A school-based intervention ('Girls Active') to increase physical activity levels among 11- to 14-year-old girls: cluster RCT

Physical activity levels of young people

It is well known that participation in regular physical activity (PA) has numerous short- and long-term physiological, psychological and social benefits for young people. 1–3 For example, reviews1,2 have shown that PA can improve blood pressure and bone mineral density, can lead to improvements in self-esteem and depression and, in the longer term, it can prevent chronic disease. Despite these benefits, a significant proportion of adolescents are not meeting the recommended guideline of participating in ≥ 60 minutes of moderate to vigorous intensity PA (MVPA) each day. 4 A recent review of pan-European studies5 found that, across all countries, less than 30% of adolescents achieved the recommended levels of MVPA. Furthermore, data from two of the most comprehensive surveys (the Global School-based Student Health Survey6 and the Health Behaviour in School-aged Children study7) showed that an even larger proportion of adolescents (80%) were not meeting the guideline levels of MVPA. 2 However, this proportion has been shown to be even higher when measured using accelerometers, with approximately 95% of 5- to 17-year-olds failing to meet the recommended level. 8 Data specific to the UK (collected in 2015) suggest that only 17% of 11- or 12-year-olds and 12% of 13- to 15-year-olds are sufficiently active. 9 Furthermore, it has been consistently shown that, as children get older, boys are more active than girls. 2,5,8,9 For example, data from the Health Survey for England 2015 showed that, while both boys and girls exhibit low levels of PA, only 16% and 9% of girls aged 11/12 years and 13–15 years, respectively, meet the guideline levels, compared with 15% of boys in each of the same age groups. 9 Worldwide data2 reinforce these findings, with the proportion of adolescents not achieving ≥ 60 minutes of MVPA per day being ≥ 80% in 56 out of 105 countries (53%) for boys and in 100 out of 105 countries (95%) for girls. Girls have therefore been identified as a key target for behaviour change in an effort to increase PA levels and improve current and future population health.

Physical activity interventions in adolescents

A plethora of PA interventions aimed at young people have been conducted over the past two decades. 10–13 These interventions have been varied in their approaches to increasing PA levels but they can be grouped into interventions targeting changes in the school cultural or physical environment, interventions targeting changes in the school curriculum, interventions incorporating extra PA within the school day, education-only interventions, after-school interventions, and interventions with a community or family component. Published reviews10,13 have found some evidence that interventions focused on increasing PA levels in adolescents can be effective, although only modest effects have been found. van Sluijs et al. 11 found that over half of the studies (n = 38, 67%) included in their review showed a positive intervention effect ranging from an increase of 2.6 minutes during physical education (PE) classes to a 42% increase in participation in regular PA and an increase of 83 minutes per week in MVPA, with multicomponent interventions providing the strongest evidence of an effect. A more recent review of interventions in older adolescents13 reported more promising results, with 7 out of 10 studies showing a significant difference in PA levels between intervention and control groups. Effects were generally small and short term (no longer than 1 month post intervention). Two reviews10,14 have been conducted that focus solely on the effectiveness of PA interventions aimed at adolescent girls. In a narrative review, Camacho-Miñano et al. 10 reported mixed results, with 10 out of 21 interventions showing a modest favourable intervention effect for PA levels. More recently, a meta-analysis quantifying the effectiveness of PA interventions among adolescent girls14 concluded that PA interventions had ‘small’ but significant effects. Compared with girls receiving the control conditions, those who received the intervention participated in 13.7% more PA than those who did not. Subgroup analyses showed that multicomponent interventions, those focusing on both PA and sedentary behaviour, theory-based interventions and intervention designs of high quality showed small trends in effecting the primary outcomes. 14 Furthermore, significant trends were found for interventions that were aimed at girls only, targeted at younger adolescent girls or delivered within school-based settings. 14

These reviews10–14 have consistently reported several major methodological limitations of previous research that have had an impact on the ability to draw robust conclusions. The vast majority of studies have suffered from weak methodological quality. This can be attributable to small sample sizes, which can lead to high rates of bias, and a reliance on PA outcomes that are assessed by self-report. 10,13,15 Furthermore, a high proportion of studies have been conducted in the USA. 10 For example, 18 of the 24 studies included in one review11 were conducted in the USA. This makes generalisability to other countries difficult, particularly because of differences in school infrastructure, curriculum, policies and practices and culture. These reviews10,13,15 have also noted a lack of long-term follow-up to evaluate the effectiveness of interventions in the longer term.

Importance of peers

Social support from parents, family and friends has been identified as a key determinant of young people’s PA level. 16,17 However, it has been observed that, as children move into adolescence and spend more time with their peers, support for PA from peers is reported to be more important than support from other sources. 18 Furthermore, numerous reviews have demonstrated a positive link between peer support and PA levels,19,20 and specifically for adolescent girls’ PA levels. 21 A 2014 review17 noted that peer support showed the greatest consistent association with PA level when compared with other sources of support. Greater social support from peers has also been associated with higher levels of self-efficacy, which, in turn, has been associated with higher PA levels. 16 Peers have been found to influence PA levels in a number of ways, including through the provision of encouragement and praise, joint participation in activity, the modelling of being active and by simply being present during PA. 17,19,20

This evidence from previous studies suggests that interventions that incorporate the influence of peers might be a useful avenue for increasing adolescent PA levels. Despite this, very few interventions have used peers as an intervention strategy. One review10 identified only two interventions that had focused on utilising the peer influence for promoting PA among adolescents and concluded that peer interventions showed promising evidence for being effective. More recently, further evidence has been published22 that supports the effectiveness of peer inclusion in interventions. Spencer et al. 22 reported that training peers to act as mentors by promoting PA during lunchtimes was successful in increasing PA levels during the school day. 22

The extensive amount of evidence highlighting the important role that peers play in adolescent PA and the limited intervention evidence conducted to date using the influence of peers highlights the need for future research to focus on the effectiveness of peer-focused interventions for increasing PA levels during adolescence.

Primary objective

To investigate whether or not the Girls Active programme leads to higher levels of objectively measured MVPA in adolescent girls compared with the control group at 14 months after the baseline assessment.

Secondary objectives

• To investigate whether or not Girls Active results in changes in the following outcomes among adolescent girls in the intervention group compared with those in the control group at 7 and 14 months’ follow-up:

  • objectively measured overall PA level (mean acceleration)

  • the proportion of girls meeting the recommended PA target of 60 minutes per day of MVPA

  • time spent on MVPA (at 7 months)

  • time spent sedentary (assessed objectively and via self-report)

  • measures of body composition

  • psychological factors that may mediate PA participation (health-related quality of life; PA self-efficacy; PA motivation; social support from peers, family and teachers; PA enjoyment; perceived importance of PA; and physical self-perceptions).

• To undertake a full economic analysis of the Girls Active programme, from a multiagency public sector perspective, at 14 months’ follow-up.

• To conduct a mixed-methods process evaluation throughout the intervention implementation period (using qualitative and quantitative measures) with teachers, students and peer leaders and the YST to provide insights into the ways in which, and extent to which, the intervention was implemented and into participant experiences of the intervention.

Trial steering committee

There were two lay members in the TSC to ensure that at least one would be available for each meeting. One was a former PE teacher (who attended one out of four meetings) and the other was a member of the public health team at the local government level (who attended two out of four meetings, with an in-person and e-mail debrief supplied for the missed meetings). The lay members contributed valuable insights on the school/education landscape and the local policy context in which this research sits. Certain questions for the lead teacher interviews were added or modified based on the input of the TSC, including a question on the impact of Girls Active on boys. A focus group with boys was also included on their recommendation.

Cost diaries

The feedback from the lead teachers was instrumental in refining our research methods in relation to the cost diaries. Feedback from lead teachers that came via the YST and direct from the teachers themselves was that the Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) cost diaries were difficult to manage. The research team took this feedback on board and, at the peer leader event, cocreated a simpler version of the cost diary with the help of all teachers in attendance. The lead teachers who found the original Microsoft Excel diary usable gave advice to the other lead teachers, meaning they supported the data collection process by helping their peers to understand and navigate the diaries.

School reports

At the 7-month interviews, lead teachers were asked which variables and measure would be of most interest to them and their school. This allowed the research team to create a personalised report to give to each school; this report contained the school’s baseline data, which was aggregated to school means (the reports were presented to schools in September 2016). Teachers felt that it was important to capitalise on the data and to use them within their schools to advocate for better or continuing provision of PA, PE and school sport.

Intervention group: Girls Active programme

Girls Active is a school-based programme aimed at girls in KS3. It has been designed to provide a support framework for schools to review their PE, sport and PA culture and practices to ensure that they are relevant and attractive to their 11- to 14-year-old female pupils. Although the intervention was multicomponent and designed to be flexible in delivery, it had several key components that were considered integral to the Girls Active process.

• Self-evaluation and mission analysis. Schools completed a mission analysis document, which helped them to review their existing culture and practice for girls’ PA within the school and to set out an action plan tailored to their girls’ needs. This process drew on review, planning and evaluation frameworks used by UK sporting associations internally to generate sporting success at the 2012 London Olympic and Paralympic Games. This analysis was an exercise that was carried out twice by lead teachers at their schools: the first time as a ‘pre-course’ task ahead of the initial training and again after the peer review day.

• Initial training for school leads. Lead teachers attended a 1-day face-to-face group training event at the hub school to introduce all school leads to the Girls Active programme. This event was facilitated by a YST national tutor. This training event had the following objectives:

  • to support teachers to explore their role and effectiveness in engaging girls in PE and sport in school

  • to challenge teachers to consider girls’ motivations in relation to PE and sport

  • to help teachers look at how a marketing approach can increase girls’ participation

  • to enable teachers to review a range of case studies and support resources

  • to provide teachers with the opportunity to share and develop practice with their peers

  • to challenge teachers to develop an action plan to support the Girls Active project with the support of other teachers and members of the senior leadership team.

The training objectives were covered in seven sessions during the training day, which involved a combination of presentations, discussions, practical activities and opportunities for teachers to share challenges, successes and ideas with each other and the tutors.

Figure 2 presents the Girls Active delivery model.

FIGURE 2.

