Mental disorders constitute some of the leading causes of the global burden of disease [1]. Not only do they have implications for the health care system and economy, but they also affect the functioning and quality of life of the affected individuals and their families. Increasing efforts have been made to synthesize evidence and scale up early interventions in community settings [2].

Over the past 2 decades, the field has undergone significant transformations in the delivery of psychological and pharmacological treatments, shifting from face-to-face to incorporating digital technologies as part of mental health care. Such interventions range from the use of web-based platforms and online video conferencing to smartphone apps, the experience sampling method (ESM), and ecological momentary intervention (EMI), as well as chatbots and virtual reality, either as an adjunct to or replacement for in-person care. Each approach may vary in its degree of human contact and intensity [3-7]. Although the application of digital technologies to mental health care has been relatively slow compared to efforts in physical health and behavior change, the emergence of the COVID-19 pandemic has greatly accelerated this process.

Among the various digital tools, the use of smartphone-based mental health apps (MHapps), specifically blended care, has received increasing scientific attention and support. Some benefits include their ability to save clinicians time, which is crucial given the burden of mental disorders and the persistent challenge of insufficient manpower, as well as to improve treatment accessibility and reduce dropout rates.

As in pharmacotherapy, adherence to care plans is a crucial determinant of treatment outcomes in psychotherapy. Using cognitive behavioral therapy (CBT) as an example, adherence includes not only regular therapy attendance but also the completion of homework. Homeworks in CBT are cognitive or behavioral activities (or their combination) that are collaboratively planned between the clinician and service user for completion outside of therapy, which involves the direct application of adaptive strategies into everyday life [8,9]. Such activities are personalized and may include scheduling meaningful activities (eg, “leaving the house for 30 minutes every day”), monitoring daily activities and mood states, and identifying cognitive biases while documenting their relationships with triggering events, mood, and physiological reactions [9,10].

Despite the importance of applying these practices to everyday life for more sustainable treatment outcomes [8], homework nonadherence is profoundly common. Beyond difficulties among the general population [11,12], the maintenance of health-related behavior change can be even more challenging in people with subclinical and clinical mental health symptoms, wherein motivational deficits and anhedonia reflect cardinal symptoms of various psychiatric conditions, such as major depression, schizophrenia spectrum disorders, eating disorders, and posttraumatic stress disorder [13]. Blending MHapps with traditional in-person interventions has the potential to fill this gap.

The portability of smartphones and the lack of geographical limitations in their use have opened new opportunities for more person-centered, holistic, and context-relevant interventions [9,14-16]. The EMI is a novel intervention approach that was built upon the ESM—a self-report diary method that enables the collection of real-time data on experiences within individuals and their interactions with one’s real-world context—by additionally allowing real-world delivery of interventions [17]. The availability of active and passive data collected using smartphone-based EMI provides clinicians with information about service users that is not possible in traditional weekly in-person therapy sessions, such as in-the-moment mood states across everyday life contexts; these data can then be used to identify patterns in mood states and their fluctuations in relation to the context (eg, where the person is, who they are with, and what they are doing) and inform more personalized clinical interventions [17-19].

Of note, although standalone EMIs are associated with reduced psychiatric symptoms compared with those not receiving EMI or other forms of psychotherapy [20], the presence of human contact and support has consistently been observed to yield larger and more robust effect sizes across a range of mental health conditions [3,21], thereby supporting the potential of blended mental health care in improving outcomes.

There is now evidence to support the acceptability and effectiveness of digital mental health interventions in not only high-income but also low- and middle-income countries [22], as well as in those previously thought to be less digitally literate, such as older adults [23]. However, several challenges remain. First, although barriers related to the use of MHapps among users have been explored (eg, limited customizability and technical issues [24,25]), little work has been done to pinpoint the barriers to implementation and its facilitators, particularly with consideration of the wider organizational and societal contexts, as well as the possible strategies to facilitate implementation [26]. The involvement of multiple stakeholders throughout the development and design process of digital interventions can help overcome challenges to their uptake in real-world settings.

