Descriptive analyses were conducted to determine whether the tools achieved benchmark values on NPS and SUS (acceptability dimension), as well as to assess platform usage (demand dimension). Inferential analyses were run as a means to explore the potential of these tools for inferring individual behavior and thus for supporting early detection and treatment response in MH.

For continuous variables, a normality test was conducted. Considering the sample size, medians and IQRs were preferred as measures of central tendency. For nominal variables, absolute and relative frequencies were described. For ordinal variables, absolute frequencies, relative frequencies, and medians were calculated.

The statistical analysis was conducted using R software version 4.3.2 (The R Foundation for Statistical Computing). We performed the Shapiro-Wilk test (P>.05 for normality), calculated central tendency measures, skewness, and kurtosis. Additionally, a bivariate analysis was carried out, with qualitative interpretations of the K6 scale and well-being scale as dependent variables. We used the chi-square test (P<.05) for associations among qualitative variables and both the Wilcoxon test (P<.05) and Mann-Whitney U test (P<.05) for quantitative comparisons. Correlation matrices were also created for survey results from MB and YMC with sensor activation data, and significance was determined with P<.05. Additional correlation matrices were produced between MB and YMC variables, also with P<.05. This paper presents only the results that reached statistical significance.

Other feasibility dimensions (implementation and practicality), as well as information extracted from interviews related to acceptability and demand, were evaluated through qualitative data analysis. For qualitative data, a thematic analysis using a deductive framework was conducted. The process began with data familiarization through transcription and repeated readings. A research assistant transcribed the interviews and developed an initial set of codes. A senior psychologist with extensive experience in qualitative research reviewed these codes and organized the material according to the dimensions proposed by Bowen et al [33]—acceptability, demand, implementation, and practicality. The themes were subsequently reviewed for internal validity. Both psychologists collaborated on the final analysis and the writing of the results. NVivo version 14 (Lumivero) was used for qualitative data analysis.

Based on survey responses, more than half of the participants considered the YMC service to be good (20/35, 57%), useful enough most frequently (20/35, 57%), and assured that they would use it again (29/35, 83%). These findings are consistent with interview results, for example, about half of the sample (16/35) is completely satisfied with the platform.

When asked about what they liked most about the platform, the possibility of receiving personalized and easy-to-understand resources was frequently mentioned (23/35 participants).

Filling out the MH surveys and seeing progress charts was considered useful by 26 of 35 participants. Additionally, 10 of 35 participants highlighted the telecounseling option as one of the most useful features of YMC, as they claimed that often they do not know where to seek help and having this option available made help-seeking easier and more accessible. Among these 10 participants, fewer than half (4/10) had no prior or current experience with psychiatric or psychological care. This suggests that their positive views on the telecounseling feature were likely independent of past service use. In general, the surveys were seen as relevant, easy to answer, short, and the possibility of choosing their topics was emphasized as a positive feature.

Usability is a key aspect to consider when studying a technology. Bowen and colleagues’ framework [33] for designing feasibility studies is not intended for technologies specifically, but for interventions more broadly. However, since the focus of this study is on 2 ICTs, usability measures were also included as part of the more general acceptability category. For YMC, participants gave this platform a score of 72.86 on the SUS, rendering it acceptable but requiring adjustments to improve its usability, mainly due to the technical difficulties that arose. Participants considered that these issues were the main cause of discouragement in using the platform. In interviews, dissatisfaction with the platform was primarily related to technical issues that limited access to the YMC web page for 3 days and generated difficulties logging in to their accounts. Nevertheless, a considerable percentage of participants (29/35, 83%) expressed the intention of using it again, and overall, they rated it as user-friendly, self-manageable, and free of contradictions. Participants considered that they could easily grasp how to use it without requiring any training or previous knowledge to start using it.

A score of 2.85 was obtained on the NPS scale: 34% (12/35) of the participants were promoters, 34% (12/35) passive, and 32% (11/35) as detractors. Despite this, the majority of participants (30/35) in the interviews expressed their intention to use the platform again, citing the personalized psychoeducational resources and the ability to track their MH through surveys as key factors for their decision.

