Synopsis

Background

Adolescence as a critical period in the development of lifelong health

Poor diet costs the NHS £6B annually; physical inactivity £1B.1 UK adolescents have poorer diets than other age groups, and fewer than 20% meet physical activity (PA) guidelines.2,3 Health behaviours track into adulthood, implying that suboptimal diet and body composition in adolescence will affect immediate health but also raise the risk of obesity and non-communicable diseases later and in the next generation.4–7 Body composition at conception and during pregnancy have profound impacts on the long-term health of the offspring.8–10 Adolescence is a period of rapid growth and physiological change and may be a critical period for the development of later health and disease.11 Because of this, the Lancet commission on adolescent health claims: ‘investments in adolescent health and well-being bring a triple dividend of benefits now, into future adult life, and for the next generation of children’.12

Improving public health

Adolescents are not simply young adults or older children but have unique physiological and psychological characteristics. They are an underserved population, with largely unrecognised and unmet needs, perhaps because they discount health and are reputedly difficult to engage in behaviour change.11 The potential for improved dietary quality and PA to enhance the current and future health of adolescents was recognised in the 2014 Chief Medical Officer for England report written shortly before the application for Engaging Adolescents in Changing Behaviour (EACH-B) funding was first drawn up, suggesting that the personal, social, health and economic education (PSHE) curriculum could be used to ‘empower children with the information that they need to plan healthier lives’.10 Adolescents require interventions tailored specifically to their needs, however, through a process of cocreation.

At the time this proposal was submitted, there was some question about whether a public health programme was ‘in scope’ for the PGfAR scheme. As a result, our application had to be approved as in remit by the Programme Director and CCF Deputy Director. The award recognised the public health relevance of EACH-B as coming from: (1) the potential to improve diet and PA levels in adolescence, (2) amplification of this benefit over the life course and into the next generation and (3) the resulting greater cost-effectiveness of interventions in adolescence versus in adulthood.

Challenges of adolescent engagement in behaviour change

Despite the fact that adolescents find it difficult to engage with the long-term consequences of health behaviour, systematic reviews suggest motivated and engaged adolescents can improve their health behaviours. Unfortunately, little is known about how to motivate and engage them in sustaining changes in the long term.13 Systematic reviews also identify significant opportunities for engaging adolescents in behaviour change.14

School-based interventions: a scoping review carried out in preparation for EACH-B showed interventions based in schools may be effective, perhaps because schools can provide a supportive environment and because there is less attrition. School-based interventions offering peer modelling, social support and choice increased PA.13,15 Less is known about interventions effective in improving adolescent diet, but parental and teacher involvement may be important.13,16,17 Peer support is effective for health behaviours other than diet and PA.18 More work is needed to understand how these influences can be harnessed to be optimally acceptable and persuasive to adolescents, who will vary in their response to interventions.

Digital platforms: these offer another opportunity for engagement; 95% of 12- to 15-year-olds owned smartphones in 2024.19 At the time EACH-B began, evidence for effectiveness of digital interventions with adolescents was limited.20 More focus was needed on incorporating effective behavioural techniques (e.g. self-monitoring; gaining approval; modelling; intention formation), and on better methods of engagement as previous digital interventions have suffered from low usage and high attrition.21 Our review of digital interventions suggested that the most unhealthy participants show more significant behaviour change and rate interventions more highly than their healthier counterparts, although they are less compliant.22 Formative interviews with 13- to 14-year-olds suggest younger adolescents know of existing commercial diet and lifestyle smartphone applications but do not feel that they are appropriate for their age group.

Incorporating gaming: our patient and public involvement (PPI) groups said social support and competition are important in helping them maintain healthy lifestyle behaviours. Incorporating these features through game elements into digital interventions may increase engagement. Games may have superior capacity to engage over other digital interventions and to sustain behaviour changes because of their potential to be highly enjoyable and intrinsically motivating.23,24 Serious games are best tailored to a particular demographic, benefit from a strong focus on game and/or psychological theory and can be effective on their own or in multicomponent programmes.25

Formative work for an intervention to maximise engagement and sustain change

Evidence reviewed above, two scoping reviews and our previous work led us to propose a complex, three-component intervention: (1) engaging and motivating adolescents in behaviour change through a school-based education programme and (2) initiating and sustaining change through a combination of face-to-face support from teachers trained in skills to support behaviour change and (3) a personalised digital intervention that involves social support.

1. LifeLab is a 2-week science module for 12- to 14-year-old secondary school students, which delivers part of the National Curriculum for Science, helps them understand the science underlying their health, engages them in thinking about their health and the health of their future children and motivates them to improve health behaviours and health outcomes.26,27 Since 2009, 93% of schools approached have participated, involving over 16,000 students.
2. ‘Healthy Conversation Skills’ (HCS) supports health and social care practitioners to engage and motivate individuals to change their health behaviours (Southampton IP).28 HCS are accessible skills, shown to increase the confidence and competence of practitioners in addressing diet and PA in their clients.29 Evidence from our research group suggests face-to-face support improves the effectiveness of digital interventions in adults.30 We adapted HCS for teachers.
3. Design of the digital intervention built on 20 years of expertise gained from the LifeGuide research programme, Southampton (www.lifeguideonline.org/; Director: Yardley; Deputy Director: co-applicant Morrison). Numerous complex digital behaviour change interventions for public health promotion and illness management have been shown to be effective in randomised controlled trial (RCT).30–32 The development of the gamified digital intervention that we developed as part of EACH-B drew on award-winning game design expertise.

Aim of the programme

The aim of this programme was to develop and test a complex intervention designed to engage, motivate and support adolescents aged 12–14 years in improving their diets and PA levels. The question we were asking was ‘how could adolescents be effectively and cost-effectively engaged and supported to make and sustain positive changes to their diets and physical activity levels?’ Our approach to answering this question is summarised in Figure 1.

FIGURE 1 Logic model justifying and summarising the EACH-B programme and its intended outcomes. (Figure not reproduced here; see the original article.)

Objectives

We planned to develop and evaluate in secondary school students an intervention using the three components described above. We set out to work with adolescents and other stakeholders to determine the most effective and cost-effective methods of integrating this intervention into the school curriculum.

To meet these objectives, we undertook the following activities organised under three research streams (RSs). We:

1. used evidence synthesis and qualitative research to identify key features of interventions to improve adolescent diet and PA levels likely to be (1) effective and cost-effective and (2) engaging, feasible and acceptable to adolescents, parents and teachers (RS1)
2. developed a health economic model to evaluate costs and health effects of diet and PA interventions for adolescents (RS1)
3. used a person-based approach to develop an intervention that is feasible, acceptable and persuasive to students, parents and teachers (RS2)
4. undertook a RCT of the effectiveness and cost-effectiveness of this intervention to improve quality of diet and PA levels of secondary school students (RS3)
5. with stakeholders, planned integration of learnings from this programme of research into the school curriculum and broader public health activities (RS1).

The relationships between these research activities are illustrated in Figure 2.

FIGURE 2 Research pathway diagram showing relationships between the elements of the programme. (Figure not reproduced here; see the original article.)

Programme management

Robust programme management was ensured through a structure of delivery, scientific and strategic accountability and input as depicted in Figure 3. The programme delivery team was organised into RSs, led by RS leads designated in the grant application. Each RS was divided into work packages (WPs) with discrete deliverables and assigned teams, led by WP leads. Depending on the type of work, deliverables and delivery timelines, team members could be part of more than one WP team, could lead two WPs or could have a combination of leadership roles.

