Youth’s social network structures and peer influences: study protocol MyMovez project – Phase I

This project uses approaches and theories centered around the influence of the social environment to investigate how social influence agents can be identified, the relative importance of selection and influence processes, and what distinguishes social influence agents compared to other peers (i.e., ‘followers’). Ample empirical studies have proven the strong influence of peers on young people’s consumption behavior and physical activity [13‐15]. For example, studies showed that youth follow and model each other’s eating behavior and that peers set guidelines or ‘social norms’ for others’ food choice and intake [16, 17]. In addition, peer groups are found to play a role in shaping physical activity behavior and youth are more active when they are together with peers [15, 18]. Social network interventions integrate peer influence processes in their campaigns to change behavior throughout the social network [8]. Yet, more research is needed to unravel how social influence agents can be identified and selected.

To date no systematic examination has been conducted on the selection process of social influence agents in peer-driven intervention studies. For example, most research on social networks are based on selecting influence agents who receive the most ‘friendship’ nominations; however, it is unclear whether this selection method is justified in the context of energy intake and expenditure [17]. Selecting influence agents based on ‘advice’ or ‘respect’ nominations (or a combination between these questions) [19] might identify other persons holding influential positions in the network compared to friendship queries. Also, these types of questions can be regarded as general and not directed to specific health behaviors, whereas it might be more useful to ask specific nomination questions (e.g., ‘with whom do you talk about PA?’ or ‘who is modeled a lot for their eating behavior?’) to identify social influence agents that exert power on health behaviors. In addition, there are different ways of analyzing the position or ‘centrality’ of individuals within networks. For example, a person receiving the most nominations has a high ‘in-degree’ centrality. Although this method is commonly used, it might be that ‘closeness’ centrality is a more effective measure because it describes how long it takes to spread information from that person to others (i.e., it takes into account the shortest paths a health message has to travel to reach the entire network). Another option would be to focus on ‘betweenness’ centrality, which indicates whether a person is important for linking different individuals or subgroups/clusters of individuals [20, 21]. To determine which individuals are the ideal social influence agents for the targeted behaviors of the MyMovez project, the selection process of social influence agents is investigated by focusing on different network centrality measures and by examining traditional (advice, leadership, friendship, etc.) and specific nomination items with regard to communication strategies, energy intake and expenditure.

It is also essential to identify which factors determine a person’s network position and which determinants mediate or predict engagement in energy behaviors. For example, research has shown an important role of ingratiation with regard to social normative and modeling behavior [13, 14]. That is, people try to influence each other to become part of a group and to be liked by conforming to social norms or modeling behavior. However, not all individuals are equally susceptible in doing so. Therefore, psychological determinants related to affiliation purposes and uncertainty reduction are collected such as need to belong, fear of negative evaluation, self-esteem, and athletic competence [13]. Nevertheless, research has focused mainly on characteristics of the followers and less on those of the social influence agents. There are limited indications that socially warm persons exert larger social influence than persons acting cold and that social influence agents have leadership status [8, 22]. To examine the characteristics of the influencers, Phase I also examines factors such as opinion leadership, public individuation, pro-social behavior, and participants are asked to profile themselves by choosing seven personality traits that describe themselves best [8, 21, 23, 24]. In addition, information about the social and physical environment is gathered that co-determine youth’s network positions and health behavior such as socio-economic status, social support, accessibility and availability of foods, and PA facilities [25].

The MyMovez project also uses a combination of renowned approaches and theories that predict behavioral change and provide insights in the performance of health behaviors and dissemination in social networks. For example, the constructs of the widely used Theory of Planned Behavior (TPB) are examined. The TPB has the general idea that when people have a positive attitude toward behavior and their social environment is supportive while they also feel they are capable and in control of their behavior, their intention to perform the behavior will be strong [26]. Importantly, people often fail to carry out healthy behaviors despite their positive intentions to eat healthy and engage in physical activity [27]. Therefore, additional constructs contributing to young people actually carrying out targeted behaviors (e.g., motivation [28, 29], self-persuasion [30], injunctive and descriptive norms [31], social modeling and impression management [32, 33]) stemming from the Diffusion of Innovation Theory, Social Normative approach [34], Self-Determination Theory [28] and the Fogg Behavior Model [29]), are examined to deepen the understanding of youth’s energy intake and expenditure. These insights from Phase I will contribute to develop an effective social network campaign for Phase II.

