Wageningen University – Department of Social Sciences Chair Group Strategic Communication

The Effect of Physical Activity Apps on Physical Activity Behavior and Users’ Evaluation of Physical Activity Apps

31 March 2016

MSc Thesis Strategic Communication (CPT-81333) MSc Applied Communication Science

Laura Ploumen 1st Supervisor: Jorinde Spook 2nd Supervisor: Edith Feskens

Abstract Background: A new development in the promotion of health and physical activity (PA) is the use of PA apps. This development brings along a new field of research. Despite the broad range of research in the previous years, there are still research gaps with regard to the effect of PA apps on behavior determinants and with regard to the users’ evaluation of PA apps. Objective: Determine what the effect is of the use of PA apps on PA and its’ determinants self-efficacy, outcome expectations, socio-structural factors, and self-regulation. In addition, an objective is to find out how Dutch adults evaluate PA apps and why they use it or do not use it. Methods: The Social Cognitive Theory (SCT) was the theoretical framework for this study. A cross-sectional study design was used, with 251 participants. Differences in determinants, PA and PA enjoyment between app users (N=63) and non-users (N=188) were measured using ANCOVA’s, adjusting for the covariates age and education. As exploratory research, mediation analyses were performed to get insight into the underlying mechanisms of the SCT model. Several apps were evaluated using a system usability score, an evaluation of behavior change techniques and open questions. Results: App users scored significantly higher than non-users on self-efficacy, outcome expectations, and selfregulation. Furthermore, app users scored higher on PA enjoyment. No significant difference was found for PA. The PA apps and their behavior change techniques in this study were evaluated positively in general. Selfmonitoring was the main reason why PA apps were used and was evaluated as a very positive aspect of PA apps. Negative aspects and possible improvements for PA apps mainly related to user-friendliness. The main reasons for not using PA apps were next to a lack of interest and need, the unawareness of the existence of such apps. Conclusion: The current study demonstrated that app users have a higher self-efficacy level, higher outcome expectations, a stronger self-regulation level, and they enjoy PA more than non-users. Nearly one-fifth of smartphone users is not aware of the existence of PA apps. Considering the differences in determinants and PA enjoyment, it is important to raise awareness of the existence of PA apps. In addition, it is important for app developers to focus on self-regulation, on user-friendliness, and the option to compare PA performances over time and with others.

I

Preface In your hands (or on your screen) is the study ‘The Effect of Physical Activity Apps on Physical Activity Behavior and Users’ Evaluation of Physical Activity’. This study is my master thesis for the master program Applied Communication Science at Wageningen University. I was engaged in writing this thesis from October 2015 to March 2016. In Applied Communication Science, the focus lies on the use of communication strategies relating to problem solving and innovation to improve the quality of life. During the program, I chose health as the central life science. I am interested in innovations related to health promotion, as these innovations have the potential to improve people’s overall health. I wondered how smartphones, a relatively new technology, could improve physical activity by the use of physical activity apps. Supervised by Jorinde Spook from the chair group Strategic Communication and by Edith Feskens from the chair group Human Nutrition, I studied physical activity apps both from a theoretical perspective and an evaluative perspective. I would like to thank Jorinde Spook for her meaningful insights and her excellent guidance and support during the process and I would like to thank Edith Feskens for her feedback from a life science perspective and for giving me the chance to distribute my questionnaire in the research panel of the chair group of Human Nutrition. Furthermore I would like the second reader, Emely de Vet. I also would like to thank the respondents for their contribution to this thesis, and thank my family and friends for their support during this period. I hope you will enjoy reading this thesis!

Laura Ploumen Wageningen, March 7, 2016.

II

Contents Abstract .................................................................................................................................................................... I Preface .................................................................................................................................................................... II List of Tables............................................................................................................................................................ V List of Figures .......................................................................................................................................................... V 1. Introduction ........................................................................................................................................................ 1 2. Theoretical Background ...................................................................................................................................... 4 2.1 State of the Art ............................................................................................................................................. 4 2.2 Social Cognitive Theory ................................................................................................................................. 4 2.3 Social Cognitive Theory and Physical Activity ............................................................................................... 5 2.4 Quantified Self .............................................................................................................................................. 6 2.5 Research Gap ................................................................................................................................................ 6 3. Research Objective, Aim and Research Questions .............................................................................................. 8 4. Methods .............................................................................................................................................................. 9 4.1 Participants & Study Design .......................................................................................................................... 9 4.2 Procedures .................................................................................................................................................... 9 4.3 Measures ...................................................................................................................................................... 9 4.3.1 Demographic Variables ......................................................................................................................... 9 4.3.2 Environment ........................................................................................................................................ 10 4.3.3 Behavior .............................................................................................................................................. 10 4.3.4 Self-Efficacy ......................................................................................................................................... 10 4.3.5 Outcome Expectations ........................................................................................................................ 11 4.3.6 Self-Regulation .................................................................................................................................... 11 4.3.7 Socio-Structural factors ....................................................................................................................... 11 4.3.8 Physical Activity Enjoyment ................................................................................................................. 12 4.3.9 Evaluation ............................................................................................................................................ 12 4.4 Analysis ....................................................................................................................................................... 13 5. Results ............................................................................................................................................................... 14 5.1 Descriptive statistics ................................................................................................................................... 14 5.1.1 Demographic Characteristics of the Study Sample ............................................................................. 14 5.1.2 Environment ........................................................................................................................................ 14 5.1.3 Physical Activity ................................................................................................................................... 15 5.1.4 App Use ............................................................................................................................................... 16 5.2 Differences between App Users and Non-Users ........................................................................................ 16 5.2.1 Differences in Demographic Characteristics ....................................................................................... 16 5.2.2 Group Comparison: App Users vs Non-Users ...................................................................................... 17 5.3 Exploratory Research: Mediation Analyses ................................................................................................ 19 5.4 Results Evaluation ....................................................................................................................................... 21 III