FIGURE 3 Engaging Adolescents in Changing Behaviour programme management structure and types of management and oversight relationships. (Figure not reproduced here; see the original article.)

Delivery was overseen by the Programme Leads, supported by the Scientific Programme Manager and the Operational Programme Manager who were members of the Programme Operational Management Group alongside the Programme Leads.

Strategic and scientific oversight was provided by the Programme Steering Committee (PSC) comprising three scientific and one public contributor and the National Institute for Health and Care Research (NIHR) Programme Manager.

Co-applicants also provided strategic and scientific input. Stakeholders and public contributors provided strategic input. Details of our stakeholder engagement and PPI activities are given in sections that follow.

Stakeholder engagement

Stakeholder engagement was essential to ensure relevance and sustainability of the intervention. By building a programme of stakeholder engagement which involved a diverse range of stakeholders, including young people, parents, teachers, local authorities, non-governmental organisations and charities, we tapped into a wealth of insights and expertise.

Patient and public involvement and engagement

Our public contributors have supported us throughout the programme. This activity has had two tracks. Engagement with young people has been organised as and when a specific contribution was needed. Our experience is that standing groups of young people are difficult to sustain for longer than about 3 months. We have therefore established a rolling programme of young people to inform us and collaborate with us. This LifeLab Youth Panel runs three times a year with 6–8 young people who are employed by the University of Southampton for 12 weeks a time to work as part of the research team. These groups are designed to be diverse and reflect experience from young people across a spectrum of ages from 13 to 17 years and of different backgrounds. Every group addresses a new issue, usually one that they identify as important for young people’s health. In addition to carrying out their own work, this group has provided EACH-B with specific support and advice on, for example, the design of outward-facing research materials or the wearability of activity trackers. The second track involved three to four parents of young people who were paid to make a public contribution to meetings of our co-applicants and the PSC. New public contributors were recruited as others left maintaining a core group who support the activity of EACH-B. Alongside the two ongoing patient and public involvement and engagement (PPIE) streams, between 2018 and 2024, EACH-B ran a series of events and projects that involved young people and other members of the public in our research. As a matter of principle, young people recruited to these activities drew from groups that were diverse in experience and background. The large numbers who took part in game jams in schools generally reflected the make-up of those schools. In total, over 1000 young people play-tested the app during different stages of its development. We ran an extensive range of PPIE activities involving young people and their parents between 2017 and 2024 when the programme ended. A full list of these is available in Appendix 1.

Understanding existing evidence for effective and cost-effective interventions with young people: research stream 1

Evidence synthesis and systematic reviews

This WP (Figure 4) aimed to address the following research questions (RQs) in order to inform the development of the EACH-B intervention, details of which are published in the protocol paper.33

FIGURE 4 Showing the organisation of the WPs within RS1. (Figure not reproduced here; see the original article.)

• Research question 1. How can digital intervention approaches, including games, engage young people in improving their diet/PA? What features are associated with effectiveness/cost-effectiveness?
• Research question 2. What interventions are effective in improving young people’s diet/PA? What features are associated with effectiveness/cost-effectiveness?

The first systematic review addressed RQ1.22 We identified effective intervention components from 27 studies and assessed the cost-effectiveness of these interventions. Due to heterogeneity of studies, meta-analysis was not feasible. Most were website interventions. Other delivery methods were text messages, games, multicomponent interventions, e-mails and social media. We found that digital interventions most likely to be effective are those that incorporate education, goal-setting, self-monitoring and parental involvement. We found too little data on the cost-effectiveness of digital health interventions to be able to draw any conclusions, and future trials should report these data.

The systematic review and meta-analysis of school-based interventions focused on diet, and weight outcomes suggested that school-based health education interventions have the potential to lower body mass index (BMI) towards a healthier range in adolescents.34 This review focused on quantitative evidence from studies reporting outcome and process evaluation findings. Multicomponent interventions involving key stakeholders such as teachers and parents and digital components were most effective and offered a promising strategy for improving the health behaviours of young people. To address these questions for PA, we identified recently conducted systematic reviews of PA interventions, and these informed the work of RS2. We sought evidence on process measures as well as outcomes, including engagement, compliance, implementation, costs and benefits. Findings from these reviews informed the economic modelling and review (WP1.2) and intervention development (RS2).

A third systematic review was carried out as part of Sarah Shaw’s PhD studies.35 The review found no clear relationship between access to supermarkets – classified as healthy food outlets – and healthier diets, possibly due to the availability of both nutritious and unhealthy foods within these stores. Evidence on the role of the consumer environment, including food pricing, availability and in-store marketing, was limited and inconsistent. The review highlighted the need for further high-quality intervention studies to better understand how food environments influence adolescents’ independent food choices.

Cost-effectiveness modelling

A Markov model was developed. The model was designed for use in extrapolating short-term observed effects on diet, activity and/or BMI to estimate future effects on the incidence of type 2 diabetes (T2D) and cardiovascular disease (CVD). The health outcomes were quantified using quality-adjusted life-years (QALYs), including direct effects of diet, activity and/or BMI on quality of life (QoL) EuroQol-5 Dimensions as well as survival and QoL effects of T2D and CVD events. Costs were estimated from a public sector perspective, including costs to schools, local authorities and the NHS. The intention was to estimate costs and QALYs over 10 years, 20 years and lifetime horizons, discounted at UK-recommended annual rates. In addition to cohort averages, results were to be presented for socioeconomic subgroups (Index of Multiple Deprivation quintiles) to show potential impacts on inequalities.

While the intention was to apply the model to estimate the cost-effectiveness of LifeLab Plus to improve diet and PA in adolescents using data generated in WP3.2, our findings did not warrant such an analysis. Main trial describes problems with data collection and analysis in the main trial, including having to stop measurements required for BMI assessment (a key driver of effectiveness in the economic model), missing data and a large imbalance in diet quality at baseline. The trial results show no apparent effect for the primary outcomes; and differences for the outcomes used in the economic model [Child Health Utility 9 Dimensions (CHU9D), fruit and vegetable portions per day, and minutes of moderate/severe PA] are small and inconsistent (Table 1). We did consider whether subgroup analysis to explore differences by socioeconomic status (SES) would be useful but concluded that this would be subject to bias and difficult to interpret.

The model has been published with intervention effects on BMI z-score and moderate/vigorous PA taken from meta-analyses of multicomponent school-based obesity prevention interventions, including PA, health education and dietary improvement.36 The model, therefore, has use beyond EACH-B, and application to other complex intervention research.

Stakeholder engagement

Stakeholder engagement meetings were held throughout the programme, evolving in response to insights and changing contexts. These meetings built a strong network across Wessex, ensuring lasting commitment to adolescent health and effective knowledge translation beyond EACH-B. Early discussions encouraged cocreation of the intervention, fostering ownership and increasing the likelihood of successful implementation.

A key milestone in this strategy was securing an Innovation-to-Commercialisation fellowship for Woods-Townsend, providing £50,000 funding for 3 months of intensive national and international stakeholder engagement. This deepened understanding of implementation needs and the commercial potential of EACH-B outputs while strengthening long-term partnerships.

The programme also contributed to policy development, with several team members actively involved in Southampton City Council’s Childhood Obesity Policy scrutiny panel. This led to direct policy impact, including recommendations to scale up LifeLab. EACH-B also supported ukactive’s ‘Generation Inactive 2’ report and played a central role in the UK Research and Innovation (UKRI) Arts and Humanities Research Council-funded Pathways to Health Consortium, a collaboration of over 30 partners reimagining adolescent health and well-being services in Southampton and beyond.

