Nathan 2012.

Study characteristics
Methods	Trial name: Good for Kids. Good for Life Study design: non‐randomised Intervention duration: 11–15 months (duration of treatment) Length of follow‐up from baseline: baseline; November 2006 to April 2007, follow‐up: October 2008 to March 2009 Differences in baseline characteristics: no significant difference in the prevalence of vegetable and fruit breaks between intervention and comparison schools. Relative to comparison schools, intervention schools were more likely to be small, and located in rural and lower socioeconomic areas. Unit of allocation: schools Unit of analysis: schools
Participants	School type: K‐6 and K‐12 central schools Region: HNE, NSW, Australia Demographic/socioeconomic characteristics: region: a demographically and socioeconomically diverse population of approximately 121,000 children aged 5–14 years (14% of the state population of 5–14 year olds). Sample: schools were primarily government, urban and lower SES schools. Inclusion/exclusion criteria Inclusion ‐ All primary schools (children 5–12 years of age) and central schools (children 5–18 years of age) across the state were eligible for participation. Exclusion ‐ Special purpose schools catering for students with special needs ‐ Juvenile justice schools ‐ Schools serving children who are hospitalised Number of schools allocated: 828 (422 intervention, 406 control) Numbers by trial group n (controls baseline) = 316 n (controls follow‐up) = 258 n (interventions baseline) = 407 n (interventions follow‐up) = 388 Recruitment: principals of both groups of schools were sent a letter inviting them to participate in the study. 2 weeks after receipt of the letter, principals were telephoned by a trained research assistant who confirmed school eligibility, sought consent to participate and scheduled a time for a telephone interview. Recruitment rate: 96.4% of intervention schools and 77.8% of control schools consented to participate in the baseline data collection.
Interventions	Number of experimental conditions: 2 (1 intervention, 1 control) Policies, practices or programmes targeted by the intervention ‐ Prevalence of vegetable and fruit breaks Implementation strategies EPOC: local consensus processes ‐ Consensus processes EPOC: local opinion leaders ‐ Leadership support and endorsement EPOC: educational meetings ‐ Staff training and professional development EPOC: educational materials ‐ Programme materials (curriculum resource and materials, information to parents) EPOC:other ‐ Incentives (material goods) EPOC: tailored interventions ‐ Follow‐up support EPOC: monitoring the performance of the delivery of the health care ‐ Implementation feedback (performance monitoring and feedback) Theoretical underpinning A structured multi‐strategy intervention was developed based on theoretical frameworks of practice change and recommendations from reviews and implementation studies conducted in schools and other settings. Description of control: comparison schools were not offered the multi‐strategy intervention described above, but were offered access to information‐based support provided by a non‐government organisation. Information regarding the programme was provided to schools via a website, newsletters and events. If a school chose to register for the programme, teaching resource materials were forwarded to the school, with schools able to receive e‐mail and telephone information‐based support if desired. If the school provided evidence of having adopted the programme, they were eligible to be 'certified' as such and to receive additional resource materials and obtain access to ongoing e‐mail and telephone support. In some areas of the state, schools could access additional support provided at the discretion of local health promotion teams (EPOC: educational material).
Outcomes	Outcome relating to the implementation of childcare service policies, practices or programmes ‐ Prevalence of fruit and vegetable breaks Data collection method: principal reported computer‐assisted telephone interviewing Validity of measures used: self‐report; however, method has been validated. The accuracy of principal‐reported implementation of vegetable and fruit breaks in schools was assessed in a convenience sample of intervention schools (42 schools; 10%). Based on observations made in these schools over a 9‐week period, pre‐service teachers located in schools reported in a pen–paper survey if classes at the school had specific breaks or if students had permission to eat vegetables or fruit (or both) during class time ('yes all classes', 'yes some classes', 'no classes', 'don't know'). The pre‐service teacher surveys were completed within 1 month of the principal telephone survey. Comparison of principal and pre‐service teacher report of vegetable and fruit breaks revealed perfect agreement (Kappa = 1.0). Outcome relating to cost: not reported Outcome relating to adverse consequences: not reported Outcome relating to child diet, PA or weight status: not reported
Notes	Research funding: New South Wales Health ASSIST programme. The project also received infrastructure support from the Hunter Medical Research Institute (HMRI) and Hunter New England Population Health. Conflicts of interest: authors reported no conflicts of interest.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Non‐randomised trial. High risk of selection bias as intervention services were recruited from a selected area and control services from a comparison region.
Allocation concealment (selection bias)	High risk	Non‐randomised trial. Intervention services were recruited from a selected area, therefore, high risk of selection bias as no concealment of allocation.
Blinding of participants and personnel (performance bias) Implementation outcome	High risk	Outcome group: all due to nature of the intervention, school staff and study personnel delivering the intervention were not blind to study allocation.
Blinding of outcome assessment (detection bias) Implementation outcome	High risk	Outcome group: prevalence of vegetable and fruit breaks. No blinding of outcome assessment and the outcome measurement is likely to be influenced by lack of blinding, self‐report considered high risk.
Incomplete outcome data (attrition bias) Implementation outcome	Low risk	Outcome group: although differential response rate for intervention vs comparison schools at baseline and follow‐up (96.4% vs 77.8% and 95.3% and 81.6%, respectively) appropriate analyses to address this were conducted. All schools lost to follow‐up were included in the generalised estimating equation model (using last value carried forward method).
Selective reporting (reporting bias)	Unclear risk	No study protocol, therefore, it was unclear if there was selective outcome reporting.
Other bias	Low risk	Appeared free from other bias.
Potential confounding	High risk	Baseline differences in intervention and control schools not adjusted for in analysis. Relative to comparison schools, intervention schools were more likely to be small, and located in rural and lower socioeconomic areas.
Overall risk of bias assessment	High risk	Most domains were at high risk of bias.