5.4.1 Quantitative Evaluation per App ......................................................................................................... 21 5.4.2 Qualitative Evaluation of Physical Activity Apps ................................................................................. 22 5.4.3 Physical Activity Enjoyment ................................................................................................................. 25 6. Discussion .......................................................................................................................................................... 26 6.1 Principal Findings ........................................................................................................................................ 26 6.1.1 Differences between App Users and Non-Users ................................................................................. 26 6.1.2 Health Norms ...................................................................................................................................... 27 6.1.3 Exploratory Research........................................................................................................................... 27 6.1.4 Evaluation of Physical Activity Apps .................................................................................................... 28 6.2 Limitations .................................................................................................................................................. 30 6.3 Practical Implications and Recommendations ............................................................................................ 31 6.4 Theoretical Contributions and Further Research ....................................................................................... 32 6.5 Conclusion................................................................................................................................................... 33 References ............................................................................................................................................................ 34 Appendix A: Questionnaire ................................................................................................................................... 39 Appendix B: MET values ........................................................................................................................................ 52 Appendix C: ‘Other Apps’ ...................................................................................................................................... 54

IV

List of Tables Table 1. Characteristics of the study population .................................................................................................. 14 Table 2. Experiences of the study population with their environment ................................................................ 15 Table 3. The amount of minutes per week per activity ........................................................................................ 15 Table 4. Percentage of participants who meet the health norms ........................................................................ 15 Table 5. Evaluation of behavior change techniques in Runkeeper and Strava ..................................................... 22 Table 6. Participants’ reasons to use a physical activity app and reported positive aspects of physical activity apps. ........................................................................................................................................................ 23 Table 7. Participants’ reasons for not using a physical activity app. ..................................................................... 24 Table 8. Negative aspects of physical activity apps .............................................................................................. 24 Table 9. Proposed improvements for physical activity apps. ................................................................................ 25

List of Figures Figure 1. Screenshots of Runkeeper ....................................................................................................................... 2 Figure 2. Screenshots of Runtastic .......................................................................................................................... 3 Figure 3. Screenshots of Strava ............................................................................................................................... 3 Figure 4. The Social Cognitive Theory ..................................................................................................................... 5 Figure 5. The number of users per PA app ............................................................................................................ 16 Figure 6. The mediating role of self-regulation goals in the effect of self-efficacy on PA .................................... 19 Figure 7. The mediating role of self-regulation plans in the effect of self-efficacy on PA. ................................... 19 Figure 8. The mediating role of self-regulation goals in the effect of outcome expectations on PA. ................... 20 Figure 9. The mediating role of self-regulation plans in the effect of outcome expectations on PA ................... 20 Figure 10. The mediating role of socio-structural factors in the effect of self-efficacy on self-regulation goals.. 20 Figure 11. The mediating role of outcome expectations in the effect of self-efficacy on self-regulation goals.. . 20 Figure 12. The mediating role of socio-structural factors in the effect of self-efficacy on self-regulation plan. .. 21 Figure 13. The mediating role of outcome expectations in the effect of self-efficacy on self-regulation plan. ... 21

V

1. Introduction It is well known that physical activity (PA) has a positive effect on multiple health outcomes, such as blood pressure, cognitive functions, and body weight. PA prevents many chronical diseases such as obesity, stroke, heart and vascular disease and diabetes type 2 (Hildebrandt, Bernaards, & Stubbe, 2013). Despite the benefits PA brings, 32.9% of Dutch adults do not reach the Combinorm, the combined guidelines for physical activity in the Netherlands. This combined guideline consists of the Nederlandse Norm Gezond Bewegen (NNGB, which recommends at least 30 minutes of moderate PA per day for at least five days a week for adults) and the Fitnorm (which recommends at least 20 minutes of intensive PA for at least 3 days a week for adults) (Hildebrandt et al., 2013). In the Netherlands, 48.3% of adults are overweight, which means that they have a Body Mass Index (BMI) of 25 or higher. In addition, 12.7% of adults in the Netherlands have a BMI of 30 and higher and are therefore called obese. Among other factors, a lack of PA is one of the factors that causes of these diseases (Brink & Blokstra, 2014). Besides the health burden of overweight and obesity, the economic burden is high: in Europe, obesity leads to 7% more direct health care costs (related to medical resource utilization including labor costs and pharmaceutical costs) and 3-4% more indirect health care costs (relating to the cessation or decrease of work productivity) (Boccuzi, 2003). Next to that, obesity can lead to social impacts such as discrimination in employment (Puhl & Brownell, 2001) or other social situations which can lead to rejection, shame or depression (Wellman & Friedberg, 2002). With all those people not meeting the guidelines for PA, a need arises for new ways to promote PA (Dallinga, Mennes, Alpay, Bijwaard, & de la Faille-Deutekom, 2015). One of these new ways to promote PA and health is the use of modern information and communication technologies. The use of health services through the internet and related technologies are called E-Health (Sarkar, Sanders, Kelleher, & Chisolm, 2015). Internet and technologies related to it, such as smartphones, have big potential in terms of promoting health (Krishna, Boren, & Balas, 2009). The use of information technology has also some disadvantages such as expensiveness, potential confidentiality, ethical issues, limited proof of long-term effects and it is difficult to predict engagement accurately. Besides that, it can be difficult to understand for users (Corcoran, 2007). Despite these disadvantages, the use of information technology in health promotion has many advantages that make the use attractive (e.g., saving time, the ability to tailor information to individuals, the participatory and interactive nature, feelings of autonomy, providing tailored feedback, and the user-friendliness like touch screen or voice activation (Corcoran, 2007)). A device that is often used and is effective for health interventions is the smartphone (Krishna, Boren, Balas, 2009; Yoganathan & Kajanan, 2013). In the Netherlands, 67% of the people own a smartphone (GFK, 2013). A smartphone is often carried the entire day (Spook, Paulussen, Kok, & Van Empelen, 2013), and data can be assessed anywhere at any time (Middelweerd, Mollee, van der Wal, Brug, & te Velde, 2014). Moreover, smartphones can track someone’s PA, provide tailored feedback to the user, and allow social support (Middelweerd et al., 2014). Using a mobile app for tracking PA is popular. In the Netherlands, 19% of people who use health services, make use of such a mobile app. (Krijgsman et al., 2015). People who exercise on a regular 1