Woods-Townsend further engaged with national policy-makers in health and education through the 2020–1 Royal Society Scientist-Member of Parliament pairing scheme. This engagement reinforced the importance of integrating public health and education sectors in adolescent well-being initiatives.

The stakeholder engagement process successfully secured genuine buy-in by ensuring the intervention met both research and policy needs. This collaborative approach not only achieved the programme’s primary objectives but also created wider impact through policy influence and system-wide engagement. The partnerships formed continue to support adolescent health research, policy development and practice, facilitating young people’s engagement with Southampton City Council and the Hampshire and Isle of Wight Integrated Care Board (ICB) and Integrated Care Partnership (ICP). The EACH-B programme has delivered lasting benefits to young people across the Wessex region.

Developing the intervention to engage and motivate teenagers: research stream 2

Some text in this section has been reproduced with permission from Strömmer et al.33 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Work on developing the three-component intervention began in the first month of the programme (October 2017) and ran until October 2019 (Figure 5). We adopted a person-based approach to guide the planning, development and usability testing of the intervention and ensure that was feasible and acceptable to young people, parents and teachers.30,37,38 This approach has proven success in supporting the development of numerous effective interventions.31,39 In an innovation that was unique at that time, we combined this approach with user-centred design principles used for digital game design and a participatory design process in order to produce a behaviourally informed yet engaging game.40 This RS involved four iterative WPs that each involved systematic and rigorous qualitative research. These were overlapping to enable initial plans for the intervention to be tested early on, revised to minimise adverse reactions from students, teachers and parents, and then tested again.

FIGURE 5 Showing the organisation of the WPs within RS2. (Figure not reproduced here; see the original article.)

Identifying barriers to eating better and moving more and strategies for change

Qualitative research with students

This study was conducted in Southampton, England, between 2017 and 2018 and has been published.41 Participants were 54 13- to 14-year-old students from a mixed, non-selective secondary school, 95% of whom were White British, and 49% were girls. Convenience sampling was used, and focus groups of between three and six students were conducted during school hours. A semistructured topic guide, developed with input from a young people’s PPI panel, guided discussions about eating healthily and being active, their experiences and views of existing health information sources and levels of environmental support for behaviour change, social influences on their views and behaviours and their experiences and perceptions of the digital world. Participants completed a brief demographic questionnaire and received a £10 shopping voucher for their participation.

Findings

Thematic analysis identified four main themes to the discussions:

1. I know what healthy means and how it might benefit me.
2. I do what others do.
3. I’m busy doing other things.
4. I want healthy to fit into my life.

These themes aligned with the three basic psychological needs outlined in Self-Determination Theory: autonomy, competence and relatedness (Figure 6). Students were aware of health guidance but struggled to maintain healthy habits. Social influences, particularly peers and parents, played a significant role in their health behaviours. Participants reported busy lives filled with various commitments, often compromising between health ideals and personal preferences when making independent choices.

FIGURE 6 The overlap of the components of Self-Determination Theory with themes from interviews with secondary school students and the relationships between these themes. Some text in this figure has been reproduced with permission from Strömmer et al.41 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

The study suggested that effective health interventions for young people should align with their values and priorities, fulfilling psychological needs for autonomy, competence and relatedness. Interventions should make healthier choices more accessible, appealing and enjoyable, while addressing environmental constraints. The potential of digital platforms and social media in supporting health interventions was highlighted, with suggestions for future research to explore their positive use in engaging young people. Exploration of the role that food plays in identity development during adolescence discovered important differences in the way younger and older adolescents responded to social media. This research is described in Report Supplementary Material 2 and Report Supplementary Material 3.

Qualitative research with parents

The study was conducted in Southampton in 2018 and has been published.42 It involved 24 parents, with 18 participating in individual interviews and 6 in 2 focus group discussions. Participants were recruited via an e-mail list following their adolescent’s visit to LifeLab. All had at least one adolescent attending secondary school. All were women, and all but one was White British. Ten were educated to degree level. Interviews lasted 20–45 minutes and were held in various convenient locations. All participants were women, mostly aged 40–49 years and predominantly White (96%). Thematic analysis was conducted using NVivo (QSR International, Warrington, UK) software.

Findings

Six themes were identified:

1. Parents tried to maintain authority over health behaviours while promoting independence.
2. Parents’ health perspectives were influenced by personal experiences and barriers like tight budgets.
3. Adolescents preferred convenient food options and were influenced by peer pressure.
4. External factors like school environments and technology were seen as barriers to health.
5. Changing dynamics of control led to compromises and conflicts.
6. Parents suggested rewards and positive use of technology to support adolescent health.

The study highlighted the shifting dynamics between parents and adolescents regarding control over health behaviours. Parents’ influence diminished as adolescents gained autonomy, often leading to communication breakdowns. The study also emphasised the need for public health interventions to address the broader food environment and marketing influences outside the home. Direct methods like parent training were suggested, but concerns about feasibility of this with busy schedules were noted.

The three papers published on the work of WP2.1 are described below.

Review of health behaviour change interventions

The authors drew on their experience developing health behaviour change interventions with diverse populations globally.43 They analysed data from several studies, including qualitative interviews and focus groups with pregnant women, adolescents and parents/caregivers in the UK and low- and middle-income countries. Thematic analysis was used to explore participants’ values, motivations and responses to interventions.

Findings

The authors proposed three main reasons why conventional behaviour change interventions often struggle to demonstrate effectiveness:

1. Interventions focus on mean effects, ignoring individual differences in response. One study, for example, showed variations in how pregnant women engaged with interventions based on their ‘health identity’ and confidence in overcoming barriers.
2. Interventions assume everyone values health equally. Research with adolescents reveals they prioritised an immediate need for autonomy, competence and relatedness over long-term health outcomes.
3. Interventions appeal primarily to rationality rather than emotions. Evidence suggested positive emotions experienced during healthy behaviours create unconscious or internal motives that sustain change.

The authors argued for a move towards acknowledging individual differences in intervention design. They proposed using creative approaches involving the arts and culture to engage people emotionally in health and well-being, and to build trust through sustained engagement. Future interventions may benefit from strategies that empower communities, appeal to emotions as well as logic, and align with people’s values and priorities. The authors concluded that engaging emotions and using creative approaches could lead to more effective and sustained behaviour change than conventional interventions, which involve information giving and education and therefore appeal only to people’s rational and cognitive motivations and reasoning.

Mixed-methods exploration of energy drink consumption in United Kingdom adolescents

The study used a mixed-methods approach.44 Repeated cross-sectional quantitative dietary data, information about consumption of energy drinks and demographic data from the National Diet and Nutrition Survey (NDNS) (n = 2587 adolescents aged 11–18) were analysed using logistic regression. Qualitative data from semistructured interviews included discussion of energy drink consumption with 74 adolescents, 95% of whom were aged 13–14 years, 24 parents and 15 teachers in Hampshire, UK, the majority of whom were women and White British. Data were analysed using thematic analysis.

Findings

Overall, 7% of young people drank energy drinks over the 4-day period of NDNS study. Consumption was higher among older adolescents, those from more deprived areas and lower income households, and those with poorer dietary quality. Consumption of energy drinks by adolescents from the most deprived areas increased between 2008 and 2016, while it decreased among the most affluent adolescents.

