Peer-led physical activity intervention for girls aged 13 to 14 years: PLAN-A cluster RCT

Ethics

This trial was granted ethics approval by the School for Policy Studies Ethics and Research Committee at the University of Bristol (Bristol, UK; reference SPSREC17–18.C22) on 24 May 2018. The trial was funded by NIHR Public Health Research Programme (PHR) (project number 17/50/01). The intervention costs were funded by lottery funding from Sport England (London, UK). The University of Bristol acted as the trial sponsor (sponsorship number 2898). The occurrence of any adverse events during the project was recorded and reported to the chairperson of the School for Policy Studies Ethics and Research Committee and the chairperson of the Trial Steering Committee (TSC).

Sampling and participants

State-funded secondary schools located in south-west England with girls in Year 9 were invited to take part. Special education and independent schools were excluded. Priority was given to schools whose local pupil premium indicator was above the median (i.e. those that were more deprived) by inviting them to participate first. Schools were required to sign a trial agreement form to take part. All Year 9 girls were eligible to participate; therefore, the trial aimed to recruit all Year 9 girls in each of the 20 participating schools. Research staff members gave a standardised recruitment briefing to Year 9 girls in attendance on that day in each school to explain the purpose of the trial, the intervention, randomisation and the data collection process. The girls were given the opportunity to ask questions and were informed that they could withdraw their participation. Each girl was given a written information pack for herself and her parents, which also contained a pupil consent and parent opt-out form. Participation in the trial was contingent on providing a signed pupil consent form. Parents could opt their daughter out of the trial by providing a completed opt-out form.

Sample size

The trial was designed to achieve 90% power for the primary outcome of weekday MVPA at follow-up and was informed by the results of the feasibility trial conducted in a slightly younger cohort of girls. 51 Based on those results, the sample size calculation assumed a fixed cluster (i.e. school) size of 70 girls, an intracluster correlation coefficient of weekday MVPA of 0.01, standard deviation (SD) of 20 minutes and coefficient of variation in cluster size of 0.22. The sample size calculation allowed for 30% of girls not providing primary outcome data and the alpha was set to 5%. Informed by the feasibility trial,52 which observed a between-group difference in weekday MVPA of 6.1 minutes, we estimated that we would need 20 schools (equating to 1400 students) to detect a difference of 6 minutes in weekday MVPA with 90% power under the assumptions outlined above. Given the inherent uncertainty in many of these assumptions, two reserve schools were recruited to mitigate school withdrawal prior to completion of baseline measures.

Data collection

The PLAN-A feasibility trial52 found that including a measurement point at the end of the 10-week intervention period increased participant burden and did not allow sufficient time for behavioural changes to take root, thus showing nothing. Therefore, to maximise the possibility of observing an intervention effect and simultaneously reducing participant burden, data were collected from Year 9 girls in all schools at just two time points: baseline [time 0 (T0); October 2018 to February 2019] and at 5–6 months post intervention, approximately 12 months post baseline [time 1 (T1)]:

• school level –

  • T0:

    • county/region

    • total number of pupils

    • total number of Year 9 pupils

    • total number of Year 9 girls

    • proportion of pupils taking free school meals

  • T1:

    • school-context measures

• participant level –

  • T0:

    • accelerometer data for 7 days

    • psychosocial questionnaire (on tablet device)

    • KIDSCREEN-10 score (on tablet device)

    • EuroQol 5-Dimensions, Youth version (EQ-5D-Y), score (on tablet device)

    • demographics and family affluence (on tablet device)

    • home postcode (on paper)

  • T1:

    • accelerometer data for 7 days

    • psychosocial questionnaire (on tablet device)

    • KIDSCREEN-10 score (on tablet device)

    • EQ-5D-Y score (on tablet device)

    • date of birth (on paper).

All questionnaire measures at T0 and T1 were collected using a Samsung^® Galaxy Tab A 10.1 (Samsung Electronics Limited, Surrey, UK) device and Research Electronic Data Capture (REDCap^®) software (Vanderbilt University, Nashville, TN, USA). 62 All demographic data were self-reported and collected using the Samsung device or reported to research staff at T0 (i.e. home postcode) or T1 (i.e. date of birth). Process evaluation measures were completed during the project at each intervention school (i.e. at observations and evaluations of training between March 2019 and June 2019) and post intervention (i.e. during qualitative work between June 2019 and July 2019). At both time points, participants who completed data collection were given a £10 shopping voucher in appreciation of their contributions to the trial. Following completion of the T1 data collection, all participating schools received £500 for taking part in PLAN-A and were provided with a summary report of the findings.

Measures

The specific measures used for variable generation are detailed in Table 2.

Randomisation

The school was the unit of randomisation. Following the completion of T0 data collection, schools were randomised to the control (n = 10) or intervention (n = 10) arm, stratified by county (Avon, Devon or Wiltshire) and the England Index of Multiple Deprivation (IMD) score for the lower-layer super output area in which the school was located (dichotomised as less than the median of sampled schools in the county versus greater than or equal to that median) to ensure balance in each stratum. Schools were allocated, using a computer-generated algorithm, to either the control or intervention arm by a member of the Bristol Randomised Trials Collaboration (BRTC) who was blinded to the school identity and worked separately to the fieldwork team. Also blinded to the allocation were the lead statistician and all other team members, other than the trial manager, research associate and fieldworkers.

Intervention

The intervention design was adapted from the ASSIST model37 and comprised four components: peer nomination, train the trainers, peer supporter training and a 10-week diffusion period. The intervention components and theoretical enactment that were developed and tested in the feasibility trial have been described in detail elsewhere. 52

Control group provision

The eligible participants of the control schools participated in both data collection and peer nomination at T0. A total of 10 schools were randomly assigned to the control condition. These schools continued with normal practice [normal physical education (PE) provision and physical activity opportunities during teaching and co-curricular activities], before completing T1 data collection.

Baseline characteristics

The baseline characteristics of schools and pupils were compared between arms by reporting relevant summary statistics to identify any potentially influential imbalance. The baseline characteristics were summarised using means and SDs, medians and interquartile ranges (IQRs), or numbers and percentages, depending on the nature of the data and its respective distribution. Where the baseline characteristics of the groups differed by > 10% or half a SD, the effect of this variable on the outcome was investigated in sensitivity analyses.

Analysis of intervention effect

Two-tailed tests were used, with effect estimates, 95% CIs and p-values presented. No adjustment for multiple testing took place. Analyses using regression models adjusted for stratification variables, as well as the baseline values of the outcome studied. The primary approach for analysis was on an intention-to-treat (ITT) basis, defined as analysing participants as randomised.

The primary outcome was the accelerometer-determined minutes of MVPA on weekdays collected at T1. This was described in each group using means and SDs. Comparisons between arms used a multivariable, mixed-effects, linear regression model to allow for clustering within schools, adjusting for baseline MVPA scores (i.e. random school effects to account for clustering), and the number of valid days of accelerometer data and randomisation variables.

Secondary outcomes consisted of mean weekend and weekday minutes of sedentary time, mean weekend minutes of MVPA and self-esteem. These were continuous measures and were analysed using the same modelling approach as that used for the primary outcome.

Sensitivity analyses

Exploratory sensitivity analyses that were part of the pre-agreed statistical analysis plan,74 which was approved by the TSC prior to any analyses being conducted, were presented alongside the primary analysis to assess the sensitivity of the primary analysis to the following conditions:

• Imbalance between groups. Where imbalance between the arms was evident at baseline, the primary analysis was repeated, adjusting for the variables showing an imbalance.

• Missing outcome data. Patterns of missing MVPA data were explored and missing values were imputed by using the MI command in Stata to perform multiple imputation by chained equations (MICE). Analyses using imputed data were compared with the primary analysis.

• Month of data measurement (i.e. seasonal bias). The primary analysis was repeated, adjusting for the month of the year in which the MVPA measurements were taken.

Subgroup analyses

Pre-agreed subgroup analyses were performed to estimate whether or not the intervention was differentially effective in subgroups of SEP, nominated peer supporters (peer supporters vs. non-peer supporters) and the proportion of sedentary time at baseline. This involved including interaction terms between the arms and moderator (i.e. SEP, peer supporter training or proportion of sedentary time at baseline) in the primary analysis models and using the likelihood ratio test for hypothesis testing.

As the trial was not powered to detect effectiveness in subgroups, these analyses were treated as exploratory and interpreted with caution.

Mediation analysis

Mediation analysis was conducted to explore whether or not any effect of the intervention was mediated by self-determined physical activity autonomous motivation,68 autonomy,46 competence69 or relatedness. 46 The MEDEFF command in Stata was used to estimate the causal mechanisms. Mediators were treated as continuously measured variables and described using the mean scores, stratified by arm.

Compliance and missing data

The following criteria have been used in previous studies. 75 Accelerometer compliance was calculated at each time point using the number of valid days of accelerometer data provided by each girl enrolled in the trial at that time. For participants to be considered compliant, valid accelerometer data for at least two weekdays were required. At each time point, the number and percentage of girls enrolled in the trial who provided questionnaire data were used to calculate questionnaire compliance. Rates of provision of accelerometer data (missing, invalid and valid) and questionnaire data (missing and not missing) were recorded for T0 and T1. Data were classified as missing when the valid wear time condition was not achieved or if an accelerometer was not returned. (See Table 12 for details of data provision at each time point.)

Measures

The process evaluation measures with pupils, trainers and teachers were conducted during or straight after the intervention period in each intervention school (March–July 2019).

Interviews and focus groups

At the end of the 10-week diffusion period, members of the research team conducted focus groups and interviews with all informant groups (i.e. peer supporters, non-peer supporters, peer supporter trainers and school contacts). Interviews with public-sector stakeholders (e.g. public health commissioners) were conducted later, once the trial results were known.

