This paper is in the following e-collection/theme issue:

Use and User Demographics of mHealth (344) Engagement with and Adherence to Digital Health Interventions, Law of Attrition (505) Evaluation and Research Methodology for mHealth (210) Hypertension Prevention and Treatment (229) mHealth for Wellness, Behavior Change and Prevention (2922) mHealth for Symptom and Disease Monitoring, Chronic Disease Management (1428)

Published on in Vol 10, No 2 (2022): February

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/33189, first published .
Exploring and Characterizing Patient Multibehavior Engagement Trails and Patient Behavior Preference Patterns in Pathway-Based mHealth Hypertension Self-Management: Analysis of Use Data

Exploring and Characterizing Patient Multibehavior Engagement Trails and Patient Behavior Preference Patterns in Pathway-Based mHealth Hypertension Self-Management: Analysis of Use Data

Exploring and Characterizing Patient Multibehavior Engagement Trails and Patient Behavior Preference Patterns in Pathway-Based mHealth Hypertension Self-Management: Analysis of Use Data

Authors of this article:

Dan Wu1 Author Orcid Image ;   Xiaoyuan Huyan2 Author Orcid Image ;   Yutong She1 Author Orcid Image ;   Junbin Hu3 Author Orcid Image ;   Huilong Duan1 Author Orcid Image ;   Ning Deng1, 4, 5 Author Orcid Image
Article Authors Cited by (3) Tweetations (3) Metrics

    Original Paper

    1College of Biomedical Engineering and Instrument Science, Ministry of Education Key Laboratory of Biomedical Engineering, Zhejiang University, Hangzhou, China

    2The First Health Care Department, The Second Medical Center & National Clinical Research Center for Geriatric Diseases, Chinese People's Liberation Army General Hospital, Beijing, China

    3Health Community Group of Yuhuan People's Hospital, Kanmen Branch, Taizhou, China

    4Alibaba–Zhejiang University Joint Research Center of Future Digital Healthcare, Hangzhou, China

    5Binjiang Institute of Zhejiang University, Hangzhou, China

    *these authors contributed equally

    Corresponding Author:

    Ning Deng, PhD

    College of Biomedical Engineering and Instrument Science

    Ministry of Education Key Laboratory of Biomedical Engineering

    Zhejiang University

    38 Zheda Rd, Zhouyiqing Bldg 512

    Yuquan Campus

    Hangzhou, 310000

    China

    Phone: 86 571 2295 2693

    Email: zju.dengning@gmail.com


    Background: Hypertension is a long-term medical condition. Mobile health (mHealth) services can help out-of-hospital patients to self-manage. However, not all management is effective, possibly because the behavior mechanism and behavior preferences of patients with various characteristics in hypertension management were unclear.

    Objective: The purpose of this study was to (1) explore patient multibehavior engagement trails in the pathway-based hypertension self-management, (2) discover patient behavior preference patterns, and (3) identify the characteristics of patients with different behavior preferences.

    Methods: This study included 863 hypertensive patients who generated 295,855 use records in the mHealth app from December 28, 2016, to July 2, 2020. Markov chain was used to infer the patient multibehavior engagement trails, which contained the type, quantity, time spent, sequence, and transition probability value (TP value) of patient behavior. K-means algorithm was used to group patients by the normalized behavior preference features: the number of behavioral states that a patient performed in each trail. The pages in the app represented the behavior states. Chi-square tests, Z-test, analyses of variance, and Bonferroni multiple comparisons were conducted to characterize the patient behavior preference patterns.

    Results: Markov chain analysis revealed 3 types of behavior transition (1-way transition, cycle transition, and self-transition) and 4 trails of patient multibehavior engagement. In perform task trail (PT-T), patients preferred to start self-management from the states of task blood pressure (BP), task drug, and task weight (TP value 0.29, 0.18, and 0.20, respectively), and spent more time on the task food state (35.87 s). Some patients entered the states of task BP and task drug (TP value 0.20, 0.25) from the reminder item state. In the result-oriented trail (RO-T), patients spent more energy on the ranking state (19.66 s) compared to the health report state (13.25 s). In the knowledge learning trail (KL-T), there was a high probability of cycle transition (TP value 0.47, 0.31) between the states of knowledge list and knowledge content. In the support acquisition trail (SA-T), there was a high probability of self-transition in the questionnaire (TP value 0.29) state. Cluster analysis discovered 3 patient behavior preference patterns: PT-T cluster, PT-T and KL-T cluster, and PT-T and SA-T cluster. There were statistically significant associations between the behavior preference pattern and gender, education level, and BP.

    Conclusions: This study identified the dynamic, longitudinal, and multidimensional characteristics of patient behavior. Patients preferred to focus on BP, medications, and weight conditions and paid attention to BP and medications using reminders. The diet management and questionnaires were complicated and difficult to implement and record. Competitive methods such as ranking were more likely to attract patients to pay attention to their own self-management states. Female patients with lower education level and poorly controlled BP were more likely to be highly involved in hypertension health education.

    JMIR Mhealth Uhealth 2022;10(2):e33189

    doi:10.2196/33189

    Keywords

    hypertension; mobile health; patient behavior; engagement; data analysis 


    Background

    Hypertension is a serious medical condition that affects health-related quality of life and increases the risks to the heart, brain, kidney, etc [1]. Controlling hypertension requires patients to achieve management goals by insistently adhering to long-term self-management plans, including regularly checking blood pressure (BP), taking medication, being physically active on a regular basis, eating more fruit and vegetables, and reducing alcohol consumption, etc. These plans should be established based on hypertension management guidelines and guidance from health care providers [2].

