A stepwise approach, consisting of 4 steps, based on the Intervention Mapping Approach and the developmental steps for mHealth interventions, was used in the development of the new app [34,39]. In step 1, determinants important to promote an active lifestyle among low-educated working young adults were selected. In step 2, evidence-based BCTs [40] were selected to convert the determinants into practical applications. In step 3, a new smartphone app called “Active Coach” was developed. In step 4, the Active Coach app was tested on errors, acceptability, and feasibility and accordingly adapted for the final version. In this chapter, only the methods of the 4 steps will be described. The results (including the content of the app) will be described next in the Results section. This study was approved by the ethics committee of the university hospital of Ghent University (B670201525362) and by the ethics committee of the Vrije Universiteit Brussel (BUN 143201112745).

To develop an evidence- and theory-based app, determinants important to promote an active lifestyle in lower-educated working young adults need to be selected. The selection was based on the existing literature, previous studies from our research group, a theoretical health behavior change model, and an exploratory qualitative study among lower-educated working young adults (see Results step 1 [4,41-57]).

In the qualitative study, focus groups were conducted among lower-educated working young adults to assess determinants of an active lifestyle (PA and AT). In addition, opinions about mobile technologies (eg, smartphones, tablets, apps, fitness trackers) and valuable features of PA apps and their use to promote an active lifestyle were explored. Eligible participants had to be employed, aged between 18 and 26 years, and without a university or college degree (lower educated). Participants (n=34, mean age: 24.0 [SD 3.0] years, 59% [20/34] men, mean years of employment: 3.0 [SD 2.0]) were recruited throughout Flanders (northern part of Belgium). Snowball and convenience sampling, two nonprobability approaches often used in qualitative research [58], were used to recruit participants via the personal network of researchers and assistants and via social media. Five focus groups (6-10 participants per group) were conducted at places that were most convenient for the participants. Focus groups were conducted until saturation, which is the point at which all questions have been thoroughly explored in detail and no new concepts or themes emerge in subsequent sessions [59]. All focus groups were held in Dutch and lasted approximately 60 min. A focus group protocol and a semistructured discussion guide (Table 1) were developed consistent with the recommended focus group methodology [60]. The guide consisted of several questions, including an introduction question, a transition question, key questions, and an ending question. For some questions, participants were asked to write down an answer. This method allows participants to think and reflect about a question before starting a group discussion and not to copy other participants’ answers or opinions. Before the discussion started, the participants provided informed consent and completed a brief questionnaire obtaining sociodemographic information. The discussions were led by a moderator. Notes were taken by an observer. With permission of the participants, all conversations were audiotaped for transcription. The focus group interviews were transcribed verbatim, and the texts were incorporated into a qualitative processing program (NVivo 9 qualitative software, QRS International). The data were analyzed based on grounded theory. Grounded theory is a method of analyzing qualitative data without preconceived theories and is characterized by intensively analyzing data, often sentence by sentence or phrase by phrase [61]. During the transcription of the conversations, we developed codes according to the responses and the themes that arose frequently and were relevant to the aim of the study. Data obtained by the questionnaire were entered into SPSS (version 23.0) to calculate descriptive statistics.

BCTs were used to translate the selected determinants into practical applications that will be used in the new app [40]. BCTs are the active component of an intervention designed to change behavior [62]. Michie et al [40] developed a BCT taxonomy of 93 hierarchically clustered techniques. BCTs were selected based on their previously demonstrated effectiveness and on the focus group results (see Results step 2 [25,39,40,63-70]).

On the basis of the translation of the BCTs into practical applications and on the basis of the results from the focus groups, a new app was developed in cooperation with a commercial mobile app development company (Cucumber Apps). It is a native Android app (only developed for the Android operating system), which means that it is completely compatible with the smartphone’s native features and hardware (eg, accelerometer, camera, GPS) and ensures the best user experience [71]. In 2015, Android was the operating system with the highest penetration rate in the smartphone market worldwide (80% Android vs 19% IOS) [72] and among 15- to 35-year-olds in Belgium (57% Android vs 37% IOS) [23]. Moreover, developing an app for iPhone (IOS) would be too expensive and too time-consuming.

