF competency levels in delivering the intervention associated with patient outcomes or mediators of any treatment effects?

Owing to the network outage and the number and quality of data that were subsequently recovered, we were unable to address question 5. In addition, although the HLFs were trained to a certain level of competency in two other domains (group skills and time management; see Chapter 3), we were not interested in assessing their competency in these domains for this fidelity analysis.

Original methods

Revised methods

Results

Discussion

We coded 395 recorded intervention sessions for fidelity. We observed that nearly all (11/13) of the HLFs achieved high levels of competency on at least one dimension of MI but fewer did so with the BCT, in which only 2 out of 13 HLFs achieved competency. The HLFs received extensive training in delivering the MI and BCT elements of MOVE IT and were supervised throughout the trial; our analysis of a non-random and probably underpowered sample of tapes indicates that the intervention was administered with acceptable levels of competencies in MI but the competency of HLFs in delivering the BCTs is not known.

When combining MITI scores across HLFs, we found that the MI aspects of the intervention were not delivered at the desired competency level. At the HLF level, none of the HLFs reached the minimum MITI-based proficiency level and only three met the competency criteria adapted for the study. By further stratifying by session type, however, we found that two HLFs delivered the MI elements in individual sessions at a level deemed competent for our study.

We observed treatment arm differences in several of the MITI domains. Most of these differences were in domains with count scores, hence these were expected given the longer length of the group session format (i.e. HLFs had more opportunity to use the technique repeatedly). Of particular interest, HLFs in group sessions demonstrated a greater percentage of open questions, whereas in individual sessions they made more reflections relative to questions. This may be because in group sessions a large proportion of time was spent facilitating group discussion, which lends to repeated use of open questions. In addition, in group sessions there may have been less opportunity for HLFs to ask closed questions, which naturally follow open questions in individual sessions (e.g. to clarify what the individual has just said or encourage further conversation). In terms of the reflection-to-question ratio, this could be explained by the fact that reflections require a deeper level of listening and understanding, which the group session format does not lend itself to as easily as individual sessions do.

Most HLFs were ‘partially proficient’ in delivering BCTs. Only three BCTs (‘prompt intention formation’, ‘prompt specific goal-setting’ and ‘agree on behavioural contract’) were administered in > 70% of sessions. Furthermore, two BCTs (‘self-talk’ and ‘relapse prevention’) were delivered in approximately < 30% of sessions. Across the BCTs defined for the trial, we observed treatment arm differences in only one (‘provide information on consequences’); however, this is possibly a false positive owing to the number of tests conducted. Finally, only two HLFs delivered, on average, ≥ 70% of the BCTs across the sessions assessed (however, this was only five sessions for one HLF and three individual sessions for the other). These results should be interpreted cautiously as we used an unvalidated BCT coding framework developed for this study; thus, we may have underestimated HLFs’ competency in delivering the BCTs.

Background

Randomised controlled trials provide a reliable and rigorous method of assessing the effectiveness of an intervention. 125 However, owing to their complex nature there are a number of additional challenges that may arise when reproducing the intervention. Therefore, it is important to evaluate the delivery and processes of the intervention from the patient’s perspective. 126

Focus groups give participants the opportunity to provide in-depth reflections of their experiences and views of the intervention. 127,128 Understanding the experiences of participants is vital to provide an insight into potential barriers and improvements that often occur in lifestyle interventions. 129 Therefore, the use of focus groups can enable researchers to review the different behavioural or social processes that may not be evident from the use of quantitative analysis only, leading to the successful implementation of an intervention in the future. 126

A recent systematic review and meta-analysis revealed the effectiveness of structured lifestyle interventions that target both diet and PA as being successful in reducing CVD risk factors. 130 Studies have found that the delivery of MI for health-related behaviour change in lifestyle interventions to be beneficial, particularly in the short term. 49,131,132 However, few have investigated the maintenance of behaviour change in the long term with the use of qualitative methodology.

The aim of this qualitative analysis was to explore the views of participants of MOVE IT in terms of reasons for engagement, factors enabling behaviour change and the potential barriers to engagement.

Methods

Two groups of participants were approached for process evaluation: (1) those who attended the intensive phase of the intervention, that is, six or more sessions (completers), and (2) those who withdrew before the intensive phase was completed or did not attend at all (non-completers). Participants were invited, based on a purposive sampling method to form heterogeneous groups. Only those who had completed all three follow-up appointments were approached.

Participants were separated into six focus groups: four groups of completers (labelled as 1A–1D) and two groups of non-completers (2A and 2B). Each group had two facilitators, who were research assistants from the MOVE IT programme and had not met the patients before, to avoid response bias. For each group, we aimed to include at least one female participant, an individual with an ethnic minority background and a mix of those who attended individual and group sessions.

Focus groups using a semistructured interview schedule were carried out with the use of a topic guide (see Appendix 4). The topic guide was informed by preliminary analysis from the HLF contribution, input from the research team and investigators, and feedback from PPI members. It consisted of open-ended questions relating to (1) the invitation to take part in the study, (2) lifestyle changes made since joining the MOVE IT programme, (3) most and least helpful aspects of the sessions, (4) participant diversity and (5) potential online interventions.