The Girls Active delivery model.

Control group: usual practice

Schools randomised to the control arm were not given any specific guidance or advice and were assumed to continue with usual practice. A questionnaire was administered at each of the measurement time points to capture the school environment, including what the schools offered to girls in KS3 outside the typical PE and school sports clubs. Locally organised activity days available to girls in Leicestershire schools and the This Girl Can campaign were mentioned.

Primary outcome measure

The primary outcome measure was accelerometer-assessed mean number of minutes per day of MVPA at 14 months for participants.

Secondary outcome measures

Data for the following secondary outcome measures were collected from participants at 7 and 14 months:

1. accelerometer-assessed overall PA level represented by mean acceleration per day (overall and split by weekday and weekend day, and during school hours and after school hours)

2. accelerometer-assessed mean minutes per day of MVPA on weekdays and weekend days split, and during the school hours and during the after-school hours split

3. the proportion of girls meeting the recommended PA guideline level of 60 minutes per day of MVPA

4. time spent on MVPA at 7 months

5. accelerometer-assessed mean number of minutes per day of sedentary behaviour (overall and split by weekday and weekend day, and during school hours and after school hours)

6. self-reported (via a validated questionnaire) mean time spent in different types of sedentary behaviours

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

8. psychological factors that may mediate PA participation (health-related quality of life; PA self-efficacy; PA motivation; social support from peers, family and teachers; enjoyment of PA; perceived importance of PA; and physical self-perceptions).

Accelerometer measurements

Participants were asked by the measurement team to wear the wrist-worn GENEActiv™ (Activinsights Ltd, Kimbolton, UK) accelerometer on their non-dominant wrist continuously (i.e. 24 hours/day) for 7 days to assess PA and sedentary time. This lightweight research-grade device resembles a sports watch and is waterproof. Published cut-points are available to classify sedentary time and MVPA based on children’s observations and indirect calorimeter data. 29,30 The GENEActiv accelerometer was selected for these advantages and, based on the team’s previous experience and feedback from secondary school pupils at a patient and public involvement event, continuous wrist wear helps maximise compliance and reduce missing data in this age group. The GENEActiv devices were initialised with a sampling frequency of 100 Hz and were set to start recording at midnight on the first day of data collection and stop recording at midnight 7 days later. The accelerometers were distributed at the school measurement visit and collected from the school between 8 days and 3 weeks after the measurement visit. GENEActiv data were downloaded using GENEActiv software version 2.2 (Activinsights Ltd, Kimbolton, UK) and saved in a raw format as binary (.bin) files. Participants were given a £5 gift voucher as a thank you for providing full data. Data were included if participants had over 16 hours of wear time recorded during each 24-hour day.

Accelerometer data processing and analysis

The GENEActiv .bin files were analysed with GGIR version 1.2-11 [R Project; The R Foundation for Statistical Computing, Vienna, Austria; http://cran.r-project.org (accessed 5 June 2018)]. 31,32 Signal processing in GGIR includes the following steps:

1. autocalibration using local gravity as a reference31

2. detection of sustained abnormally high values

3. detection of non-wear

4. calculation of the average magnitude of dynamic acceleration {i.e. the vector magnitude of acceleration corrected for gravity [Euclidean norm minus one (ENMO) g]} over 5-second epochs with negative values rounded up to zero as:

Files were excluded from all analyses if post-calibration error was greater than 0.02 g33 or if less than 16 hours of wear time was recorded by either monitor during the 24-hour day of interest. Detection of non-wear has been described in detail in ‘procedure for non-wear detection’ in the supplementary document in van Hees et al. 32 In brief, non-wear is estimated based on the SD and value range of each axis, calculated for 60-minute windows with 15-minute moving increments. If, for at least two out of the three axes, the SD is less than 13 mg or the value range is less than 50 mg, the time window is classified as non-wear.

The average magnitude of dynamic wrist acceleration (ENMO), time accumulated in sedentary behaviour, light-intensity PA, MVPA and time spent sleeping were calculated over the entire 24-hour day. Thresholds for determining PA intensity categories were as follows: 0–40 mg for sedentary behaviour (minus sleep), 41–199 mg for light activity and ≥ 200 mg for MVPA. 29,30 Each school reported the start and end times of each school day. These times were then used to extract activity variables (sedentary time, light activity and MVPA) during and after school hours. The after-school period was until 21.00 hours.

In order to partition sleep from total daily sedentary time, sleep was determined using the nocturnal sleep detection algorithm incorporated in GGIR. 34 Periods of sustained inactivity are defined as no changes in arm angle of greater than 5° for ≥ 5 minutes during the 12-hour window, centred at the least active 5 hours of the 24-hour period. Individual nights were excluded from the sleep analyses if the sleep duration recorded for a night was < 6 hours or > 12 hours and if visual examination suggested non-wear or erroneous data. Variables extracted were time of sleep onset to wake (called ‘time in bed’ in the software output but actually time from sleep onset to wake and waking time determined from accelerometer data), sleep duration (total sleep duration determined from accelerometer data excluding detected waking episodes during the night) and sleep efficiency (sleep duration/time in bed × 100). Sleep variables were averaged for weekday nights (Sunday to Thursday) and weekend nights (Friday and Saturday).

The following accelerometer variables were calculated:

• average time spent in MVPA (overall and split by weekday and weekend day, and during school hours and after school hours)

• average time spent in sedentary and light PA (overall and split by weekday and weekend day, and during school hours and after school hours)

• overall PA level (average acceleration) for the whole measurement period and split by school days and weekend days.

School (cluster) characteristics

Participant demographic characteristics

Anthropometric measurements

Owing to the relationship that PA has with adiposity, and the potential for school-based programmes to have an impact on the prevention of obesity,39 assessment of height and weight was undertaken by the measurement team for the calculation of BMI. Sitting height was measured as it is used, in addition to height and weight, to predict age at peak height velocity (APHV) using a sex-specific multiple regression equation. 40 APHV is an indicator of physical maturity, reflecting the maximum growth rate in stature during adolescence. All anthropometric measurements were taken in a private, screened-off area, and only female research team members took the measurements. Weight was measured using scales with a covered remote display so that the participants did not see their own weight. The girls were reassured that all measurements were confidential and that research data would be linked only using a non-identifiable participant number. Participants removed their shoes, any large items of clothing, such as jumpers and any items from their pockets. In order to assess body composition via impedance, participants were asked to also remove socks or tights.

Self-reported behaviours

A number of questionnaires were used to assess physical activities and sedentary behaviours that have proven relationships with current and future health outcomes and that may be affected by the Girls Active programme. These questionnaires were administered by the measurement team.

Psychosocial measures

A variety of psychosocial outcomes were assessed using existing questionnaires that have demonstrated reliability and validity for use with this age group, which were administered by the measurement team. A measure of internal consistency (Cronbach’s alpha) for these questionnaires is reported for the Girls Active baseline questionnaire results, calculated using IBM SPSS Statistics version 20.0 (IBM Corporation, Armonk, NY, USA). Psychosocial outcomes were chosen as they could be included in a SCT-grounded logic model describing variables that could mediate any effects on the primary outcome. The words or terms in some of the questionnaires originating in the USA were adapted, when necessary, to make the wording more applicable in a UK setting. At each measurement visit, participants were given specific guidance on certain questions, words or phrases that had been identified in the pilot testing as being potential causes of difficulty for participants. A questionnaire standard operating procedure (SOP) was developed and each measurement team member was required to read it before the measurement session. This SOP contained a ‘live’ list of frequently asked questions, which came from the participants, and a set of standardised answers that the measurement team were instructed to give in order to avoid the measurement staff interpreting the questions in their own way.

Other health behaviours

Data on potential negative health behaviours that might develop over time were also collected by the measurement team in an effort to assess any potential unintended consequences of the Girls Active programme.

Microcosting methods

Microcosting methodology64 was applied to calculate the costs of delivering the programme over one school year for the intervention schools taking part in the trial and to provide a mean cost per school and per pupil. In this economic analysis, we fully costed the delivery of the Girls Active programme and its constituent costs, such as teacher time, and the cost of other consumables and materials used. Costs were collected from a local authority (school) perspective, accounting for oncosts, and using the cost year 2015/16. Costs were measured in Great British pounds sterling (GBP).

A cost diary was administered by one member of the research team (JMC) to the school leads responsible for delivering Girls Active. The diary asked school leads to complete a record of the additional time, or displaced time, taken to offer the Girls Active programme, and they described activities undertaken and items purchased (e.g. sports equipment or sports clothing such as hoodies). The diary was developed with input from the wider Girls Active research team and the YST, which was responsible for training and supporting teachers to deliver the trial. Three versions of the diary were used throughout the programme implementation. The diary was originally sent to teachers by e-mail as a Microsoft Excel file, with rows for each activity and a space for ‘other’ activity and costs that were not covered by the other rows. The diary was sent with instructions and an example of a completed sheet to provide the respondents with the information and level of detail required when competing the diary. We asked the teachers to complete the diary weekly. A member of the research team requested the Microsoft Excel diary at 2-month intervals while the intervention was being implemented (April 2015 until May 2016). Any queries regarding the information that was provided were e-mailed to staff, and researcher contact details were provided in e-mails. After receiving the final Microsoft Excel diary, any final queries were sent by e-mail.

The research team received feedback that some teachers were struggling to keep the Microsoft Excel diary and were not comfortable with the format. In response, we created a paper logbook with sections to complete for activities and costs. These logbooks were printed and administered by the Girls Active research team. It was requested that teachers keep these logbooks with them in their own diaries and that they note activity and costs related to Girls Active as and when they happened. As a final method to ensure that we had information from Girls Active lead teachers, a survey was produced and administered during a peer review day that was organised by the research team and the YST. The survey used fields from the Microsoft Excel diary and the logbook and was followed up with a telephone call from the researcher responsible for the microcosting to provide context for the responses. Each of the three methods (Microsoft Excel diary, survey and logbook) requested demographic information about the teacher completing the diary, such as school name, job title and salary band.