Three themes emerged from the qualitative data: (1) adolescent consumption of energy drinks is driven by peer influence and low cost of unbranded drinks; (2) energy drink consumption is associated with other unhealthy behaviours, but adolescents lack understanding of specific health risks; (3) voluntary bans on sale of energy drinks in retail outlets and schools are perceived as ineffective.

The findings support introducing legal restrictions on energy drink sales to minors. However, the proposed ban for under-16s would miss the highest consumers aged 16–18 years. Additional policies like minimum pricing, placement restrictions and public health messaging could enhance effectiveness. Legislation may help address the widening socioeconomic inequalities in energy drink consumption among adolescents.

Qualitative work with adolescents examining their social and food environments45 (and Report Supplementary Material 4)

This qualitative study involved 13 online focus groups with 45 adolescents aged 11–18 years attending secondary school or college in England, UK. Participants, 29% of whom were girls and were from a range of SES groups, took part in semistructured discussions about their experiences making independent food choices. Data were analysed using thematic analysis. The study was conducted between October 2020 and April 2021 during the COVID-19 pandemic.

Findings

Five key themes were identified:

1. Social experiences involving eating were more important than the food itself, with fast-food outlets uniquely suited to facilitating social interactions within this age group.
2. Adolescents wanted to spend money on foods they considered worthwhile, not always the most affordable options.
3. Participants preferred making quick, low-effort food decisions, appreciating factors like prominent placement and promotions.
4. Chain food outlets were valued for offering familiar, frequently advertised foods.
5. Adolescents’ search for a sense of autonomy was central to all themes.

The findings highlighted the interplay between social and food environments in shaping adolescents’ choices. Fast-food outlets were seen as ideal spaces for socialising independently of family. Participants were aware of some government strategies around food but felt mandatory policies across all outlet types were needed. The study suggested current UK food environment policies may need refinement to effectively impact adolescents, such as including smaller retailers, addressing meal deals and restricting brand advertising. Overall, the research emphasised the need to incorporate adolescents’ lived experiences when developing food environment interventions and policies to ensure they align with young people’s values and are effective for this age group.

Planning the intervention

Intervention planning was conducted collaboratively within our multidisciplinary team with oversight from our expert co-applicants at every stage. A person-based approach was adopted for development of all three components of the EACH-B intervention. The process was as follows:

1. Data from qualitative and quantitative systematic review (WP1.1) was triangulated with insights from the primary qualitative research and stakeholders (WP2.1/WP1.3) and theoretical models of behaviour change.
2. Based on this process, decisions about components and characteristics required as part of the intervention.
3. A detailed and rigorous description of the intervention was produced, ensuring that potentially important theoretical determinants of behaviour change were not overlooked.
4. The Behaviour Change Wheel was used to define and characterise the identified barriers/facilitators and guide the selection of appropriate behaviour change techniques (BCTs) from the 93v1 taxonomy.46
5. The final step was to translate the design principles developed through this process into tangible intervention components.

Following the person-based approach, a set of ‘guiding principles’ for the intervention was developed (Table 2). These defined the key behavioural, psychosocial or contextual issues that needed to be addressed by the intervention, that were not being met within the adolescents’ existing support structures, and distinctive ways in which the intervention could address those issues in a persuasive, feasible and acceptable way, both in terms of content (specific BCTs) and delivery (tone, structure, timing).

These guiding principles ensured that priority was given to crucial intervention components and that they were implemented and communicated in ways that were intended to be optimally acceptable, persuasive and likely to support sustained behaviour change. Where there was disagreement between sources of evidence and expert opinion, discussions were held within the team until consensus was reached.

Intervention content

This process produced an intervention that combined three linked components:

1. Professional development involving communication and HCS training for teachers to support their students’ health behaviour change. HCS training has been shown to support behaviour change, is accessible to practitioners in different health and social care settings, and can produce sustained changes in how practitioners communicate and support others to change.29 The process of adapting HCS training for teachers to deliver the intervention’s face-to-face support component is described in WP2.4. Teachers were trained online or at LifeLab to conduct healthy conversations with their students whenever the opportunities arose in school.
2. An educational module for secondary school students comprising nine in-person science lessons linked to the English National Curriculum and a practical day visit to a unique educational lab facility (LifeLab) at a large UK teaching hospital (www.southampton.ac.uk/lifelab/index.page). This educational component was based on the programme previously developed and extensively tested at LifeLab. Drawing on this experience, it was redesigned and rebranded to form a component of the EACH-B intervention. LifeLab is a hands-on programme tailored to give the adolescents insights into the impact on their lives of a healthy diet and the benefits of PA.
3. A game-based smartphone app (LifeLab Plus) for young people. This was a self-guided digital tool which was designed to build adolescents’ knowledge and motivations for eating better and being more active. Development of this component is described in WP2.3. The app was designed and tested by adolescents themselves to ensure that it was tailored to the way they use their phones. For more details of the app, see Figure 7.

FIGURE 7 Logic model describing the problems and evidence base and the way that the four games within the app are hypothesised to address these problems and generate desirable outcomes. (Figure not reproduced here; see the original article.)

Logic models

Figure 1 – the EACH-B logic model – summarises the whole programme. The model proposes that completion of the LifeLab module would motivate young people to eat better and move more. Engagement with HCS-trained teachers was intended to build on and sustain the motivation to improve their health behaviours that students gained from LifeLab. The digital intervention, based on proven BCTs and game features, would support integration of these healthy habits into self-identity, leading to improvements in self-efficacy, diet and PA.

Development and usability testing of the digital intervention

Guiding principles from WP2.2 were used to inform design of the digital intervention content, structure and delivery. Key behaviour change components of the digital component were identified based on the guiding principles. The narrative and game-based elements and the translation of the draft content into a first prototype of the digital intervention was carried out by the game design team at Glasgow Caledonian University’s (GCU) digital intervention and game development lab but supported by placements of psychologists from the team at Southampton in Glasgow.

The EACH-B intervention app development process is described in a manuscript currently being prepared for publication entitled, ‘Enhancing the relevance of nutrition and physical activity interventions for adolescents: Design and development of the LifeLab Plus app’ (this is described in Report Supplementary Material 5). The app was designed using combined principles of behavioural science and game design. A blend of theoretical methods (e.g. behavioural analysis, logic modelling), evidence-based strategies (e.g. systematic reviews, qualitative research) and coproduction activities, such as interactive workshops, were used to develop a set of guiding principles and identify opportunities to incorporate behaviour change components into the app. The game design and content were mapped on to 27 BCTs derived from the qualitative research, and user testing carried out in WPs 2.1 and 2.3 to ensure that all issues emerging from primary qualitative work were addressed within the intervention design (see Table 3 for an illustration of this process).

To support a sustained commitment to diet/PA change, the app was designed to:

• provide tailored feedback and self-monitoring support in combination with goal-setting as these are known to be associated with positive changes in behaviour47
• offer support for habit-formation, since evidence is accumulating in adults that this produces sustained behaviour change48
• include game elements, since these have been found to motivate players to engage with boring tasks and may be associated with positive behaviour change23
• address adolescents’ specific dietary/PA needs.

