Design

This study used a cluster randomized control trial (RCT) with schools as the unit of randomization. The participating schools were randomly assigned to one of two clusters: the intervention group (Kies Awa intervention) or the control group (no intervention). A randomized block design ensured balanced sample sizes among the groups. In addition, randomization was restricted based on the proximity of school locations to prevent contamination between the groups in the relatively small island setting. In the intervention group, participants were exposed to trained classroom peer influencers (PIs) promoting water consumption over SSB consumption (Franken et al. 2018; Smit et al. 2021a), while no intervention occurred in the control group. The primary outcome measure was water consumption and the secondary outcome was SSB consumption.

Using G*Power 3.1 (Faul et al. 2009), the sample size for this study was calculated based on the pilot SNI in Aruba (Franken et al. 2018), which found a small intervention effect size for water and SSB consumption (β = 0.12 and β = −0.12, respectively) that was then converted to Cohen’s f (f = 0.12). For a repeated-measures ANOVA with an interaction between two groups and two repeated measures (power = 0.80, α = 0.05), 140 participants were needed. To anticipate non-response, attrition, and missing data, a much larger number of participants were recruited (see Fig. 1 for the CONSORT flow diagram of participants).

Fig. 1

CONSORT Flow Diagram of Participants

Procedure

The intervention lasted eight weeks, from January to March 2019. Pre-measurements were conducted one week before the intervention started. Participants completed paper-and-pencil questionnaires at their schools before and after the intervention, available in both Papiamento and Dutch, the official languages in Aruba. The pre-measurement questionnaire contained questions related to their demographic information, water and SSB consumption, as well as questions regarding their descriptive and injunctive norms of water and SSB consumption. At pre-measurement, the intervention group participants also answered sociometric nomination questions to identify PIs.

The same consumption and behavior-related questions were answered at post-measurement in the eighth week of the intervention. At post-measurement, participants were also asked to explain the purpose of the research to determine their awareness of the social network component. Among them, none expressed awareness of the purpose of incorporating PIs to promote water consumption. The procedures of this study were approved by the Ethics Committee of the Faculty of Social Sciences at Radboud University (ECW2014-1003–203) and followed the data management protocol of the Behavioural Science Institute of Radboud University. This RCT study was preregistered (2018-12-20) and its main ID number is NL-OMON26157 (https://trialsearch.who.int/Trial2.aspx?TrialID=NL-OMON26157).

Participants

Figure 1 illustrates the CONSORT flow diagram of participants in this study. Eligible participants were 5th and 6th-grade public primary school children in Aruba whose schools met the criteria of not being involved in other curriculum-based health programs and not having participated in the pilot SNI. Eight schools (27 classrooms) were invited to participate. Active consent was obtained from the head of the educational inspection, school board, principals, and parents/caregivers. One principal and one teacher from another school declined participation, leaving seven schools. Four schools (255 children) were randomly assigned to the intervention group, and three schools (269 children) were assigned to the control group.

Of the 524 invited children, 19 parents did not give active consent, and 138 did not submit consent forms. Before the pre-measurement questionnaires were distributed, the children provided assent by signing a form, and five declined to participate. At pre-measurement, 12 children were absent. Similarly, due to absenteeism, 12 children could not complete the questionnaire at post-measurement. After data collection, six out of the 24 classrooms had a participation rate of less than 60%, which may hinder the examination of the effects of the SNI (Marks et al. 2013; Smit et al. 2021a). For this reason, two classrooms from the intervention group and four from the control group were excluded from the analysis. The final analytical sample consisted of 7 schools with 18 classrooms and 300 participants aged 9 to 14. Of these, the intervention group included 4 schools with 10 classrooms and containing 156 participants (M = 11.08, SD = 1.00; 53.8% girls), while the control group included 3 schools, 8 classrooms, 144 participants (M = 11.32, SD = 0.96; 52.8% girls).

The social network intervention

This study aimed to examine the effectiveness of an adapted SNI for promoting the healthy behavior of choosing water consumption over SSBs. The Aruban intervention, Kies Awa (Choose Water), was based on the Dutch Share H2O intervention (Smit et al. 2021b), which utilized self-determination theory (SDT) techniques (Deci and Ryan 1985; Ryan and Deci 2017). We made two key adaptations for Aruba. We provide a brief overview of the intervention’s approach before describing the two key adaptations.

Intervention approach

The Aruban intervention resembles the Dutch intervention in that it involved identifying PIs from each classroom, providing them with training, and offering follow-up support sessions. The training, facilitated by the first author, lasted 90 min. The training had two objectives. The first objective was to help PIs develop intrinsic motivation for water consumption through training materials that emphasize the benefits of water and drawbacks of SSBs, as well as by supporting their personal reasons for choosing water (Smit et al. 2021b). The second objective was to empower PIs to motivate their classmates to consume more water through two strategies: modeling the behavior (i.e., descriptive norms) by consuming more water themselves and communicating (i.e., injunctive norms) about consuming water (Cialdini et al. 1991). By practicing these skills, they also learned skills to address potential difficulties they may face in their role. Follow-up sessions were conducted during Week 3 and Week 6 to provide general support, facilitate sharing experiences, and review information. For a description of the SDT-based techniques and training components, refer to Smit et al. (2021b).