basis use these applications more than those who do not exercise on a regular basis and higher educated people are more frequent users than lower educated people and elderly (Krijgsman et al., 2015). PA apps are mostly used by runners (i.e., 46%) to measure their training. Currently (October 2015), the most popular PA apps are Runkeeper, Runtastic, Strava and 7 Minute Workout (Google Play Store & iTunes App Store). The PA apps Runkeeper, Runtastic and Strava will be the focus of this study. These apps make use of a GPS tracking system, by which they can show the user how long and how far their workout was and how many calories were burned. Additionally, they have possibilities such as settings goals, sharing workouts via social media and competing with friends. The app ‘7 Minute Workout’, which is also part of the top 10, differs from the other apps. Runkeeper, Runtastic, and Strava make use of GPS tracking and are able to measure PA (e.g. time, distance, calories). In contrast to the other three apps, 7 Minute Workout does not make use of a GPS tracking system, it only leads the user through a 7-minute training. It can track whether the user has done his or her workouts every day, however, it cannot see the progress of PA. Therefore, this study will only focus on Runkeeper, Runtastic, and Strava. To give an idea of the layout and the content, these apps will be described shortly.

Runkeeper Runkeeper is a mobile app which can be found in the Google Play Store and in iTunes. The app description indicates that Runkeeper is ‘’the simplest way to improve fitness, whether you’re just deciding to get off the couch for a 5k, biking every day, or even deep into marathon training’’. It can measure speed, route distance, elevation, calorie burn and more and makes use of GPS. The user can connect with friends and see their statistics. Besides running, there are multiple other sports to choose. Runkeeper Elite has more possibilities than the free version (Runkeeper, 2016). In the present research, the free version will be used as this is easier to assess. Figure 1 shows screenshots of the app.

Figure 1. Screenshots of Runkeeper (Runkeeper, 2016)

2

Runtastic Running Runtastic Running can be found in the Google Play Store and in iTunes. Just like Runkeeper, it can track speed, route, distance, calories and more by using GPS. The user can connect with friends and see their statistics. Besides running, there are multiple other sports to choose. Runtastic Running is available for free, but there is a paid version, Runtastic Running PRO, which has additional features (Runtastic, 2016). In the present research, the free Runtastic version will be used (referred to as Runtastic from now on). Figure 2 shows screenshots of the app.

Figure 2. Screenshots of Runtastic (Runtastic, 2016)

Strava Running and Cycling This app is available for free in Google Play Store and iTunes. Strava is meant to track running activities and cycling activities. Just like Runkeeper and Runtastic, it can track statistics like the pace, distance, elevation, and calories burned during a training. The user can connect with friends and motivate them (Strava, 2016). Strava Running and Cycling is referred to as Strava from now on. Figure 3 shows screenshots of the app.

Figure 3. Screenshots of Strava (Strava, 2016)

3

2. Theoretical Background 2.1 State of the Art Until today, research has shown that the use of mobile apps to promote health and PA can be effective (Van Drongelen, Boot, Hynek, Twisk, Smid, van der Beek, 2014; Fanning, Mullen, & McAuley, 2012; Dallinga et al., 2015; King et al., 2013; Kirwan, Duncan, Vandelanotte, Mummery, 2012). Furthermore, by using the Extended Unified Theory of Acceptance and Use of Technology (UTAUT2), researchers have found predictors of continued use of a fitness app (Yuan, Ma, Kanthawala, & Wei Peng, 2015). Health and fitness apps are continuously used when the individual thinks the technology is going to help to perform an activity (performance expectancy), when the technology gives pleasure and fun (hedonic motivation), when the price value is good and when the technology is integrated into the persons’ habits (Yuan et al, 2015). More content-specific research has shown that students prefer apps that coach and motivate, that provide tailored feedback towards their own goals and apps that allow competition with friends (Middelweerd et al., 2015). Besides research on the effectiveness of PA apps and about preferences for PA apps, research has been done on the inclusion of behavior change techniques and theories in PA apps. Research on behavior change techniques is relevant because health behavior change interventions are proven to be more effective when they are based on health behavior change theories (Webb, Joseph, Yardley, & Michie, 2010). Results showed that the use of evidence-based behavior change techniques in PA apps is minimal (Middelweerd et al., 2014; Yang, Maher, & Conroy, 2015; Cowan et al., 2012). Feedback (Middelweerd et al., 2014), self-monitoring (Middelweerd et al., 2014; Payne, Lister, West, & Bernhard, 2015) and social support (Payne et al., 2015; Yang et al., 2015) were pointed to as the most used behavior change techniques in PA apps. These techniques are in line with the Social Cognitive Theory (SCT, Bandura, as cited in Yoganathan & Kajanan, 2015).

