Using a gamified monitoring app to change adolescents’ snack intake: the development of the REWARD app and evaluation design

As earlier mentioned, the ‘Snack Track School’ intervention includes behavioral change strategies to influence both the automatic pathway as well as the reflective pathway. The systematic, stepwise, iterative, and collaborative principles of the Intervention Mapping protocol [33] were used to put this idea in practice. The intervention was preceded by an extensive problem analysis (i.e., needs assessment) using existing and newly generated evidence, including a cross-sectional survey, experimental and focus group research ([5], De Cock Unpublished data).

Following the Intervention Mapping protocol, different stakeholders were involved in the intervention development: adolescents, teachers, principals, stakeholders from food industry, professional organizations active in health promotion, community members, umbrella school organizations. Adolescents were involved in both the conceptual (i.e., a large scale focus group research among 101 adolescents) as well as the pretesting phase of the intervention (i.e., via test labs in smaller groups of four to eight students to regularly gather feedback regarding feasibility, usability, and attractiveness of the app). Monthly teacher contacts (two per participating school) were also included during the process to ensure that the app format was fully compatible with existing pedagogic guidelines, expectations, and school programs. Other stakeholders (i.e., food industry, professional organizations active in health promotion, community members, umbrella school organizations) were frequently consulted through stakeholder meetings to explore if and affirm that the current intervention can be disseminated to the wider society. A strong participation of these actors was used in the different phases of intervention development to ensure a culturally-, age- and community-relevant intervention which might increase the likelihood of program success.

The dual-system model [11], a model explaining health behaviors as a joint function between conscious and unconscious processes, was chosen as the broad framework of our intervention, as it allows to include other models like personality models, operant learning models and the Control theory. First, the dual model suggests that behavior change has to take into account the bottom-up impulsive/automatic system (i.e., habits) in which certain stimuli or cues are linked to certain behaviors based on earlier learned associations. Moreover, this fits with personality theory that describes a high peak of reward sensitivity in adolescence and its important and understudied role in nutrition behavior. However, this also fits with operant models of behavior change that argue that reward-based strategies (i.e., provision of rewards and positive reinforcement) are needed as the main behavioral change strategies.

However, the model plaids also to include top-down determinants including elaboration and cognitive efforts to build beliefs and decisions. Based on the extensive needs assessment, the following correlates at the individual level were found to be associated with the intake of snacks: knowledge, attitude, and self-efficacy. More knowledge about the healthiness of snacks, more positive beliefs towards healthy snacks, and a higher self-efficacy to eat healthy snacks, were associated with a higher intake of healthy snacks (REWARD unpublished data, [34‐37]). These significant cognitive correlates will be targeted in the intervention via behavior change techniques like active learning, advance organizers, mere exposure, goal-setting, monitoring, and feedback to influence the reflective pathway (see also Fig. 1 and below for more details).

The ‘Snack Track School’ app is a four week intervention and  constitutes of a virtual high school environment with typical school locations (i.e., classrooms, gym hall, bike area, bathrooms, etcetera) in which all participants have their own locker, their own Snack Track Tool, and their own ‘anonymous’ avatar, which they can customize as preferred based on several options (i.e., many options for head of character, clothes, shoes, gadgets). During the four-week intervention, every week of snack monitoring has its own story line and challenges.

The central idea of the ‘Snack Track School’ intervention is that participants earn credits/points based on the nutritional value of the snacks they consume (i.e., positive reinforcement/providing rewards to influence automatic behavior). For every snack the adolescents enter in the app via the ‘Snack Track’ tool, they will be awarded by points. The healthier the snacks they consume, the more points they will receive. These credits will be responsible for the progress in the weekly challenges of the four-week app. No negative points will be provided for unhealthy snacks as punishment can have contra-productive effects [38]. A large snack database was constructed based on the Internubel Trade Name database [39]. Points will be awarded according to the UK Ofcom Nutrient Profile model [40].

In order to stimulate a well-balanced snacking pattern and not merely the tracking of as many snacks as possible, some gratuities and limitations are built in the app related to the credit system. Participants are able to track as many snacks as they want, however, only the credits of the first ten snacks are included. Three gratuities are included: 1) a bonus for a snack intake ≤ six snacks per day, 2) a bonus for a snack intake of ≥2/3 healthy snacks per day, and 3) a bonus for non snackers that are involved in the app (logging in ≥3 times in the app per day). The cut-offs of six snacks and more than 2/3 healthy snacks were based on Flemish guidelines of necessary food and nutrient intake [41]. It is advised to consume maximum 10 % energy from snacks (defined here as unhealthy snacks). Fourteen to 16- year-old girls and boys have a general advised total energy intake per day of 2100 and 2600 kcal, respectively [41]. This corresponds with a maximum intake of 210–250 kcal from unhealthy snacks (i.e., one or two unhealthy snacks per day depending on the size and nutritional content). Additionally, it is advised for adolescents to eat three pieces of fruit and four dairy products per day (i.e., not only via snacks but also via meals) [41]. The cut-off of maximum six snacks and more than 2/3 healthy snacks per day was deduced from these guidelines, an intake of three or four healthy snacks together with one or two unhealthy snacks is acceptable for adolescents as they need a lot of energy to grow.