Further, most existing MHapps have been either developed for use by specific research or clinical teams [27], for profit, or lack empirical evidence and a theoretical basis [28]. The development of a new MHapp is often not practical, not only in many low- and middle-income countries but also in teams with lower resources and in time-limited settings (eg, context of war, social unrest, and pandemic outbreaks). Free-of-charge and tailorable MHapps or platforms that can be easily adapted are crucial for improving access to personalized health care and contributing to reduced health disparities.

To facilitate advances in the field of digital interventions for mental health, the replicability of both the development and design process is crucial. This paper aims to describe the process of developing a theory-driven, person-centered smartphone-based EMI for use as part of a blended CBT intervention for depression (blended CBT-EMI for depression). Three major phases of the design procedure are described: (1) a review of existing MHapps and “no-code” app development tools for potential adaptation and customization for specific research and clinical purposes, (2) an iterative cocreation process with stakeholders via design thinking workshops, and (3) usability testing. Given that the majority of existing MHapps have been designed for young people, this study involved 2 core groups of older adults as stakeholders in the design process (peer supporters and previous service users), serving as examples for future replications.

Before involving stakeholders in the smartphone-based EMI development process, we conducted an in-depth review of existing digital MHapps to determine whether any platform was readily available for adaptation across clinical populations and settings.

Two reviews of MHapps have been published in the past 10 years (one providing recommendations for future developments of MHapps [16], and another a recent meta-analysis on the efficacy of existing MHapps for depression and anxiety [28]), although neither examined their availability for public use nor their degrees of adaptability. We therefore hand searched all MHapps cited in these 2 reviews on the iOS App Store (Apple Inc) and Android Google Play Store (Google LLC), from which the following information was extracted (as of May 6, 2025): name of the MHapp, availability for download and use, whether it could be adapted by individual researchers or clinicians, key functions, developer, date of last update, and whether clinical trials have been conducted to evaluate its efficacy.

In recent years, the gradual rise of “no-code” app development tools [29,30], which are platforms that allow researchers with no previous software programming experience to build apps via a graphical user interface, has reduced the time and cost required for development. We also reviewed their potential suitability for use compared with readily available MHapps.

After identifying a possible platform for the EMI app, a series of cocreation workshops and usability testing were conducted to ensure the design of the CBT-EMI was person-centered and to minimize implementation barriers in practice.

The IDEAS (Integrate, Design, Assess, and Share) framework [31] was adopted to guide this process. Consolidating strategies from behavioral theory, design thinking, and user-centered design, the IDEAS framework was intended to provide researchers and clinicians with a more systematic, step-by-step approach to developing effective digital interventions that target health behavior change. A total of 10 steps are specified, which can be broadly categorized under 4 stages: Integrate, Design, Assess, and Share ([31] refer to Figure 1).

Similar to a previous study on the development of a mobile app for hypertension prevention [32], this study focuses on the first 2 stages (Integrate and Design) of the IDEAS framework, with preliminary findings evaluated using a minimum viable product for the third stage (Assess). Further evaluation of its feasibility and efficacy will be conducted in the next phase and is beyond the scope of this study.

After Session 1, participants were sent a preferred number of prompts per day (as discussed during the workshop) over the subsequent 6 days to test several CBT-based homework assignments in their everyday lives. During Sessions 2 and 3, discussions were then held regarding their experiences with enacting the homework activities over the past week (eg, what they liked about the MHapp design, and what barriers they faced).

The “think-aloud” method was used to obtain participants’ immediate thoughts and feedback on different features of the platform and specifically the following: (1) sampling scheme (including the optimal frequency, length, and schedule of homework activities), (2) preferred wordings and notification settings, and (3) other features considered essential to promote MHapp use (eg, expiration windows and conditional logics). A brief questionnaire was also administered alongside the MHapp demonstration, in which participants rated their preferences for adopting certain features in the final MHapp on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree).