In surveys, this platform received ratings of good and very good from 49% (17/35) and 29% (10/35) of participants, respectively. In terms of usefulness, 25 of 35 participants considered it useful enough and 25 of 35 assured that they would use it again. When it came to overall satisfaction, 24 of 35 participants emphasized that the interface made it easy to use. Additionally, over a quarter of participants (16/35) appreciated the push notifications, which remained visible on their home screens and served as reminders to complete the surveys. Overall, participants found the app to function consistently well.

Moreover, 23 of 35 participants found the MB app beneficial, with many highlighting the ability to monitor their MH as a key advantage. In addition, 18 of 35 interviewees noted that completing daily surveys helped them become more aware of their habits and MH.

However, 20 of 35 participants expressed dissatisfaction with the app, citing the lack of feedback and resources after completing the surveys as a key issue. A lack of satisfaction that MB was only a data collection-focused, rather than MH care-focused app providing psychoeducational resources and other forms of support, led to a perception that the app became dull and monotonous over time. One participant stated:

Considering the NPS, 40% (14/35) of individuals were promoters of MB, 29% (10/35) passive, and 31% (11/35) detractors. This resulted in an NPS score of 8.6 (SD 2.61). In addition to the above, 20 of 35 participants found the MB surveys quick and easy to answer and 12 of 35 participants found the app to work well overall.

Regarding the use of sensors, participants stated that they were not aware of their activation and tracking, even though they granted permissions when first downloading and using the app. In general, participants trusted the app, and the majority of the participants (26/30) had no concerns regarding the handling of their personal data or MH information. They considered that the data requested were not sensitive and mentioned that trust in the researchers and the institution made them feel secure with sharing their data. In general, participants felt that the data were protected, and they prioritized their well-being when using the tools over concerns about data management.

On the SUS, we obtained a score of 81.35, which implies that the platform was within the acceptable range and therefore was well received. Overall, 25 of 35 of the youth said that they would use the MB platform again, mainly because they felt that answering daily questions about aspects of their MH allowed them to become aware of their well-being and the changes they were experiencing. Ease of using the platform obtained the highest score of 4.6. Participants highlighted the easy access and use of the app and that they considered that questions could be answered in a short time and in a simple way. Regarding specific questions of the SUS, the lowest score of 3.91 was obtained for whether they would use this tool frequently. In interviews, the main reason cited by participants was the lack of feedback after answering the surveys and not being able to visualize progress during the week.

We evaluated both demand and perceived demand. Regarding the first, the actual use of YMC and MB functionalities was assessed. In relation to YMC, the most used features were psychoeducational resources, which were used at least once by 100% (35/35) of participants, followed by avatars (27/35, 77%), and completing the well-being plan (18/35, 51%; Table 4). Notably, the least used feature was telecounseling used by 8.6% (3/35) of participants.

During the 3 weeks of the study, each participant used MH resources (mean 18.48, SD 14.54; median 13, IQR 10-22.5).

Regarding the monitoring functionality of YMC, some surveys were programmed to be shown daily, weekly, or every 2 weeks (Table 5). Regarding daily surveys, they were all completed by at least half of participants. The most answered ones were emotions (30/35, 86%) and stress (28/35, 80%), and the least used was physical activity with 57% (20/35) of the sample answering it at least once. Considering the total number of responses, the most used daily surveys were sleep and emotions. For weekly surveys, the most used were self-esteem, relationships, and loneliness; for biweekly surveys, the total number of responses was similar for well-being and sexual and reproductive health. Among these, follow-ups were completed by the majority of participants for the emotions (26/35, 87%) and sleep (19/35, 83%) surveys.

Regarding the MB surveys, all participants completed the surveys on demographic information, physical activity, and sleep at least once (Table 6). Surveys of positive emotions (97.1%) and relationships (97.1%) were answered by more than 90% (32/35) of participants. It is noteworthy that all participants (35/35, 100%) who responded to the physical activity survey answered it 2 or more times, showing better engagement with this survey topic on MB than on YMC. This survey was the most frequently presented to the participants.

