The development of the mobile app was guided by the Center for eHealth Research (CeHRes) road map, an established framework for the development of eHealth tools [19,34] in which evidence and theoretical insights are intertwined with design methods. This road map emphasizes a human-centered participatory design process and iterative approach to eHealth tool development [19]. Following this approach, the mobile app was designed with university students consistently at the center of the process and involved at all stages, from the analysis of the problem and context (ie, contextual inquiry) to the technical operationalization of the intervention. Students acted as informants and co-designers of the intervention. At the early stages of the intervention development, focus groups were conducted with other relevant stakeholders, that is, university psychologists and study advisors who are in close contact with this population. The development process involved a thorough review of scientific literature and multiple rounds of expert sessions involving MH professionals and behavioral scientists to lay a strong evidence- and theory-based foundation for our intervention. In collaboration with serious game and interaction designers, all intervention concepts and designs were tested with end users using qualitative and quantitative methods and iteratively adjusted in line with the results.

The intervention incorporates a transdiagnostic approach to MH (ie, Unified Protocol for Transdiagnostic Treatment of Emotional Disorders [35]) and implements techniques from various therapeutic approaches, such as positive psychology [36], mindfulness [37], acceptance and commitment therapy (ACT) [38], self-compassion [39], and cognitive behavioral therapy [40], to accommodate the diverse needs and content preferences of the student population. Using a transdiagnostic approach implies operating outside the traditional psychodiagnostics boundaries (ie, Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition categories), focusing on the factors that cause or maintain problems that students present with. Instead of restrictedly following protocols aimed at specific disorders, this app will provide evidence-based interventions in a flexible fashion and in line with the needs and preferences of individual students [41].

Targeted psychological factors include negative self-referential thoughts, rumination, experiences of prolonged negative emotional states, and lack of MH literacy. These factors have been highlighted in the scientific literature [35,42-45] as well as by students, university psychologists, and study advisors interviewed at the start of this project. Targeting these transdiagnostic factors has been shown promising in the reduction of depressive and anxiety symptoms [35], both of which are prevalent among the university student population [3,4].

The current version of the mobile app consists of a suite of MH interventions in the form of short exercises. The goal of these exercises is to teach students how to use adaptive ER strategies such as acceptance, self-soothing, and reappraisal instead of maladaptive strategies such as avoidance, suppression, and rumination. In the short term, this will help them manage their emotional responses to internal (eg, unhelpful thoughts) or external (eg, failing a test) events by improving their ability to adjust or change the intensity and duration of emotional states [46]. In addition, students are frequently prompted to evaluate their emotional states through ecological momentary assessments (EMAs), which enable them to practice self-awareness and self-monitoring. In the long run, improving self-awareness and adopting adaptive ER strategies can work preventively by decreasing the likelihood of developing serious MH problems such as anxiety and mood disorders [47].

For the final version of our MH app, we envision that students will be able to access multiple exercises at the same time and in any order and consume the content that is most relevant to their unique MH needs. Such a strategy has been previously proven successful in decreasing stress in university students in an Australian program for students called “thedesk” [48]. In addition, our app will include a recommendation system that will suggest exercises to students considering their unique characteristics, such as their goals, intervention preferences, MH state, momentary emotional state, the time of day, and individual app use history. The recommendation system introduces a potential solution to the low engagement and impact of the mobile intervention because efficiency, system responsiveness, and content relevance to users’ needs are highly valued among this population and have been reported as lacking in available MH apps [18,49].

To design an adaptive system of interventions responsive to users’ needs, the optimization of their contents and delivery is crucial. This study will combine an MRT with user experience interviews, a novel study design that can provide information on which intervention is effective for whom and in what context [30].

An MRT can evaluate how different intervention components affect an individual in real time (ie, directly after they have used it), and it assesses how this effect varies over time (eg, an intervention may be effective at the beginning, but its effectiveness may decrease with time). In addition, this method can evaluate the moderating role of contextual factors (eg, current mood or the time of day) on the intervention effect, which is especially helpful when optimizing rules for intervention delivery [50,51]. Although it is a young method, an MRT has been successfully applied to evaluate dynamic and user-tailored eHealth interventions focused on weight management [52], smoking cessation [53], and MH [54]. Within the field of MH, NeCamp et al [55] investigated the optimal timing for delivering a mobile MH intervention to medical interns. The authors found that delivering an intervention when individuals experienced lower levels of MH indicated by low mood, poor sleep, and reduced activity was beneficial, whereas intervening when the individuals were doing well had a negative effect [55]. Another MRT study evaluated the impact of an ACT-based mobile intervention on distress symptoms (ie, depressive and stress symptoms) in first-generation university students [56], and the results suggest that interventions prompting mindful awareness and acceptance of internal experience (ie, thoughts, emotions, and physical sensations) were the most impactful for decreasing depressive symptoms. In addition, interventions delivered in the evening were more predictive of decreases in distress symptoms than those delivered in the morning [56]. Both studies signal the importance of considering contextual factors to optimize the impact of digital MH interventions.