    The development of mobile technology has promoted the implementation of out-of-hospital mobile health (mHealth) services [3-6]. Extensive evidence supports the effect of mHealth services in disease control [4,5,7-11], including promoting patient engagement in health care services and helping patients develop positive behavior in their daily self-management [12-16]. In comparison with transitional hypertension management methods, mHealth services can effectively improve patient engagement in hypertension self-management [17,18]. Cechetti et al [19] designed an mHealth app with a gamification method for hypertension management, which can effectively promote patient engagement in their self-management. However, while recent studies have demonstrated the effectiveness of some mHealth services, others have performed poorly [12-15]. There have been mixed results for using mHealth services to support patient self-management of hypertension in the community [20,21]. Thus, hypertension mHealth services pose new design challenges in management strategies, partly because the mechanism by which patients engage in their self-management is not clear.

    Patient engagement is a broad concept that combines patient activation with interventions of health care services designed to increase activation and promote positive patient behavior [22]. Positive self-management behavior of patients engaging in mHealth services is essential for bringing an improvement in health outcomes [12,17,18,23-26]. Goyal et al [27] found a significant relationship between an increased number of daily blood glucose readings and improved glycated hemoglobin. Toto-Ramos et al [28] found that hypertensive patients with sustained engagement in mHealth services experienced significant reduction in BP. Thus, understanding self-management behavior of real-world patients in their daily lives can help to reveal patient behavior in natural settings. Compared to attracting patients from clinical trials who are more likely to overcome the burden associated with research work [29], this provides deeper insight into real patient self-management behaviors.

    Hypertension management requires long-term efforts, and patient behavior in hypertension management is dynamic and continuous [30,31]. Understanding the longitudinal characteristics of patients engaged in self-management is important for long-term successful management. Moreover, there are many different dimensions of behavior with patients who engage in mHealth services [12,30,32,33], such as measuring, viewing, and recording, which are associated with individual inherent preferences and habits of patients [34-36]. Patient engagement behavior has been measured as amount, duration, breadth, and depth of using mHealth services [12]. Rahman et al [37] measured patient engagement by 3 key use features: duration and frequency of using the mHealth app plus the number of use records. Sanatkar et al [33] measured 5 use features of patient engagement: number of user log-ins, number of daily trackers used, numbers of learning activities started and completed, and number of reminders received. However, the multidimensional and dynamic behavioral processes that change over time cannot be captured simply by analyzing count data captured at 1 time point in the cross-sectional data analysis. Longitudinal change in patient multibehavior can be identified through analyzing the time series data. More comprehensive understanding of the multiple self-management behavior and individual behavior preferences of patients engaging in mHealth self-management over a long period requires further research. Data mining techniques have been successfully adopted in the analysis of longitudinal events, such as human behavior navigation [26,38-40], information search behavior [41], phase-type distribution in the health care industry [42], and lifetime health care costs [43]. These would allow us to infer the characteristics of patient behavior throughout the entire period of self-management.

    Objective

    The aim of this study was to explore the trails of patient multibehavior engagement in the pathway-based mHealth hypertension self-management and identify the characteristics of patients with different behavior preferences. This included 3 objectives. The first objective was to discover the trails of patient multibehavior engagement. The second objective was to explore patient behavior preference patterns. The third objective was to identify the association between behavior preference patterns and the demographic and physiological characteristics. Identifying the multidimension and longitudinal characteristics of patient behavior within the mHealth hypertension management offers new opportunities for personalizing management goals and plans to reduce nonadherence and enhance possible effectiveness.


    Description of the mHealth Hypertension Management App

    We used data from the Blood Pressure Assistant, a pathway-based hypertension self-management app in the Digital Care Study for Hypertension Management [44]. The app was designed in accordance with a customized care pathway in compliance with the Chinese guideline for hypertension management [45] and was available for patients in the General Hospital of Ningxia Medical University. The care pathway involves 2 roles in hypertension management: health care providers and patients. The care pathway defines 3 goals for patients—improve self-management ability, enhance self-management motivation, and receive self-management support (see Figure 1)—and comprises 9 modules and 28 behavioral states (see Multimedia Appendix 1 for the detailed behavioral states). The 9 modules generate intervention plans and patient self-management plans. The term of state comes from the Markov decision process, a mathematical model of sequential decision. The behavioral state is a description of the patient’s behavior in a hypertension self-management environment or scene and is expressed by pages in the mHealth app. In this paper, we chose to use state to represent the patient behavior.

    ‎
    Figure 1. Components of pathway-based mHealth hypertension self-management. BP: blood pressure.
    View this figure

    In this care pathway, patient actions on different pages of the app reflect specific patient behaviors. The behavioral states are represented by different pages in the app. Through the app, each patient registers and enters basic demographic information, after which the patient is assigned to a health care provider who enrolls patients they manage into the mHealth hypertension management program at an online community [45]. The health care provider is responsible for formulating a tailored management plan, reviewing patients’ uploaded data, and conducting follow-up. Patients can use the app to seek help from health care providers, check their management plans, and record self-management data including BP monitoring, medication, physical activity, and diet. These patient behaviors are represented on the appointment, main, task BP, task drug, and other pages of this mHealth app. Health care providers can track current BP readings of patients through a web-based platform to adjust management plans and use mobile phones for patient follow-ups to assist in BP control.

    Data Collection

    Ethics

    Ethical approval was granted by the ethics committee for conducting human research at the General Hospital of Ningxia Medical University (NXMU-GH-2017-273). All patients in this study signed the informed consent forms for their anonymized data to be used in routine evaluations to monitor and improve health care services.

    Sample

    Since this app was launched in December 2015, 1159 patients have used it to self-manage their BP. We selected patients based on the following inclusion criteria: aged 18 to 80 years, diagnosed with hypertension, performed self-management between December 28, 2016, and July 2, 2020 (the main functions of the app were consistent during this period, ensuring patient behavior was not affected by the changes in app functions).

