Guided by self-management theory, this study utilizes the HBM and BCW to select variables representing the self-management characteristics of mHealth users. Data were collected through a questionnaire survey, and unsupervised clustering analysis was used to identify user clusters. Subsequently, the characteristic distributions of different user clusters were analyzed to construct user profiles. Through intergroup difference analysis, this study compared satisfaction levels across user types with respect to various mHealth functional designs and examined the correlations between user characteristics and satisfaction with functional designs. The overall workflow of the study is illustrated in Figure 1.

mHealth mainly improves health outcomes by empowering self-management abilities [28]. Therefore, this study selects variables representing user characteristics guided by self-management theory. In individual and family self-management theory [29], self-management refers to the process by which individuals and families use knowledge and beliefs, self-regulation skills and abilities, and social facilitation strategies to achieve health-related outcomes in their real-life environments. Among these components, knowledge and beliefs refer to information and perceptions related to a health condition or health behavior. The HBM is one of the most widely used models for explaining beliefs and perceptions underlying health behaviors [30]. HBM posits that individuals’ perceptions of their susceptibility to a health issue and the potential severity of its consequences, together with perceived benefits of adhering to health recommendations and their intention and ability to engage in health management, are the most direct drivers of health behaviors [31]. HBM operationalizes these factors through perceived susceptibility, perceived severity, perceived benefits, perceived barriers, and self-efficacy [32]. Accordingly, this study adopts these 5 variables from HBM to characterize the knowledge and beliefs component of self-management theory.

Self-regulation within self-management theory is an iterative process in which individuals utilize a variety of skills and abilities to achieve healthy behavior change [29]. The BCW, a classic theoretical model for explaining behavior change—particularly in the contexts of health promotion and disease prevention [33]—provides an important reference for selecting self-regulation–related characteristic variables in this study. The BCW posits that healthy behavior change can be achieved only when individuals possess capability, motivation, and opportunity [34]. Capability refers to the knowledge and skills required to engage in a particular behavior [35]. In the context of using mHealth for health management, essential knowledge and skills are reflected in 2 aspects: health management and the use of digital technology. eHealth literacy is among the most commonly used indicators for assessing an individual’s ability to acquire health management information and utilize electronic health information to address health issues [36,37]. Accordingly, this study selects eHealth literacy to characterize self-regulation skills and abilities. In addition, motivation in the BCW refers to the beliefs that trigger behavioral change, encompassing reflective motivation and intrinsic motivation [34]. Motivation for health management can be divided into proactive health management intention and passive health management intention triggered by perceived health status, corresponding to reflective motivation and intrinsic motivation, respectively. Therefore, this study selects health management intention [38] and perceived health status [39] to further characterize the motivational dimension of self-management behavior. In summary, this study selects a total of 8 variables to analyze self-management characteristics among mHealth users. The definitions of each variable are presented in Table 1.

The main purpose of this study is to delineate individuals’ health management characteristics and classify users based on their levels of health management characteristics, and to further analyze differences in satisfaction with various mHealth app functional designs across user categories. Therefore, it is necessary to clarify the main functional designs of mHealth apps. Previous studies have summarized common mHealth app functional designs through systematic reviews and content analyses [42-44], categorizing functions according to their primary implementation purposes, such as education, tracking, social interaction, gamification, and motivational features [42]. Building on prior research, this study reorganizes the functional characteristics reported in the literature to propose a classification of mHealth app functional designs for analysis (see Table 2).

To achieve the aforementioned research objectives, this study selected appropriate objects of study for empirical analysis. First, the comparability of different mHealth app functional designs was considered. To minimize the influence of differences in core service content across app types on comparability, this study selected a homogeneous category of apps as the empirical focus, thereby reducing potential confounding effects. Second, the application domains of different mHealth app categories were considered. Exercise, fitness, and weight management align with substantial health needs. These domains represent important modalities for primary prevention, and exercise therapy also serves as a key approach for tertiary prevention (disease management) in various conditions, such as cardiac rehabilitation, pulmonary rehabilitation, tumor rehabilitation, and metabolic disease management [49]. Given that exercise, fitness, and weight management apps account for more than half of all mHealth apps [50], and that these functions represent 48.37% of users’ motivations for using mHealth apps [51], we selected exercise, fitness, and weight management apps as the units of analysis to represent a broader and more heterogeneous user population.