The salary band information was used to calculate teacher costs in the microcosting, using the National Union of Teachers pay structure for qualified classroom teachers in England and Wales (from 1 September 2015 for cost year 2015/16). 65 A school year consisting of 39 weeks was used to calculate the teacher costs per hour, taking into account sickness, continuing professional development and annual leave. Salary calculations were inclusive of employers’ oncosts. Oncosts include National Insurance and pension charges, as well as costs for annual increments and allowances.

When collating information from the diaries, any costs relating to research (e.g. time to complete diary or measurements or to undertake interviews) were not included in the final microcosting calculations. These were considered as research costs. This decision was made in order to provide local authorities with information pertinent for future rollout (training and delivery costs), rather than costs specific to conducting a research trial.

Cost–consequences methods

Analysis

In order to analyse clustered data appropriately and in line with the main outcomes analysis, the main statistical analysis plan and methods as described in Chapter 3, Analysis of the primary outcome, were used for this cost-consequences analysis’. The generalised estimating equation (GEE) model was used to determine the difference in mean number of minutes per day of MVPA, CHU-9D utility index scores and service use frequencies and costs between pupils from schools allocated to the intervention group and those from schools allocated to the control group, taking account of clustering in the trial design. The variables included in the analysis and the GEE model specification were the same as those outlined in Chapter 3, Analysis of the primary outcome. For service use frequency and cost models, generalised linear model diagnostic tests were conducted because the models failed to converge using the GEE model structure. From the results of the diagnostic tests, the family was amended from Gaussian to gamma and the link was amended from identity to power –1. After changing these specific parameters, the models achieved convergence and these models were used for the variables of service use frequencies and costs in the xtgee models.

For the analysis of the CHU-9D, the utility index score was calculated using the syntax provided by the questionnaire developers. 70 We calculated the differences in marginal mean CHU-9D utility index scores between the groups and produced 95% confidence intervals (CIs) around these differences with 1000 bootstrapped replications. The change in marginal mean number of minutes per day of MVPA for the intervention group and the control group and the differences between the groups were calculated and 95% CIs around these differences with 1000 bootstrapped replications were produced.

Sensitivity analysis

As the intervention follow-up period was > 1 year, as part of the sensitivity analysis service use costs at 14 months post baseline were discounted at 3.5% (the base rate recommended by NICE, 201366), and the model was re-run with these discounted costs.

Missing data for mean number of minutes per day of MVPA, total frequencies of service use and total costs of service use were imputed as part of the sensitivity analysis, following the methods employed in the main analysis. The CHU-9D utility index scores were not imputed as the developers of the measure state that utility values cannot be calculated when questions have missing answers. 63 The developers advise that only observations with complete data should be included; therefore, in line with these recommendations, imputation was not conducted on the CHU-9D scores.

Exploratory subgroup analysis

As part of an exploratory subgroup analysis, the effects of year group and level of implementation, which were based on the levels described in the microcosting, were tested on mean number of minutes per day of MVPA, CHU-9D utility index scores and frequencies and costs of service use for the complete-case sample, by including them in the xtgee model run in the main analysis.

Analysis of costs

We compared the frequencies and costs of health care and school-based service use over 14 months between the intervention group and the control group. We calculated the marginal mean total service use for the intervention group and the control group. We went on to calculate the differences in marginal mean total service use between the groups, and produced 95% CIs around these differences with 1000 bootstrapped replications. Marginal means are presented throughout; this reports the mean following the xtgee model and takes account of clustering in the trial design.

Training resources

All resources used at the training events were collected. These included tutor notes, Microsoft PowerPoint^® (Microsoft Corporation, Redmond, WA, USA) presentations, the event schedule and any sessional or post-training resources that were circulated. This was to help understand what training was delivered to lead teachers. An observer from the research team attended the peer review day and took notes in an effort to explore what was delivered to teachers compared with what was planned and how the teachers responded to what was delivered. A copy of the resource folder given to each intervention school was also obtained.

Feedback forms

At the end of the initial training day, peer review day and peer leaders event, participants completed evaluation forms that included questions on content, perceived learning, delivery and venue. These forms were designed and utilised by the YST for these types of programmes. Attendance records and feedback forms from the lead teachers and peer leaders were also collected.

Exit survey

All participants in the intervention schools were asked to complete an exit survey at the end of the 14-month measurement visit. This survey queried what the participants understood about Girls Active, whether or not they took part in any Girls Active activities, what their most and least favourite parts of Girls Active were and what their feelings and views were on a number of statements related to the programme.

School environment

Lead teachers completed a school environment questionnaire at baseline and the 7-month and 14-month follow-ups, as described in School (cluster) characteristics. This questionnaire captured any changes in the school physical and social environment that may have an impact on the delivery of Girls Active.

Lead teacher interviews

Interviews with all teachers were undertaken at the 7-month and 14-month follow-ups. Appendices 1 and 3 provide the questions for the intervention school teacher interviews at 7 months and 14 months, respectively. A series of questions was mapped out to serve as an opening question. Probes were provided to the interviewer to help allow the participant to expand on the topic. The aim of these interviews was to understand what was delivered, whether or not there were any changes in the school and PE department that might have had an impact on delivery and to understand the barriers, facilitators, challenges and opportunities that the teachers encountered throughout their Girls Active programme journey. These interviews also provided an opportunity to get further feedback about training days, resources and support. Appendix 3 contains the questions for the control school lead teacher interviews. The aim of these interviews was to get a brief understanding of what had been going on at the school and to see whether or not anything had changed in terms of PA, PE and sport provision for girls. Teachers were given a participant information sheet ahead of the interview and they signed a consent form.

Youth Sport Trust and hub school staff interviews

Interviews with intervention delivery staff (i.e. staff from the YST and the hub school and the development coach) were undertaken at 7 and 14 months. The aim of these interviews was to understand what was delivered to teachers by the YST and what support was given to schools by the hub school and by the development coach. We were interested in understanding what support was given, in knowing whether the school lead initiated the request for support and how the interactions between the teachers and the available support worked in practice. It was relevant to understand the barriers, facilitators, challenges and opportunities that were faced by the intervention delivery staff. Staff were given a participant information sheet ahead of the interviews and signed a consent form. The questions used in these interviews can be found in Appendix 4. The opening questions were followed by probes to prompt the participant to expand on the topic, if necessary. All interviews were audio-recorded and transcribed verbatim.

Focus groups

Interviews and focus groups

All interviews and focus groups were recorded, transcribed verbatim and then analysed using the framework analysis approach. This approach develops a hierarchical thematic framework that is used to classify and organise data according to key themes, concepts and emergent categories. It was used to explore, compare and contrast key themes arising from the data, using the elements of the interview/focus group topic guide as a starting point. Analysis was undertaken by two researchers from the evaluation team who read and reread each transcript to identify the themes. Each group of interviews or focus groups was analysed separately and the emergent themes from each data source (e.g. teachers, peer leaders, boys, etc.) were triangulated to build up a picture of what happened in each school and across schools. School summaries for each intervention school were developed and these summarise how peer leaders were selected, the activities implemented in the school, perceptions of impact, challenges to implementation, facilitators of implementation and any contextual factors that might have influenced Girls Active.

Per-protocol population

The per-protocol population comprises those participants who engaged with the key components of the programme over 14 months and for whom complete data were available for the analysis of concern on a ‘by analysis’ basis. In the control arm, the per-protocol population includes all schools/pupils randomised to that arm. In the intervention arm, the per-protocol population was defined as schools that engaged with ≥ 70% of the programme. All pupils within these schools were included in the per-protocol population. There are 10 key components of Girls Active, of which seven are considered core components (Table 4). A school that engaged with any five out of these seven components (i.e. ≈70%) is considered to have met the minimum criteria of the programme. The per-protocol analysis adhered to the same steps as the primary analysis.

Intention-to-treat population/full analysis set

The intention-to-treat (ITT) analysis involved all of the randomised schools and the recruited pupils in these schools. Schools and pupils were analysed in the group to which they were randomly allocated, regardless of whether or not they received the assigned intervention and regardless of any protocol deviations. Missing data were imputed using multiple imputation. To allow for ITT analysis, missing data were imputed using a multiple imputation procedure that substitutes predicted values from a regression equation. The imputation was carried out by the MI (multiple imputation) command in Stata. 76 MI replaces missing values with multiple sets of simulated values to complete the data, performs standard analysis on each completed data set and adjusts the obtained parameter estimates for missing data uncertainty using Rubin’s rules to combine estimates. 77 The following procedures were followed:

• The multiple imputation was set up as wide.

• The multiple imputation registered imputation of the MVPA at 14 months and MVPA at baseline.

• A regression method was used, in which the registered variables were individually adjusted for the randomisation stratification categories (school size and percentage of BME pupils), with 100 imputations to avoid biased estimates, rseed (2232) and the force option.

• The multiple imputation estimate provided the final results using the same GEE model as the primary analysis and covariate adjustments.

Effects of the number of valid days

Five separate models were run regarding valid wear time, including participants who wore the accelerometer (with a minimum valid wear time of 16 hours each day) and who had a usable file at baseline and 14 months: (1) at least any 1 day, (2) any 2 days or more, (3) any 3 days or more, (4) any 4 days or more and (5) 3 weekdays + 1 weekend day.

Seasonal adjustment

Seasonal data were obtained retrospectively78 and matched to individual participant data. The season was based on which season the majority of days fell in when the accelerometer was worn during the assessment period. Seasons in 2015 were defined as:

• spring – 20 March to 20 June

• summer – 21 June to 22 September

• autumn – 23 September to 21 December

• winter – 22 December to 19 March.

For the 2016 assessment dates, seasons were defined as:

• spring – 20 March to 19 June

• summer – 20 June to 21 September

• autumn – 22 September to 20 December

• winter – 21 December to 19 March.

They were coded as dummy variables 1–4 and added to the GEE model.

Main complete-case analysis

Overall, 1361 participants (77.7%) were seen at baseline and the 14-month follow-up. Of these, 1211 participants (69.1%) completed the accelerometer assessment with a minimum of 2 valid days, including at least 1 school day at baseline and the 14-month follow-up. This group represents the analytical sample for the primary outcome analysis. In the complete-case analysis at the 14-month follow-up (primary outcome), the fully adjusted model showed a non-significant difference of 1.7 minutes per day (95% CI –0.8 to 4.3 minutes per day; p = 0.178) between groups. At the 7-month follow-up, a significant difference in MVPA change of 2.4 minutes per day (95% CI 0.1 to 4.7 minutes per day; p = 0.039) between groups was found. Both of these results are presented in Table 10.