Adolescents were heavily involved in the cocreation of the digital intervention. Secondary school students aged 13–17 years (n = 49) from schools in Glasgow and Southampton took part in ‘game jams’ to develop characters, ideas and concepts. Four game jam events involving 233 university game design students generated ideas for features and quests for the digital intervention. Once there was a ‘minimum viable product’, panels of these students then played the games and fed back on game features as they were developed. These activities were followed by small group and ‘think aloud’ interviews with 10 adolescents aged 13–17 years recruited from a scout group; interviews were conducted to elicit more detailed feedback. Student evaluations and feedback collected during these interviews was used to inform the in-development digital intervention; students were asked about usability, acceptability and persuasiveness of support offered for behaviour change. Qualitative data was triangulated with objective records of digital intervention usage gathered from use of the intervention outside the lab to examine and measure engagement with the digital intervention as a whole, but also to identify which features of the intervention were not being used as intended. These issues were addressed in subsequent iterations, and the process repeated until the first full feature complete version of the app was ready. This was then used in the pilot study described in WP3.1.

The four games that formed LifeLab Plus are presented in the digital intervention logic model in Figure 7. The key findings of this exercise highlighted that theoretical and evidence-informed approaches, while crucial for initial development, were not sufficient to deliver an app that was meaningful and relevant to adolescents. Findings from the qualitative research and coproduction activities informed the integration of behaviour change components into each element of the app, for example, the ‘Gutsy’ game, to maximise engagement. Examining barriers to engagement in real-world contexts, beyond technical or navigational issues, researchers were able to enhance user engagement and increased the theoretical likelihood of sustained behaviour change. The process of app development combined the speed and agility of game design processes with the rigour of academic intervention design methods.

During the design process, consideration was also given to features that involved parents, since modelling effects of parental behaviour on adolescent behaviour are highly significant,49 and improved social connectedness, since this is known to predict adolescent health risk behaviour.50 Consequently, a website was designed for parents to accompany the EACH-B intervention. Time and resources prohibited development of a game component that enabled young people to compete with each other or play together, as they also prohibited development of an app for parents, though these would have enhanced the intervention.

Development of Healthy Conversation Skills curriculum

Building on findings from WP1.3/WP2.1, a HCS curriculum was developed for teachers from the existing HCS training. HCS-trained teachers provided the face-to-face component of the intervention.

Curriculum development was informed by interviews with teachers who took part in a pilot test of HCS during 2015–7. Interviews with 15 teachers from Hampshire secondary schools were carried out to explore how they had used the skills, any barriers to their use, what additional training needs they had and how an extended training programme might be delivered in schools. Findings from the interviews formed the basis for amendments to the HCS curriculum so that it was ready for use as part of the EACH-B intervention.

Healthy Conversation Skills training was delivered by the research team during Teacher Professional Development Days attended by science teachers whose pupils were due to undertake the LifeLab programme, where they received a short (1.5 hour) introduction to HCS. Two focus groups involving 14 teachers were convened during these professional development days to explore teachers’ perceptions of how they could use the skills to support pupils to make lifestyle changes (including digital tool use) and how the HCS training programme could best be delivered in schools.

A reflexive thematic analysis was conducted using transcripts from these interviews and focus groups along with those from two interviews conducted with teachers following their participation in the EACH-B pilot RCT. Teachers from a range of different subject departments, and those representing different levels of experience, seniority and roles, were recruited.

The analysis found that teachers were enthusiastic about talking to their students about health in supportive and encouraging ways. They found opportunities to have these sorts of conversations. Conversations that happened in lessons were more generalised and focused on the practicalities of health and behaviour, whereas less-formal conversations that happened at lunchtime, in corridors or after school were more likely to provide opportunities for teachers to speak to students individually about the specific barriers they faced. Teachers had ideas about strategies that could help them to support their students’ health, but, in reality, they felt powerless because they witnessed, on a daily basis, the breadth of environmental, individual, financial and social factors that made it more difficult for students to make healthy choices. Teachers also felt that school cultures and systems did not facilitate health behaviour change but, in fact, actively made it harder for teachers to support their students.

These findings led to amendments to the basic HCS curriculum that supported teachers to make the best use of the brief opportunities during lunchtime on in passing to speak to their students. It was also concluded that there was no obvious mechanism or time for delivering HCS training during school hours and in school, at least not for the duration of the EACH-B intervention trial. These findings were fed into discussions with wider stakeholder groups about methods of embedding support for health and well-being into education.

Pilot and main trial of LifeLab Plus intervention: research stream 3

Some text in this section has been reproduced with permission from Strommer et al.33 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Research stream 3 focused on assessing the three-part intervention in a cluster-randomised trial informed first by a pilot trial4 (Figure 8).

FIGURE 8 Showing the organisation of the WPs within RS3. (Figure not reproduced here; see the original article.)

Pilot trial

In the autumn term of 2018, we recruited six schools to take part in the pilot trial, randomised such that three schools received the intervention and three the control. Data collection was successful in all schools, both at baseline and follow-up 4–6 months later. Consent rates differed marginally between the intervention arm (92%) and the control (87%). Over 85% of students in the intervention arm attended the LifeLab days in January/February 2019. The students’ demographic data were transferred to the research team by all six schools. Methods for completing questionnaires, recording height and weight measurements and distributing GENEActivs were effective. Some challenges were identified in getting the teachers to participate in the teachers’ professional development day or the online training that was made available to them. We identified various modifications to the procedures for recruiting schools, managing the data collection and downloading the app, and these were taken forward to the main trial.

Main trial

The main trial was a cluster-RCT to assess the effect of engagement with LifeLab, ongoing support from HCS-trained teachers and the smartphone app on the diet/PA of LifeLab students in publicly funded schools in South Central England, UK. Students in mid-level ability classes were recruited. The control group had none of these three interventions. Each school represented 1 cluster, and we aimed to recruit 50 schools and 2 classes in each school. The trial was powered to detect 0.25 standard deviation (SD) differences between the two arms for both of the two primary outcomes of diet quality score and total activity (minutes per day of low, moderate and vigorous activity). From previous data from schools interacting with LifeLab, we estimated an intraschool (class) coefficient of 0.035. With 46 schools, each sending 2 classes, ≈ 50 students (n = 2300), this provided 90% power at 0.025% significance level (accounting for 2 primary outcomes) to detect a 0.25 SD difference in diet quality score or number of steps between intervention and control schools. We aimed to recruit 50 schools allowing for dropout.

Outcome measures

Primary outcome measures: data were collected at baseline and follow-up. Coprimary outcomes of dietary quality and PA were assessed. Dietary quality was assessed using a 20-item food frequency questionnaire (FFQ) developed specifically for use with adolescents using data for boys and girls aged 11–18 years who took part in the NDNS.51 Principal component analysis was applied to these data in order to identify 20 indicator foods which best describe better and poorer dietary quality of UK adolescents.51,52 This FFQ has shown good comparison with important nutritional biomarkers, including 25-hydroxy vitamin D, total carotenoids, serum folate and vitamin C.52 From the FFQ, a diet quality score was calculated, which was expressed as SD units. Using GENEActiv accelerometers, PA was assessed as minutes of daily low, moderate and vigorous PA, also described as total PA. At baseline and again at the 12-month follow-up, GENEActiv was worn for 7 days, and the output data averaged over this period, or the maximum period of valid data.

Secondary outcomes measures: these included self-reported frequency of PA from a modified version of the Youth Physical Activity Questionnaire (YPAQ) validated for use in 12- to 13-year-olds;53 behavioural self-regulation and self-efficacy for healthy eating and PA (unpublished); and QoL and well-being measured by two age-appropriate tools: the CHU9D54 and the Cantril Ladder.55 Behavioural regulation and self-efficacy for PA will be assessed by the Behavioural Regulation for Exercise Questionnaire56 and the PA section of the Self-Efficacy for Healthy Eating and Physical Activity measure (SE-HEPA).57 Behavioural regulation and self-efficacy for diet will be assessed using the recently developed Confidence and Behavioural Autonomy scale. This is age-specific and has been validated against the healthy eating scales of the SE-HEPA and the Treatment Self-Regulation Questionnaire.58 The original plan was to collect height and weight and calculate BMI z-scores, with and without adjustment for pubertal status;59 this had to be abandoned immediately post-COVID-19 when the research team were no longer allowed into schools to collect data.