Using MRT research design allows for the gathering of more granular and proximal insights about a given intervention, which cannot be captured through traditional RCT designs. In an MRT, unlike in a traditional RCT where individuals are randomized to an intervention or control group at the beginning of the trial, different intervention components (or no intervention) are repeatedly randomized to the participants at prespecified decision points. Thus, in an MRT, the randomization unit is not the participant but the intervention, and participants act as their own controls. Therefore, compared with an RCT, which focuses on between-participant effects, the MRT measures within-participant effects aggregated across participants. The decision point refers to the time of day at which an individual may or may not receive an intervention. Furthermore, the proximal outcome refers to the short-term goals of the intervention (ie, intervention targets) and is seen as the facilitator of the desired distal outcome of the intervention (ie, a mediator of the long-term outcome); for example, if the focus of an intervention is on increasing physical activity, a proximal outcome would be the number of steps taken within 30 minutes after an intervention was delivered, and the distal outcome would be the overall increase in physical activity at the end of the entire intervention period [51].

Finally, owing to repeated randomization, an MRT allows for the evaluation of the moderating role of contextual factors, such as current mood and time, on the proximal effects of the intervention [22].

This study will experimentally evaluate whether the intervention as a whole and its separate intervention categories (ie, groups of exercises categorized into 5 therapeutic approaches) have the intended effects on participants’ subsequent emotional states, which, in the long term, may have an impact on students’ MH (eg, depressive and anxiety symptoms) and acquisition of ER skills [47], and whether these effects deteriorate over time. Increases in positive affect and decreases in negative affect when an intervention is received compared with when a control intervention is received are expected and considered effective. For exercises targeting unhelpful thoughts (ie, cognitive defusion [CD]), changes in thought believability and discomfort with the thought before and after the exercises will be observed, and unhelpful thoughts are expected to subside as a result of exercise completion.

The feasibility of the system will be examined by evaluating objective (ie, the pattern of interaction with the exercises) and subjective (ie, participant ratings and experience with the exercises) engagement behavioral patterns.

ER skills and level of distress (ie, stress as well as anxiety and depressive symptoms) will be evaluated before and after the intervention period for the purpose of safety monitoring and to inform the design of a follow-up RCT. In terms of safety, the results will be used to indicate whether negative changes in individuals’ MH (ie, distress symptoms) occurred between baseline and the end of the intervention period. As there is no comparator group, the results will provide only an indication of a potential link between worsened (or improved) distress symptoms and the intervention, which will be further tested in a traditional RCT. In terms of RCT design, given the innovative nature of this intervention, there is no prior data available to estimate its effect size. Therefore, we will use the pre-post evaluation of distal outcomes in this study as a pilot evaluation of changes in the distal outcomes to estimate the potential effect size of the intervention. This estimate will be used to calculate the sample size for the standard RCT, where between-group changes in distal outcomes (ie, ER skills and distress symptoms) will be evaluated.

The role of the time of day, momentary emotional state, personality type, preferences, and intervention type as moderators will be explored to determine the context under which the intervention and its categories exercises are the most effective.

From the 200 participants included in the MRT study, 21 (10.5%) will be invited to participate in the semistructured interviews that will occur within 2 weeks of the completion of the MRT study. We expect that 21 participants will provide adequate saturation of themes [62,63]. The subsample will consist of participants with different intervention completion rates (low [<30%], medium [30%-70%], and high [>70%]) to gain a better understanding of engagement patterns and user experiences with the intervention. We aim to balance the sample across gender and study level (ie, bachelor’s or master’s course) using a purposive sampling strategy. Each adherence group would thus consist of 7 participants: 3 (43%) women, 3 (43%) men, and 1 (14%) nonbinary individual. We will attempt to ensure that this subsample comprises roughly equal proportions of bachelor’s students and master’s students. In case thematic saturation is not reached after interviewing 21 participants, we will attempt to recruit additional participants. However, should saturation be achieved before interviewing 21 participants, we may opt to stop recruitment earlier.