Informed consent was obtained and each group was audiotaped. Data were anonymised and transcribed verbatim. The transcripts were analysed using thematic analysis. 133 The analysis was descriptive and interpretative and facilitated by the use of NVivo version 10 (QSR International, Melbourne, VIC, Australia) to code emerging themes and subthemes.

Results

A total of 74 participants were invited to attend a focus group, with 26 ultimately participating. There were four groups of completers and two groups of non-completers. The characteristics of those who took part are outlined in Appendix 5.

A thematic analysis of all six focus groups resulted in three generic themes. A fourth theme was generated from only the non-completers’ responses, as shown in Table 37.

Summary

This qualitative analysis explored the views of participants allocated to the intervention arms of MOVE IT. The main themes that emerged were (1) perceived benefits of the study, (2) factors enhancing behaviour change and (3) perceived risk of CVD. One further theme that emerged solely for the non-completers was (4) potential barriers to change and overcoming these barriers.

We found that participants described having positive experiences during the intervention sessions, citing the benefits of increased health awareness, positive lifestyle changes and the opportunity to learn from others. We identified several factors that might have affected the maintenance of participant behaviour change. Primarily, participants noted that the continuity of sessions and having the staff involved throughout were very important. The changes to the HLFs who carried out the sessions were disruptive for some and led to reduced engagement with the intervention. The continuity and frequency of sessions also had an impact on participants’ motivation to take part and make healthy lifestyle changes. Owing to the pragmatic design, duration of the intervention and short-term contracts of the HLFs, there was a natural staff turnover that could not be prevented, except with a vast increase in funding to potentially hire more HLFs or offer longer contracts.

Non-completers had expressed that there was a lack of feedback and engagement throughout the intervention. To improve this, it would be beneficial to provide more positive feedback from HLFs or to have an alternative method of self-reflection during the sessions. This would increase motivation for patients to continue participation and for behaviour change to be sustained. Participants’ perceptions of their own CVD risk may be viewed as a barrier to engagement.

Most participants did not believe CVD risk to be an issue and often attributed the risk to their age rather than their own lifestyle choices. In addition, some GPs were not aware of their patients’ risks or the study itself, which may have altered individuals’ perceptions of CVD risks. Therefore, for future interventions, it is important to work more closely with GPs to emphasise the benefits for patients’ health and potential to ameliorate CVD risk. There may be other reasons why the non-completers felt a lack of engagement, as they may have higher levels of psychological distress or social problems that were not articulated during the focus group.

MOVE IT emphasised the importance of the use of qualitative methodology in providing detailed experiences of participants for assessing the effectiveness of a RCT. As previously reported, studies have supported the benefits of lifestyle interventions in the short term but few have supported them in the long term. 49,131,132

The detailed reflections from participants involved in a long-term intervention support the use of qualitative methods for gaining greater understanding of the factors facilitating engagement and barriers to engagement. The responses from participants highlight the benefits of lifestyle interventions and the maintenance of behaviour changes for diet and PA levels. Finally, this study reveals areas of improvement for the future delivery and implementation of the study, including continuity of care and duration of the intervention. Although attendance at group sessions was lower than attendance at individual sessions, those who attended group sessions appeared to value the peer support aspect.

Background

The HLF team were trained by a CP to deliver MOVE IT. Training in MI involves achieving a number of core competencies,134 and HLFs were also trained to incorporate BCTs into intervention delivery. 52 There is no fixed length for the training of MI; however, a review of 28 studies found that most training schedules required 9–16 hours of training. 135 The training methods for MI involve didactic supervision and learning activities, such as role play, group exercises, discussions and feedback. 42

The application of MI is considered helpful in working with patients towards lifestyle changes in primary care by nurses who incorporate it into their practice. 136,137 Previous studies of the application of MI in practice have suggested that it can be demanding but rewarding and that the spirit of MI and simple competencies are frequently achieved, but that the application of MI skills reduces over time and further support is required for more complicated competencies. 136,138 The ability to reflect on practice, continuing to develop skills and ongoing supervision are thought to be crucial to enhancing the long-term benefits of a MI approach. 139 There is evidence that enhancing MI with CBT techniques is achievable for nurses supporting diabetes mellitus control140 or alcohol cessation,62 and that intervention providers perceive both positive and negative impacts of the intervention on patient care depending on the patient’s ability to engage. 141 Evaluating any novel intervention, from the perspective of stakeholders such as the intervention providers, is important for translation to policy. 142

The views and experiences of intervention providers are vital to assessing the feasibility of widespread implementation into practice, by exploring the challenges in training and delivery. The HLF role was newly developed for the MOVE IT trial, drawing on the promising preliminary findings of the NHS health trainer programme, and adapted to be in line with MI, which underpins the intervention. Therefore, as the intervention and job role are both novel, we sought to gain detailed feedback on HLF training and intervention delivery from both the CPs who were responsible for training and supervision and the HLF team.