Intention-to-treat analysis

The ITT analysis presented in Table 11 was consistent with the main analysis. When all participants and schools were considered, there was no significant difference in change in MVPA between groups at 14 months. However, at the 7-month follow-up, a significant difference in MVPA change of 2.3 minutes per day (95% CI 0.2 to 4.3 minutes per day; p = 0.028) between the intervention group and the control group was found.

Per-protocol analysis

The per-protocol analysis (see Table 11) at 14 months’ follow-up showed no significant difference in MVPA change between groups (p = 0.246). A significant difference in MVPA change of 3.1 minutes per day (marginal mean number; 95% CI 0.93 to 5.35 minutes per day; p = 0.005) between the intervention group and the control group was found at 7 months. This suggests that when schools followed the core components of the Girls Active programme (as defined in Table 4), it resulted in a difference in MVPA at the 7-month time point only. When observing the mean change in MVPA from baseline values, it would seem that the Girls Active programme prevented a significant decline in MVPA (i.e. the control schools had a decline of 6.2 minutes/day, whereas the intervention schools had a decline of 1.9 minutes/day, between baseline and 7 months).

Sensitivity analysis

Objective GENEActiv variables

No significant differences in MVPA change between groups were found for school days (95% CI –0.5 to 4.2 minutes/day; p = 0.126), weekend days (95% CI –2.4 to 7.92 minutes/day; p = 0.290), during school hours (95% CI –0.8 to 2.3 minutes/day; p = 0.360) or after school hours (95% CI –0.8 to 1.9 minutes/day; p = 0.443) at 7 months’ follow-up, as shown in Table 13. Similarly, at 14 months, no significant differences in MVPA change between groups were found for school days (95% CI –1.3 to 4.8 minutes/day; p = 0.251), weekend days (95% CI –3.5 to 5.5 minutes/day; p = 0.669), during school hours (95% CI –0.5 to 2.2 minutes/day; p = 0.218) or after school hours (95% CI –1.6 to 2.4 minutes/day; p = 0.701).

The mean acceleration recorded by the accelerometer (ENMO) was also investigated. There was a significant difference in ENMO between groups at 7 months (1.39 mg, 95% CI 0.1 to 2.2 mg; p = 0.030) but not at 14 months (0.66mg, 95% CI –0.6 to 1.9 of 1.39 mg; p = 0.31). There were no significant differences in ENMO when the measurement period was broken down into school days and weekend days at either 7 or 14 months (p > 0.05).

No significant differences in change in sedentary time (minutes/day) were found overall or on weekdays, weekend days or during school hours at either 7 or 14 months’ follow-up. However, at 7 months, a significant difference in change in sedentary time of –4.7 minutes per day was found during the after-school period (95% CI –8.9 to –0.6 minutes per day; p = 0.026), whereby the intervention schools’ sedentary time increased less than the control groups’ sedentary time between baseline and 7 months.

A significant difference in time spent in light PA was found between groups for all monitoring days at 7 months (5.7 minutes/day, 95% CI 1.0 to 10.5 minutes/day; p = 0.018), but this was not seen at 14 months (1.5 minutes/day, 95% CI –5.9 to 8.9 minutes/day; p = 0.692). The median values indicated that, although light PA decreased in both groups, it decreased less in the intervention group. A significant difference was found on school days at 7 months (4.5 minutes/day, 95% CI 0.25 to 8.75 minutes/day; p = 0.038) but not at 14 months (0.3 minutes/day, 95% CI –8.49 to 9.09 minutes/day; p = 0.947). No differences were seen on weekend days at either 7 or 14 months. No significant differences in light PA were found during school or in after-school periods at 7 or 14 months (all p > 0.05).

Finally, the proportion of participants who achieved ≥ 60 minutes per day of MVPA on every day of the measurement period was investigated. To see whether or not this could have been affected by the number of valid days that the participants contributed, we also report the number of valid days provided at each time point for those who met the 60-minute target. This was 5.2 days and 6.2 days for the control group and the intervention group, respectively, at baseline; 5.0 days and 5.9 days for the control group and the intervention group, respectively, at 7 months; and 4.0 days and 5.1 days for the control group and the intervention group, respectively, at 14 months. The proportion in the control group who achieved ≥ 60 minutes per day of MVPA on every day of the measurement period was 2.3% at baseline, 1.9% at 7 months and 3.0% at follow-up. A similar pattern was observed in the intervention group. No significant differences were found between the groups in terms of meeting the guidelines at either 7 or 14 months.

Anthropometric measurements

Table 14 presents the results of the analysis for the obesity-related variables. The mean BMI z-score at baseline was within the normal range (i.e. < 1 SD) for both groups. No differences between the groups were seen for BMI z-score at either 7 (95% CI –0.06 to 0.07; p = 0.908) or 14 months (95% CI –0.06 to 0.09; p = 0.636). Similarly, no differences between the groups were seen for body fat percentage at either 7 (95% CI –0.83% to 0.52%; p = 0.656) or 14 months (95% CI –0.68% to 1.17%; p = 0.600). The mean change was in the positive direction for both groups.

Self-reported and psychosocial outcomes

Participants reported on their PA levels using eight questions from the PAQ-A tool. No differences in self-reported PA were seen at either 7 (95% CI –0.11 to 0.02; p = 0.141) or 14 (95% CI –0.15 to 0.05; p = 0.318) months (Table 14). Participants also answered questions on a range of outcomes with known associations with PA levels; these results are included in Table 15.

Overall, participants in both groups scored their intentions to being active as ‘likely’ with a mean of 5.0 out of 7.0 and 4.7 out of 7.0 for the control and intervention group, respectively. At 7 months, there was no difference in intentions to being active (95% CI –0.33 to 0.03; p = 0.094). However, at 14 months there was a significant difference (–0.21, 95% CI –0.37 to –0.05; p = 0.012) in participants’ intentions to be active, with the intervention group seeing a drop in their intention score of 0.4 (unadjusted) out of 7.0.

At baseline, participants rated regular PA as being important, with both groups scoring a mean rating of 8.0 out of 10. At 7 months, there was a significant difference in perceived importance (–0.42, 95% CI –0.66 to –0.18; p < 0.001), with the intervention group dropping by 1 (unadjusted) out of 10. At 14 months, there was no significant difference between groups (95% CI –0.41 to 0.15; p = 0.350) as by then both groups had dropped by 1 (unadjusted).

At both 7 months and 14 months, there was no difference in positive attitudes to being active (i.e. participants did not have stronger feelings about the positive aspects of PA as they were between ‘neither agree nor disagree’ and ‘agree’ at all time points). At both 7 months and 14 months, there were no differences in participants’ disagreement with the negative attitudes to being active (i.e. they did not feel any worse about the negative aspects of PA as they were between ‘disagree’ and ‘neither agree nor disagree’ at all time points). When attitudes were considered as a whole score, there were no differences in participants’ attitudes to being active at 7 (95% CI –0.10 to 0.06; p = 0.667) or 14 (95% CI –0.08 to 0.02; p = 0.267) months.

At 7 (95% CI –0.05 to 0.12; p = 0.445) and 14 (95% CI –0.11 to 0.01; p = 0.105) months, there were no differences in participants’ perceptions of PA support from their family. At all time points, the unadjusted values for both groups lay between ‘disagree’ and ‘agree’ (i.e. 2.7 out of 4.0). At 7 (95% CI –0.12 to 0.10; p = 0.880) and 14 (95% CI –0.07 to 0.09; p = 0.762) months, there were no differences in participants’ perceptions of PA support from their peers. At all time points, the unadjusted values for both groups lay between ‘disagree’ and ‘agree’.

There was no difference in how participants perceived their school physical environment at 7 months (95% CI –0.10 to 0.15; p = 0.684). However, at 14 months there was a significant difference (–0.13, 95% CI –0.24 to –0.01; p = 0.032) between groups: participants’ perceptions of their school physical environment decreased slightly more in the intervention group than in the control group. At all time points, the perceptions scores were on the positive side, lying between ‘not sure’ and ‘agree’. There was no significant between-group difference in how participants perceived their school social environment at 7 (–0.01, 95% CI –0.10 to 0.07; p = 0.758) or 14 (–0.07, 95% CI –0.15 to 0.01; p = 0.080) months. At all time points, the perceptions scores were on the positive side, lying between ‘not sure’ and ‘agree’.

There was no significant between-group difference in how participants perceived the support from their PE teacher at 7 (–0.03, 95% CI –0.23 to 0.16; p = 0.733) or 14 (–0.21, 95% CI –0.45 to 0.02; p = 0.074) months. At all time points, the perceptions scores were on the positive side, lying above ‘neutral’.

There was no significant difference in participants’ confidence in being active at 7 months (–0.02, 95% CI –0.09 to 0.04; p = 0.472). However, at 14 months there was a significant difference (–0.08, 95% CI –0.14 to –0.02; p = 0.013) between groups, with the confidence of participants in the intervention group decreasing slightly more than the confidence of participants in the control group. However, this was at a magnitude of difference marginal mean of 0.1 on a scale of 1 to 5 points.

There was no significant between-group difference in participants’ enjoyment of PA at 7 (–0.03, 95% CI –0.14 to 0.08; p = 0.592) or 14 (–0.04, 95% CI –0.12 to 0.04; p = 0.074) months. The enjoyment scores from each measurement visit were generally ‘neutral’ (i.e. 3 out of 5 points).

