A school-based, peer-led programme to increase physical activity among 13- to 14-year-old adolescents: the GoActive cluster RCT

Feasibility results

The intervention was delivered by one school to the whole of Year 9 for 8 weeks during the summer term of 2013, with limited researcher assistance. Despite initially agreeing to do so, the school was unable to provide mentors, as the older students had examinations. Year 9 form teachers were trained to deliver the intervention prior to the programme commencing. The teachers delivered the intervention with the help of one GoActive team member during tutor time once per week. A total of 234 Year 9 students were exposed to the intervention, as reported by the school, with nine parents (representing 3.8% of eligible students) and 13 students (5.6% of eligible students) opting out of participation in study measures. A total of 183 (78.2%) students assented to participate in measurements, with 29 (12.4%) students not attending a measurement session (because of absence or apathy).

Pilot results

Across the three pilot schools, 596 Year 9 students were invited to participate in the evaluation of GoActive. A total of 458 students provided valid written informed assent and were measured (76.8% response rate; an average of 153 students per school) (Figure 2).

FIGURE 2.

Pilot study recruitment flow chart. All measures include accelerometry, anthropometry and outcomes questionnaire (student-reported physical activity participation, self-efficacy, peer support, friendship quality and mood).

Of the 458 baseline participants, 87.3% attended the follow-up measurement. Of these students, 400 (55%) were available for analysis of the primary outcome [n = 220; ≥ 1 day of ActiGraph data at pre and post intervention (ActiGraph, LLC, Pensacola, FL, USA)] and all 400 completed the questionnaire-based measures that assessed the secondary outcomes. Average days of accelerometer wear were 4.9 (SD 1.8) days pre intervention and 3.8 (SD 1.8) days post intervention.

The results of this pilot cluster RCT provided an indication of the potential effect of GoActive on the main outcome measure (i.e. average daily minutes in MVPA). Change in MVPA was –6.5 (SD 14.0) minutes per day in the control group and –2.5 (SD 15.4) minutes per day in the intervention group, with a change adjusted for baseline of 5.1 [95% confidence interval (CI) 1.1 to 9.2] minutes per day in favour of the intervention group. Furthermore, the results of the questionnaire-based measures indicated tentative positive effects for some secondary outcomes, including well-being and social support.

Feasibility and pilot summary

The feasibility study and pilot trial of the GoActive intervention showed feasibility of recruitment, measurement, randomisation and the ability to deliver GoActive to a whole-school year group of 13- to 14-year-olds. Both of these stages prompted several key improvements to the intervention and to the study design, including emphasis on monitor return, mentor recruitment, adequate mentor training, clearer and more consistent intervention explanations, and an improved points recording systems. The lessons learnt from each phase of this research were taken forward to the trial reported on here, aiming to evaluate the clinical effectiveness and cost-effectiveness of the GoActive intervention to increase MVPA among 13- to 14-year-olds.

Studying mechanisms

We hypothesised that the GoActive intervention components suggested by students and teachers (including mentorship, leadership, teacher support, class-based activity sessions, competition, rewards and online activity tracking) would also influence well-being through social support, self-efficacy, group cohesion, friendship quality and self-esteem. We wanted to investigate if particular engagement with components of GoActive may be associated with mediators and outcomes, for example encouragement provided by older adolescent mentors could be associated with increased self-efficacy and social support, which may be associated with changes in physical activity and mental well-being. 57,58

Mixed-methods process evaluation

The literature on school-based physical activity interventions highlights the need to critically examine the delivery of each component and the processes of any complex intervention. 59 Process evaluation provides detailed evaluative information about the delivery of an intervention from the perspective of participants, mentors, teachers and facilitators, with the aim of contextualising and interpreting its potential effects, providing greater confidence in conclusions about effectiveness. 60

It is also important to understand the primary recipients’ experiences and perspectives of the intervention, as this can not only help to determine the effectiveness of the intervention and its individual components, but can also produce new insights regarding intervention design. There is limited published research of the experiences of those directly involved in receiving these interventions,61–63 particularly using qualitative research methods. Process evaluations of school-based interventions have mainly focused on other stakeholder groups, notably parents and teachers. The process evaluations note the importance of child engagement and autonomy over their own behaviour,64 embedding diet and physical activity across the curriculum and school,64 school resources,64 parental and community engagement,61,64 and the importance of adult role models61,64 in delivering school-based physical activity interventions. However, such stakeholder perspectives cannot replace the sharing of views and experiences of adolescents themselves.

In recent decades, literature and policy have emphasised the importance of capturing views, experiences and actions of those who are the focus of the research. 65 Moore et al. 60 states that the success of an intervention is dependent on the response of the intended audience. This is traditionally considered ‘dose received’,66 and more specifically ‘dose received: satisfaction’. 67 It is argued that ‘dose received’ is a passive term, privileging quantitative measures. 60 Moore et al. 60 advocate for the critical assessment of aspects of dose received, including ‘acceptability’ and ‘satisfaction’, to examine participants’ relationships to the mechanisms through which the intervention works. Additionally, they call for the inclusion of qualitative methods as an effective approach to illuminating participant experiences. Although GoActive was designed as a whole-population approach, aiming to overcome stigmatisation and the potential detrimental mental health consequences of targeting particular subgroups in health promotion strategies,68 we wanted to ensure that the intervention was acceptable to those least likely to engage in physical activity. As such, the intervention was developed to include the opinions of those with characteristics that were deemed as common in those hard to reach for physical activity promotion (including girls, and those with low activity levels and high shyness). 6 Therefore, participants for our process evaluation were purposively sampled to account for perspectives of individuals with those characteristics.

Overview

The intervention is titled ‘GoActive’, which stands for ‘Get Others Active’. Each Year 9 class (i.e. tutor group/home room class) are asked to choose 2 of 20 available activities each week. These activities require little or no equipment, and are designed to be appealing to a wide variety of students (e.g. ultimate frisbee, Zumba and hula hoop). Materials available on the GoActive website include activity instructions (Quick Cards), which offer an overview of the chosen activity, a short explanation and suggestions for adaptations, and provide advice, safety tips and ‘factoids’ about each activity (see Appendix 1).

The GoActive intervention is implemented using a tiered leadership system, led by a council-funded facilitator. Schools were requested to select mentors. Mentors (i.e. older adolescents within the school) and peer leaders (i.e. select pupils within each class) encouraged students to try these activities each week (Figure 3). The mentors remain paired with each class for the duration of the intervention, whereas the in-class peer leaders (two per class each week, one male and one female) change every week. Teachers are encouraged to use one period of tutor time (registration/roll call) weekly to do one of the chosen activities as a class; however, students gain points for trying these new activities in or out of school. Points are gained every time a student tries an activity. There is no expectation of time spent in the activity, as points are rewarded for the taking part itself. Individual students keep track of their own points privately on the study website and their points are entered into the between-class competition so that each class competes against each other. Class rankings are circulated each week to encourage teacher support and students receive small rewards (e.g. frisbee, sports bag) for reaching points thresholds (such as 20/50/100 points). As GoActive runs on a weekly cycle, the length of the intervention can vary as appropriate for each individual school.

FIGURE 3.

The GoActive tiered delivery system.

