The study was funded by the New South Wales (NSW) Department of Education and Training (who are responsible for all government schools) through a competitive tender process and is registered with the Australian New Zealand Clinical Trials Registry (ACTRN12610001077055). The University of Wollongong Human Research Ethics Committee approved the study protocol (HE08/296).
Setting and population
This study is a partnership between the School Sport Unit of the NSW Department of Education and Training (who is responsible for the promotion and implementation of school sport within the Department) and NSW universities including the University of Wollongong, University of Sydney, University of Newcastle, University of New England, and Southern Cross University. As the structure of school sport in NSW is unique, a brief explanation is provided. In NSW, government secondary schools are required to provide students in Years 7-10 with at least two hours of planned physical activity each week [18
]. This activity can be achieved through a combination of physical education and compulsory school sport. In regards to school sport, each school can develop and conduct their own unique sport program according to their student needs and interests, school resources and teacher expertise, availability of transport and community facilities. These school sport programs may include inter and intra-school competitive sport, recreational sport, fitness and/or skill development activities. Generally schools implement their sport program pattern in a number of ways: integrated, traditional, or staggered. Where an integrated sport pattern occurs, at least 80 additional minutes are to be incorporated into the Personal Development, Health and Physical Education (PDHPE) program. In traditional and staggered sport patterns, 80-120 minutes are allocated every week specifically to sport. Traditional patterns occur on one afternoon each week, whereas in a staggered pattern certain year groups will have sport allocated at certain times. In traditional and integrated sport patterns most of a school's teaching staff, including non-physical education trained teachers, are allocated sport as part of their teaching load [8
Aims of the study
The primary aim of the study was to test if an 18-month school-based intervention targeting school sport and physical education (through the formal curriculum), school ethos (including policies and school breaks such as lunchtime), and links with the local community, could prevent the decline in objectively measured physical activity compared with matched control schools.
School and participant selection
The NSW Department of Education and Training emailed all principals of secondary schools in NSW (N = 500), calling for Expressions of Interest to participate in the project. Interested schools were asked to contact the Department and then complete a profile for their school which was used to pair-match schools prior to randomization. This profile requested information about the school's population (number of boys and girls), proportion of students from non-English speaking and Aboriginal and Torres Strait Islander backgrounds, number of years teaching experience of the physical and health education staff, and how school sport and physical education were organised in the school. Schools were then matched based on these criteria in addition to their type of school (for example, girls only, central school [found in rural areas where primary and secondary schools are combined on the one site under the same school executive], and technology high schools [which help prepare students for the changing needs of today's society through providing specialist options in robotics, computing, and rural and marine technologies]) and their geographic location. A member of the research team was then assigned to be a 'critical friend' at each school. This involved working with the school as it collected and interpreted their own data and assisting with the development, implementation, and evaluation of the school's action plan [19
]. Within each school, all girls in Grade 8 (second Grade of secondary schooling) were invited to participate in the study.
Randomization and study design
This was a group randomized controlled trial with school as the unit of randomization. Each matched pair of schools was randomly allocated to the intervention or control group using a computer-based random number producing algorithm. This was undertaken by a researcher independent of the project and then communicated to the research team who informed each school of its allocation. Due to the timeframe of the study, formative research was required to be conducted in the intervention schools in Term 4 (October to December 2008). This meant randomization needed to occur prior to baseline data collection.
Inclusion and exclusion criteria
To be eligible for the study, schools were required to not be involved in any other physical activity-related research projects and to confirm that they consented to being randomized to either the intervention or control group if they were selected for the project. If selected in the intervention group, schools needed to commit to developing an action plan, forming a school committee, attending workshops, organizing students for data collection days and putting the project in the school management plan. Girls needed to be formally enrolled in Grade 8 within the participating schools and provided written consent from themselves and their parent(s)/guardian(s) prior to participation. If a student or their parent/guardian did not consent, they were still able to participate in the intervention activities but did not participate in data collection.