    Data Extraction

    All data were stored and extracted from the server of the Blood Pressure Assistant, which contains the self-management plans, demographic information, uploaded self-management data from patients, and follow-up records of health care providers. We extracted 3 types of data from the database: demographics, physiological records, and patient use records. Demographics included patient identification, data of birth, gender (male or female), and education level (below high school, high school, university and above). Physiological records included patient identification, systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), and uploaded date (year, month, day, minute, and second). Patient use records included patient identification, page name, time page is entered and exited, and stay time. These data help determine longitudinal and multidimensional behavioral characteristics of hypertensive patients and characteristics of patients with different behavior preferences.

    Data Analysis

    Identifying Patient Multibehavior Engagement Trails

    Patient multibehavior engagement trails were indicated by the type, quantity, time spent, sequence, and transition probability value (TP value) of patient behavioral states (see Figure 2). The time spent and sequence were calculated from the time pages were entered and exited.

    The data of patient use behavior were analyzed through first-order Markov chain analysis, a method to model stochastic processes, which is suitable for analyzing user interaction with mHealth apps [46,47]. In the mHealth field, interaction between users and apps is determined through a process of connecting multiple continuous actions together. In the case of this study, we assumed that a next state depends only on the current state and not on the history of previous states. Thus, we used the first-order Markov chain called a memoryless model, containing the state space S , action, session, and transition matrix P. The discrete state space S was defined by the n different behavioral states (represented by different pages in the app): S={s1,...,sn}. A start state s0 and an exit state sn+1 were created when the patient started and left the app. The action was defined by the transition from one behavioral state to another (si→sj). The session was discriminated by the time interval between 2 consecutive states greater than 600 seconds (si–sj>600 s). The transition matrix P contained each element on row i and column j (pi,j) and indicated the transition probability that a patient moves from state si to state sj. The transition probability pi,j was defined as

    where Ni,j was the number of transitions si→sj. The quantity and time spent of the behavioral state are represented by the width and length of the rectangle (see Figure 2).

    The transition matrix P was displayed by a heat map, which allowed us to understand the characteristics of patient multibehavior transition. Analysis was carried out using Python (version 3.8, Python Software Foundation). The 1% relative quantity of states and the 0.15 transition probability were used as cutoff values to improve readability. We then combined quantity, time, transition probability, and the pathway to plot the main trails of patient multibehavior engagement.

    ‎
    Figure 2. Example of patient multibehavior engagement trail represented by first-order Markov chain. BP: blood pressure.
    View this figure
    Cluster Analysis of Patients With Different Behavior Preferences

    Cluster analysis was conducted to group patients according to patient behavior preference features: the number of behavioral states that a patient performed in each trail. First, we normalized the behavior preference features of each patient. The sum of behavior preference features of each patient for different behavior trails was 1. A K-means algorithm was then used to cluster patients through these normalized patient behavior preference features. Euclidean distance was used to calculate the similarity of features between patients. Finally, we used a silhouette score to determine the optimal number of clusters. Higher silhouette scores indicate tighter clusters, where each cluster is completely separate from the others.

    Characterizing the Clusters

    An optimal clustering result was reached based on the silhouette score of different numbers of clusters. For each cluster, we analyzed the demographic features (age, gender, and education level), and physiological features (mean BP [SBP and DBP] and mean HR). Statistical analysis was conducted in SPSS (version 24, IBM Corp). A chi-square test was performed to evaluate the statistical significance of associations between the clusters and discrete variables (gender and education level). A z-test was used to conduct pairwise comparisons of the differences among the proportion of discrete variables between the clusters. Analyses of variance (ANOVAs) were used to evaluate the statistical significance of associations between the clusters and continuous variable (age, BP, and HR). Bonferroni multiple comparisons were further conducted to examine the differences between the clusters. The cutting value (P≤.05) was used to determine whether the difference was statistically significant.


    Patient Multibehavior Engagement Trails

    Quantity and Time Characteristics of Each Behavioral State

    We restricted analysis to the remaining 863 patients, with 295,855 records from the app. Tables 1 and 2 depict the quantity and time spent in each behavioral state. It was clear that the main state (125,487 [42.42%]; 1,745,286 s [13.91 s]) was visited most in quantity and total time spent; it was the default starting page for a session. The relative number of 14 states exceeded 1%, and the mean time spent of 12 states exceeded 20 seconds. In 5 self-management tasks, the state of task BP (48,935 [16.54%]; 1,253,924 s [25.62 s]) was visited more, and the total time spent was also longer. The mean time spent on task food (149,164 s [35.87 s]) state was more than other self-management tasks. The states of ranking (7856 [2.66%]) and knowledge content (7330 [2.48%]) had relatively high quantities of visits. The states of knowledge content (615,430 s [83.96 s]) and show appointment (261,371 s [162.75 s]) had relatively long mean time spent. In 9 modules, the mean time spent in management plans (contained 5 self-management task parts: 230.21 s) was longer than health education (contained 4 knowledge parts: 124.06 s) and appointment (contained 2 appointment parts: 168.72 s).

    Table 1. Quantity of each behavioral state.
    Behavioral stateaQuantity, n (%)
    Main125,487 (42.42)
    Task blood pressure48,935 (16.54)
    Task weight33,560 (11.34)
    Task drug26,509 (8.96)
    Ranking7856 (2.66)
    Knowledge content7330 (2.48)
    Blood pressure history6070 (2.05)
    Knowledge list4871 (1.65)
    Task food4159 (1.41)
    Questionnaire4128 (1.40)
    Health report4055 (1.37)
    Reminder item3287 (1.11)
    Appointment3209 (1.08)
    Weight history3150 (1.06)
    Reminder2434 (0.82)
    Drug history2286 (0.77)
    Show appointment1606 (0.54)
    Task uncomfortable1345 (0.45)
    Person information1267 (0.43)
    Monthly report1144 (0.39)
    Knowledge collect1065 (0.36)
    Knowledge list by tag742 (0.25)
    Blood pressure guidance721 (0.24)
    Setting207 (0.07)
    Food history174 (0.06)
    Account125 (0.04)
    Log-in113 (0.04)
    Register20 (0.01)

    aDescending by quantity (relative) of behavioral states.