2.2 Social Cognitive Theory Social Cognitive Theory suggests that behavior of individuals originates from interaction between three components: a behavioral component, personal component and the environmental component (Bandura, 1989), and that individual behavior is predicted by self-efficacy, outcome expectations, socio-structural factors and selfregulatory mechanisms or goals (Bandura, as cited in Yoganathan & Kajanan, 2015). So Bandura states that there are four determinants of behavior. Self-efficacy refers to the belief someone has about his or her capabilities and competencies to execute a certain behavior (Bandura, 1998). Outcome expectations refer to the beliefs one has about the outcomes they produce with their actions (Bandura, 2004). The way someone sets standards on themselves for his behavior and how he responds to his own actions on a self-evaluative way is called self-regulation (Bandura, 1986 in Yoganathan & Kajanan, 2015) and socio-structural factors refer to social factors that enable individuals to enhance their PA behavior (Yoganathan & Kajanan, 2013). SCT has been applied to health promotion (Bandura, 1998; Bandura, 2004). The approach tries to help people to maintain a healthy lifestyle by supporting them through self-management of health habits. The approach works 4

on the demand side of human health, instead of the supply side which mainly concerns about health care access and reducing costs (Bandura, 1998). Bandura states that knowledge is the precondition for change. However, to adopt a new lifestyle and to maintain it, additional influences are needed. Therefore, the social cognitive approach addresses both socio-structural and personal determinants of health to explain behavior (Bandura, 1998). The main construct of SCT is self-efficacy, according to Bandura. Self-efficacy influences behavior directly, but also by its impact on the other constructs. Self-efficacy influences behavior through its impact on outcome expectations (physical, social and selfevaluative), socio-structural factors (facilitators and impediments for behavior), and self-regulation or goals (Bandura, 2004). The model can be seen in Figure 4.

Figure 4. The Social Cognitive Theory. Behavior is influenced directly by self-efficacy, and by indirectly by its impact on outcome expectations, socio-structural factors, and goals. Bandura (2004).

2.3 Social Cognitive Theory and Physical Activity Research has concluded that SCT is useful to explain PA (Rovniak, Anderson, Winnet, 2002; Young, Plotnikoff, Collins, Callister, & Morgan, 2014). Respectively, 55% and 31% of the variance in PA behavior were explained by SCT. This means that SCT is considered as a useful framework for the design of interventions, according to the recommendation of Baranowski, Anderson & Carmack (1998) (Young et al., 2014). Furthermore, Yoganathan & Kajanan (2015) showed that SCT can be translated into modern technology, such as smartphone fitness apps. They used a text mining approach, in which they looked for patterns in the text descriptions of fitness apps. Based on the four constructs self-efficacy, outcome expectations, self-regulation and socio-structural factors, they found that features that influence these four constructs, such as self-monitoring, cooperation, and normative influence can lead to app success (Yoganathan & Kajanan, 2015). App success can be defined as “heavily used, greatly valued and highly recommended fitness apps that fulfill the fitness needs of users” 5

(Yoganathan & Kajanan, 2013). With their study, they indicated that an app has the ability to promote effective and positive PA change. As explained in chapter 1, the popularity of PA apps can be explained by the wide reach and the tailored feedback they can give (Middelweerd et al., 2014), for instance on health and body. Another reason why PA apps are popular could be the so-called “quantified self”, which will be explained in the following section.

2.4 Quantified Self Tracking devices like PA apps is one of the segments that characterize the Internet of Things (IOT). This is a phenomenon that connects real-world objects such as the human body to the internet with tiny sensors (Swan, 2012). A mobile health tracking app like Runkeeper is an example of a multi-sensor platform that combines several sensing techniques. IOT makes it possible to measure health and body information, which is a phenomenon called the quantified self (Swan, 2012). People who measure health and body information, called self-trackers, see data as an objective resource that gives them visibility, information and action quickly. At the psychological level, empowerment and control make self-tracking attractive (Swan, 2009). Self-regulation techniques like reflecting on data, getting meaningful insights and finally, making positive changes are the ultimate goals of self-tracking (Choe, Lee, Lee, Pratt, & Kientz, 2014). Measuring health, body and training information is also called self-monitoring. The quantified self-phenomenon is therefore relevant for this study since self-monitoring plays a big role in many of mobile health and fitness apps (Yoganathan & Kajanan, 2013).

2.5 Research Gap Research on PA apps is still in its infancy and there are many aspects that need further research. Dallinga et al. (2015) suggest that there is still a knowledge gap with regard to the impact of PA apps on behavioral determinants such as self-efficacy, attitude, and social influence. These determinants are assumed to be important factors in predicting intention of behavior and therefore it is interesting to assess the impact of PA apps on these determinants (Dallinga et al., 2015). Yoganathan & Kajanan (2015) concluded in their research that apps have the ability to promote behavior change by influencing behavior determinants. However, this does not automatically imply that the influence of PA apps on behavior determinants really exists. Their text-mining approach was based on app descriptions provided by developers, while app developers’ descriptions may differ from users’ perceptions about the apps. Besides that, they did not measure PA behavior. Instead, they measured ‘app success’, defined as ‘heavily used, greatly valued and highly recommended fitness apps that fulfill the fitness needs of users’ (Yoganathan & Kajanan, 2013). Therefore, this study will address the users’ perspective with regard to PA apps. Additionally, there is a research gap with regard to users’ evaluation of PA apps (Dennison, Morrison, Conway, & Yardley, 2013). Dennison et al. (2013) created a preliminary checklist of valuable features and characteristics of health behavior apps, and focusses on the apathy, concerns and frustrations around health apps (Dennison et al., 2013). One important finding was that features of PA apps that require much effort and burden can negatively 6