Kies Awa: Adaptations to the Aruban Context

We made two key adaptations to the training of the original Dutch Share H2O intervention for the Aruban Kies Awa intervention: (1) adding roleplay techniques and (2) incorporating educational materials appropriate for Aruba.

Regarding the first adaptation, adding roleplay techniques allowed PIs to practice promoting water consumption and apply their knowledge in real-life school settings (Lanigan 2011; Lloyd et al. 2011). We added roleplay as a behavioral change technique for two reasons. The first reason was to address the social norms associated with water consumption identified as an important behavioral determinant in previous Aruban research (Franken et al. 2018, 2023). Through roleplay, PIs were encouraged to model the behavior of water consumption (to influence perceived descriptive norms) and to communicate about it (to influence perceived injunctive norms). In doing so, they motivated their classmates to increase their water consumption. During the main roleplay assignment, PIs engaged in various simulated scenarios. Working in pairs, they customized the scenarios to their preferences and discussed possible interactions between the ‘peer influencer’ and the ‘classmate.’ They then acted out their scenario, allowing other PIs to observe. To provide a clear understanding of their role through this roleplay assignment, the trainer enacted an example script where the classmate expressed thirst after playing tag at school, and the peer influencer recommended consuming water to quench thirst and regain energy, suggesting that they fill their water bottles at the water cooler.

The second reason to add roleplay as a behavioral change technique was to emphasize the importance of intrinsic motivation for water consumption in Aruba (Franken et al. 2023). Roleplay and discussions also served to encourage PIs to initiate water consumption autonomously and by spreading this behavior, motivating classmates to do the same (Ryan and Deci 2017; Smit et al. 2021b; Soenens and Vansteenkiste 2010). Through discussions and roleplay, PIs learned to identify opportunities (e.g., feeling thirsty after exercise) and difficulties (e.g., not having easy access to water in classrooms) and develop strategies to benefit from opportunities (e.g., encourage filling water bottles at water coolers) or face difficulties (e.g., encourage carrying a reusable water bottle to school).

During the roleplay assignment, PIs also learned to consider their own perspectives and those of their classmates regarding water consumption. They also learned to provide their classmates with reasons to, for example, choose to consume water after physical exercise at school instead of SSBs. By modeling and communicating about water consumption with their classmates, the PIs fostered a social environment that promoted intrinsic motivation among their classmates to consume more water as well (Smit et al. 2021b). In addition, throughout the training, the PIs practiced their role through roleplay activities such as pouring water for each other, drinking together, communicating about it, and using positive nonverbal cues (e.g., nodding or thumbs-up).

Regarding the second adaptation, we incorporated educational materials appropriate for the Aruban context to provide PIs with more relevant reasoning to increase water consumption and to encourage PIs to promote water consumption among their classmates. The first reason to incorporate context-appropriate materials was to account for the important factors of having favorable attitudes toward water consumption and positive perceptions of behavioral control regarding water consumption that were identified in previous Aruban research (Franken et al. 2023; Ajzen 1991). For example, given that environmental preservation is a concern in Aruba, we made it more meaningful for PIs by emphasizing the environmental benefits of consuming water during the training. The materials included visual imagery and discussions emphasizing the ease of increasing water consumption, its positive impact on personal health, and the island’s natural environment, wildlife, and marine ecosystems. To achieve these positive impacts, we emphasized the use of reusable water bottles (Smit et al. 2021b) to contribute to reducing the high amounts of litter associated with SSBs (Debrot et al. 2014; de Scisciolo et al. 2016; Hartley et al. 2015).

The second reason was that the original Share H2O intervention materials were designed for a Dutch audience, requiring adaptations to align with Aruba’s specific context. To establish a stronger connection between the intervention and the local context and encourage PIs to consume more water and less SSBs, we incorporated pictures of children and nature that resonated with the PIs. Additionally, we included recent Aruban consumption data to underscore the alarming levels of sugar intake among children in Aruba (Franken et al. 2018). As part of the training, the children also learned the skill of reading nutrition facts labels, enabling them to calculate the high number of teaspoons of sugar in SSBs. This knowledge was aimed at reinforcing their understanding that water consumption is essentially the only healthy option because it does not contain sugar.

MeasuresSociometric peer nominations

During pre-measurement at intervention schools, participants nominated PIs by writing a maximum of five names of classmates on five sociometric questions. These questions asked whom in their classroom they “wanted to be like,” “looked up to,” “respected,” “regarded as good leaders,” and “went for advice” (Campbell et al. 2008; Starkey et al. 2009). The role of a PI was offered to 15% of males and 15% of females, with most nominations preserving the sex distribution in each classroom (Campbell et al. 2008; Starkey et al. 2009). Due to the absence of two nominated peers from different classrooms, the next two most nominated peers were invited to participate. A total of 38 participants from ten intervention classrooms (M = 4 children per classroom, SD = 0.21) accepted the invitation and were trained as a PI to promote water consumption (50% females; M age = 11.3, SD = 1.01).