The intervention under study is delivered through the student well-being app developed by EUR in collaboration with IJsfontein, a media company specializing in serious game design. The version of the app used in this study consists of 3 core features building the intervention flow: (1) EMAs assessing participants’ emotional states before and after engaging in the assigned exercise or control intervention, (2) a visualization of participants’ emotional states acting as immediate feedback to EMA completion, and (3) a total of 20 exercises targeting different ER skills as well as 20 health facts (control intervention; refer to Multimedia Appendix 1 for example intervention pages).

The exercises are categorized into 5 intervention categories that correspond to the 5 therapeutic approaches integrated in this app: upregulation of positive affect, mindfulness, CD, relaxation and breathing, and self-compassion (refer to Textbox 1 [36, 37, 39, 64] for a list of the exercises). Each therapeutic approach is explained in an article available to participants when they engage in the corresponding exercise. The exercises are relatively short (each lasts between 3 and 12 min) and actionable in the moment. They were selected owing to their potential to directly regulate students’ emotional states and can be completed without extensive exercise background knowledge. Each exercise includes a brief explanation of the rationale behind the exercise as well as debriefing information explaining how the exercise works in general and what the reason might be for experiencing discomfort upon its completion.

ER will be measured with the Emotion Regulation Skills Questionnaire [100], which evaluates 8 types of ER skills: awareness, sensation, clarity, understanding acceptance, tolerance, compassionate self-support, readiness to confront distressing situations, and modification of negative emotions. Each skill is assessed with 3 items rated on a scale ranging from 0 (not at all) to 4 (almost always). Apart from the scores on separate subscales, the Emotion Regulation Skills Questionnaire consists of a total score computed as the average of all items. Higher scores indicate higher ER skills.

Symptoms of depression and anxiety are measured with the PHQ-9 [61] and the Generalized Anxiety Disorder-7 (GAD-7) [101] questionnaire, respectively. The PHQ-9 consists of 9 items measuring depressive symptoms, and the GAD-7 is composed of 7 items measuring anxiety symptoms. Both questionnaires instruct participants to indicate how often they have been bothered by different symptoms over the last 2 weeks on a scale ranging from 0 (not at all) to 3 (nearly every day). The summed scores for the PHQ-9 range from 0 to 27 and for the GAD-7 from 0 to 21, with higher scores indicating higher levels of depressive and anxiety symptoms. The PHQ-9 is used both to screen for eligibility and to assess the baseline level of depression.

Stress levels will be assessed with the Perceived Stress Scale-14 [102], which consists of 14 items and instructs participants to indicate how often they felt or thought a certain way over the last month on a scale ranging from 0 (never) to 4 (very often). The summed scores range from 0 to 56, with higher scores indicating higher levels of perceived stress symptoms (refer to Table 1 for an overview of the measurement schedule).

Demographics information includes age; gender; university discipline of study (eg, law and medicine); study program (ie, bachelor’s or master’s); prior experience with MH apps (ie, use); and preknowledge and use of, and experience with, the therapeutic approaches included in this study. For the purposes of the recommendation system development, additional information on students’ MH, vulnerability (eg, care duties and functional impairments), and personality traits is collected at the baseline only. These outcomes will be used to explore whether these factors play a role in recommending exercises to students.

Students’ MH is screened with the General Health Questionnaire-12 (GHQ-12) [103]. The GHQ-12 consists of 12 items, each assessing the severity of an MH problem over the past few weeks using a 4-point scale (ranging from 0 to 3). The total scores range from 0 to 36, with higher scores indicating lower psychological well-being.

Personality traits are measured with the Ten-Item Personality Inventory [104], which assesses 5 personality dimensions: extraversion, agreeableness, conscientiousness, emotional stability, and openness to experience. Participants are asked to evaluate to what extent they agree with each statement on a scale ranging from 1 (disagree strongly) to 7 (agree strongly). An example statement is “I see myself extraverted; enthusiastic.”

To secure a diverse sample in terms of gender and students’ university discipline of study, participants will be approached through various pathways. Students who had subscribed to our user research participant pool over the last couple of years will be contacted via email. The study will be advertised on EUR’s campus screens and social media platforms (ie, well-being website and Instagram account); in November 2022, it was advertised during a Student Well-being Week event. Promotional material for the study will be sent to students via communication and marketing offices and student associations.