influence the use of a PA app (Dennison et al., 2013). A limitation of the study was that it was performed in a small sample of 19 students and staff of a university. This study sample was not generable to the Dutch adult population. Therefore, this study will address user friendliness of PA apps by doing a users’ evaluation of the apps in this study. Besides, there is some more research needed about the positive aspects of physical activity apps. Apps are mostly seen as entertainment (Seiler et al., 2015). It is interesting to find out whether PA apps make physical activity more entertaining too. Therefore, physical activity enjoyment will be taken into account in this study as an extra factor. The research objective, aim and the research questions that will guide this study will be explained in chapter 3.

7

3. Research Objective, Aim, and Research Questions We aim to find out how PA apps can influence PA behavior of Dutch adults, as adults from the age of 18 are generally less physically active than children and youngsters (Rovniak, 2002), and the amount of people who are overweight is higher in adults (Brink & Blokstra, 2014). Besides, physical activity apps are mostly aimed at adults (Middelweerd, 2014). The research objective of this study is to determine what the effect is of the use of PA apps on PA and its’ determinants self-efficacy, outcome expectations, socio-structural factors, and self-regulation. In addition, an objective is to find out how Dutch adults evaluate PA apps and why they use it, e.g., to improve their self-regulation or just because they think it is fun. The insights that are gained with that will contribute to the existing literature about PA apps. Insights about the determinants and users’ evaluation of the apps could be very important for the improvement of PA apps. This study aims to find out how PA apps influence PA behavior, and determinants of PA: self-efficacy, sociostructural factors, outcome expectations, and self-regulation. Additionally, the research aims to find out why Dutch adults use PA apps such as Runkeeper, Runtastic, and Strava.

Two main research questions will be answered in this thesis:

1)

What is the effect of physical activity apps on Dutch adults’ physical activity and determinants of physical activity? -

What is the effect of physical activity apps on Dutch adults’ self-efficacy in relation to physical activity?

-

What is the effect of physical activity apps on Dutch adults’ socio-structural factors in relation to physical activity?

-

What is the effect of physical activity apps on Dutch adults’ outcome expectations in relation to physical activity?

-

What is the effect of physical activity apps on Dutch adults’ self-regulation in relation to physical activity?

2)

8

How do Dutch adults evaluate physical activity apps? -

Which PA app components are appreciated by Dutch adults?

-

To what degree are PA apps perceived as user-friendly by Dutch adults?

-

Do physical activity apps contribute to physical activity enjoyment?

4. Methods 4.1 Participants & Study Design A cross-sectional study was conducted in 2016. Distribution of the online survey was done via a research panel of the Division of Human Nutrition, called EetMeetWeet/NQplus, recruited in collaboration with the municipality health services Gelderland-midden and region Utrecht and distributed via the researchers’ own online and offline social network, via forums (for example a forum for runners) and via the Twitter channel of Runners World Nederland (a magazine). In total, 295 adults of 18 years or older participated in the present study. The final sample was 69.7% female and 30.3% male, with an average age of 46.5 (SD = 17.7, range 18 - 90). The average BMI was 23.8 (SD = 4.0). 44 participants were excluded from the analyses, because they were under 18 or because they reported unreliable answers in the questionnaire.

4.2 Procedures Participants were contacted by social media or via e-mail (the EetMeetWeet panel). A short introduction explained the purpose of the study, the time it took to complete the questionnaire (approximately 15 minutes) and made clear that the answers were treated anonymously. The first questions existed of demographic information: age, gender, educational level, length, weight and postal code. The participants were also asked whether they were a member of a sports club or sports center. The questionnaire then continued with questions about physical activity. After that, the participants were asked questions on the determinants self-efficacy, outcome expectations, socio-structural factors, and self-regulation. Then the questionnaire continued with the items about physical activity enjoyment. Next, the participant was asked to fill in whether he or she owned a smartphone and whether he or she used one of the physical activity apps in this study: Runkeeper, Runtastic, or Strava. At the end of the questionnaire, app users were asked to evaluate the app they used by filling in questions about the behavior change techniques, the system usability and giving their overall evaluation of the app. Nonusers were asked to give a reason why they did not use a PA app. At the end of the questionnaire, the respondents were thanked for their participation in the research, and an e-mail address was added in case they had questions or were interested in the results of the study.

4.3 Measures To construct a valid questionnaire, items were used from previous studies that addressed a similar topic. For each sub-question, one or more scales were used. The exact measures are explained in sections 4.3.1 to 4.3.9. The questionnaire is included in appendix A.

4.3.1 Demographic Variables To characterize the sample, demographic information was asked in the beginning of the questionnaire. Participants were asked to fill in their age, gender, educational level and their body length- and weight, to measure their BMI according to the formula BMI = kg / m2. 9

4.3.2 Environment To get some insight into the physical environment of participants, their postal code was asked. In addition, questions were asked about the perception of the participants about green spaces, sports facilities, biking- and walking paths, the safety of their neighborhood and whether their neighborhood was clean. These items were based on previous questionnaires (Giles-Corti & Donovan, 2002; Storm, Nijboer, Wendel-Vos, Visscher, & Schuit, 2006). In addition, the participants were asked to indicate whether they were a member of a sports club or a sports center. This additional descriptive data could be useful for the analysis.