This study did not restrict respondents to users of a specific app for the following reasons. First, as the aim of this study is to examine the satisfaction of different user types with various mHealth functional designs in real-world contexts and to provide generalizable recommendations, the focus is on mHealth functional modules rather than on specific app design implementations. Accordingly, the research sample was not limited to users of a single app. Moreover, this questionnaire-based data collection approach captures multiple implementations of the same functional module (eg, health education), thereby enhancing sample diversity and improving the generalizability of the findings [52,53]. Second, this study was conducted in real-world settings, which differ from laboratory-based studies that examine factors influencing mHealth satisfaction. In laboratory environments, variables must be tightly controlled to quantify the idealized effects of specific factors, often requiring the use of a single app as the research object. By contrast, this study emphasizes user satisfaction with mHealth in real-world contexts, where the use of multiple similar apps by the same user is common. Analyzing such data provides real-world evidence that can inform future research and practical mHealth design.

Based on the constructed User Health Management Characteristics Framework, this study measures 8 independent variables and 5 dependent variables reflecting satisfaction with mHealth functionalities. To achieve the aforementioned research objectives, it was necessary to select appropriate and measurable items corresponding to the characteristics defined within the framework. To ensure the rational design of the measurement scale, this study undertook a careful process of adapting previously validated scales, making nuanced adjustments to align them with the specific research objectives.

Specifically, the User Health Management Characteristics Framework includes 4 variables derived from the HBM (perceived severity, perceived susceptibility, perceived barriers, and perceived benefits), with measurement items adapted from HBM-related studies by Ahadzadeh et al [54], Saghafi-Asl et al [32], and McArthur et al [55]. Two variables were derived from the BCW (eHealth literacy and perceived health status), with measurement items drawn from BCW-related studies [39,56,57] and the Brief Health Literacy Screen scale [58]. In addition, 2 variables integrating both models (health self-efficacy and health management intention) were included, with measurement items adapted from health management–related studies, such as Zhou et al [59] and Li et al [38]. For the measurement of satisfaction with the 5 mHealth functional designs, this study first provided definitions and specific design examples of each function in the questionnaire, followed by item-based measurement. The satisfaction items were adapted from relevant studies on mHealth satisfaction [8]. All measurement items used a 5-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree).

A multistage validation and refinement process was implemented to ensure the conceptual accuracy and cultural appropriateness of the English-originated scales within the Chinese context. First, the preliminary translation and adaptation of the scale items were collaboratively conducted by research team members proficient in both Chinese and English and with expertise in health informatics. Through group discussions, each item was refined to preserve the core meaning of the original construct while aligning with Chinese linguistic norms and the research context. Second, before the formal survey, 24 participants with backgrounds in medical informatics were invited to complete a pretest of the questionnaire. Participants were asked not only to respond to the questionnaire but also to provide feedback on the clarity and contextual appropriateness of the items. Based on this feedback, the wording of the questionnaire was further refined to eliminate ambiguities and improve readability, resulting in the final version used in the formal survey. Finally, during the data analysis phase, reliability and validity tests were conducted on the measurement model, thereby verifying the effectiveness and reliability of the measurement instrument from a data-driven perspective (see Multimedia Appendix 1).

The survey targeted individuals who had used exercise, fitness, or weight loss apps within the preceding 3 months. According to the N:q hypothesis [60], the ratio of sample size to item parameters in a scale should range from 20:1 to 10:1. In this study, a total of 51 items were included for measurement; therefore, applying the more conservative ratio of 20:1, the required sample size was approximately 1020.

From June 28, 2022, to September 8, 2022, a snowball sampling survey was conducted via the WeChat platform (Tencent Holdings Limited). The questionnaire link, generated using Wenjuanxing, was distributed through Moments (Tencent Holdings Limited) and various group chats. In total, 2518 questionnaires were collected. Of these, 902 respondents reported that they had not used the target apps within the preceding 3 months. A quality control item was included in the questionnaire: “This question is an attention test to ensure careful completion of the questionnaire. For this question, please select ‘strongly disagree’.” The correct response to this item was “1. strongly disagree.” Responses that answered this item incorrectly were considered to have failed the quality control check. In addition, given the length of the questionnaire, responses completed in less than 3 minutes or with identical answers across all scale items were deemed insufficiently attentive and were also excluded. Through this screening process, 591 responses were removed. Consequently, 1025 valid questionnaires remained for subsequent analysis. Before completing the questionnaire, all participants were informed of the study’s overall purpose and detailed procedures.