Between Sessions 2 and 3, another round of real-world usability testing of the prototype (refined based on participants’ comments) took place with another set of CBT-EMI homework. During Session 3, final discussions were made to consolidate participants’ preferred features of the MHapp (including logistics), as well as its interface. A final questionnaire, which included 12 statements similar to those in previous work [18], was given to participants at the end of the workshop to capture their thoughts on the functionality of the MHapp and the overall arrangements of the workshops. Each statement was rated on a 7-point Likert scale, from 1 (not at all) to 7 (very).

All findings gathered from the co-design workshops were synthesized to determine core features to be prioritized in the CBT-EMI platform. Instant data analysis [49]—which is a time-saving approach commonly adopted for usability testing—was performed, wherein 3 researchers (SMYW, MHCI, and JKQL) discussed their observations and field notes after each workshop to identify problems reported by the participants, review action plans, and adjust the prototype of the MHapp for subsequent testing. For any inconsistent suggestions identified from the 2 rounds of workshops held with peer supporters and service users, respectively, the needs raised by the latter group were prioritized. Additional comments were also sought from clinical practitioners and mental health researchers to ensure that the final MHapp design was clinically relevant and feasible in terms of implementation and practice. Descriptive statistics were generated for the quantitative data gathered from the questionnaires on preferences for adopting certain MHapp features, as well as perceived functionality of the overall MHapp and the workshops, using SPSS (version 29.0.1.0; IBM).

Written informed consent was obtained from all participants. Ethics approval was granted by the Human Research Ethics Committee of the University of Hong Kong (EA240144). Each participant received US $38 (HK $300) after their participation in the 3 co-design workshop sessions to acknowledge their contribution to the co-design process.

A prototype of the CBT-EMI, incorporating the various CBT-based homework activities, was developed using the m-Path platform (refer to Figures 2–3). The overall flow and user-friendliness of the platform were first iteratively tested within the research team before discussions with the 2 stakeholder groups.

To ensure all participants sufficiently understood how the CBT-EMI platform operated, all homework activities were tested in a separate group of 5 older adults before the process of iterative testing by participants between sessions, with detailed explanation of different features of m-Path in Session 2 (in line with other components of the TORCH principle, namely “on-site technical support,” “rehearsal,” and “connection with group members” [40]). Hardcopy notes were not provided in this series of workshops since all details were discussed together in the small-group setting, although they are intended to be incorporated into the implementation of CBT-EMI. To ensure equity in access to the CBT-EMI and the co-design process, smartphones were available for those whose devices were not functioning (ie, “technology provision” component of the TORCH principle [40]), although all participants used their personal smartphones throughout the 3 workshop sessions, suggesting their ubiquity even in the older adult population today.

Following the Empathy Map activity, the set of needs identified by both peer supporters and service users was respectively mapped and positioned on an Eisenhower Matrix to determine their relative urgency and importance for adoption (Figure 4). The mapping of these needs on this 2×2 matrix enabled a clear visualization of how some needs were considered urgent or important by both stakeholder groups, while some were considered only urgent or important by peer supporters and not service users, and vice versa. Examples of how the MHapp could integrate features of existing services include the use of the PHQ-9 for screening and assessing depressive symptoms, as well as the use of the Wellness Recovery Action Plan to facilitate service users in developing personalized wellness plans. Users had diverse starting points, including differences in digital literacy, mental health literacy, and mental health needs.

All needs positioned in the “urgent-and-important” quadrant were adopted and implemented in the MHapp in this study. The only exceptions were those that were raised specifically by peer supporters but considered neither urgent nor important by service users.

