The first stage of this process involved defining the purpose of each app. This required specifying the relevant public health guidelines to inform the goals for behavior change, the specific strategies for behavior change, the visuals or graphic design, and the potential data to be collected.

The fundamental purpose of the apps was to support change in the lifestyle behaviors identified. To assist young adults with improving their dietary habits, goals for these behaviors had to be defined to create rules about what is adequate [21,22]. Relevant public health guidelines were consulted for physical activity levels [23,24], intake of fruit and vegetables [25], and recommended limits for take-out meals and sugar-sweetened drinks [26,27]. For example, the physical activity app (ePASS) included a “good health” target of 30 minutes of moderate level physical activity daily based on the World Health Organization’s physical activity guidelines of at least 150 minutes of aerobic activity per week for adults [24], and a “healthy weight” target of 60 minutes of moderate level physical activity daily based on expert consensus that up to 60 minutes of moderate activity daily is required to prevent unhealthy weight gain [23] (see Figure 1B). The fruit and vegetable app (eVIP) provided users with a graphical display of the number of fruit and vegetable servings they recorded out of the two servings of fruit and five servings of vegetables recommended by the Australian Government Department of Health and Aging (see Figure 1A) [25]. The sugar-sweetened drinks app (eSIYP) presented users with a color display of their total energy, sugar, and alcohol intake from all drinks consumed, in which green, orange, and red indicated the “ideal,” “acceptable,” and “too much” threshold levels of intake, respectively. These threshold levels were based on nutritional expert opinion of the recommended limits on consumption of added sugars, as per the Australian Dietary Guidelines [26]. The take-out food app (eTIYP) also presented users with a color display of the average energy and fat content of take-out meals, in which green indicated acceptable intake and red indicated excessive intake, equating to ≤30% and >30% of the dietary intake recommended by the Australian National Health and Medical Research Council and the New Zealand Ministry of Health, respectively, according to age and gender [27].

In terms of behavior change strategies, young adults often lack the self-regulatory skills, such as self-monitoring and planning, required to adopt and maintain healthy behaviors [20]. Self-regulation was fostered in each app by providing users with a platform to create daily entries of their behavior (eg, physical activities performed or vegetables consumed) from which they were provided daily or weekly summaries of their reported behavior, in reference to public health guidelines, to enable their monitoring of and planning around these behaviors. Providing feedback that is personally relevant may also be an important strategy for changing young adults’ behaviors [18]. Encouragement from success of attaining a goal and social persuasion can enhance self-efficacy, and has been identified as important in achieving behavior change, particularly during busy or stressful situations [28,29]. All apps provided motivational tips as a source of positive encouragement that would assist the young adults in creating more positive beliefs around their ability to change their behavior (eg, “You can split up your exercise target into as little as 15-minute bursts”). These tips were also tailored to users’ self-reported behavior. For example, if a user’s reported physical activity did not meet recommended guidelines, they were shown a relevant motivational tip (ie, “Plan exercise in advance and write it down if you can—try phone reminders”).

Consideration was also given to the graphic design and how this might influence behavior. The behavior-image model suggests that through processes of social- and self-comparison, individuals will compare themselves to similar human images and create projections of themselves possessing the desired characteristics of the humans in those images [30]. This process is referred to as “self-reevaluation” in the Transtheoretical Model, which uses healthy role models and imagery to assist ones progression from contemplating behavior change to preparing for changing behavior [31]. Hence, in the apps we used images of young adults who were performing the target behaviors (eg, riding a bike or drinking water) and possessing desirable characteristics of a normal healthy appearance and lifestyle, to motivate users toward changing the target behaviors. Similarly, we displayed healthier foods and drinks, rather than “junk” foods, to model these foods as ideal for consumption. All images were purchased from a commercial graphics company to avoid potential breaches in copyright.

To facilitate future research, we enabled the following data items to be exported: user identification (ID), log-in ID, sex, age, and the date and time of log-ins into each app. Additional information that was able to be exported from each app included: physical activities performed and their duration; drinks consumed and their volume, total sugar, alcohol, and energy content; energy and fat content of take-out meals, the restaurant where the meal was consumed, and the contents of the meal; and the number of servings and types of fruit and vegetables consumed. This data could be exported into comma-separated value files from our relational database (described later), which could then be exported into statistical software for further analysis.

Traditionally, apps are developed for one specific operating system (native app), such as iOS, Android, Windows, Symbian, and BlackBerry, or developed as Web-based apps. Native apps run locally on a smartphone’s operating system in a way that is analogous to programs running on a desktop computer. Web-based apps run like a Web page, whereby the app operates on an external server and the user accesses the app through the Web browser on their mobile device. Due to these functional differences, Web-based apps may be used on all smartphones regardless of the operating system as long as the user has Internet access; however, there is less opportunity to utilize the existing hardware built into the phone (eg, the camera, geopositioning, or calendar). In our case, we developed Web-based apps because we did not require the use of existing hardware on the phone and we wanted to enable downloading of data recorded by users and allow the apps to be used on multiple operating systems and through the Internet for those who did not own a smartphone. However, Web-based apps are increasingly becoming able to perform like native apps with an offline mode that can be accessed without Internet connectivity. Similarly, there is increasing potential for native apps to possess Internet connectivity, enabling the user to download updates and data from the user to be uploaded to a server. This emphasizes the dynamic nature of mobile technologies and why the type of platform selected should be discussed with an information technology specialist.