In this study, item-level missing data were addressed using mode imputation based on each respondent’s answers to other items within the same variable. Specifically, there were 3 missing values in total: 1 in perceived susceptibility, 1 in eHealth literacy, and 1 in gamification. As the questionnaire employed Likert scales, which represent ordinal rather than strictly continuous data, each missing value was replaced with the mode of the remaining items within the same scale for that respondent. The imputed values were 3, 5, and 3, respectively. Given that the dataset comprised a total of 45,100 item responses, the proportion of missing values was 0.0067%, which is negligible and unlikely to affect the results.

To address the question “What are the common combinations of health management characteristics in users during the process of using mHealth?” we employed cluster analysis. Cluster analysis is an unsupervised machine learning technique that divides a set of observations into multiple clusters composed of similar objects and can identify the structural composition of different clusters. In clustering, similarity between observations within the same cluster is high, whereas similarity between observations from different clusters is low. Among various clustering methods, the k-means algorithm produces results that are easy to interpret, are suitable for clustering high-dimensional data, and demonstrate good scalability and efficiency when dealing with large datasets [61]. The clustering in this study was based on 8 health management characteristics (Table 1), and the sample size was moderate; therefore, k-means was appropriate for this study. The k-means algorithm requires the number of clusters to be predetermined. We determined the optimal number of clusters by integrating the Elbow Method and the Silhouette Coefficient. The Elbow Method identifies the inflection point by analyzing the relationship between the within-cluster sum of squared errors (SSE) and the number of clusters, while the Silhouette Coefficient is used to mathematically assess the compactness and separation of the clustering structure. The results showed that when the number of clusters was 6, an inflection point appeared in the SSE curve, and the Silhouette Coefficient reached a locally optimal value of 0.1514 (see Multimedia Appendix 2), indicating that this clustering solution maintains good within-cluster compactness while achieving favorable between-cluster separation. Therefore, we ultimately set the number of clusters to 6. To further verify the stability of this 6-cluster solution, we performed 10 repeated experiments with random initializations. The results showed a relative fluctuation rate of only 0.002% in the SSE values, demonstrating that the 6-cluster solution exhibits high repeatability and stability. We conducted the cluster analysis to determine the grouping of mHealth app users using sklearn.cluster in Python 3.8 (Python Foundation; see Multimedia Appendix 3).

To answer the question “Is there a difference in satisfaction with the functional design of mHealth apps among different types of users?” we performed rank sum tests to examine satisfaction and preferences regarding functionalities across user groups using SPSS 26 (IBM Corp). To address the question “How do health management characteristics affect users’ satisfaction and preferences for the functional design of mHealth apps?” we conducted a partial least squares analysis using SmartPLS 4 (SmartPLS GmbH). The significance level was set at P<.05 for 2-tailed tests.

The data collected for this study were analyzed to assess potential nonresponse bias. To accomplish this and examine nonresponse bias [11], the Mann-Whitney U test was used to compare responses from the first 20% (205/1025) and the last 20% (205/1025) of the sample. The results showed no significant differences between the 2 groups for the health guidance (z=−0.728, P=.47), health education (z=−1.171, P=.24), health monitoring (z=−0.857, P=.39), social function (z=−1.794, P=.07), and gamification (z=−0.260, P=.80) constructs. Therefore, nonresponse bias was not a concern in this study. This study followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) reporting standard (see Multimedia Appendix 4).

This study utilized an online questionnaire survey. On the first page of the questionnaire, the research purpose, content, potential risks, data usage, confidentiality terms, and researcher contact information were clearly explained to participants. By checking a box, participants confirmed that they had read and understood the above information and voluntarily agreed to participate in the study, thereby providing informed consent. Participants who did not agree with the above content were directed to exit the online questionnaire. The questionnaire materials for this study were destroyed after the completion of quality control and transcription, within 6 months after the survey concluded. Any associated information, such as source IP addresses, collection time stamps, and serial numbers, was also destroyed to ensure privacy protection. Following transcription, the data were renumbered sequentially, starting from 1. All images and data presented in the manuscript or multimedia appendices are fully anonymized and cannot be used to identify any individual participant or user. No financial compensation was provided to participants. Approval was obtained from the Biomedical Ethics Committee of the Second Affiliated Hospital, School of Medicine, Zhejiang University, China (approval number IRB-2021-841).