Methods

A qualitative approach was adopted to study HLFs’ and CPs’ views and experiences of developing and delivering the interventions. All 13 HLFs and two CPs were invited to interviews. Semistructured interviews were carried out, consisting of open-ended questions relating to the training, supervision, intervention, participants and values in psychological therapies. The topic guides used for HLFs and CPs are provided in Appendix 6. Interviewees were invited to give their feedback and the discussion was audiotaped. To gather meaning from the data, inductive thematic analysis explored any underlying concepts that were mentioned regularly.

The interviews were conducted by an independent research assistant to avoid response bias, which could exist if the interviewer was part of the MOVE IT team. Audiotapes were transcribed verbatim and the data collected were anonymised. NVivo software was used to analyse the data. Emerging themes from the interview transcripts were coded and categorised into themes and subthemes.

Results

A total of nine female health-care professionals (seven HLFs and the two CPs) were interviewed. All HLFs except one were working on the MOVE IT trial at the time of their interviews. The HLFs who no longer worked on the trial were unavailable to participate. The characteristics of those interviewed are given in Table 38.

A thematic analysis resulted in three generic themes:

1. challenges and suggested improvement for the training

2. supervision issues and peer support

3. challenges with delivering the intervention.

Theme 3: challenges with delivering the intervention

Summary

This qualitative analysis explored the views of HLFs and CPs regarding the process of implementing MOVE IT. Of all 13 HLFs invited to interviews, we gained feedback from seven; we also gained feedback from both of the CPs who trained and supervised the HLF team. The main themes that emerged were (1) challenges faced during training and suggested improvements, (2) issues with supervision and peer support and (3) difficulties in delivering the intervention.

The HLFs enjoyed learning how to implement MOVE IT in addition to working with the patients. The training was described as time consuming and arduous, and HLFs felt that it could have been completed in a shorter time allowing more time for hands-on learning. Although there was a lot of information to cover, none of the HLFs mentioned a lack of confidence in the delivery of the intervention. With regards to the relationship with their supervisors, many of the HLFs felt that the fortnightly sessions were not utilised efficiently or that they could become repetitive.

The HLFs developed a strong support network between one another, and the relationship with their superiors improved once a HLF was promoted and could mediate between them, making better use of their time. It was noted that the activity booklets were age-inappropriate for the participants involved, which became a barrier during some sessions. In addition, some participants assumed that the HLFs were medical professionals who were able to discuss other health complaints and HLFs would have to reiterate that they were not medically trained while ensuring that the participants did not lose confidence in the help that they could provide. Both CPs noted the proficiency of the HLFs in working around these difficulties.

Enhanced MI was reported as an enriching and effective approach. The training undertaken by the HLFs in this study could have been adapted to allow the HLFs to learn via experience as opposed to continual rehearsal. This analysis also demonstrated the importance of effective and useful communication and support between HLFs and those managing them. It is therefore important to consider a practical and ongoing support structure. Furthermore, it may be useful to adapt the tools used in such an intervention to the age of the target group to remove potential barriers to engagement.

Summary of the clinical effectiveness of MOVE IT

This three-arm parallel RCT study tested the effectiveness of an enhanced MI intervention, delivered by specially trained health trainers (HLFs), in a group versus individual format, versus UC for reducing weight and increasing PA in adults at high risk (≥ 20.0%) of developing CVD in the next 10 years.

For the primary objective, the main findings were that there were no significant changes in weight or PA 24 months later between group enhanced MI or individual enhanced MI compared with UC.

For the secondary objectives, we found that MOVE IT delivered in a group format was not more effective than when delivered in the individual format in reducing weight and increasing PA or in reducing LDL cholesterol or CVD risk score 24 months later. There were no differences in the number of fatal or non-fatal cardiovascular events and other recorded AEs between the three treatment arms.

The health economic results revealed that there was little difference in terms of service use and costs between the three arms other than that resulting from the interventions themselves. Total service costs over the follow-up period were highest for the individual intervention arm, followed by the group intervention arm, followed by the UC arm. Differences were not statistically significant. QALYs were very similar for each arm. The group intervention was dominated by UC in that the latter had lower costs and produced more QALYs. The individual intervention did produce more QALYs than UC but the ICER indicated a cost per QALY far in excess of the threshold commonly used by NICE. For this reason, neither form of the intervention was cost-effective. There was much uncertainty around the cost and outcome differences but the conclusions of a lack of cost-effectiveness hold.

The findings of the process evaluation were as follows.

• Mediators: we found that dietary changes did not mediate any treatment effects on the primary outcomes. We did not conduct any further mediational analyses, as there was no change in the primary or secondary outcomes.

• Participation bias: we found that there was significant evidence of reduced reach in that those with a higher CVD risk, higher level of deprivation status and of African-Caribbean ethnicity were less likely to reply to invitations to participate from their general practice.

• Fidelity analysis: there were significant methodological limitations of conducting a fidelity analysis but, based on the much-reduced highly selected sample, there was evidence that nearly all of the HLFs had sufficient competencies in at least one MI skillset.

• Patient experience: the main themes that emerged were (1) perceived benefits of the study (benefits of increased health awareness, positive lifestyle changes and the opportunity to learn from others), (2) factors enhancing behaviour change (continuity of sessions over a longer period and having continuity of the same HLF) and (3) perceived risk of CVD (this was lower than was expected). One further theme that emerged for the non-completers only was (4) potential barriers to change and overcoming these barriers, such as lack of feedback and lack of engagement by their GPs.