The formative research aimed to identify from the target group (Grade 8 girls) their needs and interests and the schools and community facilitators and barriers to their participation in physical activity. Interviews were held with relevant staff (including PE and non-PE staff and a member of the school executive), informal interviews with groups of boys, and focus groups with girls in Grade 8, during October and November 2008. A member of the research team attended along with the project manager, who collected the data. In addition, girls and PE staff were asked to map community physical activity facilities and opportunities on provided maps of the school and local community. Observations of PE lessons, recess and lunchtime activities were also conducted. Results suggested that there were two main reasons schools were interested in participating in the study. These were the chance to revitalise sport in their school and the opportunity to engage particular groups of girls who were currently not participating in PE or school sport. From the perspective of staff, the reasons for girls not participating in school sport were grouped into three themes: 1) the current structure of school sport (which lacked variety and limited choice, with those who were less skilled and confident being the last to choose a sport); 2) the lack of resources; and 3) the lack of expertise among non-PE staff who supervised school sport. Among the girls, the main reasons for their non-participation were similar to those reported by staff, with additional barriers identified including the 'dominating' behaviours of boys during PE lessons and sport and the girls' perceived lack of skills and confidence in traditional PE and school sport.
Girls were asked what they would like in a school sport program to enhance their engagement and participation. They suggested the following: the opportunity to choose some of the activities (especially non-traditional ones such as yoga, self-defence, and dance); participate with their friends; motivated and knowledgeable staff; more modern sports uniforms; greater cooperative behaviour from boys; and higher levels of activity during sessions (less time spent sitting around). It was not surprising that these factors have been cited as motivators for physical activity among adolescent girls [8
These formative data were then provided to each school as a report and school committees were asked to ensure they addressed as many of them as possible when developing their intervention strategies and action plans. For example, to give girls greater choice in the types of activities in which they could participate, staff were encouraged to survey girls to determine what activities they would like and then examine ways some of them could be integrated into their school sport programs. They were also advised of the importance of students being a part of the school committee so they had a 'voice' on the program in their school.
Individual schools, in conjunction with their critical friend, developed unique 18-month action plans to implement from mid 2009 to the end of 2010. The specific intervention strategies derived for each school was independently designed to achieve the overarching aim of the project. This aim was the same for each school, and was to prevent a decline in participation in moderate-to-vigorous intensity physical activity levels among girls over the course of the intervention. There were also six secondary aims that schools needed to work towards (see Table ). Each school followed an identical process in developing their intervention. This involved: 1) forming an action learning team (Committee) within their school community; and 2) developing school-specific action plans in three areas based on the results of the formative research and on individual needs of the school. These areas were the formal curriculum, the school ethos and environment, and home/school/community links [12
]. Schools were also asked to identify barriers to them meeting the outcomes identified in their action plans. Once these plans were written, each school commenced implementation, and were encouraged by researchers to continually reflect on their progress and modify the strategies where required. Support was also given to the schools in a variety of ways, including funding from the Department of Education and Training, an initial two-day training program, regular contact with the Girls in Sport project manager and research team, informal school surveys, as well as a two-day research symposium in February 2010.
Outcomes for the Girls in Sport Intervention and Research Project
The 12 control schools will continue with their usual program and schooling without change. At the conclusion of the project these schools will receive training and materials related to the findings of the project.
Data were collected within individual school settings. Baseline data were collected by trained measurement staff from February 2009 to June 2009. After this period, the intervention schools participated in the Girls in Sport program, while the control schools continued with their regular school sport programs. Subsequent follow-up data collection took place from July to December 2010. Data collectors were blinded to group allocation. Teachers and students at the paired intervention and control schools were also blinded towards their matched comparison school. To enhance the quality of the data across all collection sites, the research assistants were formally trained in standardized measurement procedures and protocols. Each research assistant was given a detailed manual, checklist and scripts to read when informing the participants about the measures. Data collectors checked for incorrectly completed questionnaires (i.e., pages or items not filled in) and invited participants to correct their mistakes or complete missing items. The teachers in the schools were also asked to follow up any absent participants.