Participant motivation to be physically active was queried and four domains of motivation were scored. At 7 (95% CI –0.11 to 0.07; p = 0.680) and 14 (95% CI –0.15 to 0.13; p = 0.361) months, there were no significant differences in extrinsic motivation. Values for extrinsic motivation lay between ‘not really’ and ‘not sure’ at all time points. At 7 (95% CI –0.11 to 0.102; p = 0.143) and 14 (95% CI –0.05 to 0.04; p = 0.260) months, there were no significant differences in introjected motivation. Values for introjected motivation lay between ‘not really’ and ‘not sure’ at all time points. At 7 months, there was no significant difference in identified motivation (–0.02, 95% CI –0.15 to 0.10; p = 0.700). However, at 14 months there was a slight but significant difference in identified motivation in favour of the intervention group (–0.09; 95% CI –0.18 to –0.0; P = 0.032). Values for identified motivation lay between ‘not sure’ and ‘yes a bit’ at all time points. At 7 (95% CI –0.18 to 0.18; p = 0.990) and 14 (–0.06, 95% CI –0.16 to –0.04; p = 0.243) months, there were no significant differences in intrinsic motivation. Finally, at 7 (–0.03, 95% CI –0.15 to 0.09; p = 0.578) and 14 (–0.02, 95% CI –0.14 to 0.09; p = 0.684) months, there were no significant differences in participant amotivation.

Participant physical self-perception was obtained through questions on three elements: self-esteem, physical self-worth and body attractiveness. A significant difference in self-esteem was seen at 7 months (0.06, 95% CI 0.01 to 0.11; p = 0.025) in favour of the intervention group but no difference was seen at 14 months (–0.06, 95% CI –0.13 to 0.01; p = 0.086). Scores for self-esteem were between ‘neutral’ and ‘agree’ at all time points. At 7 (–0.01, 95% CI –0.07 to 0.05; p = 0.742) and 14 (–0.03, 95% CI –0.09 to 0.03; p = 0.322) months, there were no significant differences in participants’ feelings of physical self-worth. Scores declined equally in both groups to ‘neutral’ by 14 months’ follow-up. At 7 (–0.05, 95% CI –0.12 to 0.03; p = 0.195) and 14 (0.02, 95% CI –0.05 to 0.08; p = 0.630) months, there was no significant difference in participants’ feeling of body attractiveness. Scores for body attractiveness were between ‘disagree’ and ‘neutral’ at all time points.

Unintended consequences

Participants answered ‘yes’ or ‘no’ to simple questions on smoking and alcohol consumption. The percentage of participants answering ‘yes’ to each question is shown in Figure 4. At baseline, 8.6% of control participants and 9.2% of intervention participants answered ‘yes’ to having smoked a cigarette, and 33.3% of control participants and 31.4% of intervention participants answered ‘yes’ to having had an alcoholic drink. At 7 months, 15.4% of control participants and 12.5% of intervention participants answered ‘yes’ to having smoked a cigarette, and 42.1% of control participants and 38.9% of intervention participants answered ‘yes’ to having had an alcoholic drink. At 14 months, 17.7% of control participants and 15.7% of intervention participants answered ‘yes’ to having smoked a cigarette, and 53.5% of control participants and 47.7% of intervention participants answered ‘yes’ to having had an alcoholic drink. No adverse events were reported in this trial.

FIGURE 4.

Proportion of participants who reported engaging in smoking and alcohol consumption at each time point. Adjusted intervention arm difference in MVPA (minutes/day) at 7 months.

Subgroup analyses

Tables 16 and 17 present the results of the prespecified subgroup analyses for the primary outcome; Figures 5 and 6 are the corresponding forest plots. At 7 months, the between-subgroup interaction effects revealed a significant subgroup effect (i.e. the intervention had a differing effect depending on certain baseline characteristics) in terms of participant ethnicity and biological maturity. For ethnicity, the intervention was effective in white European participants, for whom there was a significant difference of 3.3 minutes per day (p = 0.017) favouring the intervention arm, whereas there was no difference between the intervention and control arms for non-white participants. For biological maturity, the intervention was effective in early maturers, in whom there was a significant difference between randomised arms of 5.1 minutes per day (p = 0.003), whereas there was no difference between the intervention and control arms for participants in the average and late stages of maturity.

FIGURE 5.

Forest plot of the effect of the intervention at 7 months on the primary end point by subgroup. Subgroups are defined at baseline. Schools with a percentage of pupils eligible for FSMs lower than the median are considered to be the less deprived schools and those with a percentage above the median are considered to be the more deprived schools. Based on a complete-case analysis, including pupils who have worn the accelerometer for a minimum of 2 valid days with at least 1 school day at baseline and 7 months (8 control and 10 intervention schools, n = 1302). Adjusted for cluster effect, baseline MVPA value, stratification categories (school size and percentage of non-white pupils), year group and school percentage of pupils eligible for FSMs. FSMs, free school meals.

FIGURE 6.

Forest plot of the effect of the intervention at 14 months on the primary end point by subgroup. Subgroups are defined at baseline. Schools with a percentage of pupils eligible for FSMs lower than the median are considered to be the less deprived schools and those with a percentage above the median are considered to be the more deprived schools. Based on a complete-case analysis, including pupils who have worn the accelerometer for a minimum of 2 valid days with at least 1 school day at baseline and 14 months (nine control and 10 intervention schools; n = 1211). Adjusted for cluster effect, baseline MVPA value, stratification categories (school size and percentage of non-white pupils), year group and school percentage of pupils eligible for FSMs. FSMs, free school meals.

At 14 months, the between-subgroup interaction effects revealed a significant subgroup effect (i.e. the intervention had a differing effect depending on certain baseline characteristics) in terms of school size. In larger schools (≥ 850 pupils), there was a significant difference between randomised arms of 3.9 minutes per day (p = 0.001), favouring the intervention arm. In smaller schools (< 850 pupils), there was a significant difference between randomised groups of –4.5 minutes per day (p = 0.004), favouring the control arm. In addition, there was a trend towards a significant interaction with year group (p = 0.051 for interaction), with the intervention becoming effective in later year groups, reaching a between-group difference of 4.7 minutes per day (p = 0.008) for those in Year 9.

Diary response rates

Out of the 10 schools, four schools completed the Microsoft Excel diary weekly. One school did not implement any of the Girls Active components and also failed to provide any diaries. The remaining five schools provided their information through the survey. Three schools also provided logbooks when returning the survey.

Implementation-level categorisation

From the descriptions of what the schools did in the diaries and when contacting teachers, it was clear that implementation varied across the nine schools. Owing to the nature of the programme, which was not prescriptive and gave schools the responsibility to implement the programme as appropriate within their own settings, Girls Active was implemented in different ways by the intervention schools taking part. Given the diversity of implementation, which is common in public health interventions, the results of the microcosting are presented as three costings that are defined in terms of the nine schools that cost information was received from:

1. a minimum implementation costing (base case), which entailed creating new development plans and delivered the programme within the curriculum after engaging with female pupils

2. a mid-range implementation costing, which entailed creating new development plans and delivered the programme within the curriculum after engaging with female pupils, and included additional activities, such as after-school clubs and taster sessions (approximately 3 hours per week)

3. a highest implementation costing (in terms of the nine sample schools in the trial), which entailed creating new development plans and delivered the programme within the curriculum after engaging with female pupils, and included additional activities, such as after-school clubs and day trips and events (approximately 4 hours per week and four trips per year, respectively).

From a resources perspective, using the minimum levels of implementation (Table 18) by delivering Girls Active during curriculum hours would require the least amount of teacher time and may be considered the least costly option for a local authority; however, there are additional commitments required, such as engaging with pupils, developing a peer leader group and action plans and, once these are undertaken, time devoted to the programme is confined to timetabled PE lessons, thus minimising resources. In contrast, devoting additional time and resources as described in the mid-range and highest levels of implementation tables (Tables 19 and 20), for example using lunch hours and after-school clubs to deliver further taster sessions and one-off events, may be considered more costly options by a local authorities; however, these models of implementation align more with a pupil perspective and take into account the programme’s philosophy, which aims to improve pupil engagement in PE and change PE and exercise culture within schools. These implementation models are presented to demonstrate what happened in the trial, and give schools and local authorities wishing to roll out the programme an indication of the time and resources required. As the programme is not prescriptive, levels of implementation in real-world contexts will depend on number of staff members, staff time, availability of staff to share and delegate tasks related to the programme among each other and their peer leader teams, availability of funds for equipment and activities requested by pupils and support from senior members of staff to undertake new activities. A further description of implementation barriers and facilitators can be found in the results of the process evaluation in Chapter 6.

The nine diaries received were classified into the three categories: (1) three schools were categorised as having a minimum level of implementation, (2) four were categorised as having a mid-range level of implementation and (3) two were categorised as having the highest level of implementation. Please note that implementation levels are in the context of the sample of 9 schools. The mean training and delivery time based on the schools in each category is presented in Tables 18–20. A mean hourly rate of £26.00 (rounded to the nearest £1) is applied throughout the tables, which is based on the mean hourly rate of the salary information provided by the teachers completing the diaries. Tables 18–20 show a range of costing models that could be employed to deliver the programme at different levels of implementation. Costs ranged from £2054 per school (which is £23 per pupil based on an average of 90 pupils) to £8545 per school (which is £95 per pupil based on an average of 90 pupils), with incorporating Girls Active strictly in curriculum hours being the least costly option. Staff time accounted for the largest costs in the costing. Schools received £1000 of funding from the YST, and in some cases they received funding from their own school budgets. Data from the process evaluation case studies (see Chapter 6, School case studies) indicated that many (6 out of 10) of the schools did not actually spend the capacity funding (one of these sought money from internal school funds), and one spent only ‘a small bit’. When the schools did spend their funding (3 out of 10), it was generally used to purchase new equipment and deliver activities requested by the girls [e.g. to purchase dodgeballs and Zumba^® (Zumba Fitness, LLC, Hallandale Beach, FL, USA) kits]. Further funding supplied, which was not received from the YST as part of capacity building, was from within public sector funding, typically the schools’ own budget and funding. When conducting exploratory subgroup analyses, the level of implementation was not shown to have a statistically significant effect on minutes of MVPA per day, CHU-9D utility index scores or total frequency or costs of service use (see Table 29).