The full intervention ran for 12 weeks, with 6 weeks of weekly facilitator visits and a further 6 weeks with reduced facilitator support. The role of the facilitator was to provide training for mentors and teachers, provide instructions/prompts to teachers, mentors and Year 9 leaders for intervention delivery, support teachers, mentors and leaders to deliver the intervention, provide activity Quick Cards that give overviews and examples of activities, provide basic equipment to facilitate activities if relevant/needed, collect point cards, update points boards in school communal area and class rooms, and to provide prizes. As described in Chapter 1 and clarified in Table 3, several intervention improvements were identified following the pilot intervention and these were incorporated into GoActive. The main differences between the intervention delivered in the pilot evaluation and the GoActive intervention delivered in the trial reported on here were (1) the use of the GoActive website to track points and facilitate communication with the mentor, and (2) delivery was led by a local council-funded facilitator instead of by a member of the research team.

The teachers, mentors and peer leaders delivered the intervention after training from a facilitator (see Facilitator and mentor training). Quick Cards provided information that allows any of these individuals to lead the 20 activities. For example, we suggested using YouTube (YouTube, LLC, San Bruno, CA, USA) for Zumba instruction, as we want the students to be able to try activities without the barrier of needing a specific class. Similarly, we suggested doing these activities at home with a friend or relative to encourage out of school participation.

Table 4 summarises the behaviour change techniques applied in GoActive. Table 5 describes the GoActive intervention’s essential functions (i.e. components of the intervention) that schools were asked to implement. Importantly, these essential functions could be adapted to suit individual settings, offering the schools flexibility to adapt the intervention and its implementation to their context.

Facilitator and mentor training

All facilitators took part in a 1-day training session before the start of the GoActive intervention. Training was provided by members of the GoActive research team and an external consultant specialised in adapting physical activity promotion to different contexts. The contents of the training focused on providing background information on GoActive and its evaluation, a detailed overview of the intervention and the role of the facilitator in intervention delivery, the options for flexibility in intervention implementation, and the resources available to the facilitators and teachers/mentors. All facilitators additionally received a 33-page manual and the research team was available throughout the intervention delivery phase if any questions or concerns arose.

The role of the GoActive intervention facilitator included training teachers and mentors. The audience of this training included Year 9 form tutors and/or house leaders, GoActive mentors and a selected school contact [often physical education (PE) or Year 9 lead]. Training sessions were scheduled for 1 hour. The sessions first provided an overview of the GoActive programme, followed by a question and answer session, during which the audience was encouraged to ask questions or clarify details. Practical training (including role playing, if appropriate) was given for (1) daily intervention delivery, (2) ideas for motivating leaders, (3) overcoming problems and (4) ideas for planning and organising activities. A slide deck and a suggested training schedule were provided to each facilitator to support training and all mentors/teachers received a manual. It was stressed that expectations should be made clear during teacher/mentor training, and the importance of and process for risk assessments was discussed.

Overall study

The overall objective was to assess the clinical effectiveness of the GoActive intervention in increasing daily MVPA in 13- to 14-year-old (Year 9) adolescents and to establish the cost-effectiveness of the programme.

The specific aims of the project were as follows:

• To assess the post-intervention and 10-month effectiveness of the GoActive programme in increasing average daily accelerometer-assessed MVPA among 13- to 14-year-old adolescents.

• To assess the effect of GoActive on the following secondary outcomes:

  • accelerometer-assessed sedentary time, light physical activity (LPA) and overall physical activity during school time, weekday evenings and weekends

  • student-reported physical activity participation, self-efficacy, peer support, self-esteem, friendship quality and well-being

  • body composition [i.e. body fat percentage and body mass index (BMI) z-score].

• To assess the short-term (within-trial) and potential long-term cost-effectiveness of the programme.

• To assess programme acceptability, uptake, maintenance and dose.

• To investigate potential moderation of intervention effects (by gender, socioeconomic status, ethnicity, baseline activity level and weight status) and potential mechanisms of effect by proposed mediators, including peer support, friendship quality, self-efficacy and self-esteem, using a mixed-methods approach.

Mixed-methods process evaluation

Process evaluation is key to being able to assess fidelity and quality of implementation, clarify causal mechanisms and identify contextual factors associated with variation in outcomes. 69 By conducting a thorough process evaluation, combining both quantitative and qualitative data collection methods and insights from different stakeholder groups, it is possible to identify the barriers and facilitators that influence the delivery of the intervention within different contexts, investigating successes and failures of implementation to maximise the learning from the trial delivery of an intervention. 70

The implementation of the programme in each school was assessed through a mixed-methods process evaluation, which included focus groups with and questionnaires completed by students and mentors, as well as individual student, teacher and facilitator interviews and observations of GoActive sessions. The process evaluation protocol has been published previously25 and part of the process evaluation was published prior to analysing intervention efficacy to avoid interpretation bias. 21

The specific objectives of the GoActive process evaluation were as follows:

1. To assess the reach, dose and fidelity of intervention delivery, to document how the intervention was implemented and to ascertain whether or not the intervention’s essential functions (i.e. components of the intervention) were adapted to suit individual settings (see Table 5).

2. To explore the GoActive intervention from the perspective of Year 9 students, mentors, teachers and facilitators, and to describe participants’ views of the intervention (including intervention acceptance).

3. To consider the maintenance and sustainability of the intervention and, if proven effective, the possible dissemination of the GoActive intervention.

The process evaluation was designed to observe the implementation of GoActive and was not used to intervene in how schools implemented the GoActive intervention where deviations from the intervention protocol were detected. The design of the process evaluation was informed by the Medical Research Council (MRC) guidance on the process evaluation of complex interventions. 60,69,71

Accelerometer-assessed outcomes (including primary outcome)

The prespecified primary outcome for effectiveness was average daily minutes of MVPA at 10-month follow-up. We measured MVPA at baseline, post intervention and 10-month follow-up using wrist-worn activity monitors (Axivity) that assess acceleration (continuous waveform data). As a change to the published protocol, wrist-worn monitors were used instead of the originally proposed hip-worn ActiGraph monitors (Axivity Ltd, Newcastle upon Tyne, UK) because of the increased likelihood of protocol adherence in the target group compared with what was seen in our pilot work. Participants were asked to wear the monitors continuously for 7 days, 24 hours a day on their non-dominant wrist. Monitor output was processed to provide minutes spent in MVPA and secondary outcomes.

Accelerometery validity and processing

The Axivity monitor has been validated to assess physical activity energy expenditure72 and to have better wear time adherence and acceptability than commonly used hip-worn monitors, among adolescents. 73 Given the 24-hour wear time protocol of the Axivity monitors, a diurnal adjustment was used to reduce any bias caused by imbalances of protocol deviations regarding non-wear. 74 Each day of possible wear was divided into four time quadrants: morning (06.00 to 12.00), afternoon (12.00 to 18.00), evening (20.00 to 24.00) and night (24.00 to 06.00). For participants to be included in analyses, over 6 hours of wear time spread over at least 2 days was required from the possible 42 hours in each daytime quadrant (i.e. ≥ 6 hours from 7 possible mornings, ≥ 6 hours from 7 possible afternoons and ≥ 6 hours from 7 possible evenings). The ‘night’ quadrant (i.e. 24.00 to 06.00) was considered as sleep time and was included in the denominator when calculating daily averages of MVPA, for consistency across all participants. Where individuals did not wear the monitor for ≥ 6 hours at night-time, despite the protocol requesting them to wear it continuously for 7 days, average night-time values were imputed using population averages (n = 91 at baseline and n = 463 at follow-up), created from GoActive participants with 100% protocol compliance regarding monitor wear, to avoid inflation of MVPA estimates. This method was verified by running simulations that exclude night data for a subsample of participants with 100% protocol compliance. For an individual hour to be included for analysis, at least 70% of possible wear time was required, with non-wear time within the hour considered as missing. 74