• Therapist experience: the overall view was that the formal training period could have been shortened, more exposure to training cases could have been provided and the HLFs had not been prepared for the real-world challenges once in the clinical setting. They perceived themselves as competent in the MI approach and BCTs. They observed the importance of working with patients towards their goals but there were some common challenges, such as patients not engaging and some of the intervention materials not being deemed age-appropriate. The HLFs felt that support from supervisors, and administrative support, was insufficient but that they could problem-solve by supporting each other.

Strengths and limitations

This RCT was powered to detect small effect sizes and, therefore, we can be confident that our (negative) findings are statistically valid. We used accelerometers as the gold standard for measuring PA. However, there may still have been a Hawthorne effect because the participants were aware of equipment and were not blind to allocation as this was a talking therapy. If there was a Hawthorne effect of wearing the accelerometer then it was negligible as the overall effect of the intervention was minimal on PA and weight loss. We had greater attrition for complete PA follow-up data. Previous research has set a standard for the amount of accelerometer wear time required per wear-day owing to variability between days of the week. 143 We found that those not providing sufficient data at follow-up were less active and more likely to smoke, had no formal educational qualifications and had more depressive symptoms. However, in the Health Survey for England, participants who provided sufficient accelerometer wear time did not differ from those who did not wear their accelerometer for the requested time. 144 As participants in our recruited sample were older than the target population, the use of fewer days of accelerometer wear data may be acceptable, as previous work has detailed that 3 consecutive wear-days accurately predicts PA levels in older adults. 145 Analysis of sedentary behaviour can require more wear-days, but this was not a main outcome of our study.

We recruited from all 12 CCGs covering the South London region, which has > 3 million residents representing a diverse socioeconomic and ethnically diverse community, including some inner-city London boroughs with high rates of CVD mortality. This was a positive aspect of the study. On the other hand, we did not include residents from the rest of the UK where there are high rates of CVD mortality, such as the north of England, south Wales and central Scotland, and therefore this sample may not have been representative of all people at high CVD risk. 146 Another limitation is that the South London sample may not be representative of the rest of the UK (i.e. owing to the health inequalities between the north and south of the UK). In addition, those London CCGs with more deprivation were slower to respond to the invitation and therefore when we were already halfway through recruitment and by the time we had completed recruitment, less time had been spent recruiting from these CCGs. This may also have led to under-recruitment of those most at risk, namely from deprived areas and non-white ethnic groups.

A pragmatic feature and strength of this study included using the GP QRISK2 tool, which is routinely available on the electronic records software for calculating CVD risk. Although this was a strength because it is an objective measure of risk, which could potentially be easily used by GPs, the limitation was that it had a high false-positive rate, generating a very high denominator number of patients, all of whom were invited for consent for screening for eligibility. The accuracy, completeness and age of data required for the QRISK2 algorithm based on the data stored in medical records were variable. This resulted in inviting and consenting many patients who turned out to have lower and therefore ineligible QRISK2 scores when screened. We had to assess an additional 1332 patients who turned out to be ineligible to participate, which was costly and time consuming. However, there were few other options. The NHS Health Check programme was not appropriate as it achieved a lower-than-expected uptake across England. 12

A limitation of the study was that there were delays in recruitment that had repercussions in that the intervention delivery was delayed for some participants. The unexpected inefficiency in using the NHS Health Check led to delaying recruitment. Switching recruitment strategy to using the QRISK2 tool introduced another delay, which had repercussions for the delivery of the intervention. This is because the switch from Health Checks to QRISK2 required a protocol change, ethics approval and application of a cost extension from NIHR. The delays in recruitment and uncertainty of funding extension approval resulted in most of the HLFs seeking jobs elsewhere as their contracts started coming to an end. At around the same time, the co-applicant’s clinical health psychologist also resigned. We had to employ and complete the induction of a second clinical health psychologist and train a new cohort of HLFs. This meant that when recruitment returned once additional funding and ethics approval had been secured there was a waiting list for delivery of the intervention. This may have contributed to some of the dropout from the intervention.

A strength was that our intervention was embedded and delivered in the heart of the community to match the real-world setting and we were able to be flexible with timing of the sessions (including late afternoons and early evenings). Our intervention focused on supporting maintenance once changes had been made in the initial phases. However, a limitation was that we could not offer sessions immediately and there was a waiting list for patients while HLFs were trained to competency. Approximately 50% of participants randomised to the individual and group interventions dropped out of the intervention after the introductory session. As most of our participants were retired, providing flexibility in the timing of sessions around work schedules was less important. We observed that participants in both the group and individual intervention arms generally adhered well to the intervention, based on the HLFs asking participants to set targets at each session and record if these were achieved or not.

We developed a standardised health trainer manual that could be readily replicated and used for training and supervision. A limitation was that the HLFs preferred practical training rather than didactic learning and classroom-based role play. We would have preferred to have the HLFs undertake more practical training with patients and this might have improved competencies, but this was impossible in terms of resources and the inherent nature of delivering a complex intervention as a research study. The HLFs were employed as NHS staff on short-term contracts; thus, we would have had to set up a CVD risk clinic very quickly, screen patients and then offer the intervention. As this was a 12-month intervention, this was not feasible.