For instance, ensuring a “user-friendly operation of the MHapp,” together with the “use of careful and friendly word choices,” was considered both urgent and important by both stakeholder groups and were principles we followed in designing the EMI platform. This was supplemented by the suggestion of providing concise instructions, offering assistance and support on the MHapp, alongside providing guidance and encouragement to participants from service users’ perspectives. Nevertheless, whereas peer supporters considered the term “homework” to be discouraging, this was not seen as a problem by service users. Similarly, incorporating more lively messages (such as adding short videos) and features that allow interactions among users to encourage MHapp use were considered helpful by peer supporters but not service users. Since these were also features that demanded relatively more technical support, we considered these as “good-to-have” features but that could be given a lower priority at the current stage of MHapp development.

For needs identified to be urgent-and-unimportant and not–urgent-and-important, we considered their feasibility and discussed strategies to make their implementation possible. As an example, “recognizing different starting points across users” (considered urgent by service users) could be addressed by ensuring sufficient guidance and support are given to participants during the CBT-EMI, both in terms of technical and mental health support, together with an emphasis on person-centered care throughout the intervention. Further, while developing a “billboard chart” to motivate homework completion may not be fully feasible at the current stage (considered important by peer supporters), one means of achieving this aim was through providing virtual “awards” as an incentive. With the “award” function on m-Path, participants can receive virtual badges after reaching preset goals (eg, completing all beeps in a day, completing 10 surveys in total, and so on). Users can view all awards they have accumulated on their personal “award page” (refer to Figure 5).

“✓” represents that the MHapp has been adapted based on specific user feedback and needs, “x” represents that no adaptation was made after reconsideration of the feedback, “n/a” represents that adaptations could not be made by individual researchers or clinicians (eg, changes only possible by the developers, namely the m-Path team in this context), and “/” represents that the feedback could not yet be incorporated at the current stage due to manpower and resource considerations but may be considered at a later stage and by other research teams.

Table 3 presents the needs and feedback generated by peer supporters and service users during Sessions 2 and 3 of the co-design workshops and whether adaptations were made accordingly. Details of how the CBT-EMI was adapted are provided in Table S3 in Multimedia Appendix 1. A total of 40 needs were summarized and categorized into the following: sampling scheme (n=6), question types (n=6), app design and UX (n=14), homework-specific feedback (n=8), and other feedback beyond the app (eg, “Incorporate additional human support” [n=6]).

Of these, 24 were from peer supporters, 7 were from service users, and 9 were from both stakeholder groups. The majority of these needs could be directly adapted for incorporation into the CBT-EMI (62.5%, n=25), and 17.5% (n=7) could not be directly adapted by individual researchers or clinicians on the platform (ie, requires amendments by the developer). A small proportion (10%, n=4) were not adapted, predominantly due to clinical and theoretical reasons (eg, since the intention of applying EMI to CBT was to encourage the enactment of CBT-based homework across settings and mood states in the real world, the prompting of 2 EMI beeps per day was retained rather than adjusting the goal to completing 12 beeps in total before the next session). For all feedback that was not adopted, we ensured the feature was acceptable to service users in the second round of workshops. An additional 4 needs identified were considered important for future blended care developments but have yet to be incorporated into the present CBT-EMI at this stage due to manpower and resource considerations (Table 3). Based on these needs and feedback, the MHapp was refined, with examples shown in Figures 6-8 (eg, “adopt briefer question types,” “make the questions and prompts more straightforward,” and “use figures and icons that are more age-inclusive and appear less old). For instance, in the Mood Check exercise (Figure 6), illustrations were replaced with more age-inclusive and youthful designs based on participants’ feedback, with additional details provided to denote what the lowest and highest values on the slider represent. In the Hot Cross Bun exercise (Figure 7), the triggering event item was amended from a multiple-choice to an open-text response, while the emotions item was replaced by the multismiley question. The amendments made to the interface were also in line with previous UX design guidelines [42,43].

Based on the ratings given by both peer supporters and service users, the enabling of multiple answers was generally preferred in multiple choice questions over a single answer (Table S4 in Multimedia Appendix 1). Other features of the MHapp platform were generally well accepted by both peer supporters and service users, although service users generally did not show a preference for the use of the multismiley question and instead preferred separating the various mood states into separate questions in the Mood Check. This question type was therefore only adopted for the CBT-based Hot Cross Bun activity, which was recommended by service users during the workshop.