The demographic characteristics of the included sample are presented in Table 3. The sample exhibited a balanced gender distribution (female: 553/1025, 53.95%; male: 472/1025, 46.05%), with participants predominantly aged 19-30 years (500/1025, 48.78%). In terms of education, most respondents held college degrees (519/1025, 50.63%), followed by master’s degrees (269/1025, 26.24%). Income distribution showed that nearly one-quarter of participants (244/1025, 23.80%) earned ≤2500 RMB (1 RMB=US $0.14) monthly, whereas a smaller proportion (49/1025, 4.78%) reported incomes of 50,001 or higher. Regarding usage patterns, the largest group reported using the service 2-3 times per week (370/1025, 36.10%), while a minority were frequent users (>7 times per week: 59/1025, 5.76%). In terms of occupation, students constituted the largest subgroup (283/1025, 27.61%), followed by public servants/clerks (208/1025, 20.29%) and professional technical personnel (196/1025, 19.12%). Overall, 664 of 1025 (64.78%) users reported using 1 app.

A descriptive analysis of the 8 self-management characteristics and satisfaction with the 5 functional designs was conducted. First, we calculated the average score of multiple items measuring each characteristic variable to represent the level of that variable. However, because the distributions of the average values for each variable were not normally distributed, medians and IQRs were used for presentation, as shown in Table 4.

The determination of the optimal number of clusters in this study, which was set at 6, was based on the relationship between the within-cluster SSE and the number of clusters. Subsequently, cluster analysis was performed on the sample using z score–standardized values of the 8 health management indicators. The distribution across the 6 clusters was relatively uniform (140, 135, 189, 218, 180, and 163). Figure 2 shows the distribution of each indicator across the 6 clusters obtained from the analysis. A z score close to 0 indicates that the corresponding indicator in a given cluster approximates the overall average level, whereas a positive z score indicates that the indicator is above the average level in that cluster.

We further analyzed differences in each indicator across the 6 clusters using the Kruskal-Wallis test to assess the distinctiveness of each cluster. The results of the Kruskal-Wallis test demonstrated statistically significant differences in the z scores of all indicators across the 6 clusters (P<.001 for all). Specifically, significant differences were observed for the following: perceived severity, H(5)=400.947 (P<.001); perceived susceptibility, H(5)=522.425, (P<.001); perceived barrier, H(5)=673.985, (P<.001); perceived benefit, H(5)=492.630, (P<.001;) health self-efficacy, H(5)=518.298, (P<.001); perceived health status, H(5)=490.411, (P<.001); eHealth literacy, H(5)=362.910, (P<.001); and health management intention, H(5)=390.573, (P<.001). These findings indicate that the 6 clusters exhibit distinct patterns and magnitudes of health management characteristics, as reflected by the significant differences in the z scores of the respective indicators. The detailed z scores and pairwise comparison results for each indicator across the 6 clusters are presented in Table 5, allowing for a detailed examination of how each cluster differs from the others in terms of specific health management characteristics. Detailed test statistics are provided in Multimedia Appendix 5, which presents pairwise comparisons of health management characteristics among user groups.

The distribution of the 8 health management traits varied across individual users. To further understand user health management characteristics and provide references for personalized services in mHealth apps, this study categorized users based on combinations of their 8 current health management factors. Given the distinctiveness of each user group, the 6 types were named according to the indicator profiles within each cluster to reflect the main health management characteristics of users in each type. Cluster 1 was labeled “positively proactive energizers.” Users in this cluster exhibited higher perceived health status, health management self-efficacy, and health management intention than proactive intenders, along with substantially lower perceived barriers and perceived susceptibility.

Cluster 2 was labeled “proactive intenders,” characterized by users whose perceived health status, eHealth literacy, and health management intention are all above average, but who also report higher perceived barriers and perceived susceptibility. Proactive intenders (135/1025, 13.17%), the smallest group, represent users who demonstrate strong proactivity in health management, recognize their potential risk for health problems, and actively manage their health despite the presence of perceived barriers.

Cluster 3 was labeled “negatively proactive energizers.” Compared with positively proactive energizers, users in this cluster have slightly below-average perceived health status and higher perceived susceptibility, yet they exhibit higher health management intention, health management self-efficacy, and eHealth literacy. Both positively proactive energizers (140/1025, 13.66%) and negatively proactive energizers (189/1025, 18.44%) represent groups with high levels of health management cognition, literacy, and capability, along with strong health management intention, making them the primary beneficiaries of current mHealth apps. The key difference between these 2 groups lies in perceived health status, which is higher among positively proactive energizers and below average among negatively proactive energizers. Cluster 4 was labeled “low health management demanders.” Users in this cluster tend to have average perceived health status, with perceived susceptibility, perceived severity, and health management intention all below the average level. Low health management demanders constitute the largest proportion of users (218/1025, 21.27%) and represent a group with relatively good current health, limited emphasis on health management, insufficient awareness of health risks and potential harm, and consequently lower intention to engage in health management.