Water consumption

Water consumption was measured by asking participants at pre- and post-measurement how much water they drank during a normal school day. The answer options ranged from 0 = zero glasses to 7 = seven or more glasses. To facilitate participants’ quantity estimation, the questionnaire illustrated that a glass also represents a bottle, a can, or a package (Franken et al. 2023; Smit et al. 2018).

Sugar-sweetened beverages consumption

SSB consumption was measured by asking participants at pre- and post-measurement how many glasses of (a) sweetened juice drinks, (b) soda, and (c) energy and sports drinks they drank on a normal school day. The answer options ranged from 0 = zero glasses to 7 = seven or more glasses. Examples of SSB brands or names were given. To facilitate participants’ quantity estimation, the questionnaire illustrated that a glass also represents a bottle, a can, or a package (Franken et al. 2023; Smit et al. 2018). The average response across the three SSB consumption items was used to obtain each participant’s total SSB consumption score.

Descriptive norms of water and sugar-sweetened beverages consumption

Participants’ perceptions of their classmates’ water and SSB consumption were measured at pre- and post-measurement by asking, “How often do your classmates consume water/SSB?” The answer categories ranged from 1 = never to 6 = always (Franken et al. 2023; Smit et al. 2018).

Injunctive norms of water and sugar-sweetened beverages consumption

Participants’ perceptions of receiving approval from their classmates to consume water and disapproval to consume SSBs were measured at pre- and post-measurement by asking, “How often do your classmates approve/disapprove that you drink water/SSB?” Answer categories ranged from 1 = never to 6 = always (Franken et al. 2023; Smit et al. 2018).

Thirst level

Thirst level was included as a covariate given that it correlates with the beverage consumption of the participants (Franken et al. 2018; Smit et al. 2018). Therefore, before participants completed the questionnaire at pre- and post-measurement, they were asked to indicate their thirst level by answering “How thirsty are you now?” using a visual analog scale (VAS) of 16 cm ranging from 0 = not thirsty at all to 16 = very thirsty (Bevelander et al. 2012).

Statistical analyses

The mean and standard deviation were examined for each variable. Independent samples t-tests were used to determine whether the scores differed between the intervention group and control group at pre-measurement and post-measurement for all study variables. Pearson’s correlations between the study variables were examined.

For the main analyses, we executed Linear Mixed Modeling (LMM) analyses in JASP Version 0.18.2 Intel (JASP Team 2024) to examine the effect of the intervention on the study outcomes of water consumption and SSB consumption. These regression-based hierarchical models take into account the four-level hierarchical structure of our dataset (Field 2013). Water and SSB consumption represented separate dependent variables for every child in a treatment group (intervention or control) in a classroom at a school. Repeated measures (level 1) were nested within children (level 2), who were nested within classes (level 3), which were nested within schools (level 4). We constructed our model in four steps to find the best-fitting model. To compare the models and to test whether predictors contribute to the explanation of the change in water consumption and SSB consumption, model fit was evaluated with an X2 difference test, which is calculated as the difference between the deviance statistics (−2 log-likelihood) of the two models that are compared (Hox et al. 2017).

In the first step, for both outcomes, we ran a model that included four levels with time (pre- or post-measurement) as a fixed effect (predictor). With this model, we determined whether significant clustering (dependency in the data) occurred. For this purpose, we calculated intra-class correlations (ICC) for levels 2 (individuals), 3 (classes), and 4 (schools). If the ICC at a particular level was below 0.05, it indicated that no substantial clustering occurred at that level and that it could be removed from the model (Field 2013). In the second step, we removed those levels that displayed an ICC < 0.05. In the third step, we added along with time, the treatment group, and the interaction effect between treatment group and time as fixed effects. The significance of the treatment by time interaction effect was tested for our first objective regarding the main effect of Kies Awa on children’s water and SSB consumption. In the fourth step, we added the covariate thirst level to examine whether our results changed after controlling for thirst level. Then, to further examine the model findings for changes in water and SSB consumption between pre- and post-measurement for each treatment group, we conducted paired sample t-tests.

For the second objective regarding testing for the moderating effect of children’s descriptive and injunctive norms of water and SSB consumption on the intervention’s effect, we included three-way interaction effects (treatment by time by norm) for each norm in each outcome model. To further understand the effect of the intervention on participants with low (1 SD below the mean) versus high (1 SD above the mean) levels of social norms for water and SSB consumption, significant interaction effects were further interpreted using simple slope analysis.

To explore the effect of the training on the PIs, we conducted two additional analyses. Paired sample t-tests were conducted to determine the impact of the training on the water and SSB consumption of the PIs. In addition, to explore the extent to which PIs used water promotion strategies, we computed means, standard deviations, and percentages (with a positive score of 3 or higher) for their post-measurement answers to questions regarding how often they used modeling and communication strategies to promote water consumption among their classmates (scale from 1 = never to 6 = always; Smit et al. 2021b).