    Table 2. Time spent in each behavioral state.
    Behavioral stateaTime spent (total), sTime spent (mean), s
    Show appointment261,371162.75
    Knowledge content615,43083.96
    Task food149,16435.87
    Monthly report39,59034.61
    Register63931.95
    Blood pressure history185,24030.52
    Blood pressure guidance19,77227.42
    Task blood pressure1,253,92425.62
    Log-in257522.78
    Task drug600,10422.64
    Task weight746,26822.24
    Knowledge list100,99520.73
    Ranking154,46819.66
    Food history339119.49
    Task uncomfortable23,47417.45
    Account211016.88
    Weight history52,21816.58
    Reminder item52,49315.97
    Main1,745,28613.91
    Knowledge list by tag10,12913.65
    Health report53,73013.25
    Drug history28,30912.38
    Questionnaire31,5797.65
    Setting14517.00
    Person information77796.14
    Appointment19,1575.97
    Knowledge collect60945.72
    Reminder84263.46

    aDescending by time spent (mean) of behavioral states.

    Patient’s Behavior Transition Matrix

    The heatmap was used to demonstrate the transition matrix of the Markov chain (see Figure 3). The various shades of color represent the probability of transition from one behavioral state to another, and the transparent color indicate that these 2 behavioral states cannot be transitioned. Horizontal and vertical coordinates indicate behavioral states. The rows summed up to 1. In most cases, the session started from the main state (0.91), and the transition probability from other behavioral states to the main state was also high, which was intended by design. When patients were in the main state, the states of task BP (0.29), task drug (0.18), and task weight (0.20) were the most visited among all behavioral states. Patients had a high probability if exiting the app from the following states: task drug (0.34), ranking (0.44), and log-in (0.55).

    ‎
    Figure 3. Transition matrix of patient multibehavior engagement. Probability of transitioning from behavioral state A[row] to B[column]. BP: blood pressure.
    View this figure

    We found that the 1-way transition (si→sj) from one behavioral state to another had a high probability in the following states: from BP guidance (0.96) or BP history (0.90) to task BP, from drug history to task drug (0.95), from food history to task food (0.99), from weight history to task weight (0.95), from reminder item to task BP (0.20) or task drug (0.25), from knowledge list by tag to knowledge content (0.98), and from account to person information (0.95). The cycle-transitions (si⇔sj) between the 2 behavioral states had a high probability in the following states: between reminder and reminder item (0.58, 0.43), between health report and monthly report (0.26, 0.92), between knowledge content and knowledge list (0.31, 0.47), between appointment and show appointment (0.41, 0.73), between person information and setting (0.16, 1.00), and between log-in and register (0.17, 0.65). We found that the self-transition (sisi) in a behavioral state was very common, and a high self-transition can be seen in the questionnaire state (0.29).

    Four Types of Patient Multibehavior Engagement Trails

    There were 28 different types of behavior states and 283 possible behavior transitions in the original design in mHealth hypertension management app. The trails of patient multibehavior engagement in the pathway-based hypertension self-management were visualized by the main transitions between the different behavioral states (see Figure 4). The main trails of patient behavior included the type, quantity, time spent, sequence, and TP value. The size of a node indicated the quantity and time spent of behavioral states; a node or line of the same or similar color indicated a trail. Visual inspection of the transition between different behavioral states in mHealth hypertension self-management revealed several types of trails with different behavioral characteristics.

    ‎
    Figure 4. Four types of patient multibehavior engagement trails. BP: blood pressure.
    View this figure

    A first trail type can be labeled as a perform task trail (PT-T; green and dark green lines). After launching the app, patients go from the main state to the self-management task states to conduct management plans. Among the 5 self-management tasks, patients were more likely to transfer to the states of task BP, task drug, and task weight (TP value 0.29, 0.18, 0.20) than the states of task food, and task uncomfortable (TP value 0.03, 0.01). Moreover, some patients entered the states of task BP and task drug (TP value 0.20, 0.25) from the reminder item state, which happened when patients needed the app to remind themselves to measure BP and take medication on time. Patients also checked their BP history and weight history when they conduct the self-management tasks. After finishing the self-management tasks, there was a high probability that patients would return to the main state. The mean time spent was longer in the task food (35.87 s) state than that in the states of task BP, task drug, and task weight, and patients spent more time checking BP history (30.52 s) than weight history (16.58 s).

    Second, a result-oriented trail (RO-T) can be distinguished (yellow and orange lines). Patients had a high probability in the states of health report and ranking in this trail, and the purpose was to check their self-management outcomes and their ranking among all hypertensive patients. The mean time spent in the ranking (19.66 s) state was longer than the health report (13.25 s) state, which showed that when competing with other patients, patients were more willing to spend time to understand the results of self-management. Eventually, this trail usually returned to the main state.

    The third trail type was labeled a knowledge learning trail (KL-T; blue lines). In this trail, patients preferred to read the health education content. In addition, there was a high probability of cycle transition (TP value 0.47, 0.31) between the knowledge list state and the knowledge content state. This situation occurred when patients switched to a new piece of knowledge content after reading a previous piece of knowledge content. Patients spent an average of 83.96 s to reading a piece of knowledge content, and they spent an average of 20.73 s in the knowledge list state to select a piece of knowledge content to read. The end of this trail was sometimes the main state.