Cluster 5 was labeled “potential health management demanders.” Users in this cluster have below-average perceived health status and, although their perceived benefits, perceived susceptibility, and perceived severity are above average, they exhibit lower health management intention and higher perceived barriers. Potential health management demanders (180/1025, 17.56%) represent individuals with poor current health status who are aware of the harms associated with health risks and recognize the benefits of health management for improving their condition, yet face substantial barriers to engaging in health management and therefore have a lower intention to do so. This group has both a clear recognition of the importance of health management and an objective need for it due to their poor health status.

Cluster 6 was labeled “passive attitude holders,” comprising users whose perceived health status is below average and who report very low perceived benefits, health management self-efficacy, perceived severity, and health management intention. Passive attitude holders (163/1025, 15.90%) share similarities with low health management demanders, including low perceived benefits, health management self-efficacy, perceived severity, and intention for health management. However, they differ in that passive attitude holders perceive their health status as below average and face higher perceived barriers. This group perceives a decline in health status but lacks the willingness to improve it, placing them at high risk for illness.

From the perspective of age distribution, the 6 user categories exhibited distinct age-structure characteristics. Positively proactive energizers (n=140) comprised 48 (34.3%) aged 19-30 years, 48 (34.3%) aged 31-45 years, 42 (30.0%) aged 46-60 years, and 2 (1.4%) aged ≥61. Proactive intenders (n=135) consisted of 71 (52.6%) aged 19-30 years, 50 (37.0%) aged 31-45 years, 12 (8.9%) aged 46-60 years, and 2 (1.5%) aged ≥61. Negatively proactive energizers (n=189) included 2 (1.1%) aged ≤18, 65 (34.4%) aged 19-30 years, 79 (41.8%) aged 31-45 years, 41 (21.7%) aged 46-60 years, and 2 (1.1%) aged ≥61. Low health management demanders (n=218) comprised 1 (0.5%) aged ≤18, 103 (47.2%) aged 19-30 years, 69 (31.7%) aged 31-45 years, 44 (20.2%) aged 46-60 years, and 1 (0.5%) aged ≥61. Potential health management demanders (n=180) included 2 (1.1%) aged ≤18, 97 (53.9%) aged 19-30 years, 54 (30.0%) aged 31-45 years, 26 (14.4%) aged 46-60 years, and 1 (0.6%) aged ≥61. Passive attitude holders (n=163) were distributed as 1 (0.6%) aged ≤18, 116 (71.2%) aged 19-30 years, 37 (22.7%) aged 31-45 years, and 9 (5.5%) aged 46-60 years. Chi-square tests indicated that the associations between user clusters and both age (Cramer V=0.145) and education (Cramer V=0.116) groups were weak.

First, the Kruskal-Wallis test was used to analyze differences in satisfaction with the 5 functional designs of the mHealth app among all users. The Kruskal-Wallis test indicated significant variations in user satisfaction with the 5 functional designs of mHealth apps—H(4)=445.388, P<.001. Specifically, the median satisfaction levels for the 5 functional dimensions among all users are presented in Table 6. Among these dimensions, health monitoring and health guidance exhibited the highest levels of satisfaction. Detailed satisfaction scores for the 5 functional designs, as well as pairwise comparisons of satisfaction among the 6 clusters, are also presented in Table 6. These results allow for a detailed examination of how each cluster differs from the others in terms of satisfaction with specific functional designs. Detailed test statistics are provided in Multimedia Appendix 6, which presents pairwise comparisons of satisfaction with the 5 functions across user groups.

Specifically, satisfaction with health guidance was higher in clusters 1, 2, 3, and 5 than in cluster 6. Pairwise comparisons revealed that clusters 1 (z=6.246, Bonferroni-adjusted P<.001), 2 (z=8.404, Bonferroni-adjusted P<.001), 3 (z=6.887, Bonferroni-adjusted P<.001), and 5 (z=3.674, Bonferroni-adjusted P=.004) differed significantly from cluster 6. Satisfaction with health education and health monitoring was significantly higher in clusters 1–3 than in clusters 4–6, with all 9 cross-group pairwise comparisons for each domain yielding statistically significant differences (health education: z=2.77-7.95, Bonferroni-adjusted P<.001 to .02 and health monitoring: z=3.05-7.88, Bonferroni-adjusted P<.001 to .03). With respect to social functions and gamification, users in cluster 2 reported higher satisfaction than those in the other 5 clusters. Pairwise comparisons revealed that cluster 2 differed significantly from cluster 1 (z=–3.584, Bonferroni-adjusted P=.005), cluster 3 (z=4.599, Bonferroni-adjusted P<.001), cluster 4 (z=6.857, Bonferroni-adjusted P<.001), cluster 5 (z=7.392, Bonferroni-adjusted P<.001), and cluster 6 (z=5.992, Bonferroni-adjusted P<.001) in social function; and from cluster 1 (z=–3.999, Bonferroni-adjusted P=.001), cluster 3 (z=5.015, Bonferroni-adjusted P<.001), cluster 4 (z=5.880, Bonferroni-adjusted P<.001), cluster 5 (z=7.200, Bonferroni-adjusted P<.001), and cluster 6 (z=6.083, Bonferroni-adjusted P<.001) in gamification.

