This study used a descriptive qualitative research design, grounded in the philosophical tenets of naturalistic inquiry [19]. This qualitative approach uses everyday language and low-inference interpretative processes to represent reality by describing experiences or events [20]. It is suitable for informing diverse practical applications, such as clinical interventions, scoring systems, needs assessments, questionnaire development, and surveys [19]. Therefore, this method was well-suited to our aim of gaining rich, contextualized insights into the requirements and preferences of patients with DM for mHealth app features.

We collected data using semi-structured, in-depth interviews. The interview guide was developed based on prior experience and a review of relevant literature, aligned with the study’s objectives. Before the formal interviews, 2 patients with DM were pre-interviewed, and the interview outline was adjusted according to the pre-interview analysis results. The formal interview outline is presented in Textbox 1. The study outcomes were reported according to the Consolidated Criteria for Reporting Qualitative Research [21].

A purposive sampling approach was used to select participants diagnosed with DM for the qualitative interviews. Eligible patients were recruited from the Hong Kong District Health Centre in May 2024. Recruitment posters were displayed at the center to solicit participant involvement. A research assistant established initial phone contact with eligible patients, introducing the study, explaining the consent procedures, and informing them about the duration of the in-depth interviews. Patients willing to participate were scheduled for in-person interviews conducted on-site at the center. All eligible participants who initially consented completed the scheduled interview sessions without dropouts or refusals.

The inclusion criteria were as follows: (1) diagnosed with DM according to the 2019 WHO diagnostic criteria; (2) aged 18‐80 years; (3) no cognitive impairment and able to communicate effectively with health care professionals; (4) signed informed consent and agreed to participate in the study. The exclusion criteria were as follows: (1) patients with severe complications or significant dysfunction of other major organs; (2) patients with incomplete data or who withdrew from the study.

The sample size was determined based on the principle of information saturation. After the 8th interview, no new themes emerged, and 2 additional interviews were conducted to confirm saturation, ultimately resulting in a total of 10 participants. To protect patient privacy, these individuals have been labeled as P1 through P10. Demographic information was collected for each participant before initiating an interview session and is reported in Table 1.

Face-to-face interviews, each lasting 60 minutes, were conducted in an independent activity room at the Center, free from external distractions. The research team, led by a principal investigator and supported by 2 research assistants, conducted all interview sessions and collected field notes throughout the process. The team had undergone comprehensive training in qualitative research methods, equipping them with the necessary skills and expertise for this study.

Each interview commenced with a brief introduction. Interviews were recorded with participant consent and guided by a pre-tested, semi-structured protocol. No other individuals were present during the interviews. Participants were not involved in the design, conduct, or reporting of the dissemination of this research. All interviews were conducted in a single session, with no repeat interviews.

Prior to the interviews, participants were informed of the study’s purpose and assured of privacy protections. The main topics included factors influencing effective DM management and perceived useful features of mHealth apps for improving self-care. During the interviews, researchers actively listened, used appropriate questioning techniques, and promptly recorded key information. They also observed and documented participants’ nonverbal cues, such as pauses, smiles, body language, and mood changes. After each session, a reflective diary was written to identify and correct any issues for the next interview.

All interviews were audio-recorded and transcribed verbatim. Inductive content analysis was applied to examine the data at both the manifest and latent levels, following the methodology outlined by Elo and Kyngäs [22], which involved open coding, category development, and abstraction. This analytical process was completed using NVivo (version 15.0; QSR International) software. Initially, 2 researchers (NY and WAKC) iteratively reviewed the transcribed text. The analysis began with open coding, where the researchers generated descriptive notes directly in the margins, which were then transferred to a coding sheet to generate subcategories. Following the open coding, the lists of resulting subcategories were grouped into main categories based on both differences and similarities. This analytical process involved ongoing, iterative discussions with the research team to refine the codes and categories, aiming to enhance the overall trustworthiness of the study’s findings. To further ensure credibility, 2 researchers independently coded the data, followed by peer debriefing meetings with the research team to refine the coding framework. An audit trail documenting analytical decisions was maintained to promote dependability. Transferability was supported by detailed reporting of participant demographics and rich contextual descriptions. Quality assurance was incorporated through methodological rigor, including researcher training, data triangulation, and audit trail maintenance.

Data from 7 principal codes and 12 subcodes were assembled into the interest domain (categorization). Data from the remaining codes were considered not relevant to the research aim and were therefore treated as dross (Table 2). The informants were not invited to comment on the transcriptions or provide feedback on the findings. Although participants discussed the functions they valued in an mHealth app, formal prioritization techniques, such as structured ranking exercises, were not used. Instead, relative priorities were inferred based on the frequency and emphasis with which different functions were mentioned during the interviews.

This study was approved by the Institutional Review Board of a university in Hong Kong (HSEARS20230705003). All participants provided written informed consent after receiving detailed information about the study’s purpose, procedures, and their right to withdraw at any time without penalty. The present study was conducted according to the principles indicated in the Declaration of Helsinki. These procedures were designed in accordance with institutional data protection policies and ethical standards for qualitative health research. During the interviews, participants were informed that they could refuse to answer any questions they did not wish to answer. To protect their privacy and confidentiality, audio recordings were stored securely, and any identifying information was removed from transcripts. The interview data were anonymized through the coding process to further ensure participant privacy.