We conducted a comprehensive process evaluation to understand the underlying mechanisms of action. We ensured a long follow-up to capture both short-term and long-term, or maintenance, outcomes. As a result of the process evaluation, we found that we did not reach those at the highest risk of CVD, participants who did not complete the intervention expressed lack of feedback and engagement as a reason, and participants attributed their high CVD risk to their age rather than their own lifestyle choices. Although we did observe several participation biases (see Chapter 7), it is unclear if more specific targeting of the intervention would have had a clinical effect, given that the absolute differences observed were small and a large sample was recruited. In addition, participants noted a lack of engagement from their GP concerning their high CVD risk, which may have negatively influenced any treatment effects. However, given the large number of GPs recruited to the study, it would not have been feasible to incorporate a larger amount of GP engagement in the study design.

Our PPI group was interactive and involved throughout. There are a number of areas in which the methodology of the trial may be improved. For instance, the importance of the invitation process and materials were highlighted by both PPI members and the intervention providers, and appropriate targeting of those who may benefit most from participation in the trial may improve the reach of the study.

There were limitations in the cost-effectiveness analyses. We relied on self-report of service use information and this may have led to recall inaccuracy. This was largely unavoidable given the range of services we wished to include in the measure of cost. Other studies have found this approach to be reasonable and there is no reason to suppose that any inaccuracy biased the findings. 147,148 A further limitation was that the EQ-5D-3L may not have been sensitive to change in this patient group. It is the recommended measure for such analyses but further work on assessing its appropriateness is warranted. The lack of differences in QALYs does support the main clinical findings of the study. Finally, we had hoped to include lost employment costs in a societal perspective; however, data on this were not available.

Interpretation

There are several possible explanations for the results of this trial: (1) although the clinical population was at high risk of CVD, the risk factors were not modifiable by lifestyle interventions, or (2) the psychological theories were not the most appropriate and/or psychotherapeutic approaches are not sufficiently potent alone.

One explanation for the negative findings is that we may not have recruited those with high CVD risk who had modifiable risk factors. The use of a QRISK2 score of ≥ 20.0% ensured that our study recruited those at the highest risk of CVD in accordance with an objective risk algorithm that takes into account numerous modifiable risk factors. However, the risk factors with the greatest weighting in the algorithm are the non-modifiable risk factors: age, sex and ethnicity. 20 It has previously been shown that risk calculators, including the QRISK2, may not have the specificity to identify CVD risk at the individual, as opposed to the population, level. 149 Those patients of older age, male sex and South Asian ethnicity are at the highest risk, but these individuals may be otherwise healthy in terms of their modifiable risk factors that are inputted into the algorithm, and therefore the aims and contents of the intervention may not be applicable to them. The use of a QRISK2 score of ≥ 20.0% may have meant that the recruitment strategy led to a sample skewed against those with high BMIs, blood pressure and LDL cholesterol levels (exactly those participants who were most likely to benefit as these risk factors are modifiable).

Indeed, we found that older people with a normal or underweight BMI were recruited, raising challenges for the intervention team considering the primary aim of a reduction in weight. Previous research has demonstrated that participants at the highest risk benefit most from preventative interventions. In Gidlow et al. ,69 an observational study based in primary care in Stoke-on-Trent, followed three groups of patients: those with at least one modifiable CVD risk factor using the Framingham risk engine score, those with diabetes mellitus or coronary heart disease (secondary prevention) and those who were obese and had a CVD risk of > 15%. They received a patient-tailored, health-trainer-delivered, lifestyle intervention over 6–12 months with face-to-face and telephone/SMS (short message service) support. They observed modest improvements in risk factors and health-related quality of life, but not in CVD risk score, except in the subsample of patients who were at highest risk (> 20%) at baseline. In the study by Hardcastle et al. ,49 the investigators focused on CVD risk factors, such as excess weight (BMI of ≥ 28 kg/m², based on a value used in the recruiting general practice), hypertension (systolic/diastolic blood pressure of ≥ 150/90 mmHg) or hypercholesterolemia (≥ 5.2 mmol/l), rather than CVD risk score. They observed that the MI intervention was particularly effective for patients with elevated numbers of CVD risk factors at baseline.

The participants in this trial had average step counts at baseline that were in line with that previously reported for healthy older adults who were participating in exercise classes. 150 Therefore, on average, the participants recruited to MOVE IT may at baseline have been at their ceiling in terms of PA achieved, leaving little room for improvement. Previous studies of PA interventions for older adults, using accelerometer measurement as an outcome measure, tend to report significant increases up to a 12-month follow-up but there have been few psychological interventions to help maintain the increased PA. 151 Another explanation for the negative findings is that patients in our sample were not obese. The association between increasing PA and decreasing HbA_1c level following behavioural intervention is stronger in those who are obese (BMI of ≥ 30 kg/m²),152 and it may be that our sample was not at a high enough risk (the mean BMI was 28.3 kg/m²) to see significance in outcomes.