The content of the app was developed to incorporate an autonomy-supportive communication style, based on the self-determination theory [73]. Self-determination theory suggests that the content of goals (ie, intrinsic vs extrinsic) and the way goal contents are communicated (ie, autonomy-supportive vs controlling) explain variance in people’s motivation and performance [74]. Autonomy-supportive includes a more motivating language (can, may, want) instead of a controlling language (should, have to). The first version of the intervention content was read and evaluated by junior and senior researchers in the field of public health (n=19, mean age: 26.2 [SD 5.1] years, 26% [5/19] men). They were consulted because of their experience and knowledge regarding public health interventions. After adapting the first version, interviews were held with lower-educated working young adults (n=10, mean age: 23.0 [SD 2.0] years, 50% [5/10] men) to check the usefulness, applicability, and understandability of the content. Participants were recruited using convenience sampling, and the interviews were audiotaped and transcribed verbatim. Analyses were conducted as described in step 1.

First, usability of the app was tested by a small group of volunteers (n=11, mean age: 28 [SD 10] years, 55% [6/11] men, both lower and higher educated) who checked the app for faults and errors (also known as “bugs”). Nielsen et al [75] showed that conducting usability testing with only 5 participants will reveal 85% of usability problems. These volunteers (all different from the focus group participants recruited in step 1) needed to own an Android smartphone and were recruited via the personal network of researchers and colleagues in 2016 (convenience sampling). They installed the app and used it for approximately 5 weeks. All problems mentioned by the volunteers were collected by the researchers via an issue list and passed on to the developers. Accordingly, the app was adapted.

Next, the adapted version of the app was tested on acceptability and feasibility by lower-educated working young adults. A contact list of the previously conducted focus groups (see step 1) was used as a basis to recruit participants via snowball and convenience sampling. In addition, participants were recruited via a social employment business with several projects for lower-educated people (VZW Ateljee, Ghent). A total of 4 participants of the acceptability and feasibility test had participated in the previously conducted focus groups; all other participants were new. Participants (n=16, mean age: 24.4 [SD 2.3] years, 63% [10/16] men, mean years of employment: 3.7 [SD 2.3] years) had to possess an Android smartphone. During the acceptability and feasibility test, 4 interviews were conducted with each of the participants during a 9-week period (1 interview every 3 weeks; see Figure 1).

During the first interview (week 0), participants were informed about the purpose of the study. They were asked to sign an informed consent form and to complete a questionnaire about sociodemographics. A think-aloud interview was conducted during the download, install, registration, and first use of the app. In the think-aloud interview, participants were asked to use the app and say out loud any thoughts that came to mind. This is particularly useful as people give their immediate reactions to every element of the app, and it allows researchers to observe how it was used [76]. Afterward, a brief semi-structured interview (Table 2) was conducted to discuss issues that came up during the think-aloud interview and to ask additional questions. During the second (week 3) and the third (week 6) interview, the same semistructured interview was conducted to obtain intermediate information about participants’ experiences with the app (Table 2). During the fourth interview (week 9), a paper-and-pencil questionnaire, based on existing questionnaires assessing acceptability and feasibility, was used to evaluate specific elements of the app [77,78]. Moreover, 4 questions on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree) were used to assess general opinions about the app (ie, clear, fun, user-friendly, attractive), opinions about the tips and facts (ie, interesting, motivating, boring), and opinions about the goals received (ie, motivating, useful, tried to achieve it). After completing the questionnaire, a semistructured interview was held to discuss the answers of the questionnaire in more detail (Table 2). All interviews were held in Dutch and were audiotaped for transcription with permission of the participant. Analyses were conducted as described in step 1. In addition, Google Analytics [79] was used to obtain app usage statistics and evaluate how participants used the Active Coach app. Google Analytics offers free tools to measure website and app data to gain usage insights.

