We developed and tested in a previous pilot study a smartphone app targeting unhealthy alcohol use, based on a web-based intervention with demonstrated efficacy among young individuals [17,18]. We further developed this existing proactive secondary prevention smartphone app, taking into account the limitations observed during the pilot studies. The app was designed to offer additional features, taking advantage of the specificities of smartphones (ie, increased level of personalization and immediacy or access to intervention material in situations outside of the reach of face-to-face or computer interventions). As for numerous electronic interventions targeting unhealthy alcohol use, this app includes a social norms intervention [1,19-21], a type of intervention considered effective for college students [22]. The social norms intervention consists of normative feedback. The user’s alcohol consumption is compared with the alcohol consumption of people of the same age and sex in Switzerland, based on Swiss population data [23]. Normative feedback is provided for the volume of drinking (number of drinks per week) and for the frequency of heavy drinking episodes (frequency of episodes with ≥5 drinks [men]/≥4 drinks [women]). The app also provides personalized feedback on the risk of harm. Personalized feedback is considered one of the possible mechanisms of brief interventions to reduce alcohol use [24,25]. In addition, the app was designed to encourage self-efficacy through autonomous goal setting. It also provides additional information and resources to users willing or needing more: information and contact options for local addiction and mental health treatment resources, including student health centers, are listed.

The development of the app included the following iterative steps: (1) development of the initial prototype based on the app tested in pilot studies [26,27]; (2) test of the initial prototype in the target population and qualitative assessment; (3) app adjustments based on qualitative findings; (4) test of the second version (prototype 2) of the app in the target population and qualitative assessment; (5) final adjustments based on qualitative findings. Figure 1 shows the app development process.

The development of the initial prototype (ie, prototype 1 in Figure1) was conducted through regular meetings between members of the research team and developers. Throughout the development process, informal testing was performed by members of the research team, developers, colleagues, and students working in our unit to ensure usability as well as text and design appropriateness. Notably, the aim was to make the app more “active” and to send messages to participants (ie, notifications) following prespecified scenarios, knowing that students appear acceptant of receiving such messages [28] and that smartphone app users identify prompts as important [29]. Two groups of students, member of the app’s target population, were recruited to test the app in 1 (prototype 1) or 2 (prototype 2) of 2 tests and participate in in-depth interviews about the content of the app.

The app development started in October 2018 and a first version was ready for testing in February 2019. The final version was ready in November 2020 (due to the COVID-19 pandemic, the randomized trial was delayed and started in May 2021).

The initial version of the app content in prototype 1 comprised 5 modules: (1) personalized feedback on self-reported alcohol consumption, with normative feedback, feedback on the calorific content of the reported consumption, and feedback on health risks (hereafter referred as “quiz”); (2) the blood alcohol content (BAC) computation module (hereafter referred as “test”); (3) the goal-setting tool (hereafter referred as “challenge”) with the possibility of obtaining a “badge” if the goal is achieved; (4) the designated driver tool (hereafter referred as “driver”); and (5) fact sheets (hereafter referred as “pedia”). The content of the app followed the existing literature [14,30-34] and previous research involving electronic interventions conducted by our group [17,18,26,27,35,36].

Participants were students from 4 tertiary education institutions in the French-speaking part of Switzerland (ie, higher education institutions in health: Haute Ecole de Santé Vaud [HESAV]; University of Lausanne [UNIL]; Federal Polytechnic School of Lausanne [EPFL]; and Lausanne School of Hotel Management [EHL]). To be eligible (in test 1 or 2 or both), participants had to be a student, fluent in French, own an iPhone (Apple Inc; although the final version of the app is available for both iOS and Android, the development was carried out on an iOS platform), score 4 or more (for men) or 3 or more (for women) in the AUDIT-C (Alcohol Use Disorder Identification Test – Consumption) [37,38], and provide informed consent to participate. Study promotion was conducted with the support of students’ associations or communication staff of the targeted institutions. Study presentation and contact information were displayed by flyers (displayed in the various institutions) and electronically (eg, on the students’ association page on Facebook). Interested students participated in a phone call aiming to provide information about study participation, answer all questions, and screen for eligibility. Eligible students who were willing to participate completed written informed consent and were then explained about how to download the app. As the app was in the development phase, it was not available to the public and was specifically installed on each individual smartphone. Participants then tested the app for 2-3 weeks and took part in a semistructured interview to provide feedback. During the testing period, participants received reminders to keep testing the app. Prototype 2 of the app was developed based on feedback from test 1. Issues identified by multiple participants were given priority when refining the app.