    The last trail type was labeled a support acquisition trail (SA-T; purple and rese red lines). The states of questionnaire and appointment indicated the behavior of patients providing information to health care providers and seeking support from health care providers. There was a high probability of self-transition in the questionnaire (TP value 0.29) state, indicating the patients entered the questionnaire state, then switched to another app but did not exit the app, and then reopened the app on the questionnaire page. The end of this trail was sometimes the main state.

    Patient Behavior Preference Patterns

    A total of 863 patients were selected for cluster analysis. We found that the silhouette score was the highest with 3 clusters of patients (see Figure 5). Hence, we accepted the 3-cluster output of K-means for further analysis. The patient behavior preferences in 4 trails were significantly different (PT-T P<.001, RO-T P=.06, KL-T P<.001, SA-T P<.001).

    There were 3 distinctive patterns of patient behavior preferences (see Figure 6). The sum of each patient’s behavior preferences for the 4 behavior trails was 1. The first cluster (PT-T) comprised 694 patients. Their behavior preference was particularly focused on PT-T (0.81), which indicated that this patient group preferred to perform self-management tasks. The second cluster (PT-T and KL-T) comprised 96 patients, who were active in PT-T (0.37) and KL-T (0.53). They were more likely to read knowledge about hypertension than to conduct self-management tasks. The third cluster (PT-T and SA-T) comprised 73 patients whose behavior preferences were PT-T (0.30) and SA-T (0.37). These patients were more willing to seek help from health care providers than to perform self-management tasks.

    ‎
    Figure 5. Comparison of the silhouette score for different number of clusters (range 2-9).
    View this figure
    ‎
    Figure 6. Three patterns of patient behavior preference.
    View this figure

    Demographics and Physiological Characteristics of the 3 Behavior Preference Patterns

    There were statistically significant associations between the behavior preference pattern and gender, education level, and BP, but there were not associations between the behavior preference pattern and age and HR (see Table 3). In the PT-T cluster and the PT-T and SA-T cluster, there were much more male patients than female patients, but the proportion of male and female patients was equal in the PT-T and KL-T cluster. Compared with the PT-T cluster, the PT-T and KL-T cluster was characterized by significantly fewer male patients, lower education level, and higher BP. The PT-T and SA-T cluster had significantly lower BP than the PT-T and KL-T cluster.

    Table 3. Descriptive characteristics of the behavior preference pattern. Multiple comparisons of the 4 clusters (at the .05 level).
    CharacteristicPT-Ta cluster (n=694)PT-T and KL-Tb cluster (n=96)PT-T and SA-Tc cluster (n=73)P value
    Age, mean (SD)53.20 (9.93)52.23 (9.99)50.86 (10.34).13
    Gender, n (%)—d——.003

    		Male469 (67.58)48 (50.00)e49 (67.12)—

    		Female225 (32.42)48 (50.00)e24 (32.88)—
    Education, n (%)———.05

    		< High school249 (35.88)43 (44.79)34 (46.57)—

    		High school105 (15.13)21 (21.88)11 (15.07)—

    		≥ University321 (46.25)30 (31.25)e25 (34.25)—

    		Don’t know19 (2.74)2 (2.08)3 (4.11)—
    SBPf, mean (SD)132.71 (15.43)137.44 (17.96)e130.77 (13.18)g.02
    DBPh, mean (SD)86.46 (11.37)90.60 (12.88)e84.87 (11.29)g.004
    HRi, mean (SD)72.41 (14.23)73.20 (17.13)73.23 (15.83).84

    aPT-T: perform task trail.

    bKL-T: knowledge learning trail.

    cSA-T: support acquisition trail.

    dNot applicable.

    eGiven cluster is significantly different from the PT-T cluster.

    fSBP: systolic blood pressure.

    gGiven cluster is significantly different from the PT-T and KL-T cluster.

    hDBP: diastolic blood pressure.

    iHR: heart rate.


    Principal Findings

    In this study, we used a stochastic model to describe longitudinal trails of patient multibehavior engagement with mHealth hypertension self-management and further analyzed the characteristics of patient groups with different behavior preferences. In a sample of 863 patients in mHealth hypertension self-management, we identified 3 types of behavior transition (1-way transition, cycle transition, and self-transition), 4 trails of patient multibehavior engagement (PT-T, RO-T, KL-T, and SA-T), and 3 behavior preference patterns (PT-T cluster, PT-T and KL-T cluster, and PT-T and SA-T cluster).

    These insights revealed what actual patients do in daily hypertension self-management and how patients use the mHealth app to conduct self-management [32,43,47]. This may facilitate tailored and precise behavioral intervention strategies with specific content, methods, and time points for patient groups with specific behavior preferences. For example, for patient groups who measure and record BP but not take medication regularly, we will send a pop-up medication reminder in the app after completing the BP recording, rather than randomly and frequently reminding patients to measure BP and take medication. After they follow this reminder, we will provide rewards and push the knowledge about the advantages of regular medication to improve their adherence. This may promote more accurate chronic disease management strategies to achieve disease control.

    Patient Multibehavior Engagement Trails

    The dynamic behavioral characteristics were identified by the transition probability between 2 continuous behaviors [39,47], which suggested the transitions in daily self-management behavior of hypertension patients. When patients started self-management, they were more likely to focus on BP, medications, and weight conditions. Patients prefer to pay attention to BP and medications from reminders, which can help them manage their conditions. Tao et al [48] found that the use of electronic reminders was associated with a significant improvement in patient adherence to medication. The cycle transition was often performed between 2 different behaviors of same module, such as reminder, health report, knowledge, and appointment, etc. This kind of behavior transition information gives us opportunities to provide specific behavioral interventions at an accurate time between 2 consecutive behaviors. In addition, self-transition was frequent when patients completed questionnaires. Patients were willing to engage in the questionnaire, but the time spent was very short. It may be because the questionnaire was too long and patients had no patience when completing it or the content of the questionnaire was difficult to understand. The clarity and conciseness of questionnaires should be improved so patients can better engage in this management part [49].