A practical and modern way to conduct large scale research among youth is via smartphones, because it is an efficient, low-cost and less time-consuming method compared to traditional paper-and-pencil questionnaires. In addition, objective measures of health and real-time data can be collected. Furthermore, the social context can be incorporated by facilitating communications via social network functionalities. Part B of this comprehensive research project develops the Wearable Lab, existing of a highly innovative smartphone-based research application (MyMovez app) connected to an activity tracker. The research technology enables examining youth’s social network structures together with energy balance-related behaviors, media consumption, socialization processes and psychosocial determinants of behavior. The smartphone and activity tracker jointly collect self-reported and objective data on daily randomized and fixed time points. Each participant receives the Wearable Lab during each data wave for a period of 7 consecutive calendar days including the weekend.

The Wearable Lab is developed in an ongoing iterative process by means of focus group sessions as well as evaluations via the MyMovez app during the data collection waves. In 2015, six focus group sessions were conducted at primary (N = 30; 10-11 y/o) and secondary schools (N = 27; 12-13 y/o); with groups of boys or girls only and one mixed group at each school. Discussions were videotaped and led by trained researchers, and involved topics such as likes and dislikes of health, research apps and wearables, the design and attractive functionalities or features of the app, wearables and behavior tracking, and the project name. Outcomes contributed to choosing the name ‘MyMovez’ for the research project and co-creating the design and features of the app such as an avatar, game and social media platform called Social Buzz. At the end of 2016, six focus groups were conducted at two secondary schools ((N = 14; 12-14 y/o and N = 28; 12-14 y/o, respectively); with groups of boys or girls only and one mixed group at each school) to gather information about their social media use and to test the implementation of the social network functionality Social Buzz. All together, these focus group studies provided useful insights on youth’s attitudes concerning the design of the research application and practical information about the use of the Wearable Lab.

During Phase I, novel procedures and approaches to process and analyze data are explored. In addition, collaborations are set up, for example, to process photo data, analyze youth’s physical environment by connecting real-time data with Geographic Information System (GIS) measures, and analyze chat conversations from the Social Buzz. This also requires up-to-date safety and data management procedures according to legal and ethical standards. In sum, the Wearable Lab provides an innovative research methodology to collect data that increases the feasibility to implement and evaluate health campaigns.

The research project is conducted according to the guidelines described in the Declaration of Helsinki and all procedures were approved by the ethical committee of the European Research Council (617253) and the Faculty Social Sciences, Radboud University, Nijmegen in the Netherlands (ECSW2014-100614-222).

A power analysis was performed based on results from a previous cross-sectional longitudinal study that investigated energy balance behaviors among Dutch youth [35]. It was assumed that obesity inducing risk behaviors are present among approximately 40% of adolescents, and with a statistical significance level of .05 and 80% power, a sample size of 600 participants would be sufficient for the cross-sectional and longitudinal study. This sample size was also deemed appropriate for social network analyses, which require approximately 20-25 complete classroom networks. Assuming that Dutch school classes exist of an average of 25 kids, this would produce a sample between 500 and 625 students. Taking into account a conservative 40% attrition rate [36] we aimed to include 850 participants and oversampled the high school population because this age group is more likely to change schools or classes or drop out while followed during a period of three years. In the end, this led to recruitment of a sample of 953 participants.

All (sub)urban schools following a regular education program were eligible for participation ranging from primary school level, and lower vocational training to secondary school. All schools located near the Nijmegen area were invited to participate in the MyMovez project. Further, Dutch Public Health Services (GGD) of each province were contacted to promote the MyMovez project at their regions. In addition, schools were invited to participate via personal contacts of students and researchers. This random procedure resulted in the participation of 21 (sub)urban schools throughout the Netherlands for Phase I of the MyMovez project. See Fig. 1 for a flow diagram of the recruitment procedure.

After obtaining (written) consent of the school’s principals, parents/legal guardians received an information letter and/or e-mail via the school announcing the project’s goals, planning, execution and safety procedures. Parents/legal guardians could enlist their child to the MyMovez study by giving consent on either a paper-and-pencil or digital form on the MyMovez website (www.​mymovez.​nl). The MyMovez website provides easily accessible information and updates about the project, team and progress (e.g., by a video illustrating how a day looks like for a participant, Q&A page, photos, and infographics). Researchers also visited schools to explain and promote the project in school classes and to hold information meetings for teachers, student mentors, and parent advisory boards.

Despite the schools’ and researchers’ efforts to promote the MyMovez project, parents/legal guardians did not always take notice of the research (e.g., the response rate at some schools was < 5%), resulting in disappointed pupils who wanted to participate on the day the research materials were handed out to their classmates. These pupils were given the opportunity to contact their parents/legal guardians to give their consent via the website or hand in a written consent form after the school break. In addition, pupils had the opportunity to participate during each of the following data collection waves, which has thus far resulted in different sample sizes for each data wave (in combination with drop out cases): n = 792 in Wave 1; n = 852 in Wave 2; n = 824 in Wave 3 and n = 671 in Wave 4. It is expected that the participation rates for Wave 5 (February 2018) will differ as well due to attrition and new participations.