Participants included in test 1 were invited to retest the app in test 2. In addition, new participants were included to ensure a naïve feedback on the app. The inclusion criteria and procedures mirrored those used in test 1. Participants received CHF 50 (about US $50 at the time of the study) at the end of each semistructured interview. All interviews were conducted by a senior researcher (VSG), in-person, using an interview guide (see Multimedia Appendix 1). The interviews aimed at exploring the following themes: general impressions on the app, perceptions regarding the app’s usefulness, perceptions on the app functioning, perceptions on the app design, perceptions regarding the app content, perceptions regarding each module of the app, and perceptions regarding the notifications. Each interview lasted between 36 and 70 minutes. Interviews were audio-recorded and transcribed verbatim by trained research assistants. All names and identifying information were removed before analyses.

Qualitative data were subjected to conventional content analysis. This method enables description of qualitative data through a systematic process of coding and classification [39]. Qualitative data were reviewed by VSG to identify recurring categories. Initial coding was conducted using a line-by-line technique aiming to narrate the actions occurring in the interviews [40]. After the initial coding process was completed, a codebook was created, wherein incident-by-incident codes were pooled and idiosyncratic or redundant codes were collapsed or eliminated. After the codebook was created, NB reviewed the codebook and tested it with 2 transcriptions. The codebook was then refined and finalized in consensus meetings. Finally, VSG rated all qualitative data from test 1 with the final codebook. The codebook was then adapted to rate qualitative data from test 2. Specifically, VSG reviewed a subset of interviews to identify new categories that did not appear in the original codebook. Finally, VSG rated all data from test 2. We present quotes in results from both tests 1 and 2 for illustration of categories emerging from the analysis. ATLAS.ti 7 (Scientific Software Development GmbH) was used to code qualitative data.

All procedures were approved by the local Ethics Committee (Commission cantonale d’éthique de la recherche sur l’être humain [CER-VD]; protocol number 2018–00560).

Multimedia Appendix 2 shows screenshots (by module) taken at the various stages of the app development process (prototype 1, prototype 2, and final version).

The major modifications concerned the design (change of name, icon/logo, home menu, general presentation, clarification of the alcohol use questionnaire with more specific options for standard drinks, presentation of feedback parts [normative feedback and BAC]) and the development of a new module allowing day-to-day monitoring. References for the normative feedback were added or placed in more prominent display to support the app’s legitimacy. A more readily available and more detailed description of standard drinks was included in the app with an option to access more detailed information on standard drinks. Different drinks choices were added. For example, an option for stronger beers/large beers (equivalent to 2 standard drinks) was added. The cartoon character was removed from the app’s icon. Similarly, the cartoon character that was created to help users navigate the app was removed. Within-app notifications and connecting messages (via within app pop-up messages or messages at the end of the feedback) between modules were included (ie, prompt to set drinking goals after an assessment of one’s alcohol use, indications on the evolution [increase or decrease] of one’s alcohol use since last use). How to obtain the different badges in the goal-setting module was also added (available when users click on any given badge). Table 1 presents how these badges can be earned. The full series of potentially collectible badges was displayed (with to-be-obtained badges grayed). Notifications and visual prompts were modified (notably, a red dot announcing that one should report something).

Further modifications were made on the design with modifications to the home page (icons added), ordering of modules, graphical presentation of the normative feedback results, and presentation of the monitoring data. Data entered in the monitoring module were transferred into the goal-setting module (if activated) and vice versa. The structure of the fact sheets module was modified. A 1-month challenge option was added. Presentation of the BAC calculator results was updated with presentation of possible symptoms at the BAC reached for the reported consumption. Because of financial constraints and availability of numerous calculators, we did not include the possibility to report food intake as part of the BAC computation. Screenshots of the final app are presented in Multimedia Appendix 2.