In addition to adapting specific question types, both peer supporters and service users emphasized the importance of initial demonstration and ongoing technology support. As older adults may encounter challenges when learning new technologies, we made additional effort during app deployment to facilitate this process (eg, providing a step-by-step guide for downloading and setting up accounts, a brief walk-through of MHapp’s built-in security policies, and an ongoing contact point for troubleshooting and support).

In terms of the overall acceptability and feasibility of the MHapp for CBT-EMI, both peer supporters and service users found the m-Path platform to be easy to control and use, and both the texts and instructions were clear (Table 4). Possibly because service users had previous experience in engaging in CBT or other intervention groups, their ratings were slightly lower than those of peer supporters on items related to whether the EMI questions were difficult or unclear (mean 1.40, SD 0.55 vs mean 2.88, SD 1.55), disturbance related to the number of beeps per day (mean 1.20, SD 0.45 vs mean 2.38, SD 1.51), as well as the time needed to answer the questions per beep (mean 1.40, SD 0.55 vs mean 2.88, SD 1.64). Overall, both groups of stakeholders were supportive of the integration of the m-Path platform into routine therapeutic groups to improve the application of skills learned in everyday life and also reported enjoying the co-design workshops (Table 4).

The integration of digital-assisted support, such as smartphone apps, into traditional in-person psychological interventions is a promising direction toward improving mental health outcomes. Yet, the challenges posed by the development of new MHapps for specific populations, including not only the high costs involved but also the need for specialized manpower and time, make the implementation and scaling up of blended care with smartphone-based technologies difficult. A potential option is to use existing MHapps; yet, our review showed that none of the 149 possible apps supported the full content adaptation necessary for localized, culturally responsive interventions. This observation highlights a significant gap in the market and underscores the need for highly adaptable and customizable platforms.

Leveraging existing no-code app development platforms, our study demonstrated the feasibility of integrating smartphone-based EMI with traditional CBT via a phased, methodologically rigorous approach. Specifically, the final platform incorporated core CBT-based activities, such as “Mood Check” and “Gratitude Diary,” which are aligned with established mechanisms of change (eg, behavioral activation and activity scheduling [44]) and the HAPA [52,53]. Evidence-based BCTs targeting self-efficacy were also incorporated to improve the enactment of CBT-based homework and its sustainment in everyday life contexts. By describing this iterative design and development process in detail, such culturally adapted, person-centered blended mental health care models have the potential for implementation not only for older adults in Hong Kong but for broader populations and contexts, including time- and resource-limited settings.

Using co-design activities, such as the Eisenhower Matrix to prioritize feedback, 40 user needs were identified—62.5% of which were implemented. These included simplified instructions and interface, the use of friendly language, and gamified elements (eg, virtual badges to reinforce engagement), which are also key features identified as important by users in other studies with young people [56,57] and are incorporated as part of the recommendations for MHapps in a recent review [16]. While some suggestions were deferred due to technical or clinical constraints, stakeholder evaluations rated the platform as intuitive and clinically relevant. Preferences such as separating mood-state questions (rather than combining them into a single scale) were particularly noted. Both peer supporters and service users rated the platform highly for ease of use and clarity. Importantly, the platform demonstrated strong feasibility for integration into group therapy settings to reinforce real-world application of therapeutic skills. Overall, blended CBT-EMI interventions appear promising for therapeutic engagement, homework adherence, and symptom management, particularly among older adults in our study context.

