The app handles a large amount of sensitive data, which must be stored and transferred securely. User data were handled according to good clinical practice standards, with users being given unique anonymous identifier numbers when the app is first launched on a new device. For secure data transmission, we require the use of HTTPS across all our end points, which makes tampering with requests and responses significantly more difficult. To authorize clients, we implemented a time-based token system where the server and client have a shared secret, which is then combined with a timestamp and hashed, and then sent over the network to authenticate the user.

The app was tested on multiple devices for a range of parameters to establish its suitability for and performance on a diverse range of hardware. The iOS app was tested on iPhone 6S (iOS 13.6), iPhone 6 Plus (iOS 13.2), iPhone 11 (iOS 13.7), iPhone 12 (iOS 14.6), and iPhone 13 (iOS 14.8). The Android version was tested on a range of devices including Google Pixel 4a (Android 13), Samsung Galaxy A32 (Android 12), Realme 5 Pro (Android 11), Huawei P20 Pro (Android 10), and Redmi 7 (Android 10). The app worked on all devices smoothly. Based on qualitative human experience, the app operations on the older phone models were as good as on the newer models, highlighting the low degree by which the app taxed the system's resources.

Performance testing was done using Xcode (for iOS) and Android Studio Profile (for Android) to ensure that the app does not disrupt the normal performance of the phone. During normal use, we observed that memory consumption was constant with no memory leaks, and CPU usage remained low, with the CPU only being used during tab switches and submissions.

We tested the app for battery usage on an iPhone 6 Plus and Redmi 7 for 2 months as development was carried out and for 2 weeks following installation of the finalized version of the app bundle. We deliberately chose older model phones, as they have less battery capacity, less energy-optimized hardware, and smaller batteries than more recent models. During normal use of the app, the battery consumption was mostly in the low to medium range.

Throughout the design and development process, we used a user-centered design approach [31] to ensure a user-focused, cocreation-driven app and followed a 3-step process:

By involving users early in the design process and maintaining open communication throughout, the user-centered design approach resulted in an app that is not only effective and engaging but also addresses the needs of our target audience.

Ethics approval for the study was granted by the Imperial College London Research Ethics Committee (ICREC 20IC6132).

Informed consent was obtained at the start of the study, with parents providing consent for those aged younger than 16 years. Only once the participant understood the privacy policy and accepted the terms and conditions of the study in the app were they able to proceed through the app and partake in the study. The privacy policy and information storage practices were also accessible within the app.

We ensured European Union General Data Protection Regulation compliance with data retrieval, anonymization, and a withdrawal process. All participants were informed, in the terms and conditions, of where and how the data they provided would be used. Users had full control over the data that were collected and could opt out of any single type of data collection at any time. Users had the option to delete the entirety of their data and withdraw from the study via the settings option in the app.

Participation was incentivized with a prize draw for a £30 (US $37) voucher and an educational session.

For user acceptance testing, participants were recruited via social media platforms (Twitter and Nextdoor, a local neighborhood site). The inclusion criteria included being aged 14-18 years, fluency in English, access to a video and audio-enabled device, and possession of a smartphone. The exclusion criterion was insufficient language skills. Participants were instructed to use the app for 3 days, explore its various features, and share their experience of using the app with the researchers following the evaluation period.

For the pilot study, we reached out to many secondary educational establishments via email and chose the first school that expressed interest in participating. Participants were recruited via this secondary school, which promoted the study directly to students through emails sent home. Inclusion and exclusion criteria for pilot testing were identical to those for user acceptance testing. The school forwarded information on the app to year 10, 11, 12, and 13 pupils.

Interested pupils voluntarily completed a web-based survey via a link provided in the promotional information. Informed consent was obtained digitally, with parents providing consent for those aged younger than 16 years. Upon completion of the survey, participants received a link to download the app from the App Store or Play Store, along with a unique login for the app. Participants were asked to use the app for 2 weeks. Upon completion of the 2-week testing period, user engagement data and the types of active and passive data shared by the users were analyzed to evaluate the app's performance.