Participants were invited to take part in a focus group or interview through information letters. All informants were required to provide informed consent to take part in the qualitative process measures. Focus groups and interviews were recorded using an encrypted digital recorder and audio recordings were transcribed verbatim. Transcribed data were anonymised before the analysis.

Separate interview guides were developed for each informant group. All interview and focus group topic guides were developed by Byron Tibbitts (PLAN-A trial manager, male) and Kathryn Willis (research associate, female), in consultation with Russell Jago (PLAN-A principal investigator) and Simon J Sebire (PLAN-A co-investigator). Interviews were conducted in private with one of three research team members [BT (male), KW (female) or TR (male)], who were all trained in qualitative data collection and analysis. Focus groups were conducted by two members of the research team in tandem, with one leading and the other supporting. Field notes were made by the supporting researcher. A prior rapport had been struck between the researchers conducting the interviews and the trainers and respondents in intervention schools through exposure to one another during intervention delivery and data collection phases.

At the start of each interview or focus group, the researcher (BT, KW or TR) outlined the purpose of the interview/group and disclosed their role in the project. The first interviews and focus groups served to pilot the topic guides, which were subsequently discussed and refined by the research team.

School context

Various quantitative and qualitative data were collected to establish school context and the potential effects on the implementation or efficacy of the intervention. Details of school size and pupil premium were provided by the school contact or school reception to establish pupil demographics. In addition, PLAN-A researchers made narrative records of school ‘social environment’ after data collection visits to reflect the experience of engaging with the school, the key contact(s), the pupils and any other factors that may have had an impact on implementation.

To explore any differences between schools in the physical activity environment, a quantitative school context audit (see Report Supplementary Material 6), which addressed similar constructs to those in Jones et al. ’s76 validated tool and was developed further in the PLAN-A feasibility trial52 and another school project,78 enabled members of the trial team to collect information on school-level physical activity provisions. Items in the adapted audit tool assessed the presence (i.e. count) and quality (graded on a five-point quality scale) of different physical activity facilities (e.g. courts, pitches, hard-surface playgrounds).

A second part of the school context measure asked school contacts to provide details on school policies regarding physical activity and the integration of physical activity throughout the curriculum. 77 To further explore school-level differences in physical activity provision that may affect the primary outcome of the trial, we asked all 20 school contacts to report termly on any change in physical activity provision or other lifestyle interventions in the school that were available to the girls in the trial.

Qualitative analysis

Parts of this section have been reproduced with permission from Jago et al. 75 © Queen’s Printer and Controller of HMSO 2019. This work was produced by Jago et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

To synthesise the wide range of qualitative data collected, the framework method79 was used, as it produces a matrix of the data (from different participant groups) that allows comparison. The framework method also allows a combined approach to the analysis, enabling themes to emerge inductively from participant accounts and for specific issues to be explored deductively. Data were analysed using the following steps:

1. Researchers from the trial team thoroughly read and re-read each transcript, and listened back to audio recordings to become familiar with the data set. Initial impressions of the data were recorded to inform code development.

2. Initial codes were identified inductively by each researcher; these were refined as subsequent transcripts were analysed. For the deductive analyses, broad predefined codes for each informant group (see Appendix 1) were applied, with the primary purpose of categorising relevant information, which was to be further interrogated to elicit more refined codes and interpretations. Specifically, for questions relating to the central tenets of SDT,40 predefined codes were used to examine whether or not the intervention supported autonomous motivation among the participants and the three needs-support constructs:

   1. autonomy: feelings of choice

   2. relatedness: a feeling of connection between participants and with the instructor

   3. competence.

3. An analytical framework that could be applied to all transcripts was agreed and developed to fit each informant group to avoid losing any key data using the following process:

   1. Three researchers (KW, TR and BT) independently read and analysed the same two transcripts from each informant group, and then met, discussed and created draft frameworks for each.

   2. A fourth researcher (SS) read the transcripts from each informant group alongside the draft frameworks and provided feedback.

   3. Kathryn Willis, Byron Tibbitts and Tom Reid met to discuss and refine draft frameworks.

4. Three researchers (KW, BT, TR) then applied each framework to the remaining transcripts (with some level of double coding) using NVivo (QSR International, Warrington, UK), version 11. New, emerging codes were discussed between researchers and each framework was adjusted accordingly.

5. A framework matrix was then used to map all coded data by informant group, summarising the data by category and including descriptive quotations. Summarising techniques were compared to ensure consistency within the research team.

6. Frequent meetings were held between the researchers to review the matrix-enabled data analysis and theme generation. During this process, illustrative quotations to demonstrate the nature of each theme were agreed.

7. The level of convergence was evaluated by triangulating the frameworks and comparing the set of codes for each informant group.

To optimise the credibility and transparency of the research, the qualitative analysis was reported in accordance with the COREQ checklist. 59

Overview and aims

The aim of the economic evaluation was to determine the cost-effectiveness of PLAN-A.

The primary questions most relevant to the economic analysis were:

• What is the estimated resource use and cost associated with the peer supporter training and PLAN-A intervention?

• What is the mean cost of the PLAN-A intervention per school and per Year 9 girl?

• What is the cost-effectiveness, in terms of (a) cost per unit increase in weekday MVPA and (b) cost per pupil achieving a 5-minute increase (which is considered a clinically important difference)80 in weekday MVPA, of the PLAN-A intervention in adolescent girls?

• What is the effect of the PLAN-A intervention on health-related quality of life (HRQoL), measured using the EQ-5D-Y and KIDSCREEN-10?

• What is the effect of the PLAN-A intervention in terms of the mapped Child Health Utility Instrument, nine dimensions (CHU-9D), utility scores (as derived from the KIDSCREEN-10)?

• What is the cost–utility, in terms of cost per quality-adjusted life-year (QALY), of the PLAN-A intervention? (This was an exploratory analysis using mapped CHU-9D scores.)

Furthermore, if there is evidence of promise that PLAN-A improves MVPA at T1, what is the potential longer-term cost-effectiveness of the intervention? (For our analysis, we consider that there is evidence of promise if the point estimate of weekday MVPA is positive and the 95% CI includes the possibility of a meaningful, i.e. 5 minutes of weekday MVPA, positive intervention effect.) The within-trial economic analysis was performed using individual participant-level data on physical activity and quality of life, and school-level data on intervention costs from the PLAN-A trial. The within-trial analysis reports results 5–6 months after the end of the PLAN-A intervention and 12 months after baseline (T0). Therefore, the economic time horizon corresponds to 1 year.

Methods for costing intervention resource

To estimate a mean cost per school and per Year 9 girl, the set-up and delivery of the PLAN-A intervention in intervention schools was costed. The resources used for other physical activity initiatives, used as part of the normal curriculum, in intervention and control schools were not measured. This assumed that the PLAN-A intervention was additional to normal practice (i.e. ‘physical activity education as usual’) and that schools did not adjust their physical activity initiatives to compensate for being allocated or not being allocated to the PLAN-A intervention. The costs were collected from a public-sector perspective, using the cost year 2019, in Great British pounds (£).

The main components of resource use were school staff, trainer (including lead trainer) and pupil time, travel and materials, venue hire and administration. In the trial, some intervention activities were supported by members of the research team and the trainers were employed by the research trial. To approximate the ‘real-world’ costs for if the intervention were scaled up to all local schools, we assumed that all intervention-related tasks would be conducted by trainers and that the trainers would be employed by local authorities. In brief, resources were logged using staff time logs [which documented the role of the school staff member/trainer, the activity, the purpose and the time spent on the activity (recorded in minutes)], expense claims (i.e. mileage/travel claims) and invoices (i.e. venue, refreshment and material costs). Where possible, resource use records were completed at the time of set-up/delivery events (i.e. the train-the-trainers, peer nomination and peer supporter events). Table 4 provides an overview of the resource captured as part of intervention cost. The intervention costs exclude the cost of intervention development during the feasibility trial. We assumed that this was a ‘sunk cost’ that would not be relevant to other schools and local authorities that were considering providing training and delivering the PLAN-A intervention in the future, if PLAN-A were to become standard practice.

Salary band information was used to calculate school staff costs using the national teaching union main pay scale81 (Table 5). The unit cost per hour for school staff considered was calculated based on 195 contractual days for a teacher (in line with the National Association of Schoolmasters Union of Women Teachers, the teaching union for the UK)81 and ‘in-school’ hours. This includes salary on-costs (i.e. pension and national insurance). Because the logs also identified who in the school was involved with delivering the PLAN-A intervention (e.g. qualified teacher, teaching administrator or teacher with leadership responsibilities), we estimated the national average rate of pay for that role to determine the cost. The unit cost for trainers also included salary on-costs and was based on the published local government pay scale82 and assumed working days/hours for local government employees. Staff costs were calculated by multiplying the units of staff time (i.e. hours) by the appropriate unit cost for each staff member.

Outcomes

Health-related utility score measurement in adolescents is difficult because of a lack of validated measures. Although the EuroQol 5-Dimensions (EQ-5D) is a widely used measure for calculating QALYs, there is no UK value set validated for use in children/adolescents. The youth version of the EQ-5D (the EQ-5D-Y)83,84 categorises health on five domains, but has no accompanying value set for scoring. The EQ-5D-Y also includes a visual analogue scale (VAS) that asks respondents to rate their health on a thermometer-like scale from 0 (worst) to 100 (best) health. Another instrument, the CHU-9D, has a value set for calculating utility,85 but was originally developed in primary school-aged children. This has, more recently, been validated in older children and adolescents (i.e. 11- to 17-year-olds),86 but was not collected in this trial.

General HRQoL was also collected using the KIDSCREEN-10 index. 88 Where item responses to trait parameters were complete, KIDSCREEN-10 scale raw scores were based on the sum of item scores (ranging from one to five), where a high value corresponds to a higher HRQoL. Scale raw scores can be transformed into T-values, with scale means around 50 and SDs around 10, again, with higher values corresponding to a higher HRQoL. If there is item-level missingness, the KIDSCREEN-10 can still be scored based on the remaining items. We used a published mapping algorithm,89 validated in adolescents, to indirectly estimate CHU-9D scores and QALYs from the KIDSCREEN-10 responses.