This study has important implications for researchers, clinicians, and digital health developers. A major strength lies in its systematic and theory-driven development process. The phased methodology—comprising needs assessment, platform selection, cultural and linguistic adaptation, and co-design—offers a replicable model for future interventions. Theoretical integration is another notable strength; the intervention is grounded in CBT and enriched by HAPA, aligning clinical techniques with behavioral science to support sustained change. The detailed mapping of how various BCTs were incorporated in the CBT-EMI enables a more fine-grained investigation of their independent effects on clinical outcomes in future studies, which can be conducive to efforts in the identification of “active ingredients” in psychological interventions [45].

The blended care intervention we presented further benefits from its strong cultural relevance. Stakeholder engagement was central to the adaptation process, ensuring the platform was sensitive to local needs and user expectations. By addressing user interface design, motivational features, and language preferences, we were able to develop a culturally attuned and user-friendly solution for mental health care in the real-world setting. These features are critical for maintaining long-term engagement among users. The co-design process also greatly facilitated stakeholder-driven adaptations and potentially enhanced the feasibility of implementation in routine care.

Nonetheless, we note several limitations. While the Mood Check items we adopted are similar to those commonly used in ESM and EMI among adolescents and adults, the present blended care platform was co-designed only with Cantonese-speaking older adults in Hong Kong. Additional rounds of co-design processes may be helpful to ensure its potential application to other age, cultural, or linguistic groups. In particular, we highlight the significance of considering the use of culturally sensitive mediums [58], local languages and dialects [59], as well as barrier-free design (eg, allowing audio input as an alternative to text-based options [60]). While the sample size of this study is in line with guidelines on performance-based usability testing [42] and is comparable to other studies using co-design methods in the development of health interventions [61,62], we acknowledge that perspectives beyond the 13 participants involved may not be captured (eg, those with less motivation to participate in the study). We did, nevertheless, aim to include participants with more diverse sociodemographic backgrounds and experiences in psychotherapy to capture a wider range of perspectives. Whether additional amendments may be needed to improve the usability of the platform for a majority of the population would require further exploration.

We also note that several user-identified needs were not implemented due to platform limitations, pointing to future opportunities for enhanced collaboration between researchers and technology developers. Given the lack of no-code app platforms available that were specifically designed for EMI, we selected m-Path for building the EMI for the present blended care. We acknowledge that the focus on a single vendor could have possible issues related to the long-term sustainability of the CBT-EMI. Yet, the aim of this study was to provide a transferable framework that could guide researchers or stakeholders in the co-design of a digital mental health platform, particularly in settings where the development of a new app is not feasible. The insights we gathered into the prioritization of MHapp features and user needs also offer a reference for future studies in the design of more age-friendly MHapps targeting the older adult population. Given the relatively short study period, we cannot determine whether individuals would continue using the platform beyond the intervention period. Whether individuals could successfully develop habits in applying CBT-based techniques in their everyday lives without the continued use of the platform postintervention would be an area for future investigation.

In line with the IDEAS framework [31], empirical validation through randomized controlled trials with a longitudinal follow-up would be an important next step to confirm the efficacy of the CBT-EMI in terms of symptom reduction, adherence, and long-term outcomes. In particular, it would be worthwhile to examine whether people receiving the novel blended care—as informed by both cognitive-behavioral and health behavior change theories—will show greater symptom improvements and higher rates of clinical remission and response (eg, using the PHQ-9 for evaluating DSM-based depressive symptoms) compared with those receiving regular in-person CBT without the addition of EMI. It would also be crucial to test the hypothesized mediating role of self-efficacy in these future studies to elucidate the mechanisms of change in the CBT-EMI.

Despite the abundance of MHapps developed, the observation that none could be readily adopted to target specific clinical or therapeutic needs across settings may substantially prevent the scaling up of digitally assisted mental health care across populations. By demonstrating how no-code platforms can accelerate the development of theory-based interventions and how blended CBT-EMI approaches can effectively extend therapeutic engagement beyond clinical settings, this work presents a replicable roadmap for developing scalable, person-centered digital mental health interventions. It would be essential for future work to test the efficacy of the blended care intervention in large-scale randomized controlled trial studies.