Page 33 - ON THE WAY TO HEALTHIER SCHOOL CANTEENS - Irma Evenhuis

Subsequently, the school canteen’s current overall level (silver or gold), and a level for all
three individual parts (in percentages) is indicated. Consequently, the health level of the
canteen can be defined as: the available basic conditions, the available healthier food and
drinks and meeting the accessibility criteria in the school canteen. The Canteen Scan will be
filled out in all intervention and control schools by a school canteen advisor. Intervention
schools will receive the outcome and feedback as part of the intervention. On the contrary, 2 the control schools will not receive the results or feedback from the Canteen Scan.
Sample size
The power calculation was based on the primary outcome, i.e. the self-reported purchase behaviour of healthier products per week. In this calculation we included an 80% power and a 5% significance level [110]. To detect a 10% difference in the proportion of purchasing healthier versus unhealthier products per week (dichotomous variable) between the intervention and control group, with the expected multi-level structure between schools (correlation of 0.05 between schools), and to obtain sufficient power (80%), we calculated that 1,505 students spread among 10 intervention and 10 control schools are needed. The increase of 10% in purchase behaviour of healthier products is based on results of comparable studies in schools [37]. Consequently, we aimed to recruit 20 schools and 100 students per school, based on an expected dropout rate of 10% [111].
Statistical analysis
Process evaluation: To test for differences in factors affecting implementation perceived by stakeholders (dependent variable) between the intervention and control group (independent variable) after the intervention (6 months), linear two-level regression analysis will be used. The used levels will be: stakeholders (level 1) and schools (level 2) and we will adjust for baseline measurements. This analysis will be performed for each individual (e.g. knowledge, attitude, self-efficacy) and environmental factor (e.g. need for support, innovation). When these analyses show no significant difference between school variance, a linear regression analysis will be performed [110]. We hypothesize that the stakeholders in the intervention group will positively change their perceived factors due to the support in implementation.
To investigate the quality of implementation quantitatively (dose delivered, dose received and satisfaction) of each implementation tool, descriptive statistics will be used. This information will be complemented by qualitative data about the overall experiences of stakeholders. This data will be analysed in three rounds, following the thematic content approach [97]. First, answers will be labelled with descriptive codes. Second, the codes will be split or merged and interpretative codes will be created. Third, codes will be compared and overarching themes defined.
Effect evaluation: After the intervention, differences in the primary outcome ‘purchase behaviour’ of students (dependent variable) between the intervention and control group (independent variable) will be analysed with two-level regression analysis (intention-to- treat). Here, we will correct for correlations of students (level 1) nested within schools (level 2). We will adjust for confounders related to students (e.g. groups of sociodemographic characteristics, behavioural determinants, and environmental determinants). In addition, the moderation effect of gender will be taken into account by stratifying the analyses,
31