    The trails of patient multibehavior engagement revealed the high probability sequences of patient behavior in mHealth hypertension self-management. This was a longitudinal and multidimensional process analysis rather than a cross-sectional analysis at a point in time. The 4 trails contained the main scenes of hypertension management in daily life [50] and described the behavioral conditions of patient self-management. In PT-T, for 5 types of self-management tasks of improving ability, patients preferred to start self-management by focusing on BP, medications and weight and spent more time on food. This may be attributed to diet management being more complicated and difficult to implement and record [51], which proposes a new challenge of how to improve the convenience of hypertension diet management. Samiul et al [52] proposed a network for automatic dietary monitoring that can gather food intake information through image, audio, and accelerometer sensors. By analyzing these data, the system can measure the type, volume, and nutrition of food, as well as the eating behavior of a person. Compared with weight history, patients spent more time reading BP history. In KL-T, patients usually chose different health knowledge contents from knowledge list to learn how to improve self-management ability. This knowledge may help patients broadly learn health concepts and behaviors [53]. In RO-T, patients read health reports to understand all aspects of their self-management, and knew the ranking of their self-management outcomes in an online hypertension community from the leaderboard. An interesting finding was that patients spent more energy on the rankings compared to reading health reports. This may suggest that gamification and competitive methods can effectively increase patient motivation and attention to their own self-management states [19,21,54]. In SA-T, patients sought help from health care providers through appointments, and completed the questionnaire to provide their own information and support research on hypertension management. These findings helped us dive into the daily self-management of actual patients and deeply understand patient behavioral characteristics through a longitudinal and multidimensional method. This provides an opportunity to apply some health behavior intervention techniques to promote patients to change their behavior and improve the effectiveness of chronic management results [50,55].

    Patient Behavior Preference Patterns

    The behavior preference patterns of patients represent the individual inherent habits of patients in daily hypertension self-management [34-36], which is an essential factor to the design of management strategies. In this study, we found 3 behavior preference patterns of patients engaging in the pathway-based mHealth hypertension management, reflecting which patient groups preferred which type of behavior to achieve better self-management results. All of 3 patterns preferred PT-T, 1 preferred KL-T, and the other preferred SA-T, but there was no obvious preference for RO-T.

    Patients with these 3 behavior preference patterns had significantly different demographic (gender and education level) and physiological (BP) characteristics. Compared with other patterns, patients with the preference for PT-T and KL-T had a lower education level, higher BP, and were much less likely to be male. First of all, there were more male patients than female patients, possibly because males were more inclined to use mHealth services for hypertension management [56,57]. However, we found that female patients were more willing to read knowledge content than male patients. Our finding is consistent with the finding by Al-Ansari et al [58] that oral health knowledge and health behavior were statistically significantly higher among the females than the males. Moreover, in this study, we found that patients with lower education level and poorly controlled BP were more likely to be highly involved in hypertension health education. Some scientific literature demonstrated a strong association between lower levels of education and poor health outcomes [59,60]. Health literacy was a potential pathway between levels of education and health outcomes [61]. Lee et al [62] and Nutbeam et al [63] also found that in comparison with people with higher education, people with lower education level were found to demonstrate lower health literacy and poor health. Naturally, these patients had a strong enthusiasm for learning about hypertension to promote their health literacy and improve health outcomes. The findings of patient preferences and characteristics are useful for designing personalized functions in the mHealth app to improve patient ability and engagement in hypertension self-management [34-36].

    Strengths and Limitations

    The strengths of this study were that, first of all, the study used time series data to unveil the longitudinal and multidimensional behavioral characteristics of patient engaging in mHealth hypertension management, which contained the type, quantity, time spent, sequence, and transition probability of patient behavior. The findings help provide more precise timing of mHealth behavior interventions for hypertension management. To the best of our knowledge, there has been no literature that uses longitudinal data to describe the trails of patient multibehavior engagement in hypertension self-management. Secondly, we analyzed the demographic and physiological characteristics of patients with different behavior preferences. Our findings revealed the association between behavior preferences, demographics, and physiology. This provided information for the further design of the appropriate types of interventions for specific patient groups. Finally, we used data from real-world patients, which revealed patient behavior in the natural setting rather than attracting patients more likely to bear the burden of research-related evaluation. This helped to generate practical insights for the behavior of actual patients in daily hypertension management.

    This study has limitations. One limitation was that the data were only from an mHealth hypertension app, which implies that the sample is presumably not representative of all patient groups. We also don’t know the behavior of patients who engage in self-management but not use the app, such as older patients who cannot use the mobile devices. Moreover, the behavioral characteristics of patients may be influenced and limited by the design of the app. Second, we used a first-order Markov chain, which simplified our analysis. Some other machine learning methods such as higher-order Markov chains should be applied to better represent global patient behavior. Finally, the behavior preferences can be caused by various demographic and social psychological characteristics of patients (such as marital status, profession, anxiety, depression, etc), so these factors need to be considered in the future studies.

    Conclusion

    The mHealth services are an effective way to conduct out-of-hospital health care services for chronic disease. In this study, we have found 3 types of behavior transitions, 4 trails of patient multibehavior engagement, and 3 patient behavior preference patterns in the pathway-based mHealth hypertension self-management. These findings gained insights of actual behavior of patients in daily hypertension self-management. The behavioral characteristics of patients were dynamic, longitudinal, and multidimension, which may create opportunities to design tailored, personalized interventions within a specific behavior time to change patient’s behavior, develop positive behavioral habits, and continuously improve and maintain the effect of hypertension management.