Harms/auditing: we did not expect the trial to give rise to any adverse events or harms. However, a risk assessment was completed and submitted as part of the ethical approval process. All staff involved in visiting schools had the enhanced level of Disclosure and Barring Service to work with children, were trained in safeguarding and awareness of eating disorders and had basic levels of awareness about dealing with someone who might become anxious for any reason during a data collection visit.

We started recruitment to the main trial in autumn 2019. By March 2020, baseline data had been collected from 224 students in 5 intervention schools, and 212 students in 5 control schools, a total of 436 students. Three intervention schools had attended LifeLab. School closures initiated by the COVID-19 pandemic meant that all trial activities were officially suspended on 18 March 2020. This suspension lasted a year, during which the EACH-B team continued to engage with young people and to collect insights, by holding focus groups with 10 groups of young people every 3 weeks. These data were collected with a view to informing the EACH-B trial going forward.

Changes resulting from the pandemic, not only to the trial but also in schools, students’ lives and behaviours, we were unable to use the baseline data from the 11 schools that had been collected before the school closures. The trial was restarted from scratch in March 2021 (soon after schools reopened in February 2021). Various adaptations had to be made to the data collection procedures because of the infection control practices adopted in schools. Data acquisition was transferred online, enabling it to be performed in the classroom under the supervision of the teacher, or even at home for students who were isolating. However, we had to stop the collection of height and weight measurements, as we were unable to send research staff into the schools and could not ask teachers to perform the measurements for us. Added to which, social distancing meant that it was not possible to take these measurements. Changes also had to be made to the intervention. For many months, it was not possible to bring students into the hospital to attend the LifeLab day, so we arranged for it to be delivered in school, with the EACH-B team providing instructions and delivering the equipment to schools. Changes were made to the app to reflect changes to students’ lifestyles as a result of the pandemic.

Randomisation of the schools was conducted externally by the Clinical Trials Unit in Southampton. A minimisation algorithm was used, aiming to balance the schools on free school meal provision, General Certificate of Secondary Education (GCSE) results and whether the schools had sent students to LifeLab in the previous 2 years. Over the subsequent 2 years, 49 schools were recruited and randomised to the trial, and baseline data collection took place in 45 of them. Four schools dropped out after randomisation (two intervention and two control), citing staffing shortages and pressures of the pandemic as the reasons. All 45 schools successfully completed the follow-up 1 year after baseline data collection. The statistician analysing the data was blinded to the allocation until all analyses were completed.

After the pandemic, we found that consent rates had dropped from those seen in the pilot trial. We received approval from the ethics committee to change from opt-in to opt-out consent for the data collection, which was granted, made easier by the fact that we were no longer collecting body measurements. We still required opt-in consent to obtain the demographic data for each student from the schools. Baseline data were obtained from 2065 students (81%) of those eligible. Consent rates were lower in the initial 15 schools (8 intervention and 7 control) that used opt-in rather than opt-out consent. Apart from that, the main reason for non-participation was absence from school. Follow-up took place 1 year after baseline. As this was inevitable in a new school year by which time students had moved to different classes and school absences post pandemic nationally have been much higher than pre pandemic, the follow-up rates were not as high as in the pilot trial. Of those who took part at baseline, 1127 students completed the follow-up. Thus, we achieved a follow-up rate from baseline of 55%. Details are provided in the Consolidated Standards of Reporting Trials (CONSORT) diagram (Figure 9).

FIGURE 9 The CONSORT diagram for EACH-B trial. (Figure not reproduced here; see the original article.)

All students were asked to wear a GENEActiv watch to record PA over the course of 1 week following the baseline data collection. The students were not keen to wear the watch as they did not look good. Some 1153 students provided GENEActiv data for a minimum of 3 days that could be used in the analysis (56% of those completing the baseline data). At follow-up, 732 students had sufficient activity data for use, but only 570 (28% of those recruited at baseline) of those also had usable baseline data.

Opt-in consent for the demographic data resulted in data for fewer students than had provided questionnaire data using the opt-out procedure. Of the 2072 students who provided baseline questionnaire data, demographic data were available for 1332 (64%).

Despite the minimisation algorithm used for the randomisation of schools, there were some imbalances in the baseline data between two arms of the trial. Notably, and worryingly, there was a 0.22 SD difference (adjusted for clustering) in the diet quality score between the two groups. This variable forms one of the two primary outcomes, and a difference of this size is almost as large as the 0.25 SD that the trial was powered to detect in this measure at follow-up. This complicates the interpretation of the analysis. Clearly, cluster-randomised trials have a higher risk of imbalance than individually randomised, as the randomisation is at school level, and only 45 schools took part (comprising over 2000 individual students). Figure 10 shows the imbalance.

FIGURE 10 Graphical representation of the baseline means for quality of diet and CIs by randomised groups. (Figure not reproduced here; see the original article.)

Full details of the baseline characteristics are presented in Table 4, which shows the baseline characteristics of all those students who provided data for the baseline survey.

From the table, one can see that the major difference at baseline between the two groups was in the diet quality score. The difference is greatest in those who did not take part at follow-up (data not shown), though it is still apparent in those who took part at both time points. Differences in the PA data between groups were much less marked as shown in Figure 11.

FIGURE 11 Graphical representation of the baseline means for PA and CIs by randomised groups. (Figure not reproduced here; see the original article.)

The main analysis was conducted using all the data available using an intention-to-treat approach. Multilevel models were used to account for clustering. Multiple linear regression analysis was used to compare all continuous outcomes from the follow-up data, and a Poisson model with robust standard errors and a log-link function for the one binary outcome under consideration. These simple analyses compared the outcomes at follow-up between the two arms, but further models were adjusted for the baseline responses for the outcome measure under consideration; this was particularly important for the diet quality score given the imbalance at baseline. Given the low response for the demographic data, these variables were not included in the main models so as to retain as much data as possible. Three continuous PA variables had to be logged to provide variables that were approximately normally distributed.

A crude (unadjusted and ignoring clustering) analysis of variables at follow-up showed some small differences between the two arms of the trial, but nothing of particular note. There was no apparent effect of the intervention in either of the two primary outcomes as shown in Figures 12 and 13.

FIGURE 12 Unadjusted mean diet quality scores (SD) at follow-up. (Figure not reproduced here; see the original article.)

FIGURE 13 Comparison PA (minutes/day) between the two groups at follow-up. (Figure not reproduced here; see the original article.)

The unadjusted findings are presented in the first column of differences/relative risk (RRs) in Table 1.

The multilevel models taking account of clustering and adjusting for the baseline values of the outcome variable in question showed little in the way of differences between the two groups (see second column of differences/RRs in Table 1). For the diet quality score, the difference at follow-up after adjustment for baseline was −0.10 (95% CI −0.28 to 0.09) SDs, and for total, PA was 8.65 (95% CI −7.79 to 25.1) minutes per day. These difference are small and of little consequence.