4.3.3 Behavior Behavior, PA, in this case, was measured using the Short Questionnaire to ASsess Health-enhancing physical activity (SQUASH) (Wendel-Vos, Shuit, Saris, & Kromhout, 2003). This questionnaire contains items about commuting activities, leisure time activities, household activities and activities at work and school (Wendel-Vos et al., 2003). The answers were based on days per week, time per day, and intensity or pace. This questionnaire enables the researcher to measure the total minutes of physical activity per week and the activity score (total minutes per week * intensity) and also whether participants meet the Nederlandse Norm Gezond Bewegen (NNGB, recommends at least 30 minutes of moderately intensive physical activity for at least 5 days a week) and the fitnorm (which recommends at least 20 minutes of intensive physical activity for at least 3 days a week). By combining these two norms, the researchers can check whether participants’ physical activity meets the Combinorm. As a number of participants may be students, and as students in the Netherlands often have part-time jobs (Lok & Loog, 2014), activities at school and activities at work were asked separately. In addition, it was asked whether these activities were done on weekdays or on weekend days. Another aspect that was added to the original SQUASH was the part of the day in which leisure time activities took place. This was added because it would be interesting to see in which part of the day participants practice sports or other activities. The SQUASH was measured using a guide made by Wendel-Vos and Schuit (2004). This guide consisted of a syntax with a detailed explanation. As ‘activities at school’ was added for the current study, this was added to the original syntax file where needed. The original syntax file used codes for sports activities. In the current questionnaire, no coding was included, so the researchers had to assign the MET values manually. This was done by following the Compendium of Physical Activities by Ainsworth et al. (2000). For sports activities that were not included in the original compendium, decisions about MET values were taken by the researchers themselves. The used MET values for every sport activity can be found in Appendix B.

4.3.4 Self-Efficacy Self-efficacy refers to the belief someone has about one’s capabilities and competencies to execute a certain behavior (Bandura, 1998). Self-efficacy is the foundation of human motivation and action. When someone does not have the belief that one has the power to make changes with their actions, one will not have the motivation 10

to execute the action (Bandura, 2004). Self-efficacy was measured using the Dutch translation of the Exercise Self-Efficacy Scale (ESES, Nooijen, Post, Spijkerman, Bergen, Stam, van den Berg-Emons, 2013). The scale existed of 10 items and the original 4-point Likert scale was adapted to a 5-point Likert scale ranging from 1 (=strongly disagree) to 5 (=strongly agree). Internal consistency was good, α=.88.

4.3.5 Outcome Expectations Outcome expectations refer to the beliefs one has about the outcomes they produce with their actions (Bandura, 2004). Outcomes can be physical outcomes, such as material benefits or losses, but the outcome can also be social approval or disapproval, or self-evaluative outcomes such as satisfaction or self-worth (Bandura, 2004). Outcome expectations were measured using a 3-item scale that was used in a campaign research by Wageningen University and Research Centre (WUR) and the Dutch Institute for Sports and exercise (NISB) (NISB, n.d.). The scale was originally a 7-point Likert scale, but was adapted to a 5-point Likert scale ranging from 1 (=strongly disagree) to 5 (=strongly agree) so that it was in line with the other questions. Internal consistency was good, α=.89.

4.3.6 Self-Regulation The way someone sets standards on themselves for his behavior and how he responds to his own actions on a self-evaluative way is called self-regulation (Bandura, 1986 in Yoganathan & Kajanan, 2015). Self-regulation is crucial for an active lifestyle (Bandura, 2004). Self-regulation was measured using the Exercise Goal-Setting Scale (EGS) and the Exercise Planning and Scheduling Scale (EPS), both developed for a study on the relation between social cognitive variables and PA (Rovniak et al., 2002). The scales were translated into the Netherlands separately by the researchers and combined into one translation. The scales existed of a total of 20 items on exercise goals and plans and the Likert scale ranged from 1 (=strongly disagree) to 5 (=strongly agree). The Cronbach’s alpha of these goals items was α=.89 and for plan α=.82. Therefore, both scales showed good internal consistency.

4.3.7 Socio-Structural factors Socio-structural factors refer to social factors that enable individuals to enhance their PA behavior (Yoganathan & Kajanan, 2013). This determinant was measured using three concepts (Ridder & Lechner, 2004): the perceived social norm, direct social support, and social influence. This was indicated by the participant for their partner, friends, and acquaintances, family and colleagues or manager. For every concept, a question was asked. For example: ‘’My partner thinks physical activity is important’’. The Likert scale was again adapted to be in line with the other questions and ranged from 1 (=strongly disagree) to 5 (=strongly agree). Internal consistency was good, α=.81.

11

4.3.8 Physical Activity Enjoyment As physical activity apps are often seen as entertainment (Seiler et al., 2015), the Physical Activity Enjoyment Scale (PACES) (Kendzierski & DeCarlo, 1991) was added to the questionnaire. Since there was no Dutch version available of this scale, the scale was translated by the researchers separately, and then combined into one Dutch translation. This scale exists of 18 items with a bipolar scale of 7 points. Internal consistency of these items was very high, α=.95.