The following theoretical methods were included in the app to affect the reflective pathway including knowledge, attitude, and self-efficacy. Active learning and advance organizers based on the Elaboration Likelihood Model were included to enhance the knowledge of the nutritional quality of a snack item. The Elaboration Likelihood Model [42] suggests that people only process information centrally (i.e., careful consideration which can increase knowledge) when they are motivated to do so, when information is personally relevant, and when the information is presented repeatedly. Through the provision of credits in the app and the progress in the app linked to these credits, participants will unconsciously learn about the healthiness of a snack. The credit system used in the app can be seen as an external motivator to change their snacking behavior. Insights about the (un)healthiness of snack items based on this credit system can stimulate central processing of this information as more knowledge about the specific credits per snack item and the application of this knowledge can enhance their progress in the app. These points are also personally relevant (i.e., linked to the own snacks consumed and related to progress in the app) and are repeatedly presented. In addition, this healthy-unhealthy continuum ranging from 0 to 50 and the placement of the different snack items on this continuum based on the credits, can act as a mental reminder (i.e., a mental axis) regarding the nutritional value of snack items using the method of advance organizers. Furthermore, the earlier mentioned gratuities and limitations also learn adolescents about a well-balanced snacking pattern.

The method of mere exposure is used to activate positive attitudes regarding healthy snacks. Zajonc [43] indicated that participants develop a more positive attitude about stimuli following repeated exposure. In our intervention more exposure to healthy snacks is expected as participants receive more credits for the healthy choice, i.e. the expected behavior, which subsequently can lead to increased exposure and thus more liking or positive attitude even if they are not consciously aware of the process [43]. Moreover, the association between positive reinforcement and healthy snacks can also result in more positive attitudes [44].

Self-efficacy will be targeted via goal setting, monitoring, and feedback. Goal setting has a high likelihood to improve health behaviors because persons who set goals exert themselves to a greater extent, persevere in their aims, concentrate more, and develop strategies to be able to perform the necessary behavior [16, 45]. Adolescents will choose a goal per week regarding improving their snacking habits. The chosen goal will be evaluated per day and extra bonus credits/rewards will be provided in case of success. At the end of every week, feedback will be provided via a week-report that portrays all consumed snacks per weekday with their matching credits. This week-report will also provide feedback about the fail or success of reaching their goal. In case of success (i.e., linked to the goals), it will increase self-efficacy/capability of eating more healthy snacks. In case of failure, the feedback and the pending reward can stimulate elaboration and another attempt, and a possible increase in self-efficacy after success in the future [16]. Goal setting will be applied from week 2 till week 4. At the beginning of these three weeks, participants will be provided four goals. They need to choose one specific week goal, which they need to reach every day during that week. In case of success, a bonus of 150 points will be won at the end of the day. Earlier research has indicated that self-monitoring is the most successful behavioral change technique in energy balance-related intervention research [15]. Participants need to enter every consumed snack in the app and can monitor their snack intake in a weekly report. Based on the self-monitoring, they will have more insight into and awareness of their snacking behavior. This week-report can create elaboration about the snacking behavior.

To increase participants’ motivation to play our app, the most significant game motivations and dynamics based on the cross-sectional REWARD study, focus group research and previous studies (REWARD unpublished data, [46‐48]) were included. Challenge (i.e., to push yourself to a higher level of skill or personal accomplishment), competition (i.e., proving who has the best skills and can react and think the fastest), and social interaction (i.e., playing together with others both online as offline) were found to be important motivations for adolescents to play games. Furthermore, in the cross-sectional REWARD study, adolescents indicated the following characteristics as most crucial: (1) competitive aspects/playing against other people/leader board rankings, (2) cooperation/working together to reach goals, (3) different story outcomes based on your player actions, (4) ‘levelling up’ a game character, and (5) earning points or other rewards (unpublished data). Earlier studies also highlighted the importance of cooperation, and rewards as game dynamics [46‐48]. Baranowski and colleagues [49] and Peng [50] indicated the importance of story lines to increase immersion and intrinsic motivation to play games. Moreover, the inclusion of three psychological needs based on the Self-Determination Theory [46, 49, 50], namely autonomy, competence, and relatedness, have been linked to increased intrinsic motivation to play games. From this viewpoint, it is necessary that participants can make their own choices in a game that lead to different endings (i.e., autonomy), can customize their own avatar according to their preferences (i.e., autonomy and relatedness), have indicators for their achievements as this better satisfies players’ needs of competence.