In terms of the various secondary outcomes, the largest difference was seen for consumption of crisps (a binary variable indicating consumption of more than 250 g per week or not). The RR associated with the intervention for consuming this amount was 0.74 (95% CI 0.62 to 0.89). Similarly, but not quite so strikingly, the RR for consuming more than a litre of fizzy drinks per week was 0.77 (95% CI 0.57 to 1.03). These were the most marked findings associated with the intervention, but given that these were two of many secondary outcomes, we feel we cannot conclude that this is solely due to the intervention. The baseline differences in relation to overall diet add to the complication of interpreting the results, despite adjusting for the baseline levels.

For the PA secondary outcomes, few differences were associated with the intervention. Of some interest is the finding that the intervention was associated with a 41% increase in the proportion of students meeting the 60-minute activity guideline [RR = 1.41 (95% CI 0.99 to 2.03)]. However, none of the other PA outcomes showed any effect of the intervention, so it is hard to conclude that this is other than a chance finding.

We considered a missing data analysis, but the amount of missing data was large, and most unlikely to be random. The large difference in one primary outcome at baseline is the major challenge in interpreting these data, and this raises questions about the viability of cluster-randomised trials in school settings and maybe elsewhere. This imbalance arose despite the use of a minimisation algorithm. Participation and follow-up rates were remarkably similar between the two arms.

As noted above, we altered the consent process to be opt-out part way through the trial, though in order to obtain the demographic data (gender, date of birth, home postcode, ethnicity), opt-in consent was required. The opt-in consent rate was 64%. From the postcodes, we obtained the Income Deprivation Affecting Children Index (IDACI) for each child. These provide the decile of deprivation in which each child lives. Although Southampton is in the wealthy south of the country, Southampton, Portsmouth and the Isle of Wight, from where many of the schools were recruited, are in the poorest quarter of local authority areas in England. A graph of the IDACI scores at baseline from those whose parents provided consent is below.

Figure 14 shows that the distribution of IDACI scores were skewed towards the most advantaged students, despite the underlying population, including some very deprived areas. It is likely that those parents who opted in for transfer of the demographic data were drawn from the more advantaged sections of the population. However, when we consider those who were followed up, the graph is even more skewed as shown in Figure 15.

FIGURE 14 Distribution of IDACI scores for participants at baseline. (Figure not reproduced here; see the original article.)

FIGURE 15 Distribution of IDACI scores for participants at follow-up. (Figure not reproduced here; see the original article.)

These graphs (see Figures 14 and 15) indicate some of the challenges in recruiting from schools. Consent is required from parents for key data, and those least in need of the intervention are more likely to take part. However, even with consent, the dropout disproportionally occurs among those from the more disadvantaged backgrounds. Targeting the more advantaged makes it even harder to show an impact from an intervention focusing on health behaviours; at the outset, these students are more likely to have better health behaviours, due to their advantages, so change is both less necessary and harder to observe.

Further, as shown in Table 5, those who opted in for the demographic data collection and who took part at both baseline and follow-up had slightly lower IDACI scores (less deprived) than those not followed up in both intervention and control groups. Throughout Table 5, it is clear that the intervention group is somewhat more deprived than the control group. However, none of these differences were particularly marked. Given the poor follow-up rates and the low proportion of participants with demographic data, further analyses of subgroups were not attempted, as they would be biased and hard to interpret.

In terms of compliance with the intervention, this was complicated by the fact that nine schools delivered the LifeLab day in the schools (due to the pandemic). Some 392 students who participated in the baseline survey were in those 9 schools, and we do not have attendance data at school on the day for them. The experience of LifeLab was thus different for these students than for the remaining 641 students who participated in the baseline in intervention schools. Of these, 561 (88%) participated in the LifeLab day in the hospital setting. All participants will have had some exposure to the modules of work leading up to and following the LifeLab day (unless they were absent from school for many weeks).

We cannot assess compliance with delivering the healthy conversations intervention component among the students, as that depended on opportunistic interactions with a teacher trained to use the skills to have an appropriate conversation. Teachers said that this was hard to do in a busy school day, and the conversations were obviously not recorded. Unfortunately, uptake of the app was very low. Despite most students downloading the app, often during the LifeLab day at the hospital, only 228 (22% of those in the intervention arm at baseline) opened it and logged in. These 228 students used the app an average of 2.2 sessions with a duration per session of 17.7 minutes. This low level of engagement with the app is unlikely to have impacted on the students’ health behaviours. Given the complexities of assessing compliance with the intervention, a more sophisticated analysis has proved impossible, and again the imbalance at baseline makes any analysis difficult to interpret.

Engaging Adolescents in Changing Behaviour process evaluation

The EACH-B process evaluation examined the implementation, context and mechanisms of impact of the LifeLab Plus intervention. A mixed-methods approach was employed, drawing on the Medical Research Council (MRC) guidance on process evaluation of complex interventions. Fifty-one interviews with students, teachers and parents were conducted and analysed using thematic analysis. Quantitative data reported numbers of students who attended LifeLab and downloaded the app.

The results showed that 87.9% of eligible students attended LifeLab, and 45.7% of eligible students downloaded the app. Participants had positive experiences of LifeLab and the teacher training, despite the pandemic preventing in-person delivery for some schools. Students engaged in a limited way with the app. Parents had positive views of the research but little knowledge of the intervention. Students valued learning about their own health, and teachers were enthusiastic about supporting their students’ health but struggled to find opportunities to do so.

Research describing the research team’s perceptions of the process of delivering and evaluating the intervention concluded that schools are appropriate settings in which to implement health interventions, but that research must benefit teachers and students (a description of this research is provided in Report Supplementary Material 1). Most students did not engage with the app because they were not presented it in a way that motivated them to use it. Careful consideration of the design of apps is required to encourage students to use them. Parents should be involved in research in ways that are accessible to them.

Discussion and study limitations

Conducting trials in schools is challenging at the best of times, but EACH-B was conducted through the pandemic. The trial was successfully completed, which, given the difficulties faced in getting schools to do anything extra in such unusual times, is remarkable. However, the results are disappointing in that we have been unable to demonstrate an effect of the intervention on the two primary outcome measures or a range of secondary variables (psychological measures and other assessments of PA). The process evaluation has given greater insights than the main trial itself and has enabled us to conclude that while schools are undoubtedly a good setting for delivery of public health interventions, interventions have to offer something meaningful to both students and staff to be effective.

Description of the data, findings and analysis methods

Through systematic review and meta-analysis, we have established that digital platforms and social media are key to engagement of young people and that school-based interventions can be successful in improving health behaviours and as part of an obesity prevention strategy but require engagement from key stakeholders, parents and teachers in particular. Despite primary evidence that food environments can be influential in shaping young people’s diets, findings from our review of the impact of the consumer environments were inconsistent; intervention studies are required to clarify associations with diet.

Our qualitative data formed a large part of that generated by the EACH-B programme. A thematic analysis of interview and focus group data collected as part of the intervention development identified that any work with young people had to meet their needs for autonomy, competence and relatedness. These needs were taken into the guiding principles on which the EACH-B intervention was based and have formed the basis for all the work with young people that we have gone on to do since. Game jams and think-aloud interviews generated data that helped develop game features for the smartphone application. Parents were keen to be involved and aware of what their young people were involved in at school but struggled to communicate. Teachers were committed to supporting students’ health and well-being but felt they were fighting a battle against a wider food environment that they could not win.