    Acknowledgments

    This study was supported by grants 2020YFC2003403 and 2020YFC2006405 from the National Key Research and Development Programs of China, grant 2020BFG02002 from the Key Research and Development Program of Ningxia, China, and grant 2021AB11007 from the Key Research and Development Program of Guangxi, China. This work was also supported by the Alibaba Cloud. The authors would like to thank the General Hospital of Ningxia Medical University.

    Authors' Contributions

    DW, XH, and ND were the primary authors of this paper, oversaw all aspects of the study design and execution, and led the writing of the manuscript. YS and JH led all data analysis with oversight from DW and XH. HD and ND provided critical revision of the manuscript for important intellectual content.

    Conflicts of Interest

    None declared.

    ‎
    Multimedia Appendix 1

    Detailed information about behavioral states.

    DOCX File , 15 KB
    1. Global Status Report On Noncommunicable Diseases 2014. Geneva: World Health Organization; 2014.   URL: https:/​/apps.​who.int/​iris/​bitstream/​handle/​10665/​148114/​9789241564854_eng.​pdf;jsessionid=3C0C9A267DA218A8836BDCA9EB2C5C0D?sequence=1 [accessed 2022-01-14]
    2. Reynolds R, Dennis S, Hasan I, Slewa J, Chen W, Tian D, et al. A systematic review of chronic disease management interventions in primary care. BMC Fam Pract 2018 Jan 09;19(1):11 [FREE Full text] [CrossRef] [Medline]
    3. Free C, Phillips G, Galli L, Watson L, Felix L, Edwards P, et al. The effectiveness of mobile-health technology-based health behaviour change or disease management interventions for health care consumers: a systematic review. PLoS Med 2013;10(1):e1001362 [FREE Full text] [CrossRef] [Medline]
    4. Osborn CY, Mayberry LS, Wallston KA, Johnson KB, Elasy TA. Understanding patient portal use: implications for medication management. J Med Internet Res 2013;15(7):e133 [FREE Full text] [CrossRef] [Medline]
    5. Wong A, Bhyat R, Srivastava S, Boissé Lomax L, Appireddy R. Patient care during the COVID-19 pandemic: use of virtual care. J Med Internet Res 2021 Jan 21;23(1):e20621 [FREE Full text] [CrossRef] [Medline]
    6. Cheung K, Ling W, Karr C, Weingardt K, Schueller S, Mohr D. Evaluation of a recommender app for apps for the treatment of depression and anxiety: an analysis of longitudinal user engagement. J Am Med Inform Assoc 2018 Aug 01;25(8):955-962 [FREE Full text] [CrossRef] [Medline]
    7. Chen X, Koskela M, Hyväkkä J. mHealth new horizons for health through mobile technologies. Proc Int Work Content-Based Multimed Index 2010;3:130-135. [CrossRef]
    8. Byambasuren O, Sanders S, Beller E, Glasziou P. Prescribable mHealth apps identified from an overview of systematic reviews. NPJ Digit Med 2018;1:12 [FREE Full text] [CrossRef] [Medline]
    9. Li R, Liang N, Bu F, Hesketh T. The effectiveness of self-management of hypertension in adults using mobile health: systematic review and meta-analysis. JMIR Mhealth Uhealth 2020 Mar 27;8(3):e17776 [FREE Full text] [CrossRef] [Medline]
    10. Dendere R, Slade C, Burton-Jones A, Sullivan C, Staib A, Janda M. Patient portals facilitating engagement with inpatient electronic medical records: a systematic review. J Med Internet Res 2019 Apr 11;21(4):e12779 [FREE Full text] [CrossRef] [Medline]
    11. Kruse CS, Argueta DA, Lopez L, Nair A. Patient and provider attitudes toward the use of patient portals for the management of chronic disease: a systematic review. J Med Internet Res 2015;17(2):e40 [FREE Full text] [CrossRef] [Medline]
    12. Pham Q, Graham G, Carrion C, Morita PP, Seto E, Stinson JN, et al. A library of analytic indicators to evaluate effective engagement with consumer mhealth apps for chronic conditions: scoping review. JMIR Mhealth Uhealth 2019 Jan 18;7(1):e11941 [FREE Full text] [CrossRef] [Medline]
    13. Hamine S, Gerth-Guyette E, Faulx D, Green BB, Ginsburg AS. Impact of mHealth chronic disease management on treatment adherence and patient outcomes: a systematic review. J Med Internet Res 2015;17(2):e52 [FREE Full text] [CrossRef] [Medline]
    14. Sakakibara BM, Ross E, Arthur G, Brown-Ganzert L, Petrin S, Sedlak T, et al. Using mobile-health to connect women with cardiovascular disease and improve self-management. Telemed J E Health 2017 Dec;23(3):233-239. [CrossRef] [Medline]
    15. Rothman AA, Wagner EH. Chronic illness management: what is the role of primary care? Ann Intern Med 2003 Feb 04;138(3):256-261. [CrossRef] [Medline]
    16. Edney S, Ryan JC, Olds T, Monroe C, Fraysse F, Vandelanotte C, et al. User engagement and attrition in an app-based physical activity intervention: secondary analysis of a randomized controlled trial. J Med Internet Res 2019 Nov 27;21(11):e14645 [FREE Full text] [CrossRef] [Medline]
    17. Rehman H, Kamal AK, Morris PB, Sayani S, Merchant AT, Virani SS. Mobile health (mHealth) technology for the management of hypertension and hyperlipidemia: slow start but loads of potential. Curr Atheroscler Rep 2017 Mar;19(3):12. [CrossRef] [Medline]
    18. Cheikh-Moussa K, Mira JJ, Orozco-Beltran D. Improving engagement among patients with chronic cardiometabolic conditions using mHealth: critical review of reviews. JMIR Mhealth Uhealth 2020 Apr 08;8(4):e15446 [FREE Full text] [CrossRef] [Medline]