Primary Outcome Measure
Objectively measured moderate-to-vigorous intensity physical activity was the primary outcome for the study. All participants wore an Actigraph accelerometer (7164 and GT1M models; Fort Walton Beach, FL) for seven consecutive days. This was attached to an adjustable elastic belt and worn over the right hip. Data were collected in 30-second epochs. The average number of minutes that the accelerometer was worn and the number of activity counts per minute (CPM) were calculated. Mean CPM as a summary measure of total physical activity in children has been validated against doubly labelled water [20
]. Thirty-second activity counts were uploaded to determine the amount of time spent in light (LPA; 1.5-2.9 METs) moderate (MPA; 4-6.9 METs) and vigorous (VPA; ≥7 METs) physical activity during the monitoring period. Age-specific count ranges relating to the above intensity levels were based on prediction equations for energy expenditure [21
]. Values were calculated for percentage of monitored time spent in light, moderate, and vigorous physical activity to account for variation in time spent wearing monitors. Participant data were included in analyses if accelerometers were worn for ≥600 minutes on ≥3 days [3
All participants were asked to keep activity monitoring logs for the seven-day period when the accelerometers were being worn. Participants also received two text messages during the seven-day period. These text messages reminded the participants to keep wearing the accelerometers and to return them at the end of the seven-day period.
Potential mediators and moderators of physical activity behaviour change
The questionnaires used to measure potential mediators and moderators were administered in a secluded area on two separate days to increase the accuracy of responses and reduce participant burden. The first visit was when the accelerometers were distributed and the second when the accelerometers were collected approximately seven days later. On the first visit the enjoyment of physical activity and school sport, self-efficacy in physical activity, social support for physical activity, social support during school sport, strategies to increase physical activity, and school physical activity environment scales were administered. On the second visit, the physical self-concept and perceived importance of physical activity scales were administered.
Potential mediators assessed included enjoyment of physical activity and school sport, physical activity self-efficacy, social support for physical activity, social support during school sport, strategies to increase physical activity, school physical activity environment, physical self-concept and perceived importance of physical activity. We conducted our own validity and reliability of all scales. To do this, four schools were selected at random to complete the scales, approximately one week apart: Seventy-five students from two schools completed those administered on the first visit, and another 75 from the other two schools completed those administered on the second visit. Intra-class correlations were performed to assess test-retest reliability and Cronbach alphas were used to assess internal consistency. AMOS 17.0 (Small Waters Corp., Chicago IL) was used to assess the construct validity of the different measures using the baseline data. Model fit was assessed using multiple indices, including chi-square index, comparative fit index (CFI) and root mean square of approximation (RMSEA).
Enjoyment of Physical Activity and School Sport
General enjoyment of physical activity and specific enjoyment of school sport were measured using a modified version of the Physical Activity Enjoyment Scale (PACES) [22
]. The S-PACES measure comprises seven negatively worded items from the original PACES instrument [23
]. The items are rated on a 5-point Likert-scale with semantic anchors ranging from "disagree a lot" to "agree a lot". The stems used to cue responses in this study were "When I am active..." and "When I participate in school sport...". Sample items were: "It frustrates me", and "It's no fun at all". The test-retest reliability (ICC = 0.86, 95% CI = 0.77 to 0.91) and internal consistency (α = 0.90) of the PACES were good. Similarly, the internal consistency (α = 0.91) and test-retest reliability (ICC = 0.83, 95% CI = 0.73, 0.89) of the scale were also good. The model fit indices for enjoyment of physical activity in the study population were χ2
= 142, p
< 0.001, CFI = 0.97 and RMSEA = 0.09. The model fit indices for enjoyment of school sport were χ2
= 246, p
< 0.001, CFI = 0.96 and RMSEA = 0.12.