Opportunity costs were considered in the microcosting. 82 Opportunity cost is defined as the value of benefits forgone by not using resources in their next best alternative use. As Girls Active was delivered in usual school hours and replaced previous PE activities with those specified by the pupils, this resulted in minimal opportunity costs for the schools taking part in the trial. It is worth stating that in order to deliver Girls Active, teachers’ usual activities were sometimes displaced; these included planning, preparation and assessment time, personal time outside working hours, overseeing social clubs or activities and free periods. On average across the nine intervention schools, teachers experienced 6 hours of displaced activity over the course of the year. Another factor related to opportunity costs is fees for training. Tables 18–21 state that training was provided by YST to schools without a charge to schools; however, this training does incur a charge estimated by the YST in January 2017 as £300 for tutor costs, £30 for tutor travel and £50 per school for the resources pack, although these may be subject to change. Venue hire and lunches may also incur a charge, but no estimate was provided by YST, as this will depend on the training venue and their related charges for these services.

Health economics sample

A total of 1752 participants were recruited to the trial. For the economics analysis, we excluded participants who had missing data for both costs and outcomes, as economic evaluations require complete cases from both costs and outcomes. 85,86 Of the 1362 participants who returned for the 14-month follow-up, 1163 had CHU-9D data at both baseline and 14-month follow-up, and 1157 had service-use data at both baseline and 14 months post baseline. After participants with missing data were removed, and participant ID numbers were matched across time points, the final sample for the economic complete-case analysis was 997 participants (intervention, n = 570 and control, n = 427). Table 21 presents the characteristics of the complete-case economic sample at baseline.

Tables 22 and 23 shows the results from the xtgee models, demonstrating that the intervention did not have an effect on the main outcomes of MVPA per day, CHU-9D utility index score or service use for the economic sample; however, characteristics, such as results at baseline, school size and percentage of BME pupils, did have an effect on the results, adding weight to the cluster trial design and therefore justifying their inclusion in the xtgee model.

The intervention group and the control group reported similar marginal means (means calculated following the xtgee model taking clustering into account) in terms of the three outcomes (MVPA/day, CHU-9D utility index scores and total service use). The control group reported a slightly higher marginal mean of total service use frequency and related costs; however, this cost was only £22 higher in the control group than in the intervention group.

Both the intervention group and the control group participated in an average of 42 minutes of MVPA per day. The intervention group reported a mean CHU-9D utility index score of 0.821, which is slightly lower than the mean utility score of 0.835 reported by the control group; however, these differences were not statistically significant (see Table 26). Compared with the control group, the intervention group had slightly lower frequencies of contacts with health care and school services [two contacts (intervention) vs. three contacts (control) over several months] and associated costs of service use [£94 (intervention) vs. £116 (control) over several months]; however, this difference was not statistically significant (see Table 23). After conducting the xtgee models, ICCs were calculated for the three outcome measures. An ICC of 0.198 was found for MVPA per day, an ICC of –0.000 was found for CHU-9D utility index score, and the ICCs were –0.008 and –0.005 for frequencies of total service use and costs of total service use, respectively.

Table 23 shows no significant differences between the marginal means of the intervention and control group for the three outcome measures and costs as indicated by the bootstrapped CIs (bootstrapping was used to create 1000 valid bootstrap replications).

Service use

Tables 24 and 25 report the results from xtgee models of the individual services. As in the main analysis reported in Tables 22 and 23, the intervention did not have an effect on service use when divided further into individual services. GPs were the health professionals that the participants were most likely to see. The intervention group and control group reported similar marginal means when service use was broken down into individual services. Marginal mean frequencies and costs of service use were similar between groups; however, the control group reported slightly higher marginal means of service use frequency and related costs. Differences in costs are most noticeable in the reporting of school counselling frequency and costs; however, this cost was only £10 higher in the control group than in the intervention group.

School absenteeism

Participants in the trial were asked to report the number of days in which they had been absent from school in the preceding 7 months. Table 26 reports the results from xtgee models of absenteeism. The intervention was shown to not effect absenteeism for the sample.

Sensitivity analysis

As the intervention follow-up period was > 1 year, as part of the sensitivity analysis, service use costs at 14 months post baseline were discounted at 3.5% (the base rate recommended by NICE in 201366), and the model was re-run with these discounted costs (Table 27). Table 27 demonstrates that the sensitivity analyses replicated the main analysis results, showing no effect for the intervention, but an effect of baseline costs and constant. The discounted marginal means reported are similar to the undiscounted marginal means reported in Tables 23 and 25.

Furthermore, as part of the sensitivity analyses, missing data were imputed following methods employed in the main analysis for service use, which increased the economics sample size to 1752 (intervention, n = 861; control, n = 885). Table 28 summarises the xtgee model results from the imputed data. As in the complete-case analysis, the intervention did not have an effect on total frequencies and total costs of service use, even when using imputed data. CHU-9D utility index scores were not imputed because the developers of the measure state that utility values cannot be calculated when questions have missing answers. 63 Data on MVPA per day were not imputed as this was previously done by the main analysis.

Exploratory subgroup analysis

As part of an exploratory subgroup analysis, the effects of year group and levels of implementation, which were based on the levels described in the microcosting, were tested on the complete-case sample by being included in the xtgee model run in the main analysis. Table 29 shows that year group was found to have a statistically significant effect on minutes of MVPA per day, but did not affect any other outcomes. Table 29 also shows that the level of Girls Active implementation did not affect any of the three main outcomes. Table 30 reports the marginal means of the subgroup analysis: pupils in Year 7 reported the highest marginal mean minutes of MVPA per day (47 minutes in the intervention group and 46 minutes in the control group) compared with pupils in Year 8 (41 minutes in the intervention group and 40 minutes in the control group); pupils in Year 9 reported the lowest means (36 minutes in the intervention group and 33 minutes in the control group).

School case studies

Short case studies that outline the engagement of each school with the intervention components are presented in Boxes 1–10. It is evident from these case studies that engagement varied between schools, and generally all schools were behind their set timelines for implementation and did not achieve all the actions planned during the training events. These case studies will be referred to throughout this chapter.

Training days and events

Over the course of the 14 months, there were a number of training days and events that were run by the YST and the hub school (Figure 7).

FIGURE 7.

Girls Active training timeline.

Mission analysis and action plans

A mission analysis self-review form was required to be completed prior to the initial training day for lead teachers, with the purpose of enabling schools to review their sport, PE and PA provision and identify priority areas for Girls Active implementation. Following the initial teacher training, teachers were then requested to submit an action plan to specifically outline what they planned to do and when. They were then required to submit another action plan approximately halfway through the programme. The capacity funding (£1000) that the schools were due to receive was contingent on the schools submitting their action plans (£500 was received on receipt of action plan 1 and £500 on receipt of action plan 2).

Most of the teachers mentioned that they found the first action plan challenging to complete (in terms of the Microsoft Excel document format) and time-consuming, with some teachers reporting that too much time was spent completing paperwork rather than being able to actually get on with implementing the programme. They felt that more guidance and/or support, especially within the first few weeks following the initial training session, would have been appreciated; for example, having someone look at their action plan to tell them if they were on the right lines:

Hard work to be very honest. Oh it’s taken ages to complete and sometimes we look at it and we go does that really mean that? Yeah it’s taken a lot of time to go through.

Teacher, 7 months, school 2

I didn’t understand really, what they were asking. So like, so kind of strategies into action, does your plan drive everything you do? What plan? What kind of things do you mean by that?

Teacher, 7 months, school 9

The format of action plan 2 was modified following comments made about the action plan 1, and teachers commented that they found it considerably easier to complete and felt that overall the questions flowed better; however, it did still appear that there was some confusion about the meaning of different sections:

The action plan that they gave us the second time around was a lot easier . . . that’s what I want to do and this is where we’re going and it was as easy that . . . but with the confusion of the action plan at the start to the next one you’ve got is something that we’ve been able to put on a second action plan that we didn’t put on the start, but should have been on the start. I think if they were, we’d have completed them in a better way or we’d have done them sooner.

Teacher, 14 months, school 6

The difficulties reported regarding completing action plan 1, along with some teachers being unsure of exactly what they needed to do following the training day, may have contributed to the late submission of the action plan to the YST by several schools, which had a knock-on effect on when the first capacity funding instalment was received. A YST representative also acknowledged that the delay in submission of the mission analysis may have been attributable to a lack of understanding:

I suppose the communication side of things. So when we’ve asked schools perhaps to link back in terms of the action plans I think in some cases it might have been a little . . . not lack of understanding, but not quite sure on certain tasks.

YST representative, 14 months

Appendix 11 summarises when each school submitted their action plans and when they received their capacity funding instalments. Action plan 1 was due before the end of the school summer term (i.e. July 2015) and action plan 2 was due in January 2016. Although all intervention schools did submit both of their action plans, four schools did not submit action plan 1 until after the 7-month follow-up evaluation visit (8 months after the initial training), and hence they did not receive their first funding instalment. Four schools also did not submit action plan 2 until the time of the 14-month evaluation visit. As a result, there was some delay in receiving the funding prior to the follow-up visits.

Resource folders

The teachers were given a folder of resources at the initial training day; these contained marketing plans, case studies, images and a branding toolkit. Many of the teachers mentioned how useful the resource folder was for generating ideas, particularly from the case studies, and that they used the content with the peer leadership group to help them understand what they needed to be doing and to facilitate the girls working independently:

The resources that we receive are very, very useful I think that enabled the girls that had taken part as the girls group to actually do everything themselves so therefore I’ve not . . . they’ve come to me and said ‘shall we do this?’.

Teacher, 7 months, school 7

The case studies were really good, it was nice to see all the schools were in a similar boat . . . because you are quite isolated in the school, you know, other than the bit of networking in the sense of PE, but you know, those sort of niggles with your kit, uniform, participation, those general day-to-day things. There were some nice resources to use for assemblies to start with as well.

Teacher 14 months, school 3

Although the resources were felt to have been a good starting point, the teachers reported that later into the programme they had not continued using them as they had developed their own resources or had other resources they could use. Furthermore, some of the teachers felt that, although the case studies were good, they did not necessarily reflect their school context or pupils and so were not relevant:

At the very start of the project, when you’re looking at the very first action plan and looking at kind of cases and things . . . you know, it kind of . . . OK, what can we do, but other than that, it’s not been used at all.

Teacher, 14 months, school 6

We used a lot at the start, not really at all towards the end.