In this study, confirmatory factor analysis was used to test the reliability and validity of the questionnaire. The results, as shown in Multimedia Appendix 7, indicate that all constructs had Cronbach α and composite reliability values above 0.7; the average variance extracted for each construct was greater than 0.6; and the factor loadings for all items exceeded 0.7, demonstrating good reliability and convergent validity for all constructs. The heterotrait-monotrait ratio (HTMT) method was used to assess discriminant validity. The commonly accepted criterion for establishing discriminant validity is an HTMT value below 0.85. The results indicate that all construct pairs exhibited HTMT values below 0.85, supporting the discriminant validity of the measurement model. Using 8 health management characteristics as independent variables and satisfaction with 5 feature designs as dependent variables, partial least squares analysis was conducted to examine the impact of health management characteristics on satisfaction with the functional design of mHealth apps. The R² value represents the proportion of variance in the dependent variable explained by the independent variables and is used to interpret the goodness-of-fit of the regression equation, while β indicates the correlation between variables. Table 7 displays the β values and statistical significance of R² for the independent and dependent variables. Specifically, perceived barriers (β=.177, z=3.586, P<.001), perceived susceptibility (β=.076, z=2.056, P=.04), health self-efficacy (β=.140, z=3.307, P=.001), perceived health status (β=.111, z=2.967, P=.003), eHealth literacy (β=.153, z=3.947, P<.001), and health management intention (β=.148, z=3.966, P<.001) had a significant positive impact on satisfaction with health guidance; perceived barriers (β=0.143, z=3.130, P=.002), health self-efficacy (β=0.147, z=3.491, P<.001), perceived health status (β= 0.120, z=3.072, P=.002), eHealth literacy (β= 0.131, z=3.362, P=.001), and health management intention (β= 0.175, z=4.670, P<.001) had a significant positive impact on satisfaction with health education; perceived severity (β=.118, z=3.235, P=.001), perceived barriers (β=.107, z=2.194, P=.03), perceived benefits (β=.089, z=2.428, P=.02), perceived health status (β=.088, z=2.188, P=.03), eHealth literacy (β=.170, z=4.561, P<.001), and health management intention (β=.083, z=2.064, P=.04) had a significant positive impact on satisfaction with health monitoring; perceived susceptibility (β=.098, z=2.314, P=.02), perceived barriers (β=.150, z=3.032, P=.002), health self-efficacy (β=.195, z=4.598, P<.001), perceived health status (β=.135, z=3.391, P=.001), eHealth literacy (β=.118, z=2.805, P=.005), and health management intention (β=.139, z=3.396, P=.001) had a significant positive impact on satisfaction with social functions, whereas perceived benefits (β=–.210, z=5.906, P<.001) had a significant negative impact and perceived susceptibility (β=.099, z=2.200, P=.03), perceived barriers (β=.139, z=3.074, P=.002), health self-efficacy (β=.137, z=3.076, P=.002), perceived health status (β=.122, z=2.834, P=.005), eHealth literacy (β=.119, z=2.771, P=.006), and health management intention (β=.100, z=2.354, P=.02) had a significant positive impact on satisfaction with gamification, while perceived benefits (β=–.165, z=5.011, P<.001) had a significant negative impact. These results help explain why different types of users exhibit varying levels of satisfaction with different functional designs. The statistical analysis results for each path are presented in Multimedia Appendix 8.

Understanding the heterogeneity of mHealth users and analyzing distribution differences in satisfaction with mHealth function design among user groups with distinct characteristics have significant referential value for personalized mHealth design. Therefore, guided by self-management theory and driven by self-management characteristics, this study constructed mHealth user personas encompassing 6 categories, analyzed the main characteristics and attributes of these 6 types of mHealth users, and examined differences in satisfaction with various mHealth function designs across user types. The results show that levels of user self-management characteristics are unevenly distributed and that different combinations of self-management characteristics can explain user heterogeneity. Moreover, a clear correlation exists between user heterogeneity arising from self-management characteristics and satisfaction with specific function designs. The findings of this study mutually confirm and extend existing literature on differences in mHealth intervention effects, user participation motivation, and technology design principles.