Among the 10 interviewees, ages ranged from 60 to 75. The majority were retired males with at least a high school education background, and there was only one female participant in the semi-structured interviews.

Descriptive qualitative research was conducted with patients with DM in Hong Kong to understand the value and usefulness of various mHealth app features that may improve their self-management. Participants identified key features such as accurate information resources, automatic tracking and monitoring of health metrics, reminders, and personalized customization options. They also emphasized the importance of intuitive usability, efficient data-sharing capabilities, and an interactive design.

Participants highlighted the significance of providing clinically validated, comprehensive DM-related content within mHealth apps. They envisioned the app as a centralized, authoritative resource covering key self-care aspects, such as disease knowledge, prevention of comorbidities, exercise recommendations, and medication management. Beyond informational resources, participants emphasized the need for decision support to guide users in making informed choices about diet, fitness, and other lifestyle factors. This aligns with the broader challenge faced by many patients with DM who often lack access to professional, disease-specific knowledge, as existing web-based resources tend to be fragmented and unreliable [23]. This research emphasizes drawing upon evidence-based guidelines, current literature, focus groups, and interviews to develop a robust, validated knowledge base [24]. The involvement of users, developers, and clinical experts is considered essential during the app development stage, as this multilateral, user-centric approach is crucial for empowering patients to optimize their self-care regimens [25].

Participants also expressed a desire for the app to automatically track and monitor health metrics, with the ability to recognize and convert images into corresponding data for storage. Regular tracking of health is often regarded as a burden by patients with chronic diseases [26]. Knowing one’s history of physical activity and other health-related data through a tracking function was seen as beneficial and motivating [12]. This aligns with prior research showing that over 70% of health apps are designed to support healthy eating, with features like carbohydrate counting and diet tracking [27]. Automatically recognizing images and converting them into data that can be uploaded eliminates the need for manual tracking, which is a significant deterrent to the uptake of DM apps [16].

Participants claimed that reminder functions helped them be more aware of and accountable for their DM self-management activities, consistent with findings from prior studies [28-30]. Reminders and alerts can improve safety by warning of hypoglycemia or hyperglycemia and enhance self-management outcomes by improving medication adherence, physical activity, and clinical appointments [9]. However, current reminders are often ignored due to poor design and frequent pop-ups [13]. Regular self-care behaviors are crucial for effective DM management, and a recent study found that nearly a quarter of individuals with DM report frequently forgetting to take their DM medication [31]. Including a customizable and attention-grabbing reminder function in a self-management app is therefore vital.

Consistent with prior research [10], this study found that formulating corresponding personalized management strategies based on the user’s health assessment data is significant for a self-management app. The app’s ability to dynamically update its content and functionality based on the user’s evolving needs is a critical differentiator. For example, Duan et al [10] interviewed pregnant women with at least one risk factor for gestational DM who hoped that solutions tailored to their specific characteristics could encourage them to maintain healthy living habits during app use. Similarly, a qualitative study on Saudi women found that obese individuals were restricted from performing enough physical exercise outdoors [32]. The app should recommend personalized exercise routines aligned with the user’s conditions and preferences. Personalized features in health apps can improve user compliance with lifestyle interventions, one of the most common strategies in mHealth apps [33]. The app should leverage artificial intelligence (AI) algorithms to provide personalized information and recommendations, tailoring content, reminders, and feedback based on user preferences, blood glucose control, and self-management knowledge. Participants expect personalized feedback based on their tracked data and the ability to customize reminder frequency and timing based on their needs. Excessive reminders were deemed annoying [12].

Most participants emphasized that good usability design is crucial for the adoption and continued usage of mHealth apps, especially for older adults with DM and patients with lower overall literacy, whose reduced physical or cognitive capabilities can be obstacles when effectively using apps [13,34]. This includes using plain, simple language with explanations, providing multi-language options, and presenting information in visual forms such as illustrations and videos. These design elements are key for attaining health literacy, as suggested by the German guideline for evidence-based health information [35]. Previous studies have pointed out that multimedia content can better adapt to the learning styles of different individuals, enable patients to take a more active role in learning, and encourage them to study more actively [36]. Research has found that information retained from watching videos is stronger than that acquired from browsing text or static images alone [37]. Therefore, when designing apps, developers should use visual methods to present a variety of health information and educational resources for patients with DM to meet their needs for information support.

This study also underscores the need for cultural adaptation when developing DM mHealth apps for local patient populations. Tailoring the app’s content and features to align with users’ regional or ethnic cultural, lifestyle, and dietary habits can enhance satisfaction and engagement [38,39]. For instance, the app could incorporate region-specific calorie calculation functions, dialect support in voice recognition, and integration of traditional Chinese medicine principles such as syndrome differentiation and seasonal wellness [40]. Future application development should consider co-design approaches involving target users, inclusion of multilingual interfaces, and alignment of content with culturally accepted dietary and lifestyle practices to improve engagement and usability.