Teacher, 14 months, school 2

Some of the case studies are really good and some of the ideas are good, but wouldn’t fit in this school. Some of them are good, I’ll nick that bit of that one, but I’ll take that bit of that one and then merge them together.

Teacher, 14 months, school 4

Dropbox and newsletters

Teachers had access to an online Dropbox^™ (Dropbox, Inc, San Francisco, CA, USA) folder to share ideas and resources. They were asked if and how they had accessed this Dropbox folder. It appeared that this resource was underutilised. Teachers had either forgotten about it or had accessed it but felt that it was not being updated regularly enough:

There’s a Dropbox account. I have looked in there a couple of times but no one seems to have put anything in. That’s a bit of a shame because . . . and I hold my hands up, I haven’t put anything in there.

Teacher, 7 months, school 8

I know there’s the Dropbox, again, it’s trying to get the time to even think about getting into the Dropbox to look.

Teacher, 7 months, school 6

A monthly newsletter was also sent to lead teachers from the YST, which gave links to material in the Dropbox: relevant news headlines; reports and studies about school PE, sport and PA; and tasks and challenges to prompt teachers to think about how to link materials to practice in their school.

Peer leadership and marketing group

The majority of schools (8 out of 10) established a girls’ peer leadership and marketing group; there was some variation on when this was done during the intervention programme (i.e. some schools set up these groups early in the programme, whereas other schools had only recently established them by the time of the 14-month evaluation visit). Two schools did not establish a peer leader group: one school did not engage with the intervention implementation because of confusion about the purpose and, at the other school, the teacher felt that it would be quicker to do tasks herself, although she did acknowledge that this was a mistake. More detail for each school is provided in School case studies.

What was implemented in the schools?

The individual school case studies detailed in School case studies outline what each school implemented and when implementation took place during the programme. All schools except one had implemented a series of activities for the Girls Active programme throughout the 14-month time frame. Of these, all but one had commenced some form of implementation by the 7-month visit. The implementation activities could be grouped into (1) support activities and (2) provision of PA opportunities. From the case studies, it is evident that most of the activities implemented were support activities. These activities are ones that are aimed at finding out what girls in the school wanted or are aimed at beginning to target cultural change.

By 7 months, half of the schools (n = 5) had conducted some sort of canvassing of opinion from girls within the school through questionnaires or face-to-face questioning (three other schools also did this but not until after the 7-month visit). This enabled the schools to understand what changes the girls wanted to see in the school for PE, sport and PA. During this time, six schools had set up their peer leader groups (however, two schools had established them only just prior to 14 months). As a result of the questionnaires, three schools then made changes to their PE kit policy or sports kit. Additional support activities included marketing to launch and promote Girls Active (e.g. assemblies, posters, bulletin boards, banners and purchasing branded hoodies for leaders).

By 7 months, two schools had introduced taster sessions (either one-off sessions or over several weeks) or new activities or clubs in which, in some cases, an external coach was brought in; however, in the other schools, these were introduced late in the intervention cycle, either just prior to 14 months or the teachers were planning to implement these in the next term. A number of schools also delivered ‘one-off’ events.

Five schools had also opted to run a rewards scheme for attendance at clubs: pupils could gain points for attendance and then receive rewards when they achieved certain milestones. Similarly, this tended to happen at some point between the 7- and 14-month evaluations, with the exception of one school that started it prior to 7 months and continued into 14 months and another school that was in the planning stages at 14 months. The schools that chose to deliver the rewards scheme offered this to girls only, except in one school in which it was delivered as a whole-school initiative and any extracurricular club could be attended to gain points (i.e. it did not have to be a physically active club). These rewards included a ‘queue jump’ pass at lunch, free sports kit, free trips (e.g. to Planet Bounce) and free snacks. In most of the schools, the peer leaders had decided on what type of rewards were on offer. The rewards scheme seemed to be a popular choice among schools, and teachers reported that this really encouraged girls to take part in activities/clubs (see Impact of Girls Active for more detail on impact).

Although it is positive that all schools except one had implemented some Girls Active activities, there was variation among the schools in terms of how much had been implemented within the 14-month programme. It was evident from the interviews and the mission analyses that the schools had not achieved as much as they had planned to do in the time frame. This was also acknowledged by the YST development coach:

To be honest they’re different to what I thought they were going to be and I think that’s because, through no fault they’re either one reason or another the schools haven’t been as far down the route as I would’ve expected them to have been.

YST development coach

Ongoing support available from hub school and Youth Sport Trust

Teachers had access to ongoing support and contact throughout the programme in the form of monthly newsletters, a dedicated mailbox for them to contact with queries as well as telephone calls and face-to-face visits. The telephone calls and visits were optional offerings from the hub school and YST, and not all schools accepted; in particular, the telephone call opportunities were not used by some schools.

Funding

When the schools received their funding instalments is detailed in Appendix 11, Table 39. Six schools received the first funding instalment by the end of the summer term in 2015 (July), which was 3 months after the initial training day. However, as mentioned, several schools struggled to complete their mission analysis and this delayed receipt of funding: these schools did not receive their funding until late 2015/early 2016 (8–9 months following the initial training day). One school did not receive its first instalment until May 2016. The second funding instalment was received in January 2016 by six schools; the remaining four did not receive their instalment until May 2016, which was around the time of the schools’ 14-month evaluation visit. It is evident from these dates that funding had often been received late in the programme cycle and, therefore, there had been only limited opportunity for spend and subsequent potential impact.

The capacity funding was viewed positively, especially at a time when school budgets were (and still are) tight. In some cases, being part of the Girls Active programme had even helped schools to apply and receive additional funding for the Girls Active activities. Even though the schools were positive about the funding, the majority of them spent none, or only a very small amount, of the funding during the 14-month programme. Some schools reported that because they rarely received money they wanted to keep it for the future to ensure sustainability. Others felt that they should clearly plan activities within their school before spending the money and this takes time:

It’s definitely something that it allows us to have more scope. So we’re quite tight on budget anyway and even things like printing and stuff, is getting stricter and stricter. And it’s just for me, to be able to go ‘well it could go in the Girls Active fund’. And then so, my head of department’s a lot more like ‘oh, OK, yeah, if that’s . . . you go ahead then’, rather than having to, not fight for it. It has allowed us to really, just be more accessible really, and allow us to kind of think a bit bigger.

Teacher, 7 months, school 9

If we can make that money last for a long period of time, then that is, you know, if we are only using a real small portion of it per year, we would make it last maybe 2 or 3 years.

Teacher, 14 months, school 7

We’ve been very lucky that we’ve got this money through the Girls Active, but we want it to last as long as possible. Once that money has gone, it’s not then sustainable, so we’ve only had one coach in. We’ve only spent a little bit of money, but it means that in September we can get another coach in, so it’s like spreading it out and we can sustain this now for a couple of years.

Teacher, 14 months, school 6

The schools that did spend some of the money reported spending it on printing materials for promotion, hoodies for the peer leaders, notice boards, a clock for the girls’ PE changing rooms and external coaches. Only one school spent all of the money, on external coaches and travel to Girls Active training events.

Impact of Girls Active

Facilitators of implementation

Teachers mentioned a variety of support factors that helped them either implement or reprioritise the programme. Specifically, support from the PE department and their senior leadership team helped provide an environment in which the lead teacher could implement the programme. Furthermore, a couple of the schools specifically mentioned that the development coach’s visits had acted as prompts to get things done. These visits were viewed as very helpful and useful for motivation, clarification of direction and general support. Finally, in some cases the data collection visits themselves acted as deadlines that encouraged action:

It’s not just coming from me telling them, it’s coming from an outside person going, ‘oh we need to make sure . . .‘ then I can say to them, ‘look they’re coming in, this needs to be done’.

Teacher, 14 months, school 2

If you just came to me now and kind of gave me a load of resources and the booklet, I don’t know whether I would . . . it was more the pressure of kind of having the three kind of dates coming up and having . . . and it’s coming in, so I had a bit of a . . . something to work towards. When you’ve got so many other demands within the school it’s like you kind of need those kind of guides, like dates to hit as such and potentially having something to work towards, it does definitely make a difference.

Teacher, 14 months, school 9

The assistant head’s actually very, very, good and he’s completely on our side. We proposed this whole idea to him and he said, whatever you need go for it, just make sure that you’re not too elaborate. He gave us . . . well it was about £600 for the initial trip to Planet Bounce for 30 students.

Teacher, 14 months, school 8

Just general support really within the department . . . you know, they’ve all been really supportive of what we’re doing.

Teacher, 14 months, school 6

Challenges to implementation

It is useful to identify the challenges that teachers faced in order to understand why some schools struggled to undertake some components of the programme. Some challenges may be specific to Girls Active; however, other challenges have important implications for other school-based programmes, which aim to make a culture shift within a school using a flexible approach.

Sustainability

Views on continuing the programme in the future were generally positive; however, some teachers did mention the role of the capacity funding within the programme and the possible impact on the programme when that money runs out. Others acknowledged that the programme was low cost and it was more about ensuring that the programme is embedded into the culture of the school without being reliant on the funding. Capacity funding was not always spent for logistics reasons (i.e. the delayed timing of payments), but the desire to keep the programme going in the future also influenced spending patterns. Developing a rolling recruitment of peer leaders each year and keeping the peer leaders interested were also mentioned as important for the programme moving forward:

Yes, I do think it is sustainable because it only costs such a small amount of money. If we can make that money last for a long period of time, then that is, if we are only using a real small portion of it per year, we would make it last maybe 2 or 3 years.

Teacher, 14 months, school 7

It’s down to making damn sure that you do it, if you want it to be sustainable you’ve got to . . . it’s down to us and the girls to make sure that that’s what it is and if we want it we’ve got to work hard to do it. I don’t think there’s anything tangible that we could add that would make it one way or the other, it’s just got to be down to willingness to do it.

Teacher, 14 months, school 1

It’s more about status and less about money isn’t it, it’s about the status and doing things rather than actually having the physical money to buy things. It’s more about what’s the word, presence, isn’t it?

Teacher, 14 months, school 3

I think now it’s been kind of put in place, I think the Miss Motivators are sustainable with our four going into each year. It’s just about keeping them like occupied with certain jobs . . . we’d have to make sure we actually do have set kind of meetings, set agendas and things like that to hit.