Physical Activity Self-Efficacy
Physical activity self-efficacy was measured using an eight-item questionnaire originally developed by Motl et al. [24
]. Example items on the self-efficacy measure included "I can be physically active during my free time on most days no matter how busy my day is," and "I can be physically active during my free time on most days even if it is hot or cold outside". All eight items were rated on a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). The reliability of the eight-item instrument was acceptable (ICC = 0.8, 95% CI = 0.85 to 0.94 and internal consistency
= 0.78) and the model fit indices for the sample were χ2
= 60, p
< 0.001, CFI = 0.98 and RMSEA = 0.04
Social Support for Physical Activity
Social support was assessed using the Peer Support Scale developed by Prochaska, Rodgers and Sallis [25
]. The instrument employed a 5-point Likert scale anchored by 1 (Never) and 5 (Daily). The participants were asked to report how many times during a typical week they received or gave various forms of support from/to their friends e.g. "Do you encourage your friends to do physical activities or play sport" or "Do your friends encourage you to do physical activities or play sport". The Peer Support Scale was found to have high internal consistency (α = .73) and good test-retest reliability (ICC = .86). The model fit indices for enjoyment of physical activity were good χ2
= 13, p
= 0.002, CFI = 0.99 and RMSEA = 0.07
Social Support during School Sport
Social support received during school sport was measured using a modified version of an existing scale [26
]. The scale included five items relating to students' beliefs about the instruction and social support students received from teachers and instructors during school sport. Students responded to a 5-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree). 'During school sport my teacher/instructor...' was the common stem and included the following items: i) appears enthusiastic about school sport, ii) teaches me valuable movement skills, iii) participates in physical activity or sport with me, iv) makes the activity enjoyable, v) encourages me to participate in the activity and vi) demonstrates sound knowledge and understanding of the activity. The internal consistency (α = 0.82) and test-retest reliability (ICC = 91, 95% CI = 0.85, 0.94) values were acceptable. The model fit indices for scale in the study population were good χ2
= 31, p
< 0.001, CFI = 0.99 and RMSEA = 0.05
Behavioural Strategies to Increase Physical Activity
Behavioural Strategies used to increase physical activity were measured using an adapted version of a scale derived from self-management theory for use with college students [27
]. Participants were asked how often they used various strategies to increase their motivation for physical activities e.g. "I do things to make physical activity more enjoyable", and "I set goals to do physical activities". The modified seven-item instrument used a 5-point Likert scale ranging from 1 (never) to 5 (very often) for responses. Test-retest reliability of the modified version were ICC = 0.93 (95% CI = 0.89 to 0.96) whilst internal consistency was α = 0.78. The model fit indices for physical activity behavioural strategies in the study population were χ2
= 263, p
< 0.001, CFI = 0.92 and RMSEA = 0.12.
School Physical Activity Environment
Participants were asked to rate the quality, accessibility and availability of the physical activity facilities at their school using a scale developed for the current study. An example of an item was "The physical activity facilities at my school are easily accessible to me". Options ranged from 1 (Strongly disagree) to 5 (Strongly agree). The test-retest reliability (ICC = 0.69, 95% CI = 0.51 to 0.81), internal consistency (α = 0.80) and model fit indices (χ2 = 98, p < 0.001, CFI = 0.96 and RMSEA = 0.07) were adequate.
Children's Physical Self-Perception Profile including the Perceived Importance Profile
The participants' physical self-perceptions were examined using the Children's Physical Self-Perception Profile (C-PSPP) inventory [28
]. The C-PSPP is a 30-item questionnaire which consists of five equally divided sub-domains: perceived sport competence, physical conditioning, body attractiveness, physical strength, and physical self-worth. Each item presents two alternative statements, from which the participants can select which one best represents themselves using "sort of true" or "really true". The factor validity of the PSPP has been supported across various age groups, including adolescents [29
]. Test-retest reliabilities and internal consistencies for all sub-domains were > 0.80. The model fit indices for the different sub-domains of the C-PSPP were considered acceptable to good in the study population. The Perceived Importance Profile (PIP) is integrated into the PSPP and centres on the importance the participants attach to four of the PSPP sub-domains, physical self-worth is excluding. The PIP utilises the same scoring structure as the PSPP and incorporates two items to measure the importance of each of the four sub-domains. The internal consistency (α = 0.78) and test-retest reliability (ICC = 91, 95% CI = 0.86, 0.94) values for the PIP were acceptable. The model fit indices for the PIP scale in the study population were χ2
= 216, p
< 0.001, CFI = 0.90 and RMSEA = 0.09.