Qualitative data collected during the process evaluation of the pilot and main trials illuminated the opportunities and issues with running trials in schools. Schools remain excellent locations to reach young people across the SES spectrum, and teaching staff are fully committed to supporting their students to take part in research. Young people also appear to enjoy taking part in research, particularly if the activities are meaningful to them. The issue in schools is the competing demands of the curriculum and national exams on both staff and students. It is clear from our analysis of interviews and group discussions that research has to represent a ‘gift’ as well as an ‘ask’ if it is to be run successfully in schools.

Baseline and follow-up assessments of dietary and PA outcomes pilot and main trials provided most of the quantitative data generated by the EACH-B programme. Multiple regression analyses of the data from the 2065 students involved showed no clear differences in dietary quality and levels of PA in the intervention students and the control students. Interpretation of these data was complicated, however, by the fact that the randomisation process failed to balance the two groups of students at baseline. Those in the intervention group have substantially higher dietary quality before the intervention than those in the control group. This meant that we were highly unlikely to see any effect of the intervention on young people’s diets. This trial was cluster-randomised, and the consent for individual participation had to be obtained after randomisation, which is not ideal. Cluster-randomised trials randomise small numbers of units and can therefore lead to imbalance at baseline. Usage data from the smartphone application was also inconclusive in that less than one-quarter of the young people engaged with the app features. We have no way of knowing what impact it might have had on their diet and PA had they used it, but take from this that a smartphone application has to offer more than our gamified app did to engage young people in the first place.

General limitations of the programme

The EACH-B programme has caused us to reflect on the viability of running cluster-randomised trials in school settings. The large difference in one primary outcome at baseline is the major challenge in interpreting the data from the trial. This imbalance arose despite the use of a minimisation algorithm. Participation and follow-up rates were remarkably similar between the two arms, but dropout between baseline and follow-up was particularly likely among students from more disadvantaged families. Because we required opt-in consent to the process of collecting demographic data, we only achieved this from 64% of students, with the most disadvantaged students being least likely to provide consent. These processes and the way they select out students likely to be most in need of intervention and support indicate some of the challenges of running trials in schools. Working with the more advantaged students makes it even harder to show an impact from an intervention focusing on health behaviours; at the outset, these students are more likely to have better health behaviours, due to their advantages, so that impact of the intervention is both less necessary and harder to observe.

The concept of the EACH-B intervention was supported by our review, which showed that school-based interventions can be effective in supporting healthier behaviour in young people.22 At the time of developing the EACH-B intervention, there was strong evidence from multiple reviews that interventions delivered in schools could increase PA in adolescents,60 improve sexual health and prevention bullying and smoking.61 There was less evidence that school-based interventions could improve diet, but we planned to use what was known about how to address other health behaviours. Key factors in the effectiveness of published interventions were that they were integrated into school settings and curricula and tailored precisely to the target populations, as we achieved in EACH-B. Other key factors that were not present in the delivery of the EACH-B intervention were that effective interventions were embedded within the school, families and the local community, delivered by school co-ordinators or became a feature of school policy. None of these was the case with the EACH-B intervention. These may be reasons why the EACH-B intervention may not have been successful. Alternatively, it may be that our trial was so challenging to conduct, given the impact of the pandemic and consequent school absence, that we may not have conducted an evaluation that would have detected any change. There are important lessons for us about how best to evaluate interventions in these complex settings.

In terms of compliance with the intervention protocol, this was complicated by the fact that nine schools delivered the day in the schools; pandemic restrictions meant they were unable to come into the hospital where LifeLab is situated. Some 392 students who participated in the baseline survey were in those 9 schools and had an experience of LifeLab that was different from that of the remaining 641 students who took part in the baseline in intervention schools of whom the 561 (88%) participated in the LifeLab day in the hospital setting. Our inability to conduct the planned cost-effectiveness evaluation of LifeLab Plus was caused by imbalances in a primary outcome measure at baseline and in differential loss to follow-up of groups of students. This made the outcome data impossible to draw conclusions from and observed between groups that differences in other secondary outcomes were too small to be of consequence. We have discussed this as a limitation in the study above at length. It is difficult to see how we could have prevented this imbalance produced as a result of the randomisation. Despite conducting four separate evidence reviews, we were unable to identify cost-effectiveness data from previous digital or other types of intervention with adolescents.

We had no way of assessing exposure of students to healthy conversations aimed to support behaviour change but have to assume it was low. The nature of these conversations is that they depend on opportunistic interactions, in the case of EACH-B, with a teacher trained to use the skills. Our process evaluation identified that teachers found this hard to do during a busy school day. Equally, exposure to the games, challenges and materials in the smartphone application was also low. Despite most students downloading the app, less than a quarter of them opened it and logged in. These students used the app an average of twice for less than 20 minutes per session. As such, the app is unlikely to have impacted on students’ health behaviours. Had we been able to pilot for longer or more extensively, we may have identified issues with the app that resulted in the disappointing level of engagement with it and therefore of exposure to this component of the intervention. On the other hand, issues with engagement may simply reflect the fundamental problem inherent in creating a digital experience that is sufficiently exciting and meaningful to young people that it will compete with social media for their attention.

Despite the fact that the trial did not enable any conclusion to be reached about the effectiveness of the EACH-B intervention, the EACH-B programme still represents a significant contribution to understanding adolescent health behaviour and how best to engage young people in looking after themselves. Its comprehensive approach, combining intervention development, rigorous evaluation, stakeholder engagement and extensive knowledge translation have produced a series of initiatives which are already demonstrating impact in research and policy in this crucial area of public health.

Modifications to the delivery of the Engaging Adolescents in Changing Behaviour programme, intervention and trial during the first 2 years of the COVID-19 pandemic

Recruitment and data collection for the RCT was paused in March 2020 due to the UK national lockdown in response to the pandemic and was resumed in March 2021. During this year-long period, other projects were conducted using a combination of EACH-B and additional funding.

We successfully restarted EACH-B following the lockdowns and school closures due to the pandemic. Schools did not start back properly until February 2021. Our engagement with schools throughout, our work with the saliva testing programme and our discussions with adolescents conducted in a companion project Teenagers and COVID-19 (TeC-19) led to us being ready to restart the trial quickly. The intervention and data collection procedures were successfully adapted to cope with the challenges of the pandemic.

Data collection: our plans for data collection were for members of the study team to visit each school and oversee the data collection. Schools were not willing to have visitors, so the data collection procedures were modified to be conducted by the class teacher, with the students entering the data directly online. We suspended collecting height and weight measurements. Students who were isolating at home due to the pandemic could complete the data collection from devices at home, with instruction and support from the teacher. EACH-B team members were available for teachers to consult in the event of difficulty. The GENEActivs for activity assessment were delivered to schools by the EACH-B team and were distributed by teachers and retrieved from students after use.

The intervention: we modified our procedures to minimise risk to school staff and students and our own team. (1) The teacher professional development moved online. It was adapted to explain the changes to the data collection procedures and the delivery of the intervention. More support was provided to the teachers in terms of training in HCS. (2) Through the COVID-19 saliva testing programme, an online version (https://lifelabonline.org/course/index.php?categoryid=56) of the LifeLab modules was developed. This enabled students to work through the LifeLab lessons in a self-directed manner and so allowed any students who were at home isolating to be able to participate. (3) LifeLab is situated in Southampton General Hospital, and the hospital would not allow visitors onto the site, nor did schools or parents want to send children to a hospital. The LifeLab team developed a portable LifeLab day so that the activities could be conducted in schools. The equipment was delivered to the school in large flight cases containing full instructions for their use. (4) The app that was part of the intervention was adapted to reflect the COVID-19 pandemic. This included avatars wearing face masks and provision of more support for solitary exercise.