In this comprehensive project, the innovative research tool Wearable Lab is developed to collect a unique longitudinal dataset according to the highest standards in ethical and legal safety regulations. The smartphone with the MyMovez research app is connected to an activity tracker that both generate different types of data that are transferred wirelessly to a secured data server. Data types are self-reported and peer-reported (i.e., short surveys and sociometric queries, respectively) and experience sampling data, Social Network Beacon data, real-time physical activity data (i.e., steps and cycling) and location information, photos and chat conversation data on individual and class level by the social media platform Social Buzz (added to the MyMovez research app since January 2017). These data are used to examine youth’s media and social network exposure, socialization experiences, psychosocial determinants of behavior, and energy balance-related behaviors.

Social network information is gathered by asking youth to nominate peers with various characteristics (i.e., who is a leader?) as well as by recording chat connections and reports between participants on the social media feature Social Buzz in the MyMovez research app. In addition, social network information is also collected using the Social Beacon Network, in which the smartphones detect and register the research phones of interaction partners during multiple time points a day via Bluetooth. These data collection methods collectively provide insights into micro-processes (with multiple measurement points during a day) as well as macro-processes (multiple assessments across a school year). Furthermore, these methods provide information about youths’ perceived networks (i.e., sociometric items) as well as more objective assessments of networks based on physical proximity (i.e., the Social Beacon Network).

Furthermore, the Wearable Lab allows for the collection of other novel types of data such as photo queries and global positioning system (GPS) data. For example, photo data are coded and used for machine learning purposes (e.g., coding of food or media use pictures) contributing to the examination of youth’s diets, social media behavior and other daily activities [37]. The GPS data provides information about cycling¹ (the activity tracker only counts the amount of steps and the intensity) and the physical environment, by linking these to Geographic Information System (GIS) data in which youth are exposed to food and PA related facilities [38]. All together, the Wearable Lab allows for the collection of a comprehensive set of data that provides the opportunity to address all of the project’s research goals.

Most schools were visited in advance to promote the research project among the participants before the project started. On the starting day of the initial data collection, trained researchers handed out the research materials (i.e., smartphone and activity-tracking bracelet) and instructed the participants in class or in small groups of approximately 5-8 school children about the use of the Wearable Lab. Specifically, it was explained how to turn the phone on and off, how to use the social chat and avatar, how to check whether the activity tracker is connected to the smartphone, how to check whether the batteries are full, and what to do when the phone displays an error. Participants were also told that they should wear the activity bracelet for the entire week and that the activity bracelet is water resistant. Participants were also shown a short movie clip about how a research day with the Wearable Lab looked like and they were handed a code of conduct showing how to behave in the Social Buzz social media platform. Next, participants signed an assent form and an additional agreement in which they promised to handle the research equipment with care during the 7-day period they were borrowing the research materials. Participants were informed that they could withdraw from the research at any moment, and that their data can be deleted if they or their parents want to. After the instructions they were provided with the opportunity to ask questions and could return to their classrooms. Each year (during Wave 2, 4, and 5), participants’ weight and height is measured individually in a separate room by a trained researcher. Depending on the school schedule, this is done after the instruction or participants are called a second time from their classroom.

On the starting day of each measurement period, participants receive questions after their school day ended. During the following days, they all receive questionnaires at random time points during the day, but never during school hours or at night between 7:30 PM and 7:00 AM. At the end of the measurement period, reminders are sent to the participants when and where they need to hand in the research materials at their schools. Researchers collect the research materials at the schools.

Data collection involves explicit self- and peer-reported survey measures (short digital questionnaires administered via the smartphones), objective measures (photos, real-time data, activity tracking, cycling, Social Beacon Network and Social Buzz data), and researcher-assessed measures (participants’ height and weight). Most survey measurements used in the MyMovez project are existing validated questionnaires, translated into Dutch. Some (parts of) questionnaires are adapted from validated questionnaires and tailored to identify specific behaviors. In cases in which no validated measures are available, new items or scales are developed for the MyMovez project. All measurements used in the MyMovez project described in the following paragraphs. The survey instruments are listed in Table 1 and described into detail in Additional file 1.

Outcome variables (e.g., social network characteristics, physical activity, drink and snack consumption), as well as TPB and motivational measures, are measured at each data wave assessment. Other measures are divided over the different data waves. A detailed scheme can be provided upon request to the corresponding author. Participants’ name, birth date, gender and postal code were provided by the school² to create an ID number in the back-end of the MyMovez research application.