This literature demonstrates that in-app peer support and professional health coaching are appealing features for patients, aligning with previous findings that interactive functions are the most popular in mobile health apps [41,42]. Some studies have used telephone or messaging groups to support patients with DM [43,44]. However, these approaches often face challenges in providing timely feedback between providers and patients, and the information shared can be easily forgotten. Therefore, designing an interactive DM management mHealth app is crucial.

Such interactive apps can formulate targeted behavior change plans for users, continuously monitor their behavioral patterns, and provide personalized feedback, effectively motivating users to complete desired self-management behaviors [45,42,46]. The literature indicates that 2-way mHealth communication between health care providers and patients may enhance glucose monitoring, patient adherence, and clinical outcomes, such as reduced HbA_1c levels [47]. Additionally, peer support is crucial for developing and sustaining patients’ self-management skills, as it allows for sharing experiences, attitudes, and coping strategies, increasing their ability to manage the stress of self-care [48,49]. Collectively, participants with DM recommended including both provider-patient communication and peer support in apps, which is worth considering by app developers [46].

To maintain innovation and competitiveness in the evolving mHealth landscape, apps informed by this study should incorporate adaptive features that reflect both user needs and technological advancements. Integrating AI-driven personalization, culturally tailored content, and co-designed interfaces with patients can differentiate such tools from generic solutions [50]. Moreover, building in mechanisms for real-time user feedback, continuous performance analytics, and agile updates will enable ongoing refinement in response to emerging trends. Collaborations with interdisciplinary teams—including health care professionals, technologists, and patient advocates—will be essential to ensure sustained relevance and market adaptability.

While AI-powered mHealth apps hold promise for delivering personalized diabetes care, their implementation presents several challenges that warrant consideration. Key concerns of the participants include algorithmic transparency, clinical validity, and data privacy [42]. To ensure safety and reliability, AI tools must undergo rigorous testing against established clinical benchmarks, with transparent reporting of algorithm design and performance metrics. Regular auditing and bias detection mechanisms should also be incorporated to maintain fairness and accuracy across diverse patient populations. In addition, privacy-preserving techniques—such as federated learning and differential privacy—may be adopted to protect patient data while enabling the development of adaptive, intelligent systems. These considerations are critical for future implementation and warrant further investigation through interdisciplinary research that integrates technical feasibility, ethical safeguards, and patient trust.

Previous research has primarily concentrated on users’ experiences and evaluations of developed apps after they have been used, with limited user involvement in the design process. This gap can lead to suboptimal user retention. In this study, we conducted in-depth interviews to investigate the preferences and functional needs of mHealth apps among individuals with DM. The insights gained are crucial for informing the development of future interventions and apps.

Additionally, we conducted thorough interviews with patients with DM in Hong Kong, where there is currently a lack of relevant research. Recognizing the importance of cultural adaptation for the success of mHealth apps, our findings can contribute to enhancing engagement among Hong Kong patients, thereby improving self-management levels and reducing health care costs. Furthermore, we emphasized the importance of using AI to provide personalized, timely, and culturally relevant information and feedback, which can also alleviate the workload of health care professionals. Ensuring the credibility and professionalism of AI information sources is essential to avoid misunderstandings that could delay treatment.

This study has several limitations. First, this study is limited by its relatively small sample size and recruitment from a single community health center, which may affect the generalizability and diversity of findings. Patients with different ages, disease durations, severity levels, and lifestyles may have distinct needs for mHealth apps, which were not fully captured. In addition, perspectives from health care professionals, such as diabetologists and diabetes educators, were not included, representing an important avenue for future research. Future studies should aim to recruit larger and more diverse samples from both community and hospital settings. Second, we were unable to capture data on how self-management experiences and perceptions of mHealth might differ across various identities (such as gender, age, and ethnicity), which could have provided richer insights into these subgroups. Third, while patients’ preferences for mHealth app features were explored qualitatively, the study did not adopt formal prioritization methods. Future studies should consider using structured prioritization techniques to systematically capture and rank patient preferences across diverse populations.

This study did not evaluate the actual behavioral changes or clinical health outcomes following mHealth app use, which limits our ability to assess the real-world impact of the proposed features. As the design phase preceded implementation, outcomes such as glycemic control, treatment adherence, and lifestyle modifications were not measured. This limits the ability to draw conclusions about the long-term effectiveness of such apps in supporting self-management.

Our findings highlight the importance of tailoring these apps to local cultures, particularly through the integration of local food options, recommendations, and language preferences. Key features desired to enhance self-care behaviors include credible resources, automatic tracking and monitoring of health metrics, reminders, personalized customization, user-friendliness, efficient data-sharing capabilities, and in-app support via virtual interactions with peers and health care professionals. The results of this study provide valuable insights for enhancing the self-management capabilities of patients with DM and for the future development of mHealth apps aimed at DM prevention.