Monitor output was processed to provide minutes spent in MVPA equivalent to ≥ 2000 ActiGraph counts per minute (c.p.m.). 4 Additional secondary accelerometry-derived outcomes were average daily minutes of sedentary time (equivalent to ≤ 100 ActiGraph c.p.m.), average daily minutes of LPA (equivalent to 101–1999 ActiGraph c.p.m.) and average daily activity (represented by average acceleration). In addition to daily averages, all intensity outcomes (including MVPA) were derived during school time (09.00 to 15.00), during weekday after school time (after 15.00) and at weekends. Participants who met the inclusion criteria for average daily MVPA were included in any analyses for which they had sufficient data (≥ 2 days). 75 As the criteria for deriving average daily MVPA did not require both weekend and weekdays of valid data,76 participant numbers varied by outcome.

Reported demographic data

Participant descriptive characteristics, including prespecified effect modifiers [e.g. gender, individual socioeconomic position (SEP) and ethnicity] were self-reported at baseline. Ethnicity was self-reported by participants who were given 20 response options and additional free-text completion options. For descriptive purposes, the reported values were recoded to five categories in accordance with recommendations:77 (1) ‘white’, (2) ‘mixed ethnicity’ (i.e. identifying with multiple ethnicities), (3) ‘Asian’ (including South Asian and Chinese), (4) ‘African and/or Caribbean’ and (5) ‘other’. Ethnicity was subsequently dichotomised for prespecified moderation analyses (‘white’ vs. remaining categories). Participants completed six items from the Family Affluence Scale (FAS) that relate to family car ownership, holidays, computers, availability of bathrooms, dishwasher ownership and having their own bedroom, which was used as a proxy of individual SEP by summing answers (possible range: 0–13) and dividing into predefined affluence groups (low = 0–6, medium = 7–9 and high = 10–13). 78,79 Shyness and sociability were assessed using five items in the Emotionality, Activity, Shyness and Sociability Scale. 80

Self-reported outcomes

Secondary outcomes were assessed at baseline, post intervention and 10-month follow-up using questionnaires that included measures validated for use in the population (e.g. social support for activity, self-efficacy, group cohesion, friendship quality, self-esteem, shyness, sociability and mental well-being). GoActive questionnaires are available as additional material [see NIHR Journals Library, URL: www.journalslibrary.nihr.ac.uk/programmes/phr/139018 (accessed 12 January 2020)]. Social support for physical activity was a mean derived from nine self-reported items (response range: 1–4) from the European Youth Heart Study. 81 Self-efficacy was a mean of eight self-reported items from Reynolds’ Psychosocial Predictors of Physical Activity: Self-efficacy scale82 (response categories: 1–6). Group cohesion was assessed by an adapted social network modelling tool in which participants were provided with a list of tutor group members on a laptop and were asked to select the names of up to five of their friends from the list provided. These data were used to derive in-degrees (i.e. the number of people identifying the participant as a friend) and out-degrees (i.e. the number of friends that participant lists as a friend). 83 Friendship quality was a mean score of eight self-reported items used in the ‘ROOTS’ study (equally weighted),84 with a response range of 1–5. Self-esteem was a mean score of self-reported items using the 10-item Rosenberg Self-Esteem Scale,85 which had original response options of 1–4. Self-reported physical activity was assessed using the Youth Physical Activity Questionnaire (YPAQ). 86 Activity frequency was calculated as sessions per week of all reported activities (range: 0–160).

The Warwick–Edinburgh Mental Wellbeing Scale was used to assess mental well-being. 87 The 14-item scale asks participants to indicate what best describes their experiences of a collection of statements over the past 2 weeks. Items relate to both hedonic and eudemonic experiences of mental health, including positive affect (e.g. ‘I’ve been feeling optimistic about the future’), relationships (e.g. ‘I’ve been feeling close to other people’) and emotional functioning (e.g. ‘I’ve been dealing with problems well’), each rated on a five-point Likert scale with responses ‘none of the time’, ‘rarely’, ‘some of the time’, ‘often’ and ‘all of the time’ scored 1–5, respectively. The scale has shown good content validity, and correlates well with other mental health and well-being scales, including the Positive and Negative Affect Scale,88,89 Short Depression-Happiness Scale90 and the World Health Organization (Five) Well-Being Index. 91 Cronbach’s alpha has been shown to be 0.89 among a student sample and 0.91 in a population sample. 87 Overall, the scale has shown good internal consistency, good test–retest reliability and good face validity. 92 Participants who responded to all 14 items on the scale were included in the mediation analysis and their score was the average of the item responses (out of a possible 1–5).

Anthropometry

Anthropometry (i.e. height, weight, waist circumference, bioimpedance) was assessed at baseline and at 10-month follow-up by trained staff. BMI z-score was calculated from height, weight, age and gender. 93 BMI z-score was also used to identify participants with overweight or obesity. 94 As a change to the published protocol,4 anthropometry was not assessed at post intervention to reduce the burden on schools and participants, and because no meaningful short-term effect on anthropometry was expected.

Appendix 3 lists further details regarding all outcome variables, including for the primary outcome and all secondary outcome measures.

Process evaluation data collection

As stated earlier, process evaluation measures were collected mid-intervention, post intervention and at 10-month follow-up. Details on the quantitative and qualitative data collected are provided below. An overview of the process evaluation data collected is presented in Table 6.

Analysis of effects on primary and secondary outcomes

The statistical analysis plan was approved by the Trial Steering Committee (TSC) prior to analyses being performed. The full statistical analysis plan is available from the GoActive study website [URL: www.mrc-epid.cam.ac.uk/research/studies/goactive/ (accessed 28 February 2020)]. Statistical quantitative analyses were performed using Stata.

For MVPA at 10-month follow-up (i.e. the primary outcome), the intervention effect, representing the baseline-adjusted difference in change from baseline between the intervention and control groups, was estimated from a linear regression model, including randomisation group, baseline value of the outcome [i.e. analysis of covariance (ANCOVA)] and the randomisation stratifiers (i.e. pupil premium funding and county). Robust standard errors (SEs) were calculated to allow for the non-independence of individuals within schools and the missing indicator method95 was used to ensure the inclusion of participants with a missing baseline value of the outcome variable. All secondary outcome variables were analysed using the same method.

We conducted a complete-case analysis in which participants and schools were included in the group to which they were randomised, although participants with a missing value of an outcome at follow-up were excluded from the analysis of that particular variable. This is a complete-case analysis that is valid under the assumption that the outcome is missing at random, conditional on randomised group and the baseline value of the outcome.

Per-protocol analysis

A further analysis of the primary outcome was performed in a per-protocol population, defined as intervention group participants reporting ‘being active during tutor times at least twice during the last 2 weeks’ [i.e. self-reported intervention engagement mid-intervention (week 6 of the intensely facilitated phase of the intervention)] and logging activity points on the study website at least once during the entire intervention period. This definition was based on a review of quantitative process evaluation data prior to the main analyses and reflects the group with highest intervention engagement, as opposed to delivery of the protocol with fidelity.