In a study such as Girls in Sport
, where each school will implement an intervention that is slightly different in its context, it is important to thoroughly document what is implemented, and the social context in which this occurs, as level of implementation can directly influence the outcomes of the study. In Girls in Sport
, each school was required to submit an action plan for 2009 and 2010. This action plan formed part of the overall school plan for the year. School plans provide a framework to drive change within a school over a 3-year period in areas such as student engagement and retention, teacher quality, and connected learning [31
The action plan for this study took each of the overarching outcomes and asked schools to write down specific strategies they would undertake to achieve this target and how they would measure success in achieving it, along with who would be involved. The activities also needed to demonstrate the area of the Health Promoting Schools Framework they represented.
Each school's specific action plan was then reviewed by the research team and Department of Education and Training staff. Schools participated in monthly teleconferences with their research partner to share their progress towards the study outcomes, specifically, the implementation of their strategies and any barriers they were experiencing. Possible solutions were brainstormed and, if further assistance was required, the problems were revised at the teleconference held every three weeks between the research team and Department staff and possible solutions fed-back to the school.
At the end of each year, schools were asked to document their progress towards the study outcomes based on their implementation of their specific strategies. They also evaluated any barriers to implementation. Interviews were also conducted at the end of the intervention with each school committee, other staff, executive, girls and boys, to triangulate these data and assess the extent to which the strategies were implemented. Observations of school sport and lunchtime activities will also take place.
Sample size justification
The primary analysis in this study will be conducted in SAS using a linear mixed model. The test of interest will be an F test with a 1 degree of freedom contrast therefore it is computationally convenient to use the t test to perform the sample size calculations. Murray [32
] proposed a method of sample size estimation and published relevant intraclass correlations on which to base the estimate. In lieu of a reliable estimate of the intra class correlation for the primary outcome measure of activity CPM an estimate of 0.01 was used in the a priori calculations. Effect sizes and variance estimates, 77.51SD(102.92) cpm, which was 18.4% of the baseline mean, were obtained from a previous study [33
]. Based on these figures, a model based on a critical t value of 2.228 (taking into consideration the matching of the schools) is obtained for estimates based on 12 schools per group. Variance estimates are adjusted for clustering as proposed in Murray [32
]; in brief the standard error of the estimate in the usual t estimation is replaced by
is the estimate of the unadjusted subject component of the variance,
is the unadjusted school component of the variance, m is the number of subjects per school and g is the number of schools per intervention. Sample sizes as low as 10 participants per school completing the study provided adequate power (>80% power and P < 0.05). Given that the estimate of effect could be considered optimistic for the present design a more modest effect size (10% of baseline mean, 42.07 cpm) was also considered. It was also anticipated that group sizes would vary between schools and therefore the estimates were based on a harmonic mean of 30 participants per school completing [34
]. With this conservative mean effect size and a harmonic mean sample size of 30 completing the study the power remains high (0.987).
Statistical analysis of the primary outcome variable, percentage of time spent in MVPA, will be performed using a linear mixed model (PROC MIXED) is SAS. This model accounts for the hierarchical structure of the data. This is a standard statistical procedure for analysis of clustered datasets and the use of this methodology in school-based trials has been extensively documented by Murray [32
]. Analysis of sedentary behaviour, light, moderate and vigorous activity will be performed in a similar manner. An advantage of the linear mixed model is that it incorporates all available data allowing for the analysis of partial datasets created when a participant drops out of the study or misses a study visit. Imputation of missing data will also be considered if missing data is substantial. This imputation will be performed using PROC MI and MIANALYSE. Sensitivity analyses will be performed. Mixed models will also be used to analyse all continuous secondary outcome variables.
Model Testing, Mediation and Moderation Analyses.
Two types of analyses will be conducted to explore the theoretical assumptions of the intervention. First, Social Cognitive Theory will be tested in AMOS using structural equation modelling. Hypothesized mediators of physical activity behaviour change will be examined using multilevel linear analysis and a product-of-coefficients test that is appropriate for cluster randomized controlled trials [35
]. Potential moderators of the intervention effects (e.g. ethnicity, socio-economic status and type of school) will also be explored using multi-level modelling. The baseline analyses presented in the current paper are conducted using PROC MIXED (SAS V 9.0, SAS Institute, Cary NC) to adjust for the matched and clustered nature of the dataset [32