We developed a smartphone app targeting unhealthy alcohol use for students using an iterative development process. This study indicates that a smartphone app is an acceptable way to deliver unhealthy alcohol use interventions to students, a population with notable technology skills. Although the app was generally well accepted and appeared to be a suitable mean to deliver a brief intervention for unhealthy alcohol use, qualitative interviews allowed us to identify important aspects for the target audience: the app has to have a high level of usability, its design must be simple and attractive, users must consider the app content targeted and relevant to their needs, the app should have a high level of credibility, and finally, notifications and prompts are crucial to keep users interested and engaged. Through the iterative process we were able to develop an app that incorporated evidence-based elements from other electronic interventions and that corresponds to the needs and perceptions of the targeted audience. Our results inform on app development and the need to focus on elements of relevance and scientific credibility, as well as clear and effective presentation of scientific data. In addition, independent of its content, an app has to present a high degree of usability (ie, any task required by the app that is not immediately understandable will be abandoned) and an up-to-date and targeted design. Any feedback must be easily and immediately interpretable (hence the necessity to present scientific data on readily understandable graphic format). The app should also be adapted to the local and cultural context and some features, such as a cartoon character, may be well received in some but not in other populations groups. Similarly, especially among students, some of whom may be studying computer programming and app development, there is a high sensitivity for an up-to-date design, which may limit the half-life of apps.

Currently, the evidence for efficacy of a smartphone app for unhealthy alcohol use is scarce and thus there is an urgent need for efficacy data. A key feature in being able to assess an intervention’s efficacy is giving the studied intervention the best chance to be used by its target population [43]. One of the challenges of electronic interventions is to have users involved with the intervention content, notably with repeated use of the app. This study adds that by involving members of the target population in the development of the intervention, substantial modifications can be made to the design and presentation of its components. These changes should lead to a more targeted app design. As an example, comments were made on the prototype 1 app icon which, for some users, would have led to a lack of interest in the app before even assessing its content. As such, design plays a crucial role in electronic interventions’ abilities to reach their target audience. While this qualitative study did not lead to changes in terms of the evidence-based components of the intervention, major changes were made on how the feedback was delivered, how results were presented, and how people would perceive the app, notably in terms of scientific credibility, noting that, in the studied population, the app developed by a university hospital was not sufficient and that data sources (eg, for the normative feedback) had to be prominently displayed to increase credibility and relatedness to the feedback results.

This study presents limitations. A notable limitation is linked to its own justification: while we targeted a population to develop a specifically suited intervention (ie, students from tertiary education institutions in Switzerland), the results are only relevant to this specific population and generalizability is limited. Nonetheless, broad categories are identified that are likely relevant to other populations as well. Other studies have shown the importance of credibility, ease of use, and tailored content [30,44-46] and while achieving this may differ according to different context and populations, we expect it to be relevant for other populations than just for the sample studied. As participants were specifically recruited to test the app, our study sample comprises people who were likely more motivated than the target population intended to use the smartphone app. Thus, results from this study will have to be compared with usage data that will be obtained in the second phase of the study (randomized trial).

Although our results are informative on the development process of the app, we currently do not have data on the app’s effectiveness in reducing unhealthy use of alcohol nor on how and when the app is used. This will be the subject of the ongoing randomized trial (started in May 2021).

Our results bear similarities with other studies on electronic interventions. Notably, Baumel and Kane [30] have shown the importance of design efficiency in real-world usage of eHealth interventions. In a study on a smartphone intervention targeting drinking conducted among members of the British Armed Forces, Puddephat and colleagues [45] showed the importance of credibility of information, targeted and personalized content, ease of use, and simplicity [45]. Garnett and colleagues [31] also showed the importance of the credibility of scientific sources. In a usability testing study among young adults who participated in focus groups after using a smartphone app targeting harmful drinking, Milward and colleagues [44] showed the importance of design, ease of use, and tailored content. Outside of a research setting, available evidence indicates that users will choose an app based on its look and credibility and tailored content [47].

This qualitative study conducted among students shows that smartphone apps targeting unhealthy alcohol use need to have a simple and attractive design, tailored features, scientific credibility, be easy to use, and that the app should regularly send notifications.