Scope of the within-trial economic evaluation

The PLAN-A intervention has the potential to affect extracurricular physical activity events that girls and families pay for (e.g. sports clubs, gyms). Given our public-sector perspective, we did not cost these activities. Indeed, recall bias between T1 and T0 would have hampered measurement. However, focus groups conducted with girls in intervention schools probed for cost implications of the intervention. If the PLAN-A intervention is effective in increasing physical activity, it may also have an impact on participants’ health-care use. However, we elected not to collect information on short-term health-care use in the trial as we wished to minimise participant burden and because of the high likelihood of recall bias at T1. If PLAN-A results in a sustained increase in MVPA, the long-term impact on health-care use, via the reduced incidence of long-term conditions that are associated with a lack of physical activity, is likely to far outweigh any short-term impact on health-care use.

Cost-effectiveness methods

To assess the cost-effectiveness of the PLAN-A intervention compared with the control group, individual pupil MVPA in intervention and control schools and the school-level incremental costs in intervention schools were calculated. The primary economic outcome for the within-trial economic evaluation was cost per pupil for a unit change in weekday MVPA at T1. This was expressed in terms of cost per additional minute of MVPA, which allows comparison with other, similar studies. 90,91 In addition, we estimated the incremental cost per 5-minute increase in MVPA at T1.

Cost and effectiveness data were combined to calculate an incremental cost-effectiveness ratio (ICER). A non-parametric bootstrapping approach was used to determine the level of sampling uncertainty surrounding the mean ICER by generating 10,000 estimates of incremental costs and benefits, and presenting results on cost-effectiveness acceptability curves (CEACs) and a cost-effectiveness plane (CEP). 92

Our primary analysis was a complete-case analysis. The patterns and potential mechanisms of missingness were examined. We planned to refer to the result relating to the multiple imputation carried out on the primary outcome measure (i.e. weekday MVPA) and to consider secondary analysis using these imputed effect data if imputation was likely to change the decision rule (see Chapter 2, Sensitivity analyses).

A secondary economic analysis took a cost–consequences approach. 93 This presented resource use, disaggregated costs and all trial outcomes in a tabular format, with outcomes and costs kept separate so that a decision-maker can draw their own conclusions on the breadth of evidence.

The exploratory analysis used mapped CHU-9D utilities in a cost–utility analysis (cost per QALY).

School recruitment and randomisation

Of the 33 schools that were approached to take part, 22 provided a signed school agreement (20 plus two reserve schools). Two schools withdrew before baseline measures and the reserve schools were utilised, ensuring that 20 schools were included. The two schools that withdrew cited staff capacity as the reason for doing so. Schools were grouped into three regions for the purposes of randomisation: Avon (including the local authorities of Bristol, Bath and North East Somerset, North Somerset and South Gloucestershire), Wiltshire and Devon. Table 6 presents the baseline deprivation characteristics of the schools randomised to each trial arm.

Schools in the intervention arm had higher IMD scores than schools in the control arm, with a median and IQR of 3 (1–4) and 4.5 (3–5), respectively.

Participant recruitment

Of the 1848 girls who received the recruitment briefing across all 20 schools, 1558 (84.31%) consented to take part. The mean number of eligible girls in each school was 92.04 (SD 24.23) and the mean number of girls who consented was 77.90 (SD 22.84). Recruitment percentages ranged from 46 to 97, but were comparable between trial arms. Table 7 presents participation rates by school and trial arm. Two schools (one in each arm) had particularly poor consent rates owing to school-specific issues (see Chapter 5, School and participant recruitment). In total, 290 girls did not provide consent to take part and were excluded from trial measures at T0. Of these, 230 opted out (intervention schools, n = 108; control schools, n = 122) and 60 did not participate for other reasons, such as absence, long-term illness, non-attendance or having left the school (intervention arm, n = 38; control arm, n = 22). Table 8 presents reasons for non-consent by school and measurement time point.

Peer supporter recruitment

In the intervention arm, 191 girls were invited to be peer supporters and, of those, 166 (86.91%) consented to take part in the study. Peer supporter recruitment rates are presented by school (see Table 7). Recruitment rates of Year 9 girls in intervention schools ranged from 13.41% to 23.21%. Two schools failed to meet the 15% recruitment target: one school because of a miscalculation in the peer nomination process that reduced the number of girls who were invited to be peer supporters (school 15) and the other because of unavoidable absences (school 14). Reasons for non-consent were provided by 17 of the 25 non-consenters and included not wanting to miss important lessons (n = 10) or known absence on the scheduled training dates (n = 4). Only three nominated girls specified that they did not want to become a peer supporter: one did not feel confident enough to fulfil the role, one felt that she was not active herself and the other gave no reason.

Trainer recruitment

Trainers were recruited via local authority networks to utilise a local delivery model. In total, 18 trainers expressed an interest and nine were invited to receive training after availability and suitability checks were applied. Of those nine, two were based in Wiltshire, five were based in Avon and two were based in Devon. The two Devon-based trainers withdrew prior to the training, citing changes in circumstances as a reason, leaving seven participating trainers. Participating trainers were all female, with ages ranging from 23 to 45 years and a range of experience in youth work, drama, yoga, physical activity or a combination of these: five had experience working with young people in a youth work or physical activity context, and the other two had experience of delivering drama or yoga workshops with young people and had both previously worked on PLAN-A as trainers in the feasibility trial. All had or obtained enhanced disclosure and barring services (DBS) checks to work with children for the trial.

Additional subgroup analyses

Two additional subgroup analyses (i.e. pupils meeting CMO guidelines and pupils’ mode of transport from school to home) were performed. There was no evidence of a difference in treatment effect between those who did and those who did not meet CMO guidelines (p-value = 0.138), or between those who did and those who did not use active travel as a mode of transport from school (p-value = 0.658).

Repeating the primary analysis using counts per minute as the outcome

After adjustment, at T1, the counts per minute (CPM) (weekday, weekend and average across all valid days) were lower in the intervention arm than in the control arm. Mimicking the primary outcome analysis, there was some evidence of a difference in mean weekday CPM, favouring the control arm (difference in means –22.23, 95% CI –44.43 to –0.05; p = 0.05) (see Appendix 2).

School-level change in moderate to vigorous physical activity from time 0 to time 1

Figure 3 presents a school-by-school comparison of change in mean weekday MVPA from T0 to T1. This figure shows that the mean weekday MVPA for all schools decreased between the time points, with the only exception being school 6 (mean change 2.52 minutes, SD 18.13 minutes). The largest decline was observed in school 1 (mean change –11.81 minutes, SD 19.87 minutes) and school 2 (mean change –12.27 minutes, SD 17.53 minutes).

FIGURE 3.

School-by-school change in the primary outcome between T0 and T1. Navy indicates the intervention; aqua indicates the control. Error bars depict 95% CIs.

Peer supporters versus non-peer supporters in intervention schools only

Comparison of weekday MVPA data between peer supporters and non-peer supporters (Appendix 3), showed that there is very weak to no evidence that there is a difference in weekday MVPA minutes at T1 when comparing trained peer supporters with non-peer supporters (adjusted difference –2.25, 95% CI –5.27 to 0.22; p = 0.071). At baseline, peer supporters had a higher mean level of weekday MVPA than non-peer supporters, but both groups had similar levels at T1. Both groups showed a decrease in their MVPA between the time points, with peer supporters having a greater reduction in MVPA than non-peer supporters. The adjusted difference in T1 MVPA between peer supporters and non-peer supporters also showed weak to no evidence of an intervention effect.

Train-the-trainer costs

The train-the-trainers course was a 3-day event held in one location for all of the trainers. Two trainers (the lead trainer and one member of the research team) delivered the course to a group of six trainers, with a resultant event cost of £319.09 per school and £3.53 per pupil (where a pupil is defined as a girl on the register at the start of Year 9; Table 24).

Intervention consumables

Details on intervention consumables associated with the production of the PLAN-A resource kits were captured from invoices and entered on the PLAN-A trial budget database (Table 25). Intervention consumables (print and related resources) equate to £250.76 per school or £2.77 per pupil.

Local authority staff time

Staff time associated with the co-ordination and delivery of peer nomination and peer supporter training was valued. In the real world, it is assumed that these activities would be carried out by a local authority-employed delivery agent. In total, the local authority staff time (circa 58 hours, on average) and resource to deliver the intervention equated to £1651 per school and £18.26 per pupil (Table 26).

School staff time

Within schools, school staff resource varied by the grade of the staff member involved (i.e. leadership, PE or administration/support teacher). In addition, the PLAN-A school contact time varied. At the extremes, school 10 recorded very little (240 minutes) school staff contact with the intervention, whereas school 16 recorded 1275 minutes. Overall, school staff time equated to a mean of £595.88 per school and £6.59 per pupil (Table 27). Although PLAN-A was delivered in usual school hours and these costs do not represent additional budgetary costs for the school, it is still the case that the teacher’s input in other PE or school activities (on average, around 18 hours) will have been displaced by involvement in PLAN-A activities.

Total costs of intervention delivery

Table 28 presents the total cost of intervention delivery for each of the 10 intervention schools. The variable school staff costs (per hour) affected overall costs per school. There was variation in per-pupil costs between the schools, with costs ranging from £20.85 to £48.86. As the intervention involved high fixed costs, the cost per pupil was associated with the size of the school.

Table 29 shows how the mean intervention cost per school (see Table 28) is broken down. The bulk of the costs came from the peer supporter training days and the staff time used to support the intervention.