4.3.9 Evaluation To answer the second main research question, additional questions were constructed for the questionnaire. If the participant was using a device other than a smartphone, the question was asked whether the participant was a smartphone owner. If the participant owned a smartphone, the question was asked whether the participant used Runkeeper, Runtastic, Strava, or another physical activity app. About Runkeeper and Strava, we got insights about the behavior change techniques (BCT’s) that were present in those apps. These insights were obtained from Anouk Middelweerd, who conducted a content analysis of physical activity apps in 2014 (Middelweerd et al., 2014). Her research showed that in both Runkeeper and Strava the following BCT’s were present: prompt self-monitoring of behavior, provide feedback on performance, prompt practice and plan social support or social change. Runkeeper, in addition, contained prompt intention formation, provide instruction, prompt specific goal setting, prompt practice and provide opportunities for social comparison. Strava contained ‘’provide contingent rewards’’ in addition to the four named above. Based on these BCT’s and using the taxonomy of BCT’s (Abraham & Michie, 2008), questions were constructed to see how the participants evaluate these BCT’s. An example of a question was: ‘’I like it that I can set goals in Runkeeper’’. The scale was a Likert scale and ranged from 1 (=strongly disagree) to 5 (=strongly agree). To let the participants think further about the app (the participants who used an app) and to get an insight into the overall thoughts and opinions of the participants, some open questions about the apps were asked. These items were constructed based on the process evaluation of a serious self-regulation game intervention (app) named ‘Balance it’ by Spook et al. (2015). Participants, who indicated in the beginning that they did not use a PA app, were asked why by using an open question. The System Usability Scale (SUS) by Brooke (1996) was used to measure user friendliness of the apps. This scale includes aspects such as the need for support, training and complexity of a system (Brooke, 1996). The scale is simple to fill in and has proven to be a valuable evaluation tool for usability of systems (Brooke, 1996). In total, the scale consists of 10 items, with a Likert scale ranging from 1 (=strongly disagree) to 5 (=strongly agree). The Dutch version of the questionnaire was used (Sauro, 2011). In the items, the word ‘system’ was replaced by ‘app’, to prevent that the participant got confused by the wording. SUS was scored using the guidelines developed by Brooke (1996). The SUS scores had values ranging from 0 to 100. Internal consistency was good, α=.87.

12

4.4 Analysis Descriptive statistics were performed to describe the demographic characteristics including age, length, weight, BMI, gender and educational level. Besides, descriptive statistics were used to indicate how many participants were sports club member, to indicate the amount of minutes per week for every activity, to indicate how many participants met the health norms and to see the average evaluation of each behavior change technique. Chisquare tests and t-tests were conducted to assess whether there were significant differences between the two groups for characteristics. ANCOVA analyses were conducted to see the difference in determinants and PA behavior between the two groups, adjusting for age and educational level. As a form of exploratory research, mediation analyses were performed for the SCT model. For this analyses, the four steps of Baron and Kenny (1986) for indicating mediation were followed. The total activity score was the outcome variable in the mediation analyses. A p-value of 0.05 or lower was considered as significant. The analyses were performed using SPSS Statistics 23.

13

5. Results In this chapter, the results of the data analyses will be presented. The chapter starts with a description of the study sample (section 5.1). After that, results for the first main research question will be presented. In section 5.2 the differences between app users and non-users will be presented and the following section (section 5.3) will present exploratory research in which the SCT model will be tested. The chapter ends with the results of the second main research question, the evaluation of physical activity apps (section 5.4).

5.1 Descriptive statistics 5.1.1 Demographic Characteristics of the Study Sample Demographic characteristics of the study sample and differences in characteristics between app users and nonapp users can be found in Table 1. The mean age of the participants was 47 (SD = 17.7). 30.3% of the participants were men and 69.7% of them were women. Most of the participants were educated at a higher vocational level or academic level (60.6%). Of all participants, 25.1% used a physical activity (PA) app. Table 1. Characteristics of the study population (means and standard deviations) Total (N = 251) M ± SD

App users (N = 63) M ± SD

Non-app users (N = 188) M ± SD

Age

46.5 ± 17.7

43 ± 14.7

49 ± 18.4

Height in meters

1.73 ± 0.08

1.7 ± 0.08

1.73 ± 0.09

Weight in Kilogram

71.9 ± 14.4

74.8 ± 19.3

70.9 ± 12.3

Body Mass Index (BMI)

23.8 ± 4.0

24.4 ± 5.4

23.7 ± 3.4

Sports club/center member (%)

55.8 %

66.7%

51.1%

Women

69.7%

63.5%

71.8%

Men

30.3%

36.5%

28.2%

very low

0.4%

0%

0.5%

low

8%

4.8%

9%

medium

31.1%

23.8%

33.5%

high

60.6%

71.4%

56.9%

Gender (%)

Education (%)

5.1.2 Environment To get an idea of the participants’ perceptions about their environment, questions were asked about the amount of green zones, about the cycle and footpaths, about the safety and cleanness of the environment and about the amount of sport facilities. Means and standard deviations can be found in Table 2.