This study categorizes mHealth users into 6 distinct groups based on their self-management characteristics, revealing a complex classification derived from self-management–related attributes such as knowledge, beliefs, self-regulation skills and abilities, and motivation. This classification system differs from traditional approaches that rely on broad demographic labels (such as age and education level), offering a more refined and interpretable psychographic user profile. Although the sample in this study is predominantly composed of young-to-middle-aged and highly educated users, comparison with the 50th Statistical Report on China’s Internet Development (2022) [62] shows that while individuals aged 20-49 account for 56.6% of all Chinese internet users, only 28.5% have used online medical services. This indicates that even within the broadly age-eligible internet population, active adopters of digital health services constitute a specific subgroup. Accordingly, this study employed an online snowball sampling method to reach and recruit active users with genuine interest and engagement in mHealth management. The “young-to-middle-aged, well-educated” characteristics of the sample, to some extent, reflect the profile of individuals who pay close attention to such apps. However, chi-square test results show that the associations between user classification and age (Cramer V=0.145) and education level (Cramer V=0.116) are weak. Furthermore, analysis of the age composition of each user group revealed that every category spans multiple age layers, suggesting that single demographic labels are insufficient to capture the characteristics of the 6 user categories identified in this analysis. The uneven distribution of multiple features across different user categories indicates that there is no complete positive correlation between individuals’ perceived health status and their health management abilities and motivations. The differentiation of users’ characteristic attributes is consistent with relevant theories such as BCWs. Potential health management demanders and passive attitude holders both exhibit low health management intention, but their underlying mechanisms differ: the former perceive susceptibility and severity but face higher perceived barriers, whereas the latter generally lack recognition of the perceived benefits and health self-efficacy of health management. This is consistent with phenomena observed in studies of smoking cessation apps [63], namely that when users remain at the level of understanding health information but fail to convert it into personalized risk perception and action confidence, behavioral intentions are difficult to activate. Based on the results of this study, targeted interventions in mHealth management are needed according to individuals’ motivation status and health perceptions, to facilitate the transformation of different perception characteristics into internal motivation and to address differences in health management behaviors arising from varying motivation statuses [64]. In addition, although negatively proactive energizers perceive an average health status, they exhibit high health management intention and self-efficacy. This finding indicates that perceived health status alone is not the sole determinant of mHealth service use, which is also supported by existing studies [65], showing that patients’ satisfaction with mHealth services may be influenced by individual differences in disease progression, perceived service attitudes, and understanding of diagnostic and treatment outcomes. Correspondingly, passive attitude holders exhibit low levels of perceived health status, perceived benefits, health self-efficacy, and health management intention, which aligns with a previous study [66] identifying lack of motivation and low perceived value as significant barriers to mHealth participation. In general, obtaining a positive mHealth service experience often depends on whether services can be personalized or adaptively adjusted to accommodate different user needs and motivations, and user heterogeneity is strongly correlated with self-management characteristics [67].

There are significant differences in satisfaction with mHealth function designs among different user types, providing counterevidence to the “one-size-fits-all” function design strategy and underscoring the necessity of personalized design. This study found that proactive intenders and positively proactive energizers, characterized by high self-efficacy and strong health management intention, reported higher satisfaction with goal-oriented and interactive functions such as health guidance, health education, and health monitoring. This finding is consistent with multiple studies [63,68,69], indicating that users who actively focus on their own well-being demonstrate higher levels of engagement with mHealth. Features such as progress tracking, personalized feedback, and access to educational content are often welcomed and perceived as useful by these users.

Low health management demanders and passive attitude holders report lower satisfaction with most mHealth functions. This finding is consistent with a previous study [70], which indicated that for users with low intrinsic motivation or insufficient perceived value, complex functions or interfaces may act as barriers to use; such users require designs that are concise, intuitive, provide immediate and intuitive feedback, and involve low operational barriers. Notably, prior research has demonstrated a positive correlation between eHealth literacy and mHealth usage [71]. However, our findings offer a contrasting perspective, showing that potential health management demanders, despite having lower-than-average eHealth literacy, report high satisfaction with health guidance and health monitoring functions. This may be because this group’s higher perceived benefits and perceived susceptibility compensate for their lower capability, thereby increasing their interest in tools that enhance health benefits and reduce health risks. This phenomenon may also be partially explained by the Dunning-Kruger effect, whereby users lacking specialized knowledge may be less able to critically evaluate application limitations, leading to inflated satisfaction ratings for mHealth. Individuals with low health literacy have been shown to report levels of confidence about their knowledge comparable to, or even higher than, those with high literacy [72]. Conversely, users with stronger health knowledge reserves tend to use educational materials and support functions in mHealth apps more frequently, thereby enhancing recognition and evaluation of specific functions [73]. The user-centered design concept, combined with the user types constructed in this study, can therefore provide valuable reference for mHealth designs that address the diverse needs of different user groups.