EQ-5D-Y VAS, KIDSCREEN-10 and CHU-9D scores

Overall, the percentage of girls on the register who completed the EQ-5D-Y VAS, KIDSCREEN-10 and CHU-9D instruments was 82%, 84% and 75%, respectively, at T0, and 74%, 75% and 74%, respectively, at T1. Of these, 72%, 75% and 73%, respectively, had no missing data at either T0 or T1.

Quality-of-life scores (measured using the EQ-5D-Y VAS, KIDSCREEN-10 raw scores and T-scores, and the mapped CHU-9D) deteriorated slightly from T0 to T1 in both the intervention and control arm (Table 30). Appendix 4 presents the mapped data for those with complete CHU-9D questionnaires by time point and trial arm.

Inspection of the profile responses of the EQ-5D-Y for each dimension of the instrument showed a similar pattern in both arms (Figures 4–8). A large percentage of girls (around 45%) at both time points and in both the intervention and control arms reported some or a lot of problems with pain or discomfort. The majority of girls (87–88%) in both groups reported that they were a bit or very worried, sad or unhappy; this proportion increased slightly at T1. It was unclear whether this was an age-related phenomenon or indicative of factors not measured in this trial, but consistency between arms suggests it was not related to the intervention. The intervention and control groups reported mean baseline CHU-9D index scores of 0.83; these scores declined by T1 (see Table 30).

FIGURE 4.

Profile of the population: percentage reporting problems with mobility.

FIGURE 5.

Profile of the population: percentage reporting problems with looking after themselves.

FIGURE 6.

Profile of the population: percentage reporting problems with doing usual activities.

FIGURE 7.

Profile of the population: percentage reporting problems with pain or discomfort.

FIGURE 8.

Profile of the population: percentage reporting problems with feeling worried, sad or unhappy.

Cost per unit change in moderate to vigorous physical activity and cost per 5-minute increase in moderate to vigorous physical activity

The bootstrapped estimate of mean incremental costs was £30.50 and the mean incremental benefit was –3.54 minutes of MVPA. The mean estimate of the ICER, calculated as the ratio of the bootstrapped cost-effect replicates, was –£37.24, with the negative ICER, in this instance, indicating that the intervention is dominated (i.e. more costly and less beneficial than the control). The majority of the bootstrapped replications were in the north-west quadrant of the CEP (Figure 9), indicating that the intervention is probably more costly and less effective at improving weekday MVPA than the control. The scatterplot crosses the y-axis, indicating that there is, however, a high degree of uncertainty surrounding the effect on weekday MVPA. The CEAC (Figure 10) shows that the probability that the intervention was cost-effective is low over a range of willingness-to-pay thresholds for a unit change in MVPA. In secondary analysis, estimating the cost per 5-minute increase in weekday MVPA, the estimate of the ICER was –£186.22.

FIGURE 9.

Mean cost and weekday MVPA and associated uncertainty (1000 bootstrap replications).

FIGURE 10.

The CEAC over a range of hypothetical willingness-to-pay thresholds for MVPA.

Cost–consequences analysis

As has already been reported (see Table 17), pupils in the intervention and control arms achieved, on average, 45 and 49 minutes of MVPA per weekday, respectively. After adjustment, the difference between arms was estimated as –2.84 (95% CI –5.94 to 0.25; p = 0.071) minutes, a trend suggesting that those in the intervention arm performed fewer minutes of weekday MVPA than those in the control group, on average.

There was no evidence of a statistically significant difference in incremental effects between arms for (the mapped) CHU-9D or QALYs (Table 31). In unadjusted analyses, there was some evidence that the KIDSCREEN-10 and EQ-5D-Y VAS scores were better in the intervention group (see Table 31). However, differences were small and consistent with no effect of the intervention in adjusted analyses.

Cost per quality-adjusted life-year

In the exploratory cost–utility analysis using the mapped CHU-9D scores, the point estimate of the probabilistic ICER was £35,748 per QALY, which is above the £20,000-per-QALY threshold conventionally applied in the UK. The bootstrapped replications of the cost–effect pairs in Figure 11 show a similar pattern to the primary economic analysis; however, the replications are split more evenly between the north-west (i.e. more costly, less effective) and north-east (i.e. more costly, more effective) quadrants of the CEP. Figure 12 identifies that at conventional willingness-to-pay threshold values of £20,000 and £30,000 per QALY, the intervention has a 48% and 51% chance of being cost-effective, respectively. At lower thresholds, the intervention is much less likely to be cost-effective.

FIGURE 11.

Mean cost and QALY and associated uncertainty (1000 bootstrap replications).

FIGURE 12.

The CEAC over a range of hypothetical willingness-to-pay thresholds.

School and participant recruitment

There were no stated barriers to taking part in the project for girls or school contacts. However, in two schools, the consent rate was low (46.27% and 52.78%) compared with the average consent rate of 88.58% in the other 18 schools. Very few non-consenting girls provided a reason for not taking part and so there are few pupil data to understand this issue. However, in one school with a low consent rate, the school contact described a pupil mind-set within the school that deterred girls from putting themselves forward for opportunities because of a lack of confidence and fear of the unknown.

Train the trainers

The trainers perceived their training for the intervention to be comprehensive, enjoyable and successful in preparing and motivating them to deliver the content to the peer supporters in line with the theoretical principles of SDT. The format, delivery style and bonding time with the other trainers were praised, as well as the opportunity to check understanding and ask questions about delivery style:

The trainers [who delivered the train the trainers] were really good. I think it was really effective, us pretending to be the girls, because you got to go through every activity, how it would be run, and then you can, kind of, go through it in your head, things that might be changed, or things you think that they would pick up quicker or need more time on. I think it was good to go through the theory side of it, to understand that, and I think enough time was spent on that.

Trainer D

The importance of being trained in broader teaching techniques was highlighted, including reading the audience, relating the activities to the peer supporter role and emphasising key objectives:

. . . you guys were saying it over and over again, ‘this is the one thing I want you to summarise. Make sure that everything you’re doing is in the context of peer supporting’ . . . But I had your voices in my head saying ‘always bring it back to the peer supporting’ . . .

Trainer C

Using the peer supporter training resources at the train-the-trainers event was an important learning aid. The resources and delivery aids were rated ‘excellent’ for all training days (see Appendix 5).

Spending 1 full day on the material for each of the peer supporter training days was praised by trainers as it gave them the time to focus on the content for each day. Two trainers would have liked the top-up day content to be covered closer to the actual delivery of the top-up days to maintain trainer confidence:

I think that it was the right amount of time. Also good for it to mirror the amount of time that we would have with the girls. So we had quite a real-life experience of what they would be doing, so that’s good.

Trainer A

Peer nomination

Peer supporter training

Fidelity to self-determination theory

Self-determination theory39 was woven throughout the intervention design: in the delivery style coached to the trainers, the content of the peer supporter training and the recommended peer-supporting techniques. 52 A crucial component of the theoretical underpinning was the promotion and support of the three inherent psychological needs: autonomy, relatedness and competence. The subsequent section utilises qualitative focus group data and questionnaire data to assess fidelity to the theoretical principles behind the design, and reports the perceived challenges to delivery of the intervention in line with SDT.

Views on trainers

Most of the peer supporters liked the trainers, and both the peer supporters and school contacts thought that they were nice, friendly, helpful (see Table 33), approachable, non-judgemental and ‘found a way to make us have a laugh’ (PS 14):

I think they were really, like, nice and if you ever had anything to talk about and they’d be open to listen, and that was nice.

PS 1

They were just really relaxed and had that relaxed atmosphere and it was something they had to do, but they approached it really well and it was never forced.

PS 4

I thought they were great . . . they built up a really nice rapport with all of the students that were there which, again, for our students, is quite important.

SC 5

The fact that the trainers were relatable and from different backgrounds to those the girls were used to (e.g. teachers and sports coaches) helped them to engage with the peer supporters:

I think that relationship [peer supporter and trainer] worked really well because, obviously, they’re going to be more engaged if it’s people that they can relate to . . . So that, obviously, had the girls interested from the beginning.

SC 1

Only a small number of peer supporters from three different schools made negative remarks about the trainers. These involved perceptions about the body language, temperament and interpersonal connection with a trainer:

She was not really approachable.

Kind of moody.

PS 17

Receipt of content

Peer supporters felt that the training was active, relaxed and included a wide variety of activities, making it more enjoyable and engaging than peer supporters had expected:

Yes, it wasn’t classroom work where you sit there and you have to write it. It was more, everyone was moving and active.

PS 10

Most peer supporters found the 2-day training interesting (mean score ± SD 3.15 ± 0.83; maximum score 4) across all schools. However, peer supporters from schools 5, 7, 10 and 15 perceived the training to be less interesting than peer supporters at other schools, with mean scores of < 3 (see Table 33). The trainers’ ratings indicated agreement with the peer supporters that the peer supporters at schools 7 and 10 may have lacked interest in the training (see Appendix 7); however, the trainers felt that peer supporters at school 17 had very little interest in both the 2-day (mean score ± SD 0.00 ± 0.00) and top-up day (mean score ± SD 1.00 ± 0.00) training (maximum score: 3), which contradicted the peer supporter’s rating of their interest (mean score ± SD 3.67 ± 0.49).

Overall, the training was perceived to be successful in increasing the peer supporters’ knowledge and understanding of physical activity and preparing the peer supporters to carry out their role. Role play using realistic scenarios and practising peer support were particularly valuable:

I thought we were very prepared, well, we were prepared as we could be for what was going to happen. We done loads of different scenarios.

PS 14

. . . we’ve learnt how to bring it up and how to apply it to people we may know.

PS 10

As in the feasibility trial,52 peer supporters and trainers perceived both the ‘peer supporter booklet’ and post box as useful to have in the training days, but they were not widely used.