Regarding sensors, MB logged various data types throughout the study, including the time of sensor activation and deactivation, duration, acceleration, and more. For this study, only the data described in Table 7 were used. Specifically, for activity type, MB uses a combination of sensors and algorithms to detect the most likely activity the user was performing. A detailed description of the data captured by each sensor is available on the Avicenna web page [42].

Specific to battery level and status, data from this sensor are relevant as they can serve as a measure of cellphone usage, which has been shown to be significantly associated with problematic internet usage in youth [43]. Other sensors, such as GPS, pedometer, and screen time, have been used previously in digital phenotyping research [43-45].

To assess demand for sensors, we evaluated the percentage of participants who had at least one record for each sensor, as well as the median value of the sensor (Table 7). Based on this approach, the most frequently used sensors were battery status and screen status (refer to Table 7 for sensor description), with data recorded for 100% (35/35) and 94.3% (33/35) of the sample, respectively. The activity type most frequently detected was the stationary state with a median of 1713 (IQR 840-2981) records, while the on-foot and on-bicycle states were recorded less frequently. The pedometer was active for 74% (26/35) of participants, providing step count data. Data on battery status showed that participants spent most of their time in a discharging state, while instances of full charge were infrequent.

The GPS sensor (Table 8) focused on travel and location metrics. Over the study period, the median total distance traveled by participants was 793.29 (IQR 140.43-1781.16) km, suggesting variability in mobility levels. The median total travel time recorded was 2501.13 (IQR 269.6875-3890.025) minutes, while participants made a median of 60056.5 (IQR 12725.5-77800.5) visits to different locations. In terms of unique places visited, the median was 42 (IQR 10-72), indicating diverse movement patterns among participants. Time at home had a median of 476.63 (IQR 11.7-927.7625) minutes, contrasting with a significantly higher median time spent outside the home at 1695.4 (IQR 174.1-3004.975) minutes, highlighting a tendency for more time in external environments.

These results showcase the demand for YMC and MB. During the duration of the study, usage was overall high; however, engagement varied according to the type of functionality, survey topic, or sensor.

Regarding perceived demand, 27 of 35 participants thought that other young people in Bogotá would use the platforms. Some (12/35), however, stated that they would need adjustments, including being in an app-based format (7/35), having a strong social media presence (1/35), and improving its interface (2/35). A group of participants (9/35) thought that only those already interested in MH topics would use them.

Given the high demand for MB and YMC, we also explored associations between sensors and MH-related measures. This allows us to explore the feasibility of not only engaging young people with digital health technologies for their MH, but also exploring their utility as potential indicators of MH state. The group with psychological distress had a mean battery level of 62.98 (SD 8.92), whereas the group without psychological distress had a mean battery level of 53.66 (SD 11.03). The results suggest that individuals with psychological distress had a significantly higher mean battery level compared to those without distress. This association was statistically significant, with a P value of .03, indicating that higher psychological stress correlates with a higher battery level. Previous research has linked problematic internet use with psychiatric symptoms [46]. This suggests that metrics like device usage for internet access, as well as related proxies such as screen time or battery level, could hold value for both clinical and research apps.

Likewise, a statistically significant association was found between battery charging status (indicating the phone is being charged) and psychological distress (P=.04). In individuals with psychological distress, the median number of battery charging records was 585 (IQR 321-615), while in those without psychological distress, the median was 188 (IQR 42-309). This finding suggests a relationship between higher levels of psychological distress and increased battery charging. Relatedly, Gansner and colleagues [43] found that a higher number of daily phone checks and phone sessions were related to lower scores in measures of problematic internet use and depression. Not in the direction found in previous studies, Gansner and colleagues proposed an alternative hypothesis: digital engagement may offer relief from negative affect in the short term. Similarly, in our sample, youth with psychological distress might be more prone to use their devices for short-term relief than youth without psychological distress, leading to more frequent charging. Interestingly, youth with psychological distress also exhibited higher battery levels, which could suggest they maintained these levels by frequently charging their phones.