14

Table 2. Experiences of the study population with their environment (means and standard deviations)

Environment has enough green zones Environment has attractive and safe cycle and footpaths Environment is safe and clean Environment has enough sport facilities

Total (N = 251) M ± SD

App users (N = 63) M ± SD

Non-app users (N = 188) M ± SD

4.23 ± .87

4.16 ± .99

4.25 ± .83

4.14 ± .83

4.14 ± .76

4.14 ± .85

3.78 ± .89 4.00 ± .84

3.94 ± .84 3.95 ± .83

3.73 ± .91 4.02 ± .84

5.1.3 Physical Activity Physical activity was measured by using the SQUASH questionnaire (see Chapter 4). Total minutes per week was measured by counting up minutes per week of commuting activities, activities at work and school, household activities and leisure time activities. Table 3 presents an overview of how the total minutes per week were distributed over the different activities for both app users and non-users. Leisure time existed of walking, biking, gardening, odd jobs and sports. Table 3. The amount of minutes per week per activity (means and standard deviations)

Commuting Activities at work Activities at school Household activities Leisure time Walking Biking Gardening Odd jobs Sports Total minutes per week

Total (N = 251) M ± SD

App users (N = 63) M ± SD

Non-users (N = 188) M ± SD

96.46 ± 142.49 1131.64 ± 1071.56 228.53 ± 660.37 629.82 ± 553.26 576.52 ± 438.40 150.13 ± 174.59 154.14 ± 186.21 45.54 ± 122.58 56.31 ± 140.94 170.39 ± 212.99 2662.96 ± 1168.34

102.35 ± 139.27 1723.60 ± 1027.14 208.73 ± 559.03 425.48 ± 389.27 537.41 ± 316.26 139.48 ±168.98 125.08 ±158.98 27.06 ± 69.23 20.32 ± 50.92 225.47 ± 172.92 2997.57 ± 1093.92

94.48 ± 145.19 933.27 ± 1013.74 235.16 ± 692.26 698.30 ± 583.26 598.62 ± 472.32 153.70 ± 176.73 163.88 ± 193.90 51.73 ± 135.44 68.38 ± 158.48 151.93 ± 222.19 2550.83 ± 1173.79

Besides the total minutes per week and the total activity score, both the Nederlandse Norm Gezond Bewegen and the Fitnorm were calculated, as well as the Combinorm. Results can be found in Table 4.

Table 4. Percentage of participants who meet the health norms

Meets NNGB Meets Fitnorm Meets Combinorm

Total (N = 251)

App users (N = 63)

Non-users (N = 188)

77.7% 59.0% 83.7%

76.2% 66.7% 90.5%

78.2% 56.4% 81.4%

15

5.1.4 App Use In the questionnaire, app users were asked to define what apps they used. Runkeeper, Strava, and Runtastic were pre-defined by the researchers. Besides these three PA apps, used PA apps were Endomondo, Nike Running, Stappenteller (step counter), Garmin, S Health, Sport watches, MyFitnessPal and a few other apps. The number of users per PA app can be found in Figure 5. The ‘’other’’ category existed of a wide variety of PA apps which were only named once by participants. These PA apps can be found in Appendix C: ‘Other Apps’. Note that the sum of numbers in the figure does count up to 75 while there were 63 app users; some participants filled in two or three apps.

Physical Activity Apps Runkeeper Strava Nike Running Endomondo Garmin Runtastic Stappenteller S Health Sportwatch MyFitnessPal Other

21 14 5 5 4 3 3 3 3 2 12

Figure 5. The number of users per PA app

5.2 Differences between App Users and Non-Users 5.2.1 Differences in Demographic Characteristics Chi-squared tests were conducted to test whether there was a difference between the two groups for the categorical variables: gender and sports club member. No significant association was found between the gender of the participants and their app use behavior (Χ2 (1) = 1.546, p = .27). The association between being a sports club member and being an app user was not significant (Χ2 (1) = 4.044, p = .06). However, in Table 1 it can be seen that more app users (66.7%) than non-users (51.1%) were a sports club member. For age, BMI, and educational level, independent samples t-tests were conducted to test whether there was a difference between app users and non-app users. App users were younger than non-app users (t (249) = 2.611, p = .01). The mean age of app users was 42.98 (SD = 14.70), whereas the mean age of non-users was 48.96 (SD = 18.20). No significant difference was found between app users and non-app users regarding BMI, t (249) = -.964, p = .34. For educational level, a significant difference was found between app users and non-users, t (249)= -2.279, p = .02. App users reported a higher educational level (M = 3.67, SD = .57) than non-app users (M = 3.47, SD = .68). We therefore controlled for the differences in age and education in the following analyses. 16

5.2.2 Group Comparison: App Users vs Non-Users One-way ANCOVA was conducted to compare app users and non-users for physical activity and determinants of the SCT model, adjusting for age and educational level. Planned contrasts were conducted to see the differences in means for both groups. Standard errors were chosen as the measure of variability, as standard errors take sample size into consideration (Hopkins, 2000). Levene’s test for equality was performed for PA and for every determinant. For most cases there was homogeneity of variances, except for household activities. 5.2.2.1 Physical Activity The results of the one-way ANCOVA showed that the covariates age and education were not significantly related to total activity score. Furthermore, there was no significant difference between app users and non-users for total activity score after adjusting for age and education (F (1, 247) = 1.13, p = .29). For total minutes per week, the covariate age was significantly related to minutes per week, F (1, 247) = 28.32, p < .001, a higher age was related to fewer minutes of PA per week. After adjusting for age and education, there was no significant difference between app users and non-users for total minutes per week, (F (1, 247) = 3.26, p = .07). Planned contrasts showed that app users had more PA minutes per week (M = 2881.42, SE = 139.26) than non-users (M = 2589.76, SE = 79.91). Converted to hours per week this means that app users have 48 hours of PA per week, and non-users 43 hours of PA per week.

To get more insight in the different activities, differences between app users and non-users were checked for the PA categories commuting activities, work activities, school activities, household activities and leisure time activities. No significant differences were found for commuting activities, school activities, and leisure time activities. For activities at work, a significant negative correlation was found between age and activities at work, F (1, 247) = 13.66, p < .001. In addition, a significant positive correlation was found between the covariate education and activities at work, F (1, 247) = 19.77, p