Evidence of peer support: fidelity to diffusion of innovations theory

Evidence of peer-supporting activities was obtained using a self-report questionnaire (see Appendix 9) and discussed in focus groups. The most commonly cited types of support offered (‘often’ or ‘very often’) were being active with friends (62.12%), being an active role model (48.48%) and inviting friends to be active with them (43.94%):

I was sort of saying that ‘I really want to do more sports clubs, do you want to join with me?’. So making it not about them, like they’re doing something wrong, so it was about me, ‘will you come with me?’. It’s a technique we learnt at the training.

PS 14

Some peer supporters reported using facts in conversations (others felt that this approach might bring an awkwardness to conversations; see Chapter 5, Challenges to being a peer supporter), which shocked some non-peer supporters, as well as encouraging their friends to be active by making physical activity fun:

Finding a fun way to do it, so, I find peer supporting – well, I don’t really call it that. I’ll make it into a game, so if we’re going to do running in PE, we’ll go on a mad one, like running around the field and we’re actually being active, but you don’t realise it. It’s a way of disguising it.

PS 16

I think people were quite shocked when they found out you need like an hour a day. They were like ‘that’s so much’ and I was like ‘I know, you need to get running,’ and things like that which are shocking facts that nobody really knows.

PS 14

Peer supporters preferred to adopt a subtle approach to their encouragements and felt that peer supporting was quite natural and aligned with their identity:

. . . being quite subtle about it [peer supporting] . . . So not letting them know there’s a reason why you’re telling them all of this.

PS 7

Sometimes it’s like a natural thing you do, to help them to do PE or try and encourage them to do it. It’s like part of who you already are, to try and encourage people as it is.

PS 15

When asked about the types of support provided and the number of friends they had supported, being active with friends was again the most prevalent type of support (40.15% reported providing this type of support to four or more friends); however, the strategy with the second largest reach was talking to friends about the benefits of being active (28.03% reported providing this type of support to four or more friends). The least popular types of support offered by peer supporters were those that involved encouraging friends to adopt more or new activities.

Peer supporters encouraged participation in sports, active travel, walking, running, cycling, after-school clubs and going to the gym. At least one non-peer supporter in the focus groups in each of the seven schools indicated that the peer supporters had been encouraging or supportive in these activities (see Chapter 5, Peer supporter autonomy support) or by using the strategies mentioned above:

. . . me and my friend were going to get the bus to go out, but then I convinced her to walk with me . . .

PS 5

My friend, she talks about being active more now than before she went [to the training]. She wanted me to run with her to work so she could get more active . . .

NPS 15

Peer supporting was made easier when peer supporters were in one-on-one situations with close friends and during PE. Several peer supporters expressed the importance of trust and identifying when conversations needed to be confidential:

I only really stick to my close friends, so I just basically say what I know they can take.

PS 15

I feel like if someone came to you about something, you would want that to be kept to yourself because they come to you about something, you don’t want to then tell other people. That person purposely came to you because they trust you.

PS 4

Other peer supporters felt that working together to encourage their peers to be active was a form of support for each other, boosting their confidence. This was perceived to hold more persuasive power when convincing peers to be active:

. . . when you’re in a friendship group and you’re bringing something up, if the peer supporter is there they can help that plan go forward if the person . . . they can go along with it, I guess, and help you push it . . .

PS 14

If there are more people who have that opinion, like, ‘it’s good to run,’ then it’ll be like ‘oh, it is actually good to run,’ so there’s more . . .

People take it in more.

PS 5

Although we have clear evidence of peer supporters’ attempts at encouraging physical activity among their peers, conflicting accounts from various stakeholders also suggest that peer supporter activity was inconsistent. Almost every school had peer supporters who reported that they had performed no peer supporting, with the majority of peer supporters in school 17 stating that they had done ‘not a lot’ (PS 17) to get messages to friends about being active. This is a direct contradiction of the results from the peer support fidelity questionnaire (see Appendix 9):

I haven’t done any [peer supporting].

PS 1

The non-peer supporters across all intervention schools gave the overarching impression that most peer supporters had not undertaken any work to carry out their role, or at least that they had not witnessed ‘active’ peer supporting by their friends. Some suggested that the peer supporters forgot about the project after the training. The non-peer supporters felt that the peer supporters should have done more to support their friends:

If anything, you wouldn’t notice that they’re trying to be supportive with you because, like, if they’ve been trained to be supportive with sport and anything else, they just genuinely have not been supportive at all. I know for a fact that one of my mates, like, [PS name], she just hasn’t been supportive of anything.

NPS 1

. . . I just don’t think they put any effort into trying to get people who aren’t as active or aren’t as sporty, to try and encourage them to do anything, which I thought was going to be part of it . . . I haven’t seen them making a real effort to try and encourage a broader spectrum of pupils.

NPS 14

On the other hand, it could be argued that the application of a more subtle approach to peer supporting, in line with the training that the peer supporters had received, would preclude the non-peer supporters from noticing some peer supporter activity:

They wouldn’t worry because they don’t realise, which is quite good, because it’s just like having a conversation.

PS 16

Many trainers and school contacts questioned whether or not some peer supporters had done anything to support their peers to be active. Trainers believed that, despite peer supporters providing realistic examples of peer support during the top-up day, some examples may have given the impression that they had provided peer support more than was in fact the case. Most school contacts suspected that this could be true, but believed that the majority of peer supporters would have tried to embrace their role using the strategies from the PLAN-A training:

And then it comes to the top-up day and it was like ‘How have you got on?’. Hadn’t done a thing. Hadn’t done a thing.

Trainer B

I think some of them had [peer supported]. I think some of them probably just said stuff because they knew they had to. But some stuff, like, they couldn’t have just thought that up on the spot.

Trainer D

However, it was also suggested that many peer supporters were supporting their peers naturally, without realising they were doing so; it was only through discussion with the trainers at the top-up day that peer supporters realised how they had been peer supporting:

I didn’t realise I was doing it, but I was being a peer supporter, but I wasn’t consciously thinking ‘Oh, I need to do it and then go and write it in my book’ . . . I didn’t make a conscious decision to go and peer support, I was just doing it more naturally.

PS 2

Others had said ‘we haven’t done anything. We haven’t done anything and we are desperately guilty about it’. Then, actually, when you tweaked a little bit more, they had actually done a few things, they just hadn’t considered that to be peer supporting.

Trainer E

Aside from supporting peers, some peer supporters provided support to their family members more than their friends at school, potentially because they felt more comfortable doing so. It was unclear if, by peer supporting their family, they felt that their role was being executed and, therefore, did not attempt to encourage their peers at school:

I tried to, I think, once with my sister, and I ended up making her walk around [place name] with me. But she enjoyed it, which I guess is the only time it really worked. That’s because I think it’s family.

PS 2

Some of them talked, specifically, about things they’d done with family. There were a couple that had talked quite specifically about that, and maybe because they were the examples that they could remember, or they might have been the times when they felt comfortable to try it out a bit more.

Trainer A

Effect on peer supporters

Effect on other girls in Year 9

Many peer supporters felt that they were successful at encouraging and supporting their friends to be active and could give examples of how they did this (see Chapter 5, Evidence of peer support: fidelity to diffusion of innovations theory).

Some peer supporters felt that many non-peer supporters were more open to being active and enjoyed physical activity more. Two school contacts agreed that Year 9 girls were now more engaged and supportive of each other in PE:

I think we’ve definitely seen an increase in engagement in lessons. I think they’ve [all Year 9 girls] all been a lot more supportive . . . they’ve all been so supportive of each other and cheering each other on. Some of them have run extra laps with the slow ones, to keep them going, which would never have happened before we did this at all.

SC 10

Non-peer supporters at school 7 felt that girls in their year may have become more confident in PE and in front of boys since the intervention began:

The boys, they would just play normally, but the girls, I think they would just stand back because they were too embarrassed or something if they fall over or anything like that.

And, since PLAN-A, has that changed at all?

I’d say for just half of the girls, yes.

Non-peer supporters at another school said that there is now less pressure to be good at PE. A small number of non-peer supporters and school contacts from different schools reported uptake of new activities since the intervention began, including walking, running, going to the gym and being outside together, as well as increased sign-up to school clubs. Nevertheless, they were unsure whether this was due to PLAN-A or seasonal variation. Conversely, many non-peer supporters stated that they had not started new physical activity since the intervention began:

[When asked how girls’ confidence has changed in PE after PLAN-A] it’s just that we don’t feel that we have to do well in PE. That might be a teacher or it might be just in general, the people. We don’t feel like we’re pressured to be amazing sports people.

NPS 10

After you did PLAN-A, a lot more girls will go to the gym or try it and a lot of people liked it, so they kept on going back.

NPS 4

I know there are a few Year 9 girls that now want to join the cricket after school as well, but then I don’t know whether that’s just because it’s warm weather now and they’re a lot happier to be outside than inside.

SC 1

Some peer supporters were unable to or did not carry out their role and were therefore unsure whether or not they had had any impact on their peers. Some peer supporters who did peer support did not think they had any impact and did not notice any changes in physical activity. Any impact they may have had was described as not significant or long-lasting. Some attributed this to the fact that non-peer supporters were unaware of the project and what the peer supporters were trying to do:

Not really [been successful in encouraging others to be active], just because it’s quite hard to just go up to someone and start the conversation with them if you’re not sure why they have that barrier.

PS 17

I’ve had some success, but it’s not been an overwhelming amount where people have suddenly gone up the amount of activity they’ve done. It’s normally a one-time thing that happens. And then it doesn’t happen again.

PS 2

At most schools, peer supporters going to the training days initially sparked interest among non-peer supporters, but this diminished with time. Most non-peer supporters did not know why certain girls went away for the training unless the peer supporters told them, which they found hard to explain or understand:

I think some people do know where we went, because when they go ‘Where’d you go?’ some of us went ‘Well, we did this, and these are the facts,’ but I think they just forgot about it over time.