After the app's development, user acceptance testing was conducted with a recruited patient and public involvement group. We carried out an exploratory qualitative study with this group to understand the target audience's user experience and identify barriers to adopting Mindcraft. Eight young people (including 4 female individuals) aged 15-17 years, with a mean age of 15.9 (SD 0.9) years were invited to download the app and use it for a 3-day evaluation period. Two researchers conducted 2 digital patient and public involvement groups via Microsoft Teams (Microsoft), each comprising 4 participants, lasting 180 minutes. The semistructured interview explored their experience using the app.

The qualitative feedback was predominantly positive. Discussions with study participants revealed that some viewed the app as a friendly tool promoting self-awareness and personal growth. They found the app enabled them to increase their emotional awareness and gain a better understanding of themselves, as reported in previous studies on self-monitoring tools [32,33]. Participants perceived the app interface as engaging and less clinical. Although they expressed concerns regarding data privacy and sharing (particularly passive data) and desired greater control, they were relatively more comfortable sharing passive data for mental health. Our test users did not consider social competition a motivator in mental health apps, which contrasts with the findings of other studies [34,35].

Based on user feedback, we revised the app's text prompts and updated the UI to offer more control over passive sensor data collection. The updated app was uploaded to the App Store and Play Store for the pilot study with a larger user group.

The app was advertised to students from a secondary school across 4-year groups (aged 14-18 years) for a 2-week trial period. A total of 39 students (20 male students and 19 female students), with a mean age of 15.74 (SD 1.09) years, volunteered and downloaded the app. Of these, 32 used the app on iPhones, while the rest used Android phones.

We calculated user engagement over time by analyzing the number of users who used the app over the 2-week evaluation period (Figure 4). Out of the 39 users, 16 (41%) used the app after 7 days, while 9 (23%) used it after 14 days. The mean number of days the app was used during the first 14 days was 5.76 (SD 4.48). Passive data collection allowed the app to gather more well-being metrics over time without user intervention. While the number of users sharing active data decreased after the first week probably because of the app's novelty wearing off, the number of users sharing passive data remained steady in the second week (Figure 4), indicating that passive data collection may offer a more sustainable data collection method in the long term.

Seven active data questions were mandatory, while the other 12 were optional (enabled by default). Although users had the option to disable optional active data questions, 92.5% (36/39) left them enabled. Over 90% of users (36/39, 92.5%) answered all active data questions (except for the customizable Your Measure) on the days they used the app (Figure 5A). Loneliness, appetite, and sleep quality were the most frequently answered active data questions. Concerning passive trackers, users favored sharing noninvasive data, such as step count and battery levels (Figure 5B). However, they were reluctant to share intrusive passive data, like location and background noise levels, because of privacy concerns.

The development and initial testing of the Mindcraft app have shown promising results in monitoring mental health. The pilot study found good user retention and engagement, with 23.07% of users (9/39) using the app after 14 days. While Mindcraft's effectiveness and acceptability among the target population to track their mental health may be attributed to its user-centered design, the emphasis on transparency and privacy, and the customizable use of both active and passive data collection strategies, we recognize that our study's good retention rates should be interpreted with caution. The rates may be influenced by participants' awareness of being part of a research study, potentially leading to increased engagement compared to the naturalistic app user retention rate of 4% over a 15-day period [36,37].

Our results support previous findings [38-40], indicating that passive data collection can provide a sustainable and effective way to monitor mental health symptoms over time. We found that passive data collection reduced user burden while also enabling the collection of a wider range of well-being metrics. These benefits have important implications for the design of mental health monitoring apps, as app developers can leverage the advantages of passive data collection. By prioritizing user privacy, transparency, and customizability of data sharing, developers can design apps that foster user trust and engagement.