From the existing literature, self-efficacy has been shown to be one of the most important determinants of PA [41-43] and AT [44-46] among (young) adults. Social support was positively associated with PA and AT in adolescents [42] and young adults [4,47]. A review also showed that social support and having a companion for PA were positively associated with different types of PA, including AT [57]. Attitude has been an inconsistent determinant in the literature [42]. However, important perceived benefits and barriers of PA and AT have been mentioned in recent studies. Health benefits, recreation (releasing tension), social contact, and body image have shown to be important benefits for young adults’ PA participation [52,53]. Important benefits of AT among young adults are low costs, autonomy and flexibility, and a short travel time in urban areas [47,49]. Barriers of PA are lack of time, lack of motivation, and lack of money [50,51]. Barriers of AT are bad weather, practicality (eg, how to deal with luggage), comfort (eg, sweating), and lack of facilities such as bicycle parking or showers and changing rooms at work [44,47,48]. Although knowledge might not be sufficient to change behavior, it is a necessary prerequisite to an individual’s positive motivation to engage in more PA [56]. It has been shown that knowledge of PA guidelines in Irish and English adults is very low and that lower education is associated with not knowing the guidelines [54,55].

In the focus groups conducted in this study, these results were confirmed by the target group. For example, lower-educated working young adults mentioned similar perceived benefits and barriers of PA and AT as found in the existing literature. They also indicated to have a lack of knowledge regarding an active lifestyle, and they were very interested in information and advice. Other results of the conducted focus groups (regarding mobile technologies, valuable app features, and app use to promote an active lifestyle) are discussed in steps 2 and 3 of the results section.

As a result, following determinants of an active lifestyle in lower-educated working young adults were selected: knowledge, attitude (perceived benefits and perceived barriers), social support, and self-efficacy. Therefore, the attitude–social influence–self-efficacy (ASE) model [80] was used as a base to develop an app for the promotion of an active lifestyle. The ASE model is a theoretical model that describes the processes wherein health behaviors, such as an active lifestyle, are shaped. It states that an active lifestyle is defined by intention to act, whereas intention, in turn, is determined by attitudes (ie benefits and barriers), social influences, self-efficacy, and the knowledge and skills needed to achieve an active lifestyle [80]. The ASE model has been used in the development of previous health interventions [81-84].

BCTs were used to translate the selected determinants into practical applications [40] (Table 3). Multiple self-regulation techniques were selected (self-monitoring, goal-setting, feedback on behavior, review of behavior goals, instruction on how to perform the behavior) as it has been shown that these techniques are important to target the selected determinants to increase PA in interventions [63-66]. Moreover, both in qualitative and quantitative studies, young adults rated self-regulation techniques (especially self-monitoring and goal-setting) as most valuable features to increase self-efficacy among health behavior apps [25,85]. In addition, participants of the focus groups conducted for this study also mentioned self-regulation techniques (self-monitoring, goal-setting, and instruction on how to perform the behavior) as valuable app features. Furthermore, to encourage self-efficacy regarding the (re)use of the app, “prompts/cues” was selected as a BCT. Providing prompts or cues (notifications in an app) has had a positive effect on reuse of intervention websites among adults, adolescents, and children, particularly those with low socioeconomic status [68-70]. A qualitative study among young adults also found that relevant and timely (but not too frequent) alerts and reminders are valuable features of health behavior apps [25]. Furthermore, participants of the current focus groups mentioned push notifications as necessary to not forget an app.

The BCTs “instruction on how to perform the behavior” and “information about health consequences” were selected to increase knowledge and to encourage a positive attitude toward PA and AT. A qualitative study among young adults showed that providing feedback and advice to guide people about how they can change behavior was evaluated as a valuable app feature [25]. Participants of the focus groups conducted in this study also indicated to be interested in information and advice regarding an active lifestyle:

Finally, the BCT “enhancing network linkages” was selected. This BCT focuses more indirectly on social support by advising on mobilizing and maintaining social networks (eg, tips on being active together) [39,67]. Although social support is positively associated with an active lifestyle, it has been found that smartphone users do not like apps that link and share (health) information with social network sites [25,67]. In the conducted focus groups, Facebook was indicated as the most popular social network. Participants used it daily for communication or games, but they did not want to post or share health-related information on Facebook via an app.