Sensitivity analyses

Post hoc sensitivity analyses recommended by the TSC were performed, in which the primary outcome was calculated (1) excluding time between 24.00 and 06.00 and (2) using a stricter inclusion criterion of 12 hours of wear time per quadrant.

Effect modification

For the primary outcome and secondary accelerometer-assessed physical activity outcomes, effect modification was conducted by (1) gender, (2) socioeconomic status (medium or lower vs. high, according to FAS score), (3) ethnicity (white vs. any other ethnic background), (4) baseline physical activity (≥ 60 minutes MVPA/day vs. < 60 minutes) and (5) weight status (with normal weight vs. with overweight or obesity) and was tested with an F-test of the relative multiplicative interaction parameter in the ANCOVA model. Effect modifiers were selected based on previous evidence of potential differential effects. 13,14 Subgroup analyses were performed within all categories defined by these variables.

We used baseline accelerometer-assessed activity data to classify participants’ activity level using 30 and 60 minutes of MVPA, using (average) thresholds. The 60-minute threshold was applied to be consistent with UK physical activity guidelines, whereas the additional 30-minute threshold was applied to achieve greater heterogeneity in the moderator, as activity levels were low at baseline (minutes of MVPA at baseline for whole group: 35.6 ± 18.6 minutes/day). Subgroup analyses were performed for the primary outcome within all categories defined by these variables, but only where interactions showed p < 0.05 for secondary outcomes.

Mediation analyses

We aimed to use mediation analysis as a novel approach to evaluate the potential mediating role of psychosocial factors in the association between engagement in the intervention components suggested by students in our intervention co-design process and changes in physical activity and well-being.

Mediation of effects on the primary outcome (i.e. MVPA) and well-being was assessed using linear regression models stratified by gender (adjusted for age, ethnicity, language, school, BMI z-score and baseline values), assessing associations between (1) exposures and mediators, (2) exposures and outcomes (without mediators) and (3) exposures and mediators with outcomes, using bootstrap resampling.

Economic analysis and cost-effectiveness

A within-trial cost-effectiveness analysis comparing the GoActive intervention with the control was conducted from the perspective of the school funder (assumed to be the local authority). Cost per school and per participant was calculated for intervention group participants based on facilitator and teacher time input, and materials. Quantities of inputs were based on the protocol, supplemented by study records, and unit costs reported in 2019 Great British pounds. As costings considered the additional costs associated with the GoActive intervention only, there were zero costs associated with control.

Outcomes in the economic analysis were the primary outcome (i.e. change in minutes of MVPA) and quality-adjusted life-years (QALYs). QALYs are useful as they measure overall health-related quality and quantity of life. QALYs were assessed using the Child Health Utility-9D, which has been validated for use in adolescents24 and was included in the participant questionnaire at all follow-up time points. Responses were converted to utilities using the recommended published algorithm96 and QALYs were calculated as the area under the curve, with straight-line interpolation between time periods. Total accrued time from baseline to 10-month follow-up, and hence the time horizon for the study, is approximately 2 academic years. Undiscounted and discounted QALYs (at 3.5%) are calculated and an unadjusted and adjusted imputed analysis reported. The latter analysis adjusted for age, socioeconomic status and other demographic variables. Multiple imputation with chained equations was carried out on missing utility data using the predictive mean matching method and five imputed data sets. Analysis was conducted in R (The R Foundation for Statistical Computing, Vienna, Austria). Code is available from the corresponding author on request.

Exploratory economic modelling

The study protocol stated that long-term modelling would be undertaken only if a positive effect on physical activity was observed, on the basis than it is highly unlikely for the intervention to be (cost-) effective in the long term if it is not so in the short term. However, irrespective of the point estimate, incorporating the uncertainty in study results into a previously developed, longer-term model enables analysis of where uncertainties in longer-term cost-effectiveness lie and can be used to guide further research via value-of-information analysis. For example, the treatment effect of the GoActive intervention may be uncertain, which a priori might suggest that further research was warranted. However, the modelling may reveal that greater value would be obtained from a research project investigating, for example, the link between physical activity and blood pressure, or blood pressure and risk of cardiovascular event.

We therefore conducted an exploratory analysis, inserting the results from the GoActive trial into the previously developed PACE (physical activity cost-effectiveness) model,97 adapted to this study question, reporting longer-term costs and QALYs accrued as per the protocol, but also incorporating a value-of-information analysis. Adapting the model to the study question simply comprised converting the study results from minutes of MVPA to metabolic equivalent of task (MET)-hour equivalents. 98

The PACE model has been described previously,97 but, briefly, it generates a cohort of individuals based on English demographic data. The model follows them for a defined number of years under a base case of average physical activity levels, tracking BMI, blood pressure, glucose (glycated haemoglobin) and lipids, the incidence and prevalence of type 2 diabetes and associated complications, cardiovascular events and certain cancers according to a set of risk equations extracted from the literature. Costs and health state utilities are assigned to these events, resulting in a measure of accrued costs and QALYs over the time horizon of the model. A second-order Monte Carlo simulation propagates parameter uncertainty (i.e. uncertainty in coefficients of the risk equations) to decision uncertainty (i.e. uncertainty in costs and QALYs). The model is then repeated under alternative physical activity scenarios, input as mean and SE of difference in MET-hours per day, with assumptions as to the ‘decay’ rate of the intervention effect (base case assumed 55% per annum), yielding a mean and SE cost and QALY gained from each intervention, from which incremental cost-effectiveness ratios can be calculated.

The model simulated a cohort of 10,000 14-year-olds and followed them for a period of 10 years under a ‘do nothing’/control scenario, repeated 10,000 times. This was then repeated, incorporating the distribution of intervention effect described above.

Costs included in the PACE model are health service costs, to which we added the cost to local authorities of the intervention. The cost perspective of the modelled analysis is therefore the public sector.

Results are reported as mean and 95% credibility intervals for costs and QALYs in each arm, as well as increments between the arms, net monetary benefit from each arm and incremental net monetary benefit calculated at £20,000 per QALY gained. Value-of-information analysis estimates the expected value of perfect information (EVPI) and expected value of perfect parameter information (EVPPI), providing a guide as to where future research is likely to be of most value. Population EVPI and EVPPI are calculated based on population projections of the numbers of 14-year-olds in the UK from 2020 to 2029 (i.e. 10-year time horizon), discounted at 3.5%. This yields a beneficial population of 7,169,651.

Analysis of process evaluation data

Main outcome

Tests for effect modification indicated differences in the effect of the intervention between subgroups, in particular between boys and girls, and between high and medium/low socioeconomic status (Table 17 and Figure 6). The results of the subgroup analyses suggested a negative intervention effect among boys and a positive intervention effect for those with low and medium socioeconomic status. However, the subgroup results are inconclusive, as the CIs included zero (Table 18).

Secondary accelerometry outcomes

In addition to indications of differential effectiveness by demographic characteristics, analyses of the secondary accelerometer-assessed outcomes also showed some suggestion of temporal patterning of effectiveness, with changes on weekdays (both during and after school) generally favouring the control participants, with the reverse on weekend days (see Figures 7 and 8). To examine these distributions of effect across time and subgroups in more detail, we explored potential differential effects of the GoActive intervention on accelerometer-assessed outcomes across the week by subgroups for SEP, weight status, ethnicity, baseline activity level and gender.

Within-trial cost-effectiveness analysis

The cost of delivering the intervention was estimated to be £2520 per school, compared with control schools. The average cost per student was £13.06 (Tables 28 and 29). The mean QALYs accrued was 1.242 (SE 0.005) in the intervention group compared with 1.244 (SE 0.005) in the control group (difference adjusted for baseline data –0.006, 95% CI –0.017 to 0.005) (Table 30). Therefore, the ‘most plausible’ interpretation is that the GoActive intervention is dominated by control, that is it increases costs and leads to lower QALY gains (albeit noting that we did not detect a statistically significant difference in QALYs).

Economic evaluation explanatory modelling results

At 10-month follow-up, the intervention group engaged in 1.91 fewer minutes of MVPA per day (95% CI –5.53 to 1.70 minutes) than the control group (see Table 10). The PACE model requires physical activity expressed as MET-hours. MVPA is assumed as an intensity of six METs. 98 This equates to a difference of –1.91 × 6/60 = –0.191 MET-hours per day, with a SE of 0.92 minutes of MVPA, equating to 0.092 MET-hours per day. The cost of the intervention is assumed to be £13.10 per student, incurred in the first year of the model only. These values were inserted into the model to predict longer-term costs and outcomes, and to conduct the value-of-information analysis.

As expected, the point estimate incremental net benefit is negative (given a willingness to pay of £20,000 per QALY), but with substantial decision uncertainty (95% credibility interval –£252 to £229). Therefore, the GoActive intervention is not cost-effective at a £20,000 per QALY willingness-to-pay threshold. It is certain that the intervention will impose a cost on schools/local authorities, but there is virtually no difference in QALYs accrued over the subsequent 10 years. Overall, uncertainty in cost and QALYs (and hence incremental net benefit) is due to uncertainty in longer-term incidences of cardiovascular events and the associated cost of treatment, rather than uncertainty in the ‘treatment effect’ of the GoActive intervention. This is borne out by the value-of-information analysis, which finds zero value in a further study of the GoActive intervention. The greatest gains from reducing uncertainty are in unit costs of subsequent events (accounting for 19% of the EVPI), the risks of cardiovascular and stroke events (14.3% and 7.9%, respectively) and the future pathways of blood pressure, cholesterol and glucose levels (6.5%). Given the young age of the cohort, the absolute numbers of such events will be extremely small, but they have large impacts on costs and quality of life, therefore driving the majority of the uncertainty in the results. As stated previously, this analysis should be considered highly exploratory in nature, with the emphasis placed on the value-of-information analysis, illustrating where the key uncertainties lie, rather than a robust estimate of the long-term cost-effectiveness of the GoActive intervention (Tables 31 and 32).

Reflections on the GoActive sessions

Participants reflected on a number of key GoActive components, including their enjoyment of the intervention (e.g. fun), competition, choice and novelty. Analysis of participant responses indicated that many of these overlapped and ultimately had an impact on enjoyment and participation.

Reach

The reach, or proportion of participants who attended GoActive sessions, calculated from a self-reported Year 9 student questionnaire, was 39.4%. The reported dose of at least one GoActive class session in the last 2 weeks during the distant support phase ranged from 11.2% to 63.2% between schools, as reported from Year 9 students (see Table 36). This was in contrast to 93.7% of the mentors and 84.2% of the teachers who reported that the sessions had been delivered at least once over the last 2 weeks. GoActive sessions were delivered over a median time frame of 12 weeks.

Interview data indicated that dose had changed in response to competing priorities, which had an impact on resourcing (e.g. school space availability, a lack of time or engagement issues). One contact teacher explained:

When it’s exam season it’s exam season for [Year] 9, 10 and 11, so we had to stop for a certain point . . . we had to stop because of the sports hall and gym were being used and we couldn’t get the kids out during registration because their exams started at 9 o’clock.

Contact teacher, school H

Recruitment

Procedures to attract and maintain participant involvement in the intervention included rewards. On a five-point Likert scale that ranged from ‘do not like it at all’ (1) to ‘like it a lot’ (5), 38.3% of Year 9 students reported liking rewards [mean 3.8 (SD 1.4)]. Additional ‘thank you’ gifts were provided after each measurement session. These included pens, earphones, mints and stationery, which may have facilitated sustained involvement. In turn, 87% of Year 9 students were retained at mid-intervention, 80% were retained post intervention and 76% were retained at 10 months post intervention.

The top three reasons for mentors joining the programme reported from the questionnaire data were (1) for the incentives/prizes (i.e. £20 vouchers and a hoodie), (2) to be more active themselves and (3) because a teacher encouraged them. In the focus group discussions, mentors suggested that their continued involvement was linked to ‘fun’ and enjoyment, or the social aspect of spending time with their peers. Others stated that they believed in the aims of the GoActive intervention for Year 9 students and that their involvement was linked to the perceptions that the GoActive intervention was ‘good for their [Year 9 students’] health’.

Dose delivered (completeness)

The dose delivered (completeness), or the number of GoActive intervention components implemented, differed between and within schools. The complete GoActive programme was not implemented by any school. Table 36 depicts the implementation of the four GoActive essential components3 per school, as reported by Year 9 students, website analytics, and logged by mentors and facilitators.

Intervention fidelity

From the data presented in Table 36, the fidelity, or the extent to which the GoActive intervention was implemented as planned, varied between schools. It should be noted that despite reporting that no similar programmes were running at the school pre intervention, interviews with Year 9 students, mentors and the contact teacher revealed that school D had been running a weekly ‘healthy active form time’ activity session with all year groups in the school.

The GoActive intervention used specific behaviour change techniques to increase adolescents’ daily MVPA (e.g. novelty, choice, flexibility, competition, mentorship and rewards). 3 A description of the application of the tenet in the GoActive programme will be shown to demonstrate GoActive protocol, as opposed to actual implementation.

Overall effects

Our findings are in line with emerging review-level evidence, which has identified the limited effectiveness of research-driven physical activity promotion interventions on whole-day MVPA. 14,15 Of note is that high-quality evaluations of school-based interventions targeting adolescent physical activity are still rare, and this evaluation therefore significantly adds to a limited evidence base. The absence of an intervention effect on time spent in MVPA could be partly because of inadequate implementation. The per-protocol population was small and our process evaluation findings indicate that some intervention components, such as mentorship, were not adequately implemented. However, the per-protocol analysis produced similar results to the main analyses, indicating that if the intervention had been implemented with higher fidelity then it may still not have been effective at a whole-population level. The per-protocol definition focused on website use and reported activity sessions. Use of the website was low and contrasts with the high engagement observed in the pilot trial,5 which indicated preliminary effectiveness. This pattern is common in behavioural interventions, with 75% lower effectiveness seen for behavioural interventions across various health behaviours at the full trial stage compared with feasibility and pilot testing. 26 This is thought to be at least partly due to adaptations needed to implement programmes at scale, which highlights the challenge of scaling-up health promotion interventions. 26 Since its inception, the GoActive intervention has been designed to be scalable by including a website and flexibility for use in multiple school structures. However, implementation difficulties may have arisen from the provision of implementation flexibility for schools – an issue also identified in the National Institute for Health Research-funded Girls Active study75 – as well as a lack of clarity in the conceptualisation of the mentor and teacher roles. Additionally, the delivery agent of the intervention changed between the pilot (research staff) and full trial (local authority-funded health trainer, supported by the research team), which may have contributed to the reduced effectiveness. This points to the challenge for researchers to design interventions that are scalable at the outset, which would minimise the need for major adaptions.

Secondary outcomes suggested a negative impact of the intervention on LPA and sedentary time on weekdays (both in school and out of school), with the opposite effect seen on weekends. Adolescent-focused process evaluation results indicate that, at times, the intervention may have fostered a climate that was not conducive to physical activity within school (e.g. the sessions appeared to have a lack of social cohesion and connection, and activity choice was often dominated by boys). 103 However, this may not have extended to weekends. One of the main aims of the GoActive programme was to use school time to encourage participation in activities with friends and family outside school. On a population level, most of the decline in physical activity during adolescence happens on weekends. 106,107 Therefore, it would be worthwhile teasing out what intervention components may be associated with weekend activity. The negative findings for LPA and sedentary time on weekdays were reversed for weekends. These opposing associations largely cancelled each other out, leading to no effect for daily averages, with the intervention not appearing to increase activity of higher intensity (i.e. MVPA).

The evidence base for school-based physical activity interventions for mental well-being among young people is inconsistent108,109 and often overlooked in place of physical health outcomes. 13,110–112 Although a secondary analysis, the examination of the effect of the GoActive programme on mental well-being is an important contribution to a relatively underexplored area of research. We found no overall intervention effect on mental well-being among participating adolescents. This finding was unexpected, given the utilisation of psychological techniques in the GoActive intervention to engage young people to promote physical activity. Such techniques were hypothesised to protect mental well-being through building peer support, self-efficacy, self-esteem and strengthened friendships, all of which are known enablers for positive mental well-being. 113 It is possible that in focusing on such psychological techniques, in place of physical activity levels itself, as done in this current study, we may have overlooked the physiological role of activity in promoting positive mental well-being. Indeed, this is concordant with the GoActive pilot study findings, which found both improved mental well-being and an overall positive impact of the intervention on physical activity levels of young people. 5

Effects in subpopulations

To the best of our knowledge, this is one of the first adolescent physical activity promotion trials to comprehensively investigate which population subgroups may benefit more or less from the exposure to the intervention, and whether or not this differs across activity intensity and time of week. Although the GoActive intervention was not effective in changing the main outcome measure, average daily MVPA at long-term follow-up, there was evidence that the intervention had an impact on population subgroups differentially. The differences in effects between subgroups were small, but potentially meaningful at a population level. 30 The intervention appeared to have a more negative effect among boys, as well as those reporting high SEP. These findings contrast with results from a recent review, which showed no difference between subgroups for intervention effectiveness when assessing whole-day MVPA. However, this was mainly in primary school-based studies. 20 Across subgroups, our results provide a tentative suggestion of a narrowing of inequalities in physical activity levels, as boys are often reported to have higher activity levels than girls,5 although differences in activity levels by SEP are less clear. 57,58 The unfavourable impact among boys for average daily MVPA contrasts with our insights from the mixed-methods process evaluation paper exploring satisfaction with the dose received. This evaluation reported higher intervention acceptability among boys and found that activity choice appeared to be largely driven by boys. 33 These results indicate that gender differences in intervention delivery may not have manifested as expected regarding intervention effect. These contrasting results reinforce the importance of a thorough process evaluation, including observations of delivery, and highlight the complexity of psychosocial issues surrounding activity promotion.

The GoActive intervention appeared to be more effective among individuals with low SEP, which is in contrast to a recent meta-analysis showing no differential effectiveness by SEP. 20 Exploratory moderator analyses demonstrated that participants from a high SEP background appeared to have been negatively impacted, both during and after school, and the reasons for this are unclear. Despite the fact that evidence regarding socioeconomic differences in activity levels is equivocal,58 individuals with lower SEP may do less-vigorous activity59 and may have less opportunity for a variety of structured activities. 60 This lack of equity contributes to health inequalities throughout the life course,61 and reducing health inequalities in behaviours and health is therefore a public health priority. 62 It is possible that individuals of low SEP may have particularly benefited from the chance to try a variety of activities in the GoActive intervention, as the opportunities may not have been available to them otherwise. There appears to be some utility of comprehensive school physical activity interventions for increasing adolescents’ physical activity behaviour, particularly in disadvantaged neighbourhoods, and such interventions could be particularly relevant among certain population groups. 63

The intervention effect on mental well-being for students in lower socioeconomic schools is an important finding because there are known specific mental health vulnerabilities of young people in disadvantaged settings, such as low family income, multiple insecurities (e.g. employment, safety, food provision) and increased risk of chronic physical conditions. 114,115 It is possible that students within more disadvantaged settings may have been particularly sensitive to benefits from the enhanced social support and mentoring provided in the GoActive intervention, in the context of wider home and environmental circumstances that may pose unique mental health risks.

Moreover, participants with overweight or obesity at baseline particularly benefited from the intervention in increasing their out-of-school activity in the short term. Living with overweight or obesity may be associated with increased barriers to participating in group physical activities. 116 The GoActive intervention may have increased self-efficacy among these potentially marginalised individuals, as adolescents who have tried an activity are more likely to want to do it again. 25 This could at least partly explain why the group with overweight or obesity appeared to increase their activity out of school, as they may have particularly benefited from the provision of opportunities to try new activities.

Implementation

The GoActive intervention was implemented, to some extent, in all of the schools. As the programme was often placed in the hands of one teacher within the school, usually by senior school management, there were issues with workload, lowering the GoActive intervention on the priority list. Additionally, this meant that the programme was not embedded across the whole-school year group as intended and this increased the time pressure on the contact teacher who was leading the GoActive intervention. Teachers highlighted and appreciated the support from the GoActive team; however, most teachers suggested that it would have been useful to have a member of the GoActive staff come in to run the sessions. This directly contradicted suggestions from students and teachers in the development work, which strongly advocated for in-school older mentors to take control of the programme and the request for distance from the researchers. Contact teachers and mentors voiced support for the GoActive programme and what it was trying to achieve; however, implementation of programme components was slow and often occurred late in the evaluation time frame, meaning that the potential for impact on outcomes of interest, including MVPA, was limited.

Enablers of the implementation of the GoActive intervention included school support, embedding a routine, and mentor and tutor support. Challenges that had a negative impact on implementation included school-level constraints, such as having limited space for physical activity, lack of time, uncertainty of the roles subgroups played within the GoActive intervention and sustaining Year 9 student engagement. This is supported by mediation analyses that suggest that, even though no evidence was found to support the use of GoActive intervention components in increasing physical activity, higher perceived mentor and teacher support was associated with improved well-being among boys and girls via various mediators, including self-efficacy, self-esteem and social support. 2

The context of an intervention cannot be overlooked or undervalued. Although the intervention itself was complex in nature, its interaction with its context was also highly complex. For example, once exams were scheduled, one school did not have the space for the physical activity required for the intervention. Another school already had the infrastructure and support from the SLT to embed the GoActive intervention in school practice. The disparity of resources, staffing, equipment and space between schools imposed clear barriers to the implementation of the intervention.

In a change from the feasibility and pilot studies,5,6 in the trial reported here, council-funded facilitators were the official entry point of the intervention within the school and there were several challenges surrounding this role, including facilitator staff changes between training and programme start and also during the programme. In addition, school and facilitator availability did not always align, exacerbating limited school access for some facilitators. Facilitator website logs were not fully completed and, as such, they provided only a partial picture of what happened during the facilitator visits to schools and within the facilitator–mentor meetings. There is potential for bias if only those who felt particularly strongly about the intervention returned their questionnaires or completed their website logs. A number of steps were taken to minimise these issues, including building rapport with the programme staff at meetings and over e-mail. Although facilitators stated that they understood the programme during the training, this was not consistent throughout their post-intervention process evaluation. Perhaps this could be at least partly due to the complexity of implementing the programme in varied school settings and dealing with school preferences for programme implementation. Although appropriate for potential future programme dissemination, having facilitators outside the study team added another level of substantial complexity to implementation. This role was only a small part of a busy health trainer role for the facilitators and therefore it is unlikely that they were as dedicated to the programme as the research team. Incorporating more strategies to increase buy-in from the facilitators of the programme could have been helpful.

GoActive intervention components

The proposed logic model was tested in examining the association between perception of intervention components and physical activity and well-being via a range of proposed psychosocial mediators. 5 The logic model was partly supported, with mentorship being the most consistent component leading to change in well-being, although not physical activity, via a range of psychosocial mediators for both boys and girls. We also identified different patterns of mediation for boys and girls, with teacher support and class sessions identified for boys and competition and rewards identified for girls. Many interventions use components such as mentorship, leadership, class-based activity sessions and online activity tracking to increase physical activity. The potential for these physical activity interventions to additionally target well-being are becoming increasingly salient. 21,24,117 Despite this widespread use, relatively little is known about the mechanisms by which these intervention components may target outcomes via proposed mediators. We provide a reflection of the implementation and effectiveness of the GoActive intervention components below.

Gender differences in physical activity interventions

The whole-school approach of GoActive aimed to avoid the stigmatisation of targeting particular groups;68 however, the results indicate that separate activities for boys and girls may warrant further investigation. The detrimental impact among boys for average daily MVPA contrasts with our insights from the process evaluation, indicating higher intervention acceptability among boys and that activity choice appeared largely driven by boys. 21 These results indicate that gender differences in intervention delivery experience may not have manifested as expected regarding intervention effect.

Aside from mentor support, the results of the mediation analyses differed for boys and girls, with teacher support and class-based activity sessions identified as important for boys, whereas rewards and competition were identified for girls. These gender differences could be at least partly explained by differences in attitudes to physical activity among boys and girls, the complex nature of girls’ relationship with physical activity, and the gendered societal pressures and expectations that can enable or inhibit physical activity. 133,134 We conducted our co-design work with 26 students (18 boys and 8 girls) and it is possible that these students had particular views that did not represent the majority of those in the full trial. 6 Gender differences in our results may perhaps have been exacerbated by the larger amount of boys included in the co-design phase. 6

The results showed that teacher encouragement of physical activity and co-educational class-based sessions could be more appropriate for boys. It is possible that girls may not have felt comfortable co-participating in activities, as PE in the UK is usually segregated by gender and girls may be particularly self-conscious of doing physical activity. 135 Body image is a strong predictor of MVPA in both boys and girls,136 and is also linked to well-being. 137 Concerns around masculine and feminine ideologies of health-related behaviour and body image are very relevant to physical activity promotion in a school setting, but are rarely central to physical activity provision in schools. It follows that considering social identity in physical activity promotion may have particular value in schools, as it has been proposed to have the potential to facilitate the promotion of exercise behaviour and to impact physical activity norms positively, and this has also been identified as facilitating successful mentorship. 121 We proposed that boys’ opinions may hold more weight in the class environment. 103 If this is the case, class-based activity sessions may not have fostered increased social support among girls and could be a potential explanation for why the class-based activity sessions and perceived teacher support were identified as mediators among boys only.

Reflections from theory

The process evaluation results indicate that there was high variability in intervention fidelity for some components, for example the regularity of GoActive sessions (choice and novelty), mentors and in-class peer leaders. These GoActive essential elements were linked to two facets of motivation informed by self-determination theory: (1) extrinsic motivation and (2) intrinsic motivation. 47,138 Despite intervention components aligning with the basic needs for competence, autonomy and relatedness, linked to self-determination theory, modifications or implementation issues potentially remove key underpinning principles of motivation and subsequent behaviour change. Additionally, the considerable variance in the length of GoActive activities may suggest inconsistency.

Despite intervention components aligning with the basic needs for competence, autonomy and relatedness, results suggest that implementation issues and gender differences may have limited, or even reversed, the intended effect. Gender differences were identified, potentially leading to the perpetuation of disparities in perceived competence and autonomy regarding physical activity among girls, which we had aimed to avoid with a whole-population approach. Both autonomy and competence are experiences that are readily affected by conditions in the social environment (i.e. facing non-optimal, overwhelming challenges can lead to feelings of incompetence and disengagement). 122 It appears that elements of the intervention (such as competition and choice) may have undermined girls’ autonomy and perceived competence and led to disengagement. Largely, among boys, choice may have facilitated autonomous motivation and, when carried out well, mentorship and class-based activity appeared to have a positive impact on relatedness. Rewards were generally liked as a strategy but, along with competition, these elements did not adequately support perceived competence and autonomy and risked further marginalising girls and shy/inactive individuals. The results highlight the importance and difficulty of creating activity and needs-supporting environments, and demonstrate how easy it is to inadvertently perpetuate activity-thwarting environments because of issues with intervention implementation.

Mentorship was intended to increase relatedness and social cohesion with older students who were tasked with fostering a sense of connection and positive social climate to facilitate participant interest in physical activity. 54 Both increased self-esteem and social support appeared to mediate the association between higher perceived mentor support and increased well-being, but not physical activity. The ‘quality’ of mentorship may be particularly important when considering the potential impact on psychosocial factors, as mentorship may work successfully only if the recipients are satisfied with the mentorship on offer. 103 Participants seldom reported feelings of connectedness from mentor interaction, and some of the mentors may have perpetuated social environments that were less conducive to physical activity. It is likely that these activity- ‘conducive or thwarting’ environments could influence well-being through the same mediators proposed for physical activity and our results suggest that the associations appear stronger with well-being than physical activity.

Rewards were considered a positive aspect of the intervention and this is consistent with previous findings in behavioural research. 139,140 Using rewards to influence behaviour is a controversial method in health promotion and is contrary to some elements of self-determination theory,47 as it has been suggested that all types of reward may undermine intrinsic motivation. 141 Rewards may be successfully used in behaviour change, but they need to be for behaviour that is desirable, enjoyable and perceived as important. 140 It appears that these criteria were not met in this case, as the intervention did not do enough to encourage students to perceive activity as important. For some students, the GoActive intervention seems to have constituted desirable behaviour and been enjoyable, but implementation may have been a barrier to facilitating this.

Activity choice in our intervention design was intended to target autonomous motivation, but the process evaluation findings indicated that boys dominated class discussion in choosing activities for the class. Although some girls made suggestions for activity selection, the social context of the form group and the dominant voices of the boys seemingly deterred girls from persevering with their choice. This may be due to a number of factors, including social context, comfortability and empowerment of autonomy; however, it is noteworthy that these Year 9 students did not actually express the desire to exert autonomy over the choices provided. Additionally, girls’ novel activity choices led to a lack of participation, which may have deterred others from making suggestions in case it proved an unpopular choice. In turn, rather than supporting feelings of autonomy, the process of choices becoming limited may link to feelings of incompetence and disengagement. Focusing more on empowering Year 9 students to assert autonomy and make a choice may have been beneficial. Additionally, the provision of activity choice from set options on the GoActive website may have further limited autonomy by not encouraging participant input. 142

Although we hypothesised that incorporating novel activities would improve perceived competence, participants often selected familiar activities, which may have been strategic in avoiding the demonstration of a lack of perceived competence. Overall, participants did consider the sessions fun if there was a social element; however, some girls intimated that participating in GoActive sessions kept them from being sociable elsewhere. Although social aspects of the intervention appeared important, qualitative data indicated that peer support, particularly relating to in-class leaders or mentors, was not always well received. This missed opportunity to develop a sense of belonging and connection has important implications on a participant’s intrinsic motivation.

Finally, it has been suggested that for a school-based intervention to work, it needs to include a mechanism from at least one category outlined in the Theory of Expanded, Extended, and Enhanced Opportunities,143 and the GoActive intervention targeted two of these. The ‘expansion’ mechanism suggests providing new occasions to be active by replacing sedentary time with physical activity, such as adding activity to previously sedentary tutor times. Another suggested mechanism implemented in the GoActive intervention is ‘extension’, which suggests lengthening the time currently allocated to activity, such as by encouraging students to be active out of school and in tutor times. 143 Process evaluation data revealed that the GoActive programme was not consistently implemented and therefore may not have led to sufficient expansion or extension of student activity provision.

Reflections on context

Insight gained from the current process evaluation has prompted greater reflection regarding the implementation of complex interventions and the consideration of using the school environment as a context for physical activity interventions. Our findings have a number of implications for the development and evaluation of public health improvement interventions for use in educational settings. Given the limited success of school-based physical activity promotion to date,13 we call for a step change in our approach to intervention design and implementation.

The results presented here raise the question of the appropriateness and value of standardised intervention protocols across a multisite approach. The context of an intervention cannot be overlooked or undervalued. Although the GoActive intervention itself was complex in nature, its interaction with its context was also highly complex. Our analysis highlights the importance of gaining both breadth and depth in understanding of the context of individual schools. The disparity of resources, staffing, equipment and space between schools posed clear challenges for the implementation of the intervention. Each school is different, with competing priorities, as well as resource and time constraints. Future research should consider how best to utilise schools as an intervention setting. Although schools might provide significant reach to adolescents, there are multiple other factors that create challenges, for example teacher/student rapport or school culture around physical activity. A well-resourced intervention can create time and opportunities in the current school system, while minimising the burden of delivery for schools.

Co-production is seen as an inclusive method of intervention development, which may be thought to take into consideration contextual concerns. 144 The GoActive study demonstrates that, if we are to continue to embrace schools as an intervention setting, we need to do more than co-produce interventions, that is we need to understand each school’s culture, particularly as contexts change (e.g. moving to multiacademy trust or an academy chain). We should endeavour to understand how a school’s culture can be a part of intervention design or a complex intervention, for example attitudes towards physical activity. An approach with greater emphasis on school empowerment through the selection of an appropriate and relevant programme to implement in their context, or implementing a protocol followed by input from schools to tailor the intervention, could work best. 145 However, this more practice-based intervention may require new modes of evaluation other than RCTs.

Reflections on the participatory co-design approach

Our results highlight several impasses between suggestions made by students in the co-design phase and the perception of components when implemented. In our process evaluation, students stated that they would have preferred the intervention integrated into the school timetable;103 however, this contrasts directly with the suggestion in the co-design phase to have older students run the programme with distance from teachers and researchers. 6 Although participants indicated a desire to try non-standard activities in the development phase, when implemented, students were reluctant to choose and participate in unfamiliar activities, contrasting the requests for novelty that were central to the participatory input in the earlier phase. 103

Some components developed in the co-design phase were well liked, such as mentorship. Mentorship was strongly suggested in the co-design phase and when mentorship was carried out well it was highly acceptable. Although, in theory, intervention components such as mentorship and leadership align with student and teacher requests, in reality the implementation of these components may not be easily incorporated in the school context. The logistical challenges of mentor recruitment and scheduling training became apparent during the full trial, but this was not perceived to be a challenge by students or teachers during participatory design work.

The diverse range of opinions and preferences across individuals makes it challenging to incorporate multiple student ideas into programmes that can be implemented widely at scale. A different group of students, from different schools, participated in the co-design work to those that were participants in the RCT, and this could offer one explanation for why some components in the co-design process did not lead to change in physical activity and/or well-being. Conducting a participant-led design with the same students who receive the intervention may overcome some of this incongruence, but that appears problematic when aiming to implement a consistent programme at scale and evaluate it in a RCT. Whole-systems approaches with varied, detailed and documented stakeholder input are increasingly being promoted,146 but pragmatic applications of RCTs may be required to assess such approaches.

Future studies may consider a detailed co-design phase that focuses on the practicalities of implementation, including the organisation of training and facilitator retention. It is also important to conduct honest and rigorous process evaluations (including observations) to better establish the underlying barriers to and facilitators of the use of individual intervention components in a school setting.

Publications arising from this grant

Brown HE, Whittle F, Jong ST, Croxson C, Sharp SJ, Wilkinson P, et al. A cluster randomised controlled trial to evaluate the effectiveness and cost-effectiveness of the GoActive intervention to increase physical activity among adolescents aged 13–14 years. BMJ Open 2017;7:e014419.

Jong ST, Brown HE, Croxson CHD, Wilkinson P, Corder KL, van Sluijs EMF. GoActive: a protocol for the mixed methods process evaluation of a school-based physical activity promotion programme for 13–14 year old adolescents. Trials 2018;19:282.

Corder K, Sharp S, Foubister F, Brown HE, Jong ST, Wells EK, et al. Effectiveness of the GoActive intervention to increase physical activity in adolescents aged 13–14 years: a cluster randomised controlled trial. Lancet 2019;394.

Corder K, Sharp SJ, Jong ST, Foubister C, Brown HE, Wells EK, Armitage SM, et al. Effectiveness and cost-effectiveness of the GoActive intervention to increase physical activity among UK adolescents: a cluster randomised controlled trial. PLOS Med 2020;17:e1003210.

Corder K, Werneck AO, Jong ST, Hoare E, Brown HE, Foubister C, et al. Pathways to increasing adolescent physical activity and wellbeing: a mediation analysis of intervention components designed using a participatory approach. Int J Environ Res Public Health 2020;17:E390.

Jong ST, Croxson CHD, Guell C, Lawlor ER, Foubister C, Brown HE, et al. Adolescents’ perspectives on a school-based physical activity intervention: a mixed method study. J Sport Health Sci 2020;9:28–40.

Related publications

Corder K, Schiff A, Kesten JM, van Sluijs EM. Development of a universal approach to increase physical activity among adolescents: the GoActive intervention. BMJ Open 2015;5:e008610.

Corder K, Brown HE, Schiff A, van Sluijs EM. Feasibility study and pilot cluster-randomised controlled trial of the GoActive intervention aiming to promote physical activity among adolescents: outcomes and lessons learnt. BMJ Open 2016;6:e012335.