The primary outcome consists of the Healthy Snacking Index. A Food Frequency Questionnaire (FFQ) was developed to measure average snack consumption in adolescents via 28 snack items. This new FFQ is based on an earlier FFQ instrument for children developed by Huybrechts and colleagues [51], but includes portion sizes that match adolescents’ intake patterns. Adolescents will indicate their answer in a list of six frequency categories, namely never or seldom; 1–3 days/month; 1 day/week; 2–4 days/week; 5–6 days/week; every day [51]. The FFQ contains four to six daily portion size categories per snack item and a list of common standard measures as examples. Snacks are in this study defined as all food items consumed outside (>30 min) of breakfast, lunch and dinner, in accordance to Rodriguez and Moreno’s definition of snacking [52]. The 28 snack items assessed in the FFQ are: dried fruit, fruit, raw vegetables, nuts and seeds, chocolate and pralines, candy bars, candy, dry cookies, other cookies such as chocolate cookies, breakfast rolls, pastries, breakfast cereals, unsweetened yoghurt, sweetened yoghurt, pudding, mousses, ice-cream, popsicles, sandwiches with sweet or savoury spread, cheese or meat cubes, chips and similar products, other savoury snacks such as bread sticks, sausage/cheese rolls and pizza, other fried snacks such as spring rolls and cheese croquettes, fries, kebab, hamburgers and pasta cups.

The classification of snacks and drinks into healthy and unhealthy was based on the nutrient profiling model as developed by the UK NP Ofcom model [40]. This model calculates for each food product a score that represents its healthiness In order to apply this model, an average nutrient composition per FFQ category was needed to apply this nutrient profile model. The average nutrient composition per FFQ category was calculated by averaging the nutritional composition (obtained from the Belgian food composition table [39] expressed per 100 g) of the most frequently consumed food items by adolescents within that category, as reported in the HELENA study [53]. The average energy, sugar, and fat and sodium intakes per FFQ category were then calculated by multiplying the amounts (g) of the food consumed and the average nutritional values expressed per g (the average values per 100 g divided by 100) [39]. According to the average amount of sodium, sugar, fat and kJoules, fibre, proteins and the portion of fruit and vegetables or nuts present in the product the NP model gives the food product a score, if a food product scores more than 4 points it was considered unhealthy. The FFQ items crisps, other salty snacks, sausage/cheese rolls and pizza, other fried snacks, fries, hamburgers, cheese or meat cubes, icecream, popsicles, breakfast cereals, pudding, sandwiches with sweet or savory spread, mousses, chocolate, candy bars, candy, dry cookies, other cookies, breakfast rolls and pastries were considered to be unhealthy.

The daily intake of each FFQ category will be obtained by multiplying the frequency of consumption with the quantity of consumption per week (g) divided by 7. These daily intakes, will then be summed to obtain the daily intake of healthy snacks (g), and unhealthy snacks (g). Finally a health index for snacks will be calculated: (gram healthy snacks/‘gram healthy snacks + gram unhealthy snacks))*100. The reliability and validity of this FFQ to assess the healthy snack ratio for interventions purposes were tested and reported elsewhere [De Cock unpublished]. The healthy snack index’ reliability was good and the validity was acceptable.

At the pretest, the principal of every school will be asked to complete a questionnaire assessing the food policy at school. This school management questionnaire is based on the school management questionnaire of the ENERGY project [60]. Since the REWARD project only focuses on healthy food choices only the parts of ENERGY questionnaire that were related to food facilities are used. The questionnaire comprises of the following sections: general characteristics of the school, the physical environment (opportunities to eat and drink at school), the political environment concerning regulations and practices pertaining to food and drinks and a last section specifically on vending machines (typical sale of snacks and soft drinks).

The purpose of the REWARD process evaluation is to describe 1) the reach of the intervention (incl. recruitment of sample); 2) degree of dose delivered by researchers (limited in our intervention as we provide one app to all participants) and dose received (exposure and satisfaction) by adolescents; and 3) contextual factors that may have influenced dose delivered and/or dose received; in order to get insight into intervention effects. Table 2 presents an overview of the measurements of different process evaluation concepts.

During the intervention, weekly process evaluation moments (15 min) will be included in the classroom, these will be implemented by the researchers and will be audiotaped. A process evaluation questionnaire will be added to the adolescent questionnaire at the post-test to assess participants’ engagement and use of the app, and the interaction between peers resulting from playing the app. In addition, adolescents’ dose received and satisfaction will also be assessed through logs of the app. Contextual factors will be measured through a post-test questionnaire for adolescents, the principals of the schools (both control and intervention) and involved teachers.

After every data collection wave (baseline, post), adolescent questionnaires will be scanned using the scanning software package TELEForm (version 6.1, Cardiff Software Inc., San Marcos, California, USA). Afterwards data will be translated into an SPSS file.

Descriptive analyses will be used at the individual level to compare the intervention and control group at baseline. Linear mixed models taking into account the clusters (child-class-school) will be used to evaluate differences in the primary and secondary outcome variables. In addition, subgroup analyses, and mediating and moderating analyses will be conducted to provide more insight into intervention effects. The process data (including objectively measured data and self-reported data) will be translated in a variable indicating the degree of intervention implementation and will be linked to the effect evaluation.