PS 5

Some school contacts corroborated non-peer supporter claims that taking part in data collection, specifically wearing the accelerometers, made them more aware of and thus increased their physical activity:

I think when we used the belts, definitely. I think when the belts came in, the girls were actively thinking ‘Actually, I’m being tracked, and I’m not doing very much’. I think that would’ve helped participation. I really liked that side of it.

SC 7

The consensus from non-peer supporters was that the intervention had no social effects, stating that there were no differences in friendship groups or how they interacted. However, comments from school contacts challenged this view, as some felt that conversations between different girls had increased:

We didn’t really have a change in our friendship group.

NPS 14

I think a social one in terms of them having the conversations. The freedom between the PLAN-A students and the non-PLAN-A students to have those conversations.

SC 16

Overall, all stakeholder groups were uncertain about whether the intervention had resulted in any social or attitude changes in other girls in Year 9, although this would be difficult to detect or measure robustly in any case. None had observed pervasive changes in attitude towards, confidence in or talking about physical activity since taking part in PLAN-A:

Do you think there has been an impact on the other girls in Year 9 who weren’t peer supporters?

I honestly don’t know. I don’t think I could say either way. Without talking to them, I don’t think I could say outright ‘Yes, it has,’ or ‘No, it hasn’t,’ if I’m honest.

School attitude, curriculum and policies

The project aligned with key aims within the PE departments of most schools around improving participation and leadership and was therefore considered a priority within PE in those schools. In several schools, PLAN-A aligned with the school ethos, as healthy lifestyles and physical activity were embedded in their school values; however, only one school (school 14) saw supporting the project as a school-level priority:

In terms of the PE department, our aim is to make sure we’re giving the skills and the confidence to all of our students to, hopefully, instil life-long involvement in physical activity and healthy lifestyles.

SC 16

It was ideal and it brought in everything that we want to push forward to the girls, with the added bonus that, actually, if it did work, which you guys decide, we get the added bonus of increased fitness and health of our girls as well, so it was perfect.

SC 14

In one school, the project was not seen as a priority because lessons and exams were of higher importance:

It wasn’t really very high in our priorities as a department or a school. Yes, the day-to-day running of lessons and exam results and things like that were a higher priority.

SC 17

In school 10, the leadership’s attitude to the trial as a whole was described as ‘poor’ (SC 10) by the school contact, which they contextualised by explaining that the senior leadership team (SLT) had not properly read the trial information and did not fully understand what was required of the school or the benefits to the participants:

Maybe, the head teacher and SLT, I don’t think they fully understand the benefits and what you’re trying to get out of it.

SC 10

Just because they lost the paperwork. I don’t even know who actually signed all the paperwork and who, actually, was the first person. Because our staff turnover last year was really high. I reckon at least three other people who took it forward to SLT have probably left.

SC 10

Appendix 10 shows that school policy scores (with higher scores indicating that the policy is more supportive of physical activity) were largely similar between intervention (mean score ± SD 7.50 ± 1.51) and control schools (mean score ± SD 7.40 ± 2.62; maximum score 12). However, three schools (control, n = 2; intervention, n = 1) received very low policy scores (i.e. ≤ 6).

Appendix 11 shows the level of physical activity throughout the wider curriculum of each school. The mean score ± SD for the control schools (3.80 ± 1.62) was marginally higher than that in the intervention schools (3.20 ± 1.62; maximum score 8); however, the difference was small and there was high school-to-school variability, so any difference between arms would not have been consistent across all schools in each arm.

Physical activity provision

School environment audit data (see Appendix 12) that examined the facility provision at each school to support active travel, sports and general play (see Chapter 2, School context) were highly variable between schools, but did not show meaningful differences between the intervention (mean score ± SD 81.90 ± 12.61) and control group (mean score ± SD 84.55 ± 15.00; maximum score = 164). Comparing baseline school-level physical activity data with physical activity provision scores showed that a low score for physical activity provision (e.g. schools 12 and 17) appeared to be crudely associated with low levels of physical activity (Figure 15).

FIGURE 15.

School-by-school comparison of facility provision scores over baseline MVPA minutes.

No schools took part in any other physical activity research projects during this trial, and just three (control, n = 2; intervention, n = 1) had any involvement with local initiatives that might have influenced physical activity behaviours (see Appendix 13). Termly changes to PE provision or extracurricular opportunities between T0 and T1 were balanced between trial arms; however, four control schools described a change in the focus of PE for girls, moving away from competitive sport and towards well-being and health. School contacts in those schools described a gender-specific consideration not just for what girls enjoyed, but also for what would motivate them to continue to be active beyond school. No intervention schools reported a similar change:

We also look at modern influences such as the interest girls have in exercise to music and dance. So, ‘Fitness Marshall’- and ‘Body Pump’-type exercises have been incorporated into the health and fitness sessions of our PE curriculum. For one of their 2-hour PE lessons, there is a ‘mass fitness activity’ that has been influenced by modern fitness trends, such as HIIT [high-intensity interval training] or exercise to music, with the aim of inspiring them to exercise once they leave school.

SC 3

Intervention weaknesses

When asked about potential factors limiting the effectiveness of school-based, individual-level interventions to increase physical activity in young people, respondents described a range of interconnected factors. These stemmed from the school as the setting, because the priorities of individual schools do not necessarily allow space for an agenda that facilitates increases in physical activity or a culture that promotes it throughout the school:

I think, obviously, the pressures upon schools in terms of standards, progress and achievements, obviously, has a significant factor as well, when it comes to making decisions on what schools prioritise and what resource they put behind that.

External stakeholder (ES) 8

Additional issues were highlighted, including failure to consult the end user and engage parents in design stages, and a lack of wider support for change:

We’ve made the mistake for years of putting on activities for young people, as opposed to understanding what the needs of the young people are and delivering what they want.

ES 2

There was an overall view that many individual-level interventions were reliant on isolated champions, were implemented in environments in which physical activity was not a priority and failed to consider the scaffolding that is necessary for sustained change within and outside of school.

Whole-school approach considerations

The consensus among stakeholders was that adoption of a whole-school approach to promoting physical activity is contingent on sympathetic school leadership, with leaders needing to be convinced of the value of the approach for their school and students. Of primary concern to school leadership is the criteria that the Office for Standards in Education, Children’s Services and Skills (Ofsted) in the UK use to evaluate schools, so several stakeholders recommended aligning school assessment criteria with well-being outcomes to increase adoption of whole-school approaches to physical activity:

. . . if you’re going to have a whole-school approach, it has to relate to those wider measures for schools to really buy into it, certainly at senior level.

ES 8

Stakeholders reported that any whole-school approach must work around school-specific challenges at the point of implementation, including, but not limited to, resource issues, staff capacity and confidence to adapt lessons, and pressure to deliver against existing Ofsted criteria. Consultation in design is therefore vital to creating a flexible and sustainable offer for schools that meets the needs of their community and is deliverable by their staff:

. . . making sure we can be as flexible as possible for schools, to enable them to take an approach that best suits their circumstances.

ES 8

Culture change within schools was discussed at length and framed as both a crucial component in any whole-school approach and the main objective of the approach itself. All respondents agreed that there must be a united approach within the school to create culture change, but that the support and cooperation of the wider community around each school is also crucial to creating an environment in which behaviour change could be maintained:

Because if your environment and your structures and your attitudes aren’t changing, I’m not sure you get any real meaningful change either.

ES 9

Research used and needed

The interviews highlighted that the use of research differed by organisation, and some stakeholders expressed a lack of confidence to source and apply high-quality relevant research to underpin intervention design or evaluation. Providing robust research that was easily accessible and linked physical activity to measurable educational outcomes was highlighted as a key need:

I need to present data that shows that engagement has produced that educational outcome that has improved their chances of educational success or social mobility . . .

ES 10

Respondents reported that case studies are a preferred research model for schools to see if and, crucially, how programmes work in practice. This is an important finding as the majority of research focuses on summary data and does not follow the case-study approach. There is an appetite for a united approach between research institutions and public-sector organisations to co-produce useful evidence to drive the physical activity agenda forward:

I think there has got to be a more coordinated approach to influence the government as well. I think we need to be looking at potential policy recommendations and initiatives . . .

ES 6

Impact of COVID-19 and post-COVID considerations

At an unprecedented level, schools have had to change how they operate in the face of the COVID-19 pandemic. Respondents felt that these changes had reduced the opportunities to be active at school and added barriers, thereby exacerbating existing health inequalities and increasing the complexity of implementing school-based physical activity interventions of any kind. It was also felt, however, that the importance of physical activity for well-being had been highlighted during this time and that this should be harnessed to encourage uptake from schools by reframing the benefits of physical activity to aid post-COVID recovery and linking it with school attainment objectives:

I suppose, the opportunity, you can argue, is to reframe the role of physical activity as a real positive in terms of what we’re trying to do now within schools.

ES 4

Post COVID, virtual physical activity offers could be used as a strategy for schools and organisations to engage a broader range of pupils and address provision inequalities:

We’re always going to run a virtual element because we’re attracting children that wouldn’t do that in a team.

ES 2

Comparison with the PLAN-A feasibility trial

There were several differences between the process evaluation of the feasibility trial55 and the definitive trial of the PLAN-A intervention. We think that it is useful to present the data that might inform our understanding of these differences, as well as the perspectives of one researcher (KW) who observed intervention delivery and two trainers who delivered the intervention in both trials.

The two trials produced very similar results in terms of dose, fidelity and quality of intervention delivery, suggesting that the intervention is scalable. However, certain factors in the definitive trial may have led to differences in how the intervention was received compared with how it was received in the feasibility trial. 52 The principal difference between the two trials was the age of the participants: girls were in Year 8 (i.e. aged 12–13 years) in the feasibility trial and in Year 9 (i.e. aged 13–14 years) in the definitive trial. The transition to the General Data Protection Regulations (GDPR)97 between the two trials meant that the consenting procedures and age of participants had to change to be GDPR compliant. This presented a conflict between the theoretical underpinning of the intervention (i.e. DOI, effectively, that the trial is conducted within a naturally occurring social network, e.g. a school year group) and the GDPR rules on use of opt-out consent. 97 To use an informed, opt-out consenting procedure, girls had to be aged 13 years and, therefore, in Year 9. Trainers and researchers alike felt that this had a significant impact on delivery factors and receipt of the intervention.