The objective accelerometer measurement of PA provides a more reliable measure than self-report as it is less prone to bias. However, the days for which the PA measure is collected may not represent normal PA levels of participants; they may, for example, be elevated above the usual level owing to social desirability. 153 A review of the impact of interventions combining BCTs that target diet and PA found no differences in PA outcomes with UC when self-report PA outcome measures were used. 154 These studies did find a significant impact on BMI, suggesting that beneficial effects of the intervention had taken place and that perhaps a change in PA levels could not be successfully captured by the self-report measure or at the long-term follow-up. It may be that participants in our study made some initial increases in PA levels that were not captured at the 12- and 24-month follow-ups through accelerometry or at baseline; in addition, outcome PA measures of all participants may have been artificially raised. To further investigate the impact of interventions on PA, studies would need to undertake continuous objective measurement throughout participation, as opposed to the norm of 4–7 consecutive days’ wear. 143 Low-cost and unobtrusive means to monitor PA throughout participation in a RCT, such as wearable and smartphone technology, provide this opportunity, although the reliability and validity of these methods are yet to be assessed. 155,156

A second possible explanation for the negative findings is that the intervention potency or overall dose was suboptimal. The psychological skills were based on sound well-established psychological theories: (1) the theory of planned behaviour for initiation of behaviour change, which states that in order to change behaviour, people need to form an intention,53 (2) principles underpinning MI that emphasise that the patient’s motivation to change can be used as a therapeutic tool,41 (3) CBT91 and (4) social cognitive theory,92 which emphasises the importance of peer support in shaping behaviour, which was applied to the group intervention arm of MOVE IT.

We used well-established and evidence-based psychotherapeutic techniques to apply the theoretical framework for MOVE IT. MI was used to support participants in forming healthy intentions. MI is a collaborative conversation style for strengthening a person’s own motivation, belief and commitment to change. We used principles from CBT to support the transition from intention to action and from action to maintenance93 by identifying and challenging unhelpful thoughts or thinking styles and promoting more positive emotions and behaviours. Using an iterative process, we devised the intervention curriculum and a participant workbook to support training, supervision, collaboration and replication. We had an intensive phase and the maintenance phase consisted of four sessions delivered at 3, 6, 9 and 12 months from intervention commencement. Those randomised to the group intervention arm were encouraged to use peer learning and the peer-support environment to facilitate change during both the intensive and maintenance phases. Novel methods and teaching aids were used to supplement the delivery of BCTs, such as visual aids (food labels)/cue cards, exercise demonstrations, video/audio material of patient testimonials, activity-based learning around meal planning and SMS/e-mail reminders. Cultural and religious awareness is built in to the intervention.

We employed and trained health trainers with an intensive package of didactic learning, role playing and feedback, group exercises, reading and case study discussion over 3 months. Competency in administering the intervention was assessed through a knowledge quiz and delivering several sessions to volunteers. 94–96 For HLFs that were not competent, they underwent further training until competent. HLF competency was monitored throughout the intervention by the CP regular supervision and quality assurance by reviewing audiotaped sessions.

This process is a standard approach to developing and evaluating a set of psychotherapeutic skills. It may be that for this clinical setting this approach is, or components of this approach are, insufficient. The theoretical frameworks were originally developed for mental health conditions that can remit, such as alcohol dependence, anxiety and depressive disorders. It would seem that a clinical state such as risk of CVD as a construct that could remit if one could develop and change one’s thinking patterns about lifestyles has face validity for applying similar psychotherapeutic techniques. However, perhaps a psychological model is not appropriate and intensive instruction is more apt.

Landmark studies have repeatedly shown that intensive lifestyle instruction that focuses primarily on behaviour change, such as the diabetes prevention studies,157–159 weight reduction programmes160 and, recently, studies on reversal of type 2 diabetes mellitus,161 does lead to significantly improved outcomes. Even the simple behavioural act of prescribing a pedometer leads to improved PA. 34 In these interventions, the clinically active ingredients were access to intensive, highly-structured and prescribed dietary and/or PA instruction following a counselling approach with minimal psychotherapeutic techniques. Our approach was to use talking therapies to instigate behaviour change rather than instruction, with the philosophy that intentions need to change before behaviours do and that changing people’s way of thinking (as opposed to instructing them) would also be cheaper in the long term and more likely to be maintained. It could be that with lifestyles interventions the active ingredients are intensive instruction and advice. This suggests that the theoretical models of theory of planned behaviour and the theory that everyone has a degree of motivation and intent may not be appropriate for lifestyle change.