To enable users to record their behavior, each app had to be linked with the relevant data for that behavior. The following data items were included for each app:

1. ePASS: type of activity (ie, gym, sports, recreational, or housework) and the intensity (ie, moderate vs vigorous) of 91 unique activities, where “moderate” was defined as a metabolic equivalent of task (MET) value of 3-6 and “vigorous” was defined as > 6, derived from the compendium of physical activities [32].

2. eVIP: serving size equivalents for 48 types of fruit and 61 vegetables, where 1 serving was equivalent to 150 g of fruit or 75 g of vegetables [26] (eg, 0.5 cups chopped or 4 spears of cooked asparagus are both equivalent to 1 serving of vegetables).

3. eTIYP: total energy and fat content of 504 take-out food and drink menu items.

4. eSIYP: drink category (eg, waters, vitamin waters, hot chocolate, tea/coffee, alcohol, soft drinks, sports drinks, cordials, fruit juices, fruit drinks, flavored milks, and milkshakes) and the total energy, sugar, and alcohol content of 114 unique drinks.

Nutrient composition data for eTIYP and eSIYP were sourced from the Australian government food and nutrient database, NUTTAB [33]. The foods listed in this database were chemically analyzed or, when unavailable, sourced from food manufacturer nutrient label data which may or may not be based on chemical analysis.

These data items were all contained within one relational database, which is essentially a database where the data items (or variables) are arranged into a series of tables with each table representing a different aspect or “relation” in the data. For each app, there were 1-2 tables containing the behavioral data. For example, the eVIP app required one table for the types of fruit and vegetables, such as “eggplant/aubergine (cooked),” and a second table listing the portion sizes, such as “0.5 cups diced,” “3 thin slices,” or “1 thick slice,” with their respective serving equivalents of 1 serving, 1 serving and 0.5 servings. An additional table was included for each app containing the motivational tips and another table contained user details (ie, name, age, gender, log-in ID, and user ID). The relational database management system software, MySQL (Oracle Corporation, Redwood Shores, California, United States), was used to access and query data items contained within the relational database using structured query language (SQL). For example, if a user logged into the eVIP app at lunchtime, SQL was used to identify a motivational tip about including fruit or vegetables at the lunch meal to present to the user from the motivational tips table in the relational database for the eVIP app.

Programming for the apps was written with Python programming language software (Python Software Foundation, Wolfeboro Falls, New Hampshire, United States) to communicate with the relational database and generate the user interface (ie, what the user sees and interacts with), including the HyperText Markup Language (HTML) (ie, the building blocks of a Web page) and the cascading style sheets (CSS) (ie, the visual formatting of the HTML). The HTML and CSS information was then interpreted by the Web browser on the user’s mobile device to create the user interface (see Figures 1-5). To illustrate an example of this programming, if a user recorded 0.5 servings of fruit, this data was stored in the relational database and the user was instantly presented with an image of half of one apple shaded green (representing 0.5 servings of fruit) on the home screen of the eVIP app (see Figure 1A). A separate app was created for each of the four behaviors, rather than creating a combined app, to permit the targeting of different behaviors in future intervention research, such as addressing only those behaviors that are particularly challenging for the individual. Further, there was a need to limit each app to < 6 screens, five of which were common to all of the apps (Figures 1-5), to simplify and increase the speed at which users could navigate their way through the apps.

Data presented and manipulated in the prototype version of each app were crosschecked against the relational database by two authors (LH and AC) for accuracy. This involved recording intake and activity behaviors at random in each of the four apps and checking whether the data presented to the users were correct (eg, the total energy of a take-out food item) and that the data were accurately manipulated in the apps (eg, converting fruit and vegetable portions recorded into the number of equivalent servings or summing the total energy content of all sugar-sweetened drinks recorded).

Twenty-one adults aged 18 to 35 years who were participating in a weight loss trial were provided access to the apps, and were asked for their feedback on the performance of the apps as part of an online survey. This survey included two questions assessing the usability of the apps, including: “Did you have any problems downloading the smartphone apps?” and “Did you have any problems using the smartphone apps?” with three response options: “yes,” “no,” or “did not access them.” If they responded “yes,” they were then prompted in an open-ended question: “Please tell us what problems you experienced.” Two other open-ended questions asked, “How could the smartphone apps be improved?” and “Please tell us any other comments that you have.” Re-occurring themes in the qualitative responses to these open-ended questions were identified and summarized. Procedures for collecting this information were approved by the University of Sydney Human Research Ethics Committee (approval #13698).