Teacher, 14 months, school 9

Teachers’ overall views of the Girls Active programme

Overall, the Girls Active programme was seen in a positive light. Teachers seemed to view it as a worthwhile programme and a great opportunity for their school. Teachers reported that they would take part again if the opportunity arose, while acknowledging the need for more people to help within their school:

I’d definitely get involved again. There’s been a few admin/technical issues that we probably need to iron out but, yeah I’d definitely get involved again. It’s given the girls a bit of a lift. It’s about them as well, it’s quite important that the girls have had a chance to show enthusiasm and interest for girls’ sport.

Teacher, 7 months, school 3

We thought it was a brilliant project that our girls would be on board with this and it was a good thing for the school to be honest . . . [Interviewer: would you do it again if you were . . .] Yes, definitely, yeah.

Teacher, 7 months, school 2

Oh, I would do everything again the same, and set up similar to how I have done this time. Because I do feel that it’s probably impacted them better than I thought it would do. It might just be that you have more staff, so you could spread the workload, maybe make it even more, kind of, focused. So, if there is two of you, one of you in charge of 7 and 8 [year groups], one of in charge of 9 and 10 [year groups]. So, you can really focus on what they needed. But, yeah, I definitely would do it again.

Teacher, 14 months, school 4

Honestly I think Girls Active project is such a good one. I think because anything we can do to get girls involved in PE is so important and I think it’s just the way to do it like getting the girls to do it themselves so that it spreads the word that way; really good.

Teacher, 14 months, school 5

Control schools

The teacher responsible for the study in each control school (i.e. evaluation) completed a school environment questionnaire. They were also interviewed at 7 and 14 months to explore whether or not any changes had happened in the school that might have had an impact on girls’ PA, for example any changes to PE, new sports and activity clubs at lunchtime and after school, new sports facilities or equipment, or involvement in new initiatives. All schools except one reported some changes in their school within the 14-month programme. Two schools (schools 15 and 16) reported using This Girl Can resources (e.g. posters and quotes) in the school at 7 months. Two schools (schools 11 and 16) also reported taking a small selection of girls to 1-day activity events in which they took part in cheerleading and taster sessions of different activities. One school (school 13) reported taking part in a Women in Sport and Physical Activity (WISPA) event at 7 months, at which girls in the school were offered 6 weeks of Zumba classes. Another school reported taking part in a NHS Healthy Minds project that aimed to improve the physical, social and emotional well-being of adolescents. This project appeared to have been offered to schools in the Northamptonshire county. Two schools (schools 16 and 17) reported reductions to their PE allocation for certain Year 10 pupils at the 7-month time point. This would only have the potential to have an influence at 14 months, when the pupils who were in Year 9 at baseline were followed up; 6 out of the 10 control schools included Year 9 pupils. At 14 months, four schools (schools 13, 14, 16 and 18) reported the introduction of new clubs/activities. One school reported a new girls-only after-school club for Year 8 pupils at which they could choose to go any sport/PA, another school reported a lunchtime boxing session, one school had carried on with Zumba classes because of their popularity when delivered as a 6-week taster prior to the 7-month time point and another school reported that it had started lunchtime clubs (having not previously provided any lunchtime activities). The changes noted above are evidence of the usual evolution of offerings within schools. These alterations and additions have the potential to influence the PA levels of girls in those schools and could therefore mask any effect of Girls Active in the intervention schools.

Main trial design

A major strength of this evaluation is the robust methodology. This was a fully powered cluster RCT in which randomisation took place at the school level after baseline assessments. Allowing Girls Active to be implemented at the school level (rather than at the class level) was more ecologically valid (as it is a school-based programme). For the evaluation, a random sample of 80–90 girls per schools were included. Depending on the school size, this ranged from 25.9% to 78.2% of KS3 girls (mean 37.2%). Logistics and cost made it difficult to include all KS3 girls in the evaluation. Given that teachers are encouraged to embed Girls Active within the whole school, we felt that it would have had an impact on any girl in KS3 if this objective had been met; however, the reality is that in the shorter term the programme seems to have more focus and a narrower targeted evaluation sample may have yielded different results. Although our 14-month follow-up is longer than many school-based programmes, the analysis may have benefited from a longer time frame, assuming that schools were still committed to delivering elements of the Girls Active programme in the longer term. Our sample size accounted for a loss of two schools (we were unable to measure two schools at 7 months and one school at 14 months) and a 30% non-compliance/loss to follow-up over the course of the 14-month evaluation. At 14 months, we had 69.1% of the baseline sample to analyse, which is slightly lower than our power calculation accounted for.

An objective measure of the primary outcome was a major strength. 92 We chose a wrist-worn rather than a waist-worn accelerometer to maximise wear among adolescents. The compliance with wrist accelerometer wear was excellent at each time point, with 96.0% and 93.5% of the seen/available sample providing at least 1 day of data at the 7-month follow-up and 14-month follow-up, respectively. This compares favourably to previous large studies of a similar age group93 and also to the average missingness from young people’s accelerometer studies of 37.4% (SD 20.2%, range 3.3–75.4%). 94 These previous studies used non-wrist protocols so it would seem that our variety of strategies of using the wrist placement, an incentive for device wear (£5 voucher) and a full explanation of the device to the students resulted in great compliance with the protocol.

Novel aspects of applying health economics to the Girls Active trial

There is a lack of cost-effectiveness evidence surrounding PA interventions for children and young people. 80,81 This cost–consequences analysis adds to the literature and provides results from a UK context. As mentioned previously, cost–consequences analysis is championed as a method that is particularly relevant to economic evaluations alongside public health interventions. 83,84,95

Conducting economic evaluations alongside public health interventions has been described as challenging, and often requires a pragmatic approach. 84,95 The difficulties faced by teachers in completing the Microsoft Excel version of the cost diary, which was essential for the economics analysis to quantify the resources required to deliver the Girls Active programme, raised a particular challenge in this trial. We took a pragmatic and iterative approach to designing measures to collect this information, adapting the Microsoft Excel cost diary to a survey and logbook to increase response rates. This adaptive approach to the economic data collection yielded further responses from the teachers and resulted in 9 out of the 10 intervention schools providing data for the microcosting analysis.

The intervention was not prescriptive in nature. It did not dictate how schools should deliver Girls Active nor the number of hours teachers should spend recruiting peer leaders, reviewing current activity and developing new lesson plans. This presented the challenge of displaying the costs of the intervention in a meaningful way, which reflected what happened for the intervention schools taking part in the trial, but could also be used in the future by local authorities to indicate how Girls Active could be rolled out in their locality. Owing to the adaptive approach taken to collecting cost information, we had a wealth of data to draw from. Furthermore, by undertaking follow-up telephone calls with the teachers who completed the cost diaries, we had first-hand accounts of their experiences, which provided context to the information received. This context was instrumental in helping the researchers to make the decision to present the data as three costings, using the different levels of teacher activity and additional hours outside timetabled PE lessons, which we believe provides useful information for those interested in delivering Girls Active.

Process evaluation

An extensive process evaluation, following the guidance for process evaluation of complex interventions,96 was conducted, collecting data from a variety of sources and using a variety of methods. We used feedback forms, interviews, focus groups, training records and observer notes to build up an accurate picture of the how the Girls Active training was delivered and the how the intervention was implemented. As Girls Active is designed to be a flexible programme, there is an infinite number of ways that teachers, pupils and schools could decide to implement the programmes and make PA changes in the school; our variety of methods were able to account for this flexibility. The school-specific case studies in Chapter 6, School case studies, have built up a full picture of what the programme can actually look like in schools. These case studies were informed by a variety of subjective and objective sources. Our TSC lay members provided valuable contextual information on the policy and practice climate in which Girls Active is being implemented. They suggested questions and themes that were added to the interview and focus group schedules, and the addition of a boys’ focus groups was also a suggestion of our engaged TSC.

Generalisability

The generalisability of the findings must be considered in terms of whether or not the setting (schools) and the recipients (participants) of the intervention were representative of the area and the country from which they were drawn. For this study, 82 schools were contacted to yield 20 schools that entered the trial. The area in which this study was undertaken was deemed to be suitable because it included a mix of urban and rural schools, a range of socioeconomic levels and schools with varying percentages of non-white pupils. When the characteristics from Table 5 are considered, it would seem that there was a higher proportion of larger schools (≥ 850 pupils) in the trial (70.0%) than the English average (57.8%). Girls Active schools had slightly more non-white pupils (23.5%) than the English average (20.9%) and the percentage of pupils eligible for free school meals (11.5%) was lower than the English average (14.8%).

In terms of the participant characteristics, the proportion of participants who were overweight/obese at baseline (28.2%) was lower than the average values for England (32.2%; Year 6, 2015/16). 97 In the Girls Active sample, 5.0% were classed as underweight, compared with the English average of 1.5% of Year 6 girls (at 9/10 years old, these girls are younger than those in the Girls Active sample). One of the reasons for this could be that the sample contained a large proportion of South Asian children (11.7%) who, based on National Child Measurement Programme data (2015/16), have the highest prevalence of being underweight at Year 6. 97 In the present sample at baseline, the proportion of underweight participants was 9.5% among South Asian girls compared with 4.6% among white European girls.

Sources of potential bias

We do not have any information on the girls who, although randomly chosen as part of the first 90 girls, did not provide assent or have their parents/guardians return the opt-out consent form. However, those who were the least active, older and had higher BMI z-scores at baseline were more likely to opt out of the two follow-up visits. This is unsurprising and is a limitation of many trials. The large proportion of white European participants who were lost to follow-up came from the school that was uncontactable at 14 months and the school that opted for a slimmed-down version of the measurement day. Therefore, of the 390 participants (22.3%) who did not complete the 14-month assessment, 89 were from the school that was uncontactable and 28 were from the school with the shortened measurement visit; these 117 participants actively decided not to return for a follow-up visit. There may have been selection bias in the subgroup of girls and boys chosen by the teachers for the focus groups; however, this was done in an effort to find a range of activity levels and, also, to find pupils who would be willing to speak up in these situations. Their views may not be representative of all the pupils in their school.