A related, but perhaps separate, explanation is that the techniques we used to apply these theories were too weak. Secondary to factors totally outside our control, we could not assess competencies fully, but the extremely limited evidence we had was that the competencies, although acceptable, were not highly proficient. Assuming that this was indeed the case, then despite the intensive MI training of the HLFs this in itself was insufficient and one needs to establish very high levels of competencies in these skills. The health trainers did refer to this in the process evaluation; they thought that there was too much didactic and classroom-based teaching when they would have preferred more clinical cases and more supervision during the intervention. There have been very few studies that measure the competencies of low-intensity psychotherapy in lifestyle-based interventions. In the field of supporting self-management in type 2 diabetes mellitus, there have been only three studies that have formally tested competencies in MI. 105,162,163 The original landmark studies of MI were delivered by those who had developed the models, were leaders in the field and specialised wholly in MI. 41 As champions of MI and developers of the competency materials, it is likely that the earlier studies of MI would have shown high competencies and that this led to improved outcomes. It is unfortunate that although these data were collected we are not able to measure the competencies that would allow us to examine the level of skill. It may be reasonable to assume that the competencies delivered were probably adequate even if not excellent, in keeping with what would be expected for a large-scale delivery of MI by NHS band 3 equivalent health trainers. The reasons are that the health trainers were self-selected, most already had experience in this field and were given excellent references, they received high-intensity training and supervision, and their own feedback suggested that they were very motivated to improve their skills and seek more practical support and had insight into their clinical needs.

Although we observed that the HLFs generally had lower than expected competency in delivering the BCTs, it is important to note that the employed BCT coding framework was developed by us and was not validated, potentially leading us to underestimate HLFs’ competencies in this domain. If we had the full data set of recorded intervention sessions, we believe that we would have observed much higher levels of competency. Anecdotes from the HLFs suggested that they were highly motivated, professional, committed, ambitious and knew that they were being observed; these are all the key generic therapist elements often needed to be competent in delivering talking therapies.

Finally, it is worth considering that participants in the UC arm may have received a high level of standard care over the course of the study, which could have positively affected their PA and weight, thus obscuring any treatment effects from MOVE IT. However, given that the UC arm did not have a significant increase in PA or reduction in weight, this alternative is unlikely.

Clinical implications

There are a number of clinical implications to consider. These findings suggest that low-intensity psychological intervention in the form of MI to the general population at high risk of CVD risk is not effective. The acquisition of a workforce of health trainers who have received extensive training in MI to a level of low-to-moderate skills in MI is probably also not indicated. If this kind of training was to be offered, perhaps the focus should use the manual we have developed as a basis but to focus on practical learning through case work (i.e. practical experience and case studies). There is also some consideration to be given as to whether these findings apply to other clinical settings that require lifestyle change, such as obesity, prediabetes, diabetes mellitus and people with CVD. Some older, male patients have a perception that they are not at risk of CVD or that this risk is to be expected at this stage in their lives. This view may need to be studied further to understand if it is associated with worse health outcomes.

The implication of the cost-effectiveness analyses is that introducing a generic enhanced MI-based health trainer programme for people at high risk of CVD is not a cost-effective use of resources.

Research implications and future directions

There are number of potential directions for the future. The first is to test whether or not this intervention would be more successful with people at much higher risk of CVD who have modifiable risk factors. This would involve selecting patients from primary care or community settings, and perhaps younger patients, with higher levels of obesity, lipids and other CVD risk factors that are potentially modifiable (unlike risk factors such as age, ethnicity and sex, which are not modifiable). One potential next step could be to model the intervention in different populations, as we are not aware of intensive MI interventions to support lifestyle change in which there was improved targeting.

An alternative approach would have been to increase the actual CVD risk for eligibility and to restrict the study to a younger population, those living in deprived areas and those of non-white ethnicity. Another improvement could be to change the format of the intervention or its delivery (i.e. being led by a practice nurse). 164 Setting BMI thresholds for recruitment would be another improvement to our methodology, which may mean that we recruit patients at higher risk and who may have benefited more from the intervention. However, our sensitivity analyses adjusting for different BMI thresholds did not indicate that this factor influenced the effect of intervention.

Second, there may need to be a paradigm shift in the psychological constructs underpinning lifestyle change in chronic disease self-management. Brief interventions for medical conditions that are chronic are increasingly showing to be not effective and outdated. New concepts are needed that take into account the burden, or the long haul, of living with a chronic condition. Our intervention was for 12 months but perhaps interventions that consist of bursts of support over years need to be evaluated; the dilemma is that these are expensive to fund and the results take a long time to be reported, and there may be a cohort effect as medical science might move forward in that time. MOVE IT emphasised the importance of the use of qualitative methodology in providing detailed experiences of participants for assessing the effectiveness of a RCT. As previously reported, studies have supported the benefits of lifestyle interventions in the short term but few have supported their benefits in the long term. 49,131,132 In MOVE IT, continuity of care and duration of the intervention was of value to those who attended. Although attendance to group sessions was less than attendance to individual sessions, those who attended group sessions appear to value the peer support. Future research should include more qualitative work on why those at the highest risk of CVD or those with modifiable risk factors do not attend lifestyle interventions. It was impossible to do this in MOVE IT as we could not gain ethics approval to contact patients who had not replied to our invitation sent via their GPs.

Summary

An intensive lifestyle intervention using enhanced MI skills was not associated with reduced weight or increased PA in a sample of people at risk of CVD. The reasons for this are probably attributable to the sample having predominantly non-modifiable risk factors. Future interventions should focus on those at high CVD risk and with modifiable risk factors and should be delivered at high levels of measured competency.

Contributions of authors

Khalida Ismail (Professor of Psychiatry and Medicine and Consultant Liaison Psychiatrist) led the grant application and conduct of all aspects of the trial as PI.