The new native Android app, Active Coach, aims to promote an active lifestyle in lower-educated working young adults via a 9-week program. There is no consensus on the optimal duration of app programs to ensure user engagement [18]. However, in a recent review on app interventions to improve health behavior (ie, PA), intervention durations longer than 8 weeks tended to be effective [18]. Nevertheless, very lengthy app programs might not be useful, as app use often declines rapidly because people lack commitment and use apps in a transient, casual way [25,86].

Users of the Active Coach can choose how they want to make their lifestyle more active, through general PA or through AT. Participants of the focus groups conducted in this study agreed that a smartphone app would be used more and more suitable to promote an active lifestyle than a website:

Although focus group participants used their smartphone on a daily basis, many of them indicated that they were not allowed to carry their smartphone with them during working hours. However, they clearly preferred automatic tracking of their activity behavior:

Therefore, to ensure all-day and automatic self-monitoring of the activity behavior, the Active Coach app works in combination with a wearable activity tracker, the Fitbit Charge. The Fitbit Charge is a wrist-worn activity tracker that uses a 3-axis accelerometer to track a person’s movement [87]. The wristband can track the number of steps walked, active minutes, floors climbed, the quality of sleep, and other personal metrics. For the Active Coach app, only the number of steps walked was used. Fitbit trackers are valid and reliable devices for measuring step counts in healthy young adults [20,88].

The Active Coach app includes a registration process and 7 other components, each with their own influence on the determinants (Table 3). In the focus groups conducted in this study, participants said they were willing to go through a registration process at the beginning of an app to receive more personal information:

However, participants indicated that the questions of the registration process should not be too detailed, too long, or go back too far in the past. Therefore, the registration process consists of only 3 screens. First, personal questions (name, email-address, password, gender, date of birth) are asked. Second, it is asked what kind of job the user has (mostly sitting or mostly standing or walking), and how the user would like to become more active (PA or AT). Third, perceived benefits are asked regarding the chosen behavior. If a user chooses PA, answer options are as follows: (1) being fit and healthy, (2) looking good (weight maintenance, appearance), and (3) relaxing, having distraction and/or social contact. If a user choses AT, answer options are as follows: (1) being fit and healthy, (2) saving money (no fuel costs), and (3) practical reasons (no traffic jams, not searching for parking spots).

After completing the registration, the Active Coach app consists of a 9-week program. During those 9 weeks, user’s PA behavior is being tracked by the Fitbit Charge (step count) and their AT behavior is being tracked by mobile smartphone sensors (GPS and accelerometer). Regardless of the activity choice (PA or AT), both behaviors will be tracked automatically and will be visible for the user in the app on a graphical display (steps/day [PA] and minutes/day [AT] per day, week, month, and year) in the app (Figure 2). However, the goal setting and the information received will differ according to the chosen behavior.

The first week of the 9-week program is a baseline week during which the baseline activity level of the user is measured. At the end of this week, a personal goal dependent on the chosen behavior (PA in steps/day or AT in min/day) is set by the app for the following week (eg, Your goal for next week is to try and walk 6000 steps each day). The goal was based on mean steps per day or min AT per day of at least 3 of the 7 baseline days, which was increased by 10%. Participants of the focus groups conducted in this study agreed that there should be a goal or a target to reach for. Some wanted it to be set automatically, whereas others wanted to choose their own goal.

Therefore, the app provides both automatic and individually set goals. The focus group participants also liked receiving rewards (eg, virtual rewards such as a medal) or positive feedback when they achieve a goal or target. Therefore, every day during the following 8 weeks, users receive a notification on whether or not they achieved their day goal. They can also see their daily and weekly goal progression on the graphs in the app (Figure 2). In addition, users receive feedback on their goal achievement at the end of each week (Sunday). If they achieve their goal, they can increase it with 10%, 25%, or 50% or they can maintain the same goal for the next week. If they do not achieve their goal, they can choose to decrease it with 10%, 25%, or 50% or they can maintain the same goal for the next week. Additionally, users are asked why they did not achieve their goal (perceived barriers). For a user who chooses PA, answer options are as follows: (1) lack of time, motivation, energy, etc; (2) lack of money, sports equipment, etc; (3) no sports partner; and (4) bad health (sick, injury, tired). For a user who chooses AT, answer options were as follows: (1) feeling unsafe (dangerous traffic, bike theft), (2) practical reasons (bad weather, luggage), (3) laziness or habit of taking the car, and (4) bad health (sick, injury, tired). This information is being used to give users more personal feedback.