    19. Cechetti NP, Bellei EA, Biduski D, Rodriguez JPM, Roman MK, De Marchi ACB. Developing and implementing a gamification method to improve user engagement: a case study with an m-Health application for hypertension monitoring. Telematics Informatics 2019 Aug;41:126-138. [CrossRef]
    20. Goyal S, Morita PP, Picton P, Seto E, Zbib A, Cafazzo JA. Uptake of a consumer-focused mHealth application for the assessment and prevention of heart disease: the <30 days study. JMIR Mhealth Uhealth 2016 Mar 24;4(1):e32 [FREE Full text] [CrossRef] [Medline]
    21. Edney S, Ryan JC, Olds T, Monroe C, Fraysse F, Vandelanotte C, et al. User engagement and attrition in an app-based physical activity intervention: secondary analysis of a randomized controlled trial. J Med Internet Res 2019 Nov 27;21(11):e14645 [FREE Full text] [CrossRef] [Medline]
    22. James J. Health policy brief: patient engagement. Health Affairs 2013 Feb 14. [CrossRef]
    23. Eysenbach G. The law of attrition. J Med Internet Res 2005;7(1):e11 [FREE Full text] [CrossRef] [Medline]
    24. Tyagi H, Sabharwal M, Dixit N, Pal A, Deo S. Leveraging providers' preferences to customize instructional content in information and communications technology-based training interventions: retrospective analysis of a mobile phone-based intervention in India. JMIR Mhealth Uhealth 2020 Mar 03;8(3):e15998 [FREE Full text] [CrossRef] [Medline]
    25. Kaplan AL, Cohen ER, Zimlichman E. Improving patient engagement in self-measured blood pressure monitoring using a mobile health technology. Health Inf Sci Syst 2017 Dec 7;5(1):4 [FREE Full text] [CrossRef] [Medline]
    26. Chien I, Enrique A, Palacios J, Regan T, Keegan D, Carter D, et al. A machine learning approach to understanding patterns of engagement with internet-delivered mental health interventions. JAMA Netw Open 2020 Jul 01;3(7):e2010791 [FREE Full text] [CrossRef] [Medline]
    27. Goyal S, Nunn CA, Rotondi M, Couperthwaite AB, Reiser S, Simone A, et al. A mobile app for the self-management of type 1 diabetes among adolescents: a randomized controlled trial. JMIR Mhealth Uhealth 2017 Jun 19;5(6):e82. [CrossRef] [Medline]
    28. Toro-Ramos T, Kim Y, Wood M, Rajda J, Niejadlik K, Honcz J, et al. Efficacy of a mobile hypertension prevention delivery platform with human coaching. J Hum Hypertens 2017 Dec;31(12):795-800 [FREE Full text] [CrossRef] [Medline]
    29. Sanders I, Short CE, Bogomolova S, Stanford T, Plotnikoff R, Vandelanotte C, et al. Characteristics of adopters of an online social networking physical activity mobile phone app: cluster analysis. JMIR Mhealth Uhealth 2019 Jun 03;7(6):e12484 [FREE Full text] [CrossRef] [Medline]
    30. Yardley L, Spring BJ, Riper H, Morrison LG, Crane DH, Curtis K, et al. Understanding and promoting effective engagement with digital behavior change interventions. Am J Prev Med 2016 Nov;51(5):833-842. [CrossRef] [Medline]
    31. Wu D, An J, Yu P, Lin H, Ma L, Duan H, et al. Patterns for patient engagement with the hypertension management and effects of electronic health care provider follow-up on these patterns: cluster analysis. J Med Internet Res 2021 Sep 28;23(9):e25630 [FREE Full text] [CrossRef] [Medline]
    32. Rahman QA, Janmohamed T, Pirbaglou M, Ritvo P, Heffernan JM, Clarke H, et al. Patterns of user engagement with the mobile app, Manage My Pain: results of a data mining investigation. JMIR Mhealth Uhealth 2017 Jul 12;5(7):e96 [FREE Full text] [CrossRef] [Medline]
    33. Sanatkar S, Baldwin PA, Huckvale K, Clarke J, Christensen H, Harvey S, et al. Using cluster analysis to explore engagement and e-attainment as emergent behavior in electronic mental health. J Med Internet Res 2019 Nov 28;21(11):e14728 [FREE Full text] [CrossRef] [Medline]
    34. Perski O, Blandford A, Garnett C, Crane D, West R, Michie S. A self-report measure of engagement with digital behavior change interventions (DBCIs): development and psychometric evaluation of the "DBCI Engagement Scale". Transl Behav Med 2020 Feb 03;10(1):267-277. [CrossRef] [Medline]
    35. Abd-Alrazaq AA, Bewick BM, Farragher T, Gardner P. Factors that affect the use of electronic personal health records among patients: a systematic review. Int J Med Inform 2019 Jun;126:164-175. [CrossRef] [Medline]
    36. Kruse CS, Argueta DA, Lopez L, Nair A. Patient and provider attitudes toward the use of patient portals for the management of chronic disease: a systematic review. J Med Internet Res 2015;17(2):e40 [FREE Full text] [CrossRef] [Medline]
    37. Rahman QA, Janmohamed T, Pirbaglou M, Ritvo P, Heffernan JM, Clarke H, et al. Patterns of user engagement with the mobile app, Manage My Pain: results of a data mining investigation. JMIR Mhealth Uhealth 2017 Jul 12;5(7):e96 [FREE Full text] [CrossRef] [Medline]
    38. Stragier J, Vandewiele G, Coppens P, Ongenae F, Van den Broeck W, De Turck F, et al. Data mining in the development of mobile health apps: assessing in-app navigation through Markov chain analysis. J Med Internet Res 2019 Jun 07;21(6):e11934 [FREE Full text] [CrossRef] [Medline]
    39. Borges J, Levene M. Evaluating variable-length Markov chain models for analysis