Daniel Stahl (Reader in Biostatistics and Head of the Statistics Learning Group) contributed to the grant application, protocol writing and ongoing study monitoring, developed the statistical analysis plan, advised on the collection of outcome measures and data management and conducted the primary and secondary analyses.

Adam Bayley (Clinical Trial Manager) provided oversight of the research team, contributed to the day-to-day conduct of the trial and was involved in writing and collating contributions to the report.

Katherine Twist (Clinical Trial Manager) provided oversight of the research team and contributed to recruitment of research sites and the day-to-day conduct of the trial.

Kurtis Stewart (Trial Administrator) provided administrative support to the trial and contributed to sections of the report, including intervention details and fidelity analysis.

Katie Ridge (Assistant Trial Manager) contributed to the recruitment of research sites, data collection schedule development and ongoing day-to-day running of the trial.

Emma Britneff (Assistant Trial Manager) provided oversight of the research team and contributed to the day-to-day conduct of the trial.

Mark Ashworth (Senior Lecturer in General Practice) contributed to the grant application, protocol writing and ongoing study monitoring. He networked with, advised and facilitated the study for the new CCGs in South London and advised on primary-care-related issues.

Nicole de Zoysa (Senior Clinical Psychologist) contributed to the grant application, protocol writing and intervention development. She developed the manual, curriculum and competency framework and trained and supervised the HLF team.

Jennifer Rundle (Clinical Psychologist) provided ongoing training and supervision of the HLF team and contributed to the development of the process evaluation, including preliminary fidelity analyses.

Derek Cook (Professor of Epidemiology) contributed to the grant application, protocol writing and ongoing study monitoring. He supervised setup and collection of PA data and the interpretation of outcome measures.

Peter Whincup (Professor of Cardiovascular Epidemiology) contributed to the grant application, protocol writing and ongoing study monitoring. He was responsible for senior supervision of all aspects of CVD epidemiology, Health Checks and health service aspects of the trial.

Janet Treasure (Professor of Psychiatry) contributed to the grant application, intervention development (MI approach, training and manual development) and ongoing study monitoring.

Paul McCrone (Professor of Health Economics) contributed to the grant application, protocol writing and ongoing study monitoring. He conducted the cost-effectiveness analysis.

Anne Greenough (Professor of Neonatology and Clinical Respiratory Physiology) contributed to the grant application, protocol writing and ongoing study monitoring.

Kirsty Winkley (Postdoctoral Fellow and Lecturer in Diabetes and Psychology) contributed to the grant application, protocol writing and ongoing study monitoring. As the senior project manager, she provided oversight of both the research team and the HLF team during recruitment and intervention delivery. She supervised the recruitment of staff members, the quality assurance of data collection and the process evaluation.

The MOVE IT trial team

We set up a multidisciplinary team to ensure that all scientific and practical considerations were addressed for the successful conduct of this study. This represented international reputations and expertise in health psychology, clinical psychology, nursing, clinical epidemiology of long-term conditions (obesity, depression, CVD and diabetes mellitus), clinical trials, health economics of complex interventions, statistics in behavioural sciences, academic and commissioning primary care and project management. Collectively, we have a wealth of expertise in running large cohort clinical trials of complex interventions and in the dissemination and translation into local services and training programmes, especially in the local multicultural setting. Our strength also lies in the collaboration between two teaching hospitals via the health innovation and education cluster.

Trial Steering Committee

We would like to thank the TSC members: Professor Steve Iliffe (Chairperson), University College London; Professor Tom Marshall (member of the Health Technology Assessment and Efficacy and Mechanism Evaluation boards, 2009–11), University of Birmingham; Professor James Carpenter, London School of Hygiene and Tropical Medicine, Medical Research Council Clinical Trials Unit; and Dr Tim Anstiss, independent medical doctor and consultant. We would also like to thank the PPI members for attending all TSC meetings throughout the trial.

Data Monitoring and Ethics Committee

We would like to thank the DMEC members: Professor Betty Kirkwood (previous Chairperson), London School of Hygiene and Tropical Medicine; Professor Helen Weiss (Chairperson), London School of Hygiene and Tropical Medicine; Professor Stephanie Taylor, Queen Mary University of London, Barts and The London School of Medicine and Dentistry, Blizard Institute; and Dr David Blane, Imperial College London.

Publications

Bayley A, de Zoysa N, Cook DG, Whincup PH, Stahl D, Twist K, et al. Comparing the effectiveness of an enhanced MOtiVational intErviewing InTervention (MOVE IT) with usual care for reducing cardiovascular risk in high risk subjects: study protocol for a randomised controlled trial. Trials 2015;16:112.

Bayley A , Stahl D, Ashworth M, Cook D, Whincup P, Treasure J, et al. Response bias to a randomised controlled trial of a lifestyle intervention in people at high risk of cardiovascular disease: a cross-sectional analysis. BMC Public Health 2018;18:1092.

Ismail K, Bayley A, Twist K, Stewart K, Ridge K, Britneff E, et al. Reducing weight and increasing physical activity in people at high risk of cardiovascular disease: a randomised controlled trial comparing the effectiveness of enhanced motivational interviewing intervention with usual care [published online ahead of print December 12 2019]. Heart 2019.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.