Every Monday and Friday during the 8 weeks after the baseline week, users receive a notification with a practical tip, and every Wednesday, they receive a notification with a scientific fact to help and motivate them to reach their goal. The content of the tips and facts is tailored based on information from the registration process (gender, sitting or standing job, PA or AT, selected benefits), on goal achievement, and on the selected barriers. If users achieve their goal from the previous week, they receive 1 general tip (eg, Naviki is a very practical app to help plan your route with information on slopes and the weather on the road) and fact (eg, Did you know that being active when you’re young decreases the risk of heart disease later in life?) and 1 tip adapted according to the benefits selected during the registration (eg, benefit: staying fit and healthy, tip: You can stay fit and healthy together with your colleagues by joining a fun biking contest). If users do not achieve their goal from the previous week, they receive 2 tips adapted according to the barriers selected at the end of the previous week (eg, barrier: high costs, tip: By asking your colleagues to make a walk during lunch break, you can be more active at work without any extra costs). Users can select the preferred time to receive these notifications. When pretesting the tips and facts with lower-educated working young adults, they indicated that some tips and facts were too long, too commanding, and too obvious. With their suggestions, tips and facts were adapted to be shorter, more powerful, and more supportive. New information was also added, and some written messages were replaced by figures.

The volunteers (n=11) who tested the Active Coach app for technical errors and faults all owned different brands of Android smartphones. Several problems occurred such as too small fonts, errors in texts, not receiving messages, receiving wrong messages, receiving notifications at the wrong time, and error messages. All these problems were fixed. The main change was adding an extra page in the app to collect all received notifications about tips, facts, and goals, so that the users could consult these again at any moment.

Next, lower-educated working young adults (n=16) tested the app on acceptability and feasibility. Two participants dropped out due to a damaged smartphone. On the basis of the interviews, participants were very positive about the simplicity of the Active Coach app. They used and understood the app without any problems:

Participants were also positive about the design of the app:

Participants had no problems with the registration process: they liked that it was not too long and they understood all the questions. Although they did not want more questions during the registration process, it was said that the app (particularly the tips) should be more tailored:

However, they really liked the scientific facts and thought they were very interesting. In addition, participants also liked that they received the tips and facts as notifications, because it reminded them to have a look at the app. Participants indicated that they regularly looked at the notifications page to reread some tips and facts.

Participants were positive about the goals. They particularly liked the weekly possibility to increase or decrease their goal:

They also liked the graphical display of the tracked behavior and the goals (“It is very clear, with all the different colors. And also divided in day, week, month, year. That’s good.”), although they had some suggestions for improvement:

In the beginning, participants complained about battery drainage, mostly because their Bluetooth, GPS, and/or Internet connection was switched on the entire day. However, they quickly fixed this problem themselves:

Finally, participants really liked the Fitbit wearable:

However, they found it to be a disadvantage that it is not waterproof and they could not wear it while swimming. Some participants also looked at the native Fitbit app, mainly to check additional features such as calories burned or sleep.

Results from the questionnaire during the fourth and last interview showed that 11 of the 14 participants thought that they were more active because of the Active Coach app and also 11 of the 14 would recommend the app to others. Furthermore, as shown in Table 4, almost all participants (13 of 14) agreed or strongly agreed that the app was clear and understandable and that the tips and facts were understandable. In addition, more than half agreed or strongly agreed that it was motivating to have a goal and useful to receive weekly feedback about that goal. On the basis of these results, adaptations were made to the app. Some minor problems with the notifications page were fixed, and the list of practical tips was re-evaluated.

Results from Google Analytics showed that the Active Coach app was most used on days and hours that users received notifications with tips, facts, and weekly feedback on goal achievements (Monday, Wednesday, Friday, and Sunday between 8 and 10 PM). It also showed that they used the app on average 1 min per session. The app was visited by minimum 1 and maximum 14 users on each day of the acceptability and feasibility testing period with a minimum of 5 sessions a day and a maximum of 31 sessions a day.