Study design and participants
The Thirsty? Choose Water! Study was a secondary school-based intervention conducted as a randomised controlled trial with a two-by-two factorial design. Two interventions (one behavioural and one environmental) alone and in combination were tested to determine their impact on the primary outcome of increasing water consumption in year 7 students and the secondary outcomes of changing knowledge, attitudes and consumption of SSBs; and changes in the school environment to promote water. A factorial design allowed the effects of both interventions within the same study to be tested and to explore interactions between the interventions, and increase statistical power. Full details of the study design, sample size, power calculations, setting, sample, eligibility, recruitment and study participation and interventions design are detailed in the study protocol paper.19 Ethics approval was obtained through the Hunter New England Human Research Ethics Committee (17/08/16/4.07), the NSW Department of Education (SERAP2017457) and Broken Bay, Sydney and Wollongong Catholic Diocese and study procedures were approved.
Sixty-one secondary schools (Government, Independent and Catholic) within the boundaries of three NSW Local Health Districts (LHDs) were recruited to the study in late 2017/early 2018, through the study project officers. Randomisation occurred at the school level, whereby the study statistician utilised a computerised random number generator to allocate schools to one of the four study arms. Following allocation to study group, all schools participated in baseline data collection in early 2018. Students and schools were not blinded to the group they were allocated, given the type of interventions trialled. The power and sample size calculation identified that 6370 year 7 students would need to be recruited to the study, to detect a change of 5% or more in the primary outcome measure, being increased water consumption. At the commencement of the study 72% (n= 6531) of eligible students (n=8992) completed baseline data collection, reducing to 62% post-intervention and 58% at follow-up.
Intervention
The two intervention components implemented are described below.
The Behavioural intervention - The TCW-BI aligned with the Health Promoting Schools Framework (HPSF)15 and consisted of three components including teaching, promotional and vaccination interventions. The teaching intervention aligned with the NSW Stage 4 PDHPE curriculum and included two ‘Thirsty? Choose Water!’ lessons delivered within PDHPE to educate students on the key messages regarding water and SSBs. This was reinforced through the promotional intervention utilising school newsletters, social media, posters and other promotional materials which displayed the ‘choose water’ message, including ‘toilet talkers’ (info-graphics on toilet doors), hi-vis vests worn by teachers and posters/flip charts at the canteen. The vaccination intervention was delivered during the school-based vaccination program, where the messages were delivered during the clinic through pull-up banners in the clinic area, a distractor card used by the nurse during the administration of the vaccination and the “Spouts and Straws” board game that students played in the standard 15 minutes post-vaccination waiting time. A paper with the full description of the TCW-BI, its acceptability and standalone outcomes is currently under review.
The Environmental intervention - The environmental intervention included the installation of one CWS in all schools allocated to receive this intervention. Following extensive research, a stainless steel wall-mounted station with a chilled drinking fountain, bottle refill capacity and bottle refill counter were identified for installation. Project Officers worked with the intervention schools to determine the most suitable position for the CWS, allowing the greatest access point for year 7 students, and installation occurred following baseline data collection. To promote the CWS to students a large promotional sticker was positioned near the CWS. A full description of the CWS intervention, acceptability and outcomes is currently under review.
Data Collection
Student surveys: These were completed at three time points: baseline (Term 1, 2018), post-intervention (following delivery of the TCW-BI and/or CWS installation – Term 3 and 4, 2018) and at follow-up, as students commenced year 8 (Term 1, 2019). The survey included validated questions from the NSW Schools Physical Activity and Nutrition Survey (SPANS)20 regarding students’ beverage consumption, where they purchased beverages and how often SSBs were available in their home, as well as numerous questions designed to reflect knowledge taught through the TCW-BI, including the effects of SSBs and dehydration. Students were also asked about their water consumption, whether they took a drink bottle to school, where they filled it and bubbler usage. Student surveys were predominantly completed utilising an online survey tool within PDHPE lessons. Some schools experienced difficulties accessing the online survey tool, and therefore completed a paper-based version of the survey.
Water Flow Measurement: At CWS installation a data logger and separate water meter were fitted to measure water flow from the in-built bubbler and bottle-filler of the CWS. The data logger measured water flow remotely and continuously in real-time and was installed for one year to capture water flow across four seasons. Data were downloaded to an online portal that researchers and schools could access.
School level data: School level data was collected from key school contacts via an online survey at baseline and follow-up. Variables of interest included school type (e.g. Government, Independent or Catholic), school size (total number and composition of students), the Index of Community Socio-Educational Advantage (ICSEA) (an index of educational advantage or disadvantage), availability of vending machines, the closest access point to sources of SSBs external to the school measured in walking distance, available drinking water sources for students, how water was promoted as the preferred drink of choice and feedback on CWS installation and usage.
Procedures
Consent for study participation was at the school level and provided by the school principal. Student participation in the study evaluation was through an opt-out process, whereby parents/carers were advised to provide written documentation to the school if they did not wish for their child to participate in the student surveys (with the exception of one Catholic Diocese, which requested an opt-in process, which was adhered to for those schools). A Project Officer in each LHD worked closely with schools to advise them of research processes, timeframes and study requirements. Project Officers also provided relevant training to teachers and immunisation teams who delivered the TCW-BI, and worked with schools on the installation of CWSs.
Data analysis
All statistical analyses were conducted with Stata V16.0 (StataCorp, College Station, Texas, USA). Frequency counts and percentages were used to describe study data. Employing a 2x2 factorial design, random-effects ordinal logistic regression modelling was used to estimate the main effects of BI (no, yes), CWS (no, yes). In addition, this approach not only allowed for the estimation of accompanying interaction, but also utilised all sample observations by not adopting a complete-case analysis, which would have only included students on the condition that they had provided information at all time-points. To control for confounding and ensure all models were parsimonious, a two-stage screening process was undertaken. Variables with a univariable p-value <0.05 were further assessed for model inclusion using a manual stepwise backward elimination approach (p≥0.05 for removal). To account for a lack of independence when students were observed at multiple time-points, a variance estimator was used to cluster observations from the same individual. For each multivariable model, the proportional odds assumption was assessed using the Brant test.21 The presence of multicollinearity and model specification were tested respectively by the variance inflation factor (VIF), and the Pregibon link test.22 For each model, unadjusted and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Due to the presence of heterogeneity in the implementation of BI and CWS, intervention effects are reported only for change between baseline and follow-up. The level of statistical significance for all two-sided tests was set at 5%.