Interested students will subscribe to the study via a QR code or hyperlink. Those subscribing to the study will be emailed a link to a web page with information for participants and will have 2 weeks to review the information and decide on participation. They will be invited to contact the researcher via phone or email for any study-related question. Students who are interested in participating can indicate this preference at the bottom of the web page and will be automatically directed to the digital informed consent form. Once the informed consent form is signed, participants are screened for eligibility.

Eligible participants will fill in the baseline questionnaires and be asked to install the app on their mobile phones through which the intervention will be delivered over a period of 3 weeks.

Each student will receive a unique randomly generated user ID to access the app to protect their privacy. Participants’ status as EUR students will be verified through the authentication of their student accounts. Once the authentication phase is concluded, participants will be asked to select 1 morning (7 AM to noon) and 1 evening (5 PM to 10 PM) time slot when they believe they are most likely to have time to complete the daily assessments and exercises (Multimedia Appendix 1).

During the intervention phase, participants are asked to complete the intervention flow (ie, preintervention EMA, followed by exercise or health snippet, followed by postintervention EMA) 2 times per day. The prompts (eg, “Good morning! Don’t forget to take a moment for yourself.” or “Psst, did you check your evening exercise for today?”) to complete the flow will be sent to participants at the 2 self-selected time slots. Participants can self-initiate the intervention flow during each time frame by opening the app. If they complete the intervention flow before the self-selected time slot, the reminder for this time slot is skipped. Of note, participants must complete the preintervention EMA to be able to continue and complete the intervention flow. The intervention flow will be available to participants to complete between 7 AM and 2 PM (morning time slot) and between 5 PM and midnight (evening time slot). The completion of tasks in the app is not possible outside of the predetermined time slots. In addition, to ensure that the changes in emotional states are most likely a result of exercise completion rather than other factors, a time limit was set. Participants will have 30 minutes to complete the intervention flow; otherwise, they need to start the intervention flow from the beginning, starting with the preintervention EMA.

Every time participants fill in the preintervention EMA, they have a 60% chance of receiving an exercise and a 40% chance of receiving a health snippet. The 60:40 ratio was chosen to ensure a higher chance of receiving an intervention at every decision point. This strategy has been used in previous MRT studies too [51,52]. For this study, a simple randomization with replacement will be used. The randomization is carried out automatically within the app and is as follows:  randomization of exercise (60% chance) or control intervention; if an exercise is chosen, a second randomization to the intervention categories follows (20% chance for each intervention category), followed by the random selection of 1 exercise within the randomized intervention category (equal chance across exercises); if a control intervention is randomized (40% chance), a second randomization takes place in which a health information snippet is selected. Participants will be asked to complete 42 preintervention EMAs and 42 postintervention EMAs in total during the study. They can receive an exercise 0 times, 1 time, or maximally 2 times in a day. On average, they will be prompted to engage in 1.2 exercises per day and 25.2 exercises throughout the study.

Upon the completion of either the exercise or the control intervention, the participants are asked to reassess their emotional state by completing the postintervention EMA. When assigned to an exercise, in addition to the postintervention EMA, participants will be asked to rate the likability and helpfulness of the exercise. When CD exercises are selected, participants will also complete the corresponding measures of thought believability and discomfort before and after the exercise.

Given that the primary objective of this study is to evaluate the functioning of the intervention rather than observe it in a naturalistic setting, we have implemented an adherence protocol to enhance completion rates and ensure high-quality data. To support participants in staying engaged, they will receive emails or telephone calls from researchers, depending on the number of missed sessions. The adherence protocol is presented in Multimedia Appendix 3.

Immediately after the intervention period is concluded, participants will complete the postintervention assessment, including the secondary outcome measures (refer to Table 1 for an overview of the measurements performed during phase 1 of the study).

In the second phase, semistructured interviews will be conducted with 21 (10.5%) of the 200 study participants. A purposive sampling strategy will be used to balance the sample across gender, study level, and the level of adherence during the MRT. Participants who consented to participate in the user experience interviews will be selected based on the aforementioned characteristics and contacted by the researcher to set up an interview session within 2 weeks of completing the follow-up questionnaire. The interviews will take place on the web or at the EUR behavioral laboratory and will last between 45 and 60 minutes. An overview of the study procedures is presented in Figure 1.