We acknowledge that numerous studies and apps [11,14,19,41-55] use passive sensing to support mental health and well-being. However, the majority of these studies target adult populations, while research on CYP remains limited [19,20]. Our study aims to address this gap by focusing on the mental well-being of the general adolescent population using a combined active-passive data approach. A few studies [19,43,54,55] have investigated passive sensing for adolescent mental health. Maharjan et al [55] evaluated passive sensing for predicting maternal depression in young mothers in a low-resource setting. Cao et al [43] and Mullick et al [54] examined the effectiveness of passive sensing in tracking depression symptoms in clinically depressed adolescents. The study by MacLeod et al [19] is the closest to our study, as it involved adolescents from both clinical and nonclinical settings. However, it exclusively used passive sensing, and the participants had no interactions with the study app after the initial setup. In comparison, our study focuses on monitoring the mental well-being of the general adolescent population by integrating both active and passive data. Our app enables users to decide which active and passive data they want to share, providing them with control over their privacy, a key factor for real-world deployment to the general population. We demonstrate that combining both active and passive data collection strategies, with an emphasis on privacy, transparency, and an engaging design, effectively promotes sustained engagement with the app and improves long-term data collection from the general adolescent population. Additionally, we incorporated a user-centered design approach, gathering end user feedback during the development process, an element often neglected in mental health app studies [56,57]. It is worth mentioning that, given the lack of regulatory oversight in the mental health app market [10,58], Mindcraft's development was guided by a team of experienced clinicians, ensuring evidence-based and clinically relevant content. In summary, although Mindcraft shares some similarities with other mental health apps and studies, its focus on the general adolescent population through an integrated active-passive data approach, user-centered design, and clinically relevant content highlights its potential to contribute meaningfully to the field.

Our study has some limitations that suggest directions for future research. Although positive results about the user engagement were observed, the conclusions must be considered preliminary because of the relatively small sample size, and larger samples are needed to validate the acceptability of the platform among the target population. Additionally, the pilot study users used the app for a limited period of 2 weeks, making it necessary to investigate its usage over a longer period to assess its effectiveness. Another limitation to consider is the potential impact of the research study environment on our retention and engagement rates. The heightened engagement we observed might be partly attributed to the participants' awareness of their involvement in a study. Therefore, future studies should aim to assess Mindcraft's retention rates in real-world settings outside the research context, to provide a more accurate reflection of its performance.

Throughout the study, we recognized opportunities to improve Mindcraft's user experience and capability to build a more comprehensive user behavior profile in its future version. Potential future enhancements could include gamification, which has been shown to increase user engagement in therapeutic apps [59] and mental health interventions for adolescents [60]. The app may also benefit from new passive data sources like voice recordings for speech emotion recognition and continuous accelerometer and gyroscope data for more meaningful activity measures and sleep monitoring. Integrating music listening patterns, as shown by Rickard et al [11], could further help Mindcraft better track users' emotional well-being over longer periods without being too intrusive.

Although the potential of recommended systems has been occasionally explored in health care research [61-63], the possibility of taking advantage of this technology to improve mental health care is yet to be sufficiently explored. Mindcraft presents the opportunity for an enhanced recommendation system to be developed where recommendations may be delivered based on personality, along with other factors such as demographics, behavior, and self-reported scores for specific symptoms. Future versions of Mindcraft could integrate a reinforcement learning–based personalized recommendation system, where an agent may be trained to work toward recommending activities [64,65] or mental imagery [66] to increase mood and mental health scores while being penalized and rewarded accordingly based on score variations and user feedback.

While platforms such as Kooth and Childline provide comprehensive coverage for those seeking help, these services require CYP to be motivated to engage with a counselor. Mindcraft can help overcome this motivation barrier by using passive tracking data to detect early indicators of poor well-being and mental health issues, triggering them to seek counseling help. Alternatively, the Mindcraft app could also be extended to serve as an intervention system by incorporating in-app counseling services based on the active and passive data collected. This would be a step toward achieving the goal of self-service mental health services and improving access to care for those who may not otherwise seek help.

The Mindcraft app has shown promising results in monitoring mental health and promoting user engagement among CYP during its development and initial testing. The app's user-centered design, the focus on privacy and transparency, and a combination of active and passive data collection strategies have contributed to its effectiveness and acceptability among the target population. Future enhancements could include gamification, additional sensor data, and personalized interventions to improve user engagement and mental health outcomes. By continuing to refine and expand the app, the Mindcraft platform has the potential to contribute meaningfully to the field of mental health care for young people.