Participants who perceived they slept enough had a median battery discharge of 556 (IQR 200-2968), while those who did not perceive they slept enough had a significantly higher median discharge of 2867 (IQR 1697.5-3935.5). This suggests a statistically significant relationship between battery discharge and perceived sleep quality (P=.003). As for the activity type sensors, a statistically significant relationship was found between immobility and time spent in vehicles in relation to sleep quality (P=.03). Participants who perceived they slept enough recorded a median of 1213 (IQR 218-2607) minutes of immobility, whereas those who did not perceive they slept enough reported a median of 2129 (IQR 1515-6701) minutes. This association was statistically significant.

Similarly, time spent in a vehicle also differed significantly between the two groups. Participants who perceived they slept enough had a median of 46 (IQR 17-525) recorded instances of being in a vehicle, compared to 262 (IQR 150-762) minutes for those who did not perceive they slept enough. This difference was statistically significant, with a P value of .05, suggesting that greater immobility is associated with a worse perception of sleep quality.

A Mann-Whitney U test was performed to compare Screen Status Off in the group that perceived they had not slept well and the group that perceived they had slept well. There was a significant difference in Screen Status Off between these two groups: z=−2.089; P=.04. This test indicates that participants who felt they didn’t sleep well had their phones off for a longer time.

There was a statistically significant negative correlation between the number of unique places visited by a person and their score on the K6 psychological distress test. For this analysis, a nonparametric Spearman correlation test was conducted, resulting in a correlation coefficient of −.424 (P=.05). This implies that the greater the number of unique places a person visited during the 3-week period, the lower their score on the psychological distress test. Unique places are defined as those locations that the participant visited only once during the study.

During the 3-week study, the YMC platform encountered multiple technical difficulties, including server malfunctions, web page errors (reported by 19/35 participants), slow registration processes, delays in starting sessions (22/35), and dissatisfaction with the lack of notifications (3/35). Other complaints included excessive emails, limited survey options, and issues visualizing results.

In contrast, the MB app had no major technical issues and was appreciated for its offline functionality, fast performance (18/35), and push notifications (8/35). However, some survey responses were not recorded or processed, affecting 3 participants, and inconsistent passive data recordings led to the exclusion of some data points. Both platforms faced complaints about rigid survey schedules, which often resulted in expired surveys. The screen state sensor occasionally recorded implausible durations exceeding 24 hours, resulting in 6 data points being excluded. Similarly, the activity type sensor, which categorizes activities (eg, walking and running), recorded discrepancies for one participant where “on-foot” activity data did not align with walking and running records, leading to the exclusion of that participant’s data. These issues highlight challenges in accurately capturing passive data for analysis.

Participant dropout involved 3 individuals. One completed the study but was unreachable for an interview, while the other two left due to technical issues with YMC, including unsaved responses and platform freezing.

Participants reported several positive effects from using the platforms, including increased awareness of their MH and habits (21/35 for MB and 16/35 for YMC), with specific emphasis on sleep surveys helping them recognize the importance of sleep hygiene. Many participants noted that the platforms motivated them to adopt healthier routines (28/35 for MB and 25/35 for YMC) by encouraging mindfulness about daily activities and creating dedicated time for MH. In YMC, learning was frequently highlighted as a benefit (30/35), with participants gaining knowledge, validation, and support through resources, survey results, and graphs.

However, participants expressed dissatisfaction with the lack of feedback in MB (18/35), suggesting it felt like a data collection tool without deeper engagement. A few participants (4/35 for MB and 1/35 for YMC) saw no positive effects, and boredom was noted as a potential barrier to long-term use (3/35 for MB and 4/35 for YMC). Stress from technical malfunctions (7/35 for YMC), frustration with forgetting surveys, and occasional device issues were also reported.