The findings of this study provide clear implications for the personalized design and practical application of mHealth. mHealth developers should adopt a user-centered design approach and, based on the user types constructed in this study, tailor functional modules accordingly. For positively proactive energizers, further development of data tracking, personalized goal setting, and systematic educational content can be pursued, alongside the appropriate introduction of gamification elements to enhance users’ sense of enjoyment [74]. For potential health management demanders, design efforts should focus on lowering usage barriers, providing clear and step-by-step health guidance and monitoring tools, and strengthening health-related beliefs. For passive attitude holders, the primary objective may be to establish initial engagement and trust through simple designs and light-touch interventions, while incorporating social features to help this group build social support [75], increase awareness of health risks, and enhance beliefs in health management. Furthermore, the study’s results suggest that when evaluating the effectiveness of mHealth functions, user heterogeneity must be considered, as overall average satisfaction may obscure substantial differences among user subgroups.

Although this study was carefully designed in terms of methodology and empirical analysis, several limitations should be acknowledged. First, to avoid comparability issues arising from differences in functional design across mHealth apps with different core service content, and to recruit as diverse a user sample as possible, this study selected exercise, fitness, and weight management apps as the objects of study. However, these apps cannot fully represent all types of mHealth apps. Nevertheless, by conducting empirical research on one of the most widely used and numerous categories of mHealth apps, this study verifies that individuals’ self-management characteristics have important referential value for mHealth user classification. The results demonstrate significant differences in satisfaction with various functional designs across user types, and that the effects of self-management characteristics on satisfaction differ across functional design dimensions. These results provide direct empirical reference for the personalized design of fitness and weight loss apps, and the research conclusions may also offer inspiration for user classification–related studies in other types of mHealth apps. Second, although questionnaires are an effective research method, they cannot fully represent all users. Access to feedback data from the entire mHealth user population would enable a more refined classification of user types. At present, limited research on mHealth apps has focused on users’ health cognition and self-management abilities, which constrains the analysis of comprehensive user characteristics from a big data perspective. To mitigate sampling bias, data were collected from 34 regions across China, and the sample size was substantially larger than that of comparable studies, supporting the representativeness of the findings. Third, although empirical results indicate that 8 self-management characteristics significantly influence user satisfaction, these characteristics explain only a limited proportion of the variance in satisfaction with mHealth (R²=0.094-0.183). Although the identified pathways were statistically significant, a substantial proportion of the variance in satisfaction remains unexplained by these characteristics. Given that the primary aim of this study was to examine whether the effects of heterogeneous user types on satisfaction differ—that is, to construct user profiles based on self-management characteristics and analyze differences in satisfaction across user types—we did not conduct an in-depth examination of other potential determinants of satisfaction (eg, usage context, service quality, and user experience). Future research on factors influencing mHealth satisfaction could incorporate additional predictive variables to improve the explanatory power of the model.

This study identified 6 distinct mHealth user typologies (fitness and weight management apps) derived from a comprehensive set of self-management characteristics. The results reveal an uneven distribution of multidimensional characteristics across different user categories. This granular classification provides a deeper understanding of user heterogeneity than traditional demographic-based approaches. Overall, the primary contribution of this work is an empirically validated framework that offers a structured approach to understanding mHealth user diversity. The identified user typologies and their differential functional satisfaction profiles yield actionable insights for the personalized design of fitness and weight management apps. Specifically, for health guidance and education features, proactive intenders, positively proactive energizers, and negatively proactive energizers report higher satisfaction. For health monitoring functions, potential health management demanders, proactive intenders, positively proactive energizers, and negatively proactive energizers demonstrate higher satisfaction. For social features and gamification design, proactive intenders exhibit significantly higher satisfaction than the other 5 user types. From a practical perspective, these findings underscore the importance of adapting mHealth features to the distinct self-management needs and preferences of specific user groups. Such adaptation is essential for enhancing user satisfaction, fostering sustained engagement, and ultimately improving the perceived usefulness of digital health solutions.