In the definitive trial, 9.58% fewer girls consented to be a peer supporter than in the feasibility trial52 (86.91% vs. 96.49%, respectively). The main reason for this was not wanting to miss important lessons. This may be a product of Year 9 participants (and their parents) placing greater emphasis on academic focus. Being slightly older, the peer supporters in the full trial were able to understand activity explanations and grasp the key purpose of the training more quickly than the younger peer supporters in the feasibility trial. 52 The trainers who had delivered the intervention in both trials suggested that this may have led to the peer supporters perceiving the training as repetitive and lacking sufficient content. By contrast, during the pilot phase of the feasibility trial,52 refinements were made based on participant feedback to ensure key themes and activities were reiterated. A perception that activities were repetitive may have contributed to reduced engagement and interest during the training in the definitive trial, which was found to be an issue in some schools.

Being 1 year older also seemed to impact the girls’ attitude to the project. Year 9 peer supporters appeared to be slightly more indifferent during the training than those in Year 8, as if they did not want to be seen as too ‘keen’ or to care too much about the training for fear of appearing ‘uncool’. Despite being interested in what they were learning, they appeared to lack the motivation to overcome the barriers to peer supporting, that is when their peers were stubborn or resistant; the peer supporters in Year 8 seemed more invested, and more willing to follow the trainer’s instruction and carry out their peer supporter role.

Year group sizes were larger in the schools in the definitive trial than in the feasibility trial52 (mean 92.05 vs. 75.16, respectively), which could be a result of delivering across more schools (20 vs. 6, respectively) and/or Year 9 being an intake year for many secondary schools. This resulted in larger peer supporter groups in the definitive trial than in the feasibility trial. 52 This was a key challenge to delivery for the trainers in the full trial, mainly through challenges to behaviour management that undermined the quality of delivery. It was also difficult to find community training venues that were large enough, which presented challenges with regard to space constraints.

The training being held on site in one school in the feasibility trial presented challenges of frequent disruption from other pupils and peer supporters being able to leave for other school commitments. 52 In the two schools that delivered on site in the full trial, these challenges were similar, yet more pronounced, with the addition of more disruptive behaviour from the peer supporters. This could also be related to the older age group, as they may have been more confident and familiar in their school setting to behave in such a manner.

Health economics

The direct cost of physical inactivity to global health systems in 2013 was conservatively estimated to be $53.8B (international dollars),102 with an additional $13.7B in productivity losses. 102 Effective prevention during adolescence is urgently needed to reduce health system costs, improve health and prevent premature mortality.

The mean cost of the PLAN-A intervention per school was similar to that estimated in the feasibility trial52 (£2817 vs. £2685, respectively) and in other school-based physical activity interventions (GoActive,91 The Bristol Girls Dance Project103 and Girls Active104) targeted at a similar age group. As the PLAN-A intervention was targeted at girls in Year 9 only, the cost per pupil was higher than that in the Girls Active104 or GoActive91 interventions, which were targeted at all secondary school girls and all Year 9 pupils, respectively. Nevertheless, the mean cost per pupil (£31) of the PLAN-A intervention was considerably lower than that of the Physical Activity 4 Everyone intervention, which involved more intensive classroom-based physical activity and implementation strategies ($394 per pupil in 2014 Australian dollars). 105 Providing peer supporter training to encourage physical activity in girls has the potential to be affordable for schools and cost-effective for society, if it results in a sustained improvement in MVPA.

School-based physical activity interventions might demonstrate cost-effectiveness in the short term if the initial intervention costs are justified by improvements in HRQoL or are offset by reduced use of health services. Observational and experimental studies have demonstrated the association between physical activity and higher levels of psychological well-being, and the potential for physical activity interventions to improve adolescents’ mental health. 106 Our data highlight the scope for improving the mental health of girls in this age group, with more than half of girls reporting feeling ‘a bit’ or ‘very’ worried, sad or unhappy at the baseline assessment. However, this potential has not been borne out in economic evaluations91,103,104,107 of school-based physical activity interventions that have used preference-based measures of HRQoL, such as the CHU-9D or EQ-5D. In large part, this is because of the lack of a meaningful effect of these interventions on physical activity. There are also concerns that these outcome measures might not be sensitive enough to detect the small improvements in mental health that might be expected from school-based physical activity interventions. 108

As with most other evaluations of school-based interventions, we did not assess whether or not the intervention costs were offset by reduced use of health services during the trial. In the Girls Active trial, which elicited this information from girls at 7 and 14 months post baseline, analysis was hampered by high levels of missing data and concerns about the accuracy of recall. 104 Given this, the use of routine data from electronic health records might be the most efficient and unbiased way to capture this information in future work.

Most economic evaluations of school-based physical activity interventions that have evaluated short-term cost-effectiveness in terms of cost per QALYs have concluded, as we do, that there is insufficient evidence that the intervention costs are justified by initial improvements in HRQoL. One exception to this is the Daily Mile evaluation in primary school (i.e. aged 4–11 years) children, which found that there was a high probability that the intervention was cost-effective in girls, but not boys. 109

A second pathway for school-based physical activity interventions to be cost-effective is through the long-term prevention of activity-related health conditions, such as cardiovascular disease, type 2 diabetes and some types of cancer. This would require that any short-term changes in physical activity and sedentary behaviour are sustained for long enough after the intervention period to have an impact on disease incidence later in life. For many years, work in this area has been restricted by the lack of high-quality models that would allow the intermediate outcomes measured in trials, such as MVPA or body mass index in adolescents, to be extrapolated over the life course to predict disease incidence and associated health-care costs, morbidity and mortality. More recently, decision analysis models have been developed to simulate the lifetime costs and outcomes of changes in body mass index110 and MVPA111 in adolescence. These models demonstrate the potential for interventions with a similar cost to PLAN-A to be cost-effective even if the impact on MVPA is modest (i.e. approximately a 5-minute improvement) and the effect is relatively short-lived (i.e. tapering to no effect by 7 years). 111 This is a moot point for PLAN-A, in which we did not observe any short-term effect on MVPA. However, it does demonstrate the importance of conducting long-term follow-up of participants in trials of interventions that do find evidence of short-term improvements in physical activity.

Intervention delivery versus intervention model

Neither our measurable indicators of intervention implementation success (i.e. pupil engagement and enjoyment, fidelity to intervention design and theories) nor our measures of school environment (i.e. physical activity facility provision and policy audit) converged with or could explain a lack of intervention effectiveness (as determined by improvement in weekday MVPA) at the school level. This raises the question ‘If the intervention was delivered as planned, with good fidelity to underpinning theory, did insufficient peer supporting take place to bring about positive improvements in physical activity, or is it realistic to expect a(ny) single intervention to bring about meaningful change?’.

Existing literature in this area suggests that other peer-led interventions delivered at scale had no lasting effect on MVPA levels in adolescent girls112,124 or in a mixed-gender school setting. 125 Harrington et al. 112 found evidence of a small (i.e. 2–3 additional minutes of MVPA per day), short-term (i.e. 7-month) effect of the ‘Girls Active’ intervention among adolescent girls in English schools, but this was not maintained at the 14-month follow-up. In a randomised controlled trial of an intervention that trained older adolescents (i.e. mentors) and in-class peer leaders to encourage classes to undertake two new weekly activities, Corder et al. 125 reported no difference in MVPA between the control and intervention groups at follow-up. Moreover, although a peer-led walking intervention in schools124 resulted in an increase in daily light-intensity physical activity of 8.27 minutes among adolescent girls, it had no significant effect on MVPA.

Many elements of Harrington et al. ’s112 Girls Active intervention focused on changing school culture around physical activity, but not adding new opportunities to be active, which was hypothesised to be insufficient to increase MVPA. PLAN-A relies on a similar model, changing peer culture and expecting girls to find their own ways to increase their physical activity. Indeed, in a recent feasibility trial exploring peer-mentoring in conjunction with new physical activity opportunities, Owen et al. 126 reported that, when using peer mentoring alone, physical activity declined, whereas, when used in combination with an after-school club, the same peer mentoring intervention increased physical activity by 3.2 minutes per day (p = 0.007); however, this was not a fully powered trial. Collectively, these studies highlight a need to explore alternative intervention approaches.

Peer-led delivery models and associated measure sensitivity

The process evaluation in this trial and in our earlier feasibility trial52 demonstrate that peer-led interventions such as PLAN-A can be delivered with good fidelity to intervention design and theory. However, challenges identified around behaviour management with larger groups and the intervention setting resulted in a ‘drift’ from important design elements, such as the identification of influential peer leaders. There are questions regarding the suitability of the individuals selected to be peer supporters and how well represented social groups are within the cohort,127 which reflect limitations of the peer-nomination process itself. Recent research has tested a social network analysis (SNA) approach to peer nomination using a closeness centrality selection criterion. 128 This selection criterion is predicated on establishing how closely each pupil is connected to all of their peers via a SNA and then selecting those needing the fewest intermediaries to reach everyone in the network. 129 This approach differs from the approach used in this trial and ASSIST,37 in which the overall number of nominations a person received was the criterion for selection, and presents a more data-driven approach to peer nomination that has been shown to be successful at selecting more effective agents for change, who are capable of influencing physical activity in their peers,130 namely those with higher ‘sociometric popularity’ rather than ‘prestige popularity’. 131 Nevertheless, establishing the efficacy of a peer nomination process for selecting pupils with the greatest influence over their peers is only part of the peer-led approach. Once identified, these agents for change must also diffuse new behavioural norms.

In this trial, we experienced difficulties in determining the extent to which trained peer supporters engaged in supportive conversations and other peer support strategies (see Chapter 5, Evidence of peer support: fidelity to diffusion of innovations theory). Participants in this trial and the previous feasibility work52 demonstrated that diaries are not an effective way to capture this information, and qualitative evidence suggests that the subtle nature of much of peer supporting makes it hard to recall, harder still to quantify and, potentially, very difficult for recipients to notice. In a roll-out of the ASSIST programme, it was also a challenge to measure the peer-supporting conversations that took place. 132 In contrast to ASSIST, the outcome of interest in PLAN-A (i.e. weekday MVPA minutes) was a much more complex behaviour to measure (comprising type, volume, intensity and context) than a discrete behaviour, such as smoking and, therefore, it was not possible to use a binary variable. Furthermore, the focus of ASSIST for most children was not smoking (for most this meant not starting to smoke), whereas for PLAN-A it was being more active; the latter is, arguably, comparatively amorphous and, perhaps, not as strongly socially linked to peer influence and social status among adolescents. Measurement considerations are twofold: measurement of the exposure (i.e. peer supporting) and of the outcome (i.e. MVPA).

Regarding the exposure, ecological momentary assessment (EMA) has been used in various settings and fields to capture real-time data through repeated measures of behaviour or behavioural contexts. Using EMA dramatically reduces the period of recall, meaning biases associated with retrospective recall are reduced and the need for participants to summarise events or experiences over extended periods of time is removed. 133 Thus, EMA could, theoretically, be used to improve measurement of peer-supporting actions, or at least to reduce the recall bias experienced in this and other peer-based trials.

However, as highlighted in a recent review of EMA studies on physical activity and sedentary behaviour,134 the methodology is vulnerable to several risks of bias that would present challenges to its use in capturing peer-supporting activity. The primary risk is the sampling approach, time-based versus event-based sampling, as either approach would be imperfect. If a time-based sampling approach was used, then peer supporters would be asked to recall peer-supporting activity and context at scheduled intervals during the day, which may lose the benefit of momentary assessment and be subject to recall bias. On the other hand, an event-based approach – which is usually self-initiated – would rely on participants logging details of peer-supporting activities at the time, which, based on feedback from peer supporters in this trial about the use of their diaries, is something that adolescents would not engage with reliably.

Furthermore, despite advances in technology that mean that EMA can be performed using participants’ smartphones, the method is burdensome for participants and is likely to blur the lines between measurement and providing an additional intervention, which would need to be replicated if the intervention were scaled up. Although many studies using EMA report high compliance,134–136 the measurement periods are usually short (between 1 day and 2 weeks) and the feasibility of using EMA continuously for extended periods of time (e.g. a 10-week diffusion period, as used in PLAN-A and ASSIST37) has not been tested.

Considering the outcome measure’s sensitivity, the use of objective means of measurement (i.e. accelerometers) has been widely accepted as a more robust way of capturing physical activity than self-report, which is prone to overestimation of higher-intensity physical activity and underestimation of sedentary time. 137,138 However, there are known limitations of using accelerometers, including an inability to identify time spent in different activities139 and the risk that the brief measurement period (7 days is typical) may be too short to reliably observe changes in ‘normal’ behaviour patterns, or it may be subject to confounding by external factors. In this trial, we hypothesised that participants in two schools exhibited lower-than-predicted physical activity because of adverse weather, reinforcing this concern. In addition, using hip placement of the accelerometer, as was used in this trial, and asking that the devices be removed when in water or when playing contact sports risked underestimating or missing certain types of physical activity, such as cycling,140 swimming and contact sports, that can account for significant amounts of MVPA. It is possible that the challenges associated with measurement of the exposure (i.e. peer supporting) or nuanced and imperfect measurement of the outcome (i.e. change in MVPA) relative to the exposure reduced our ability to accurately quantify the efficacy of a peer-led approach to changing physical activity.

Likewise, if such interventions can be delivered at scale, but are limited in their effectiveness at changing physical activity, then we may naturally question whether the reasons for that lie in the intervention model itself (using peer support to diffuse new norms) or in social factors that are hard to measure (e.g. pre-existing local cultural norms) that undermine the assumptions inherent in a peer-based DOI approach. One such social factor pertinent to adolescents today is the use of social media to interact and influence. 141 There is uncertainty about whether or not peer-led interventions in schools can fully capitalise on how adolescents interact and influence each other using social media because of challenges in measuring engagement and impact. 142 Thus, there are major challenges to designing an intervention that includes substantial smartphone and social media components (whether delivering messages, support, ideas to peer supporters and encouraging peer supporters to influence their peers through their usual social media channels, or in the assessment of peer-supporting activity) and is delivered during school time. It would be worth exploring how social media and mobile technology could be incorporated into peer-based interventions and their acceptability to schools, pupils and parents.

Issues with scaling up interventions

The data presented in this report are consistent with recent evidence on the challenges of scaling up physical activity interventions from feasibility studies to definitive trials. In a recent meta-analysis, Beets et al. 143 identified nine forms of bias in this area, of which two were relevant to this trial. We found that there were changes in intervention support between the feasibility trial52 and this trial as the group sizes were larger than those in the feasibility trial and, therefore, trainer–pupil contact was more diluted, and there was a target audience bias as we had to change the year group from Year 8 to Year 9. The data reported here reinforce the need to focus on external validity and the potential implementation of intervention results from the design stage. 144

Alternative approaches to interventions targeting individual-level change and the need to explore whole-school approaches

The ultimate target of public health interventions is behaviour change at a population level, which is logically driven by behaviour change among the individual members of a population. However, as advocated in recent papers,145,146 individual-level interventions are not sufficient to bring about individual- or population-level change because the determinants of our behaviour are diverse, complex and often beyond individual control (e.g. policy, physical environment, SEP). PLAN-A attempted to advance beyond a simple individual-level education-based intervention; instead, it was designed to harness the power of naturally existing peer groups and used informal (e.g. peer diffusion) rather than formal (e.g. educational lessons, additional structured physical activity time in school) methods to promote the uptake of physical activity. However, despite this, the intervention model is relatively constrained and one-dimensional: it does not embrace a broader system or ‘whole-school’ approach,146 which would intervene at multiple levels including, but not limited to, peer influence. Indeed, from a systems perspective, it seems somewhat unrealistic to expect a short-term intervention focused on a single determinant of physical activity (i.e. peers) to have a meaningful impact on physical activity behaviour. It is important to note that this was a relatively high-agency intervention, and it may be the case that an intervention that required lower agency would have been more successful. Low-agency interventions are consistent with the concept of whole-school interventions147 and, thus, in the future we suggest that lower-agency interventions are considered.

An important question that arises from this project is ‘What next?’. Whole-school approaches to physical activity have been advocated for some time,148 and the recently developed Creating Active Schools framework advances this work using insight from researchers, teachers and sport practitioners. 146 However, like PLAN-A, school-based interventions often focus on a single or a few components of the whole-school approach only. This is probably driven by a methodological need to:

• define what the intervention is and design it well

• estimate the effect of particular intervention strategies to inform future investment (e.g. active travel, enhanced PE or after-school activity clubs)

• design interventions that could be attractive to schools (i.e. interventions that are clear, discrete, simple and not resource intensive)

• design interventions that can be rolled out at scale.

However, these elements create a conflict between intervention depth and breadth. Many facets of good intervention design (e.g. detailed planning, iterative co-development, piloting, pre-testing, the use of logic models, theory, attention to behaviour change techniques) take considerable time and resources to perform well and often necessitate a narrowing of the intervention scope. This was the approach used in PLAN-A, and it arguably resulted in good intervention ‘depth’. However, a systems or whole-school approach requires greater ‘breadth’, with interventions operating across multiple domains simultaneously (e.g. active travel, school culture and ethos, high-quality PE, active lessons, active breaks, active extracurricular clubs). Nevertheless, it is not yet clear how multidimensional whole-school approaches can be designed in a way that balances and optimises depth and breadth.

The balance between breadth and depth is also an issue that affects intervention evaluation. It is, relatively speaking, straightforward to evaluate a well-defined, one-dimensional, school-based physical activity intervention such as those conducted in recent years by our group and others. 51,78,123,125 The intervention can be clearly described, fidelity can be assessed, and its effect compared using a control group within a traditional cluster-randomised controlled trial design; again, we can achieve good depth, but limited breadth. It is not clear how best to evaluate broader whole-school approaches, which act on multiple domains of physical activity, nor how to determine which elements are and are not effective, or how interventions may interact with pupil-level factors (e.g. age, gender, SEP), cohort effects and school-level factors (e.g. deprivation, facilities, size or location). Even if researchers can determine that a whole-school approach is effective, it may be unrealistic to expect the approach to be generalisable. Schools are like ecosystems in that no two are the same: they are in a state of constant flux; serve diverse communities with different needs; and have shifting priorities, variable budgets and resources, and changeable internal and external contexts. Effective interventions are likely to be those that can be designed and modified to accommodate these factors; therefore, an intervention that works in one school may not suit, work or be needed in another.

One approach is to focus further on depth by designing specific and discrete interventions (e.g. increasing the amount of MVPA time in PE) and very closely measuring the outcomes that are matched to the behaviour (e.g. MVPA time in PE), repeating this across multiple domains (e.g. active travel, classes, miles and clubs) and being content with the assumed additive effect of the small effects on MVPA across multiple domains, summing to a meaningful or health-enhancing level of physical activity. If multiple interventions are needed as part of a whole-school approach, testing discrete interventions would also lack external validity and would not provide insight as to whether or not schools are able to implement multiple interventions simultaneously. If we focus more on breadth, we risk developing weaker or more diluted interventions and it may be very challenging to identify which intervention actions are effective. It may be harder to convince schools and/or commissioners to adopt more resource-intensive approaches. This is a challenge that needs to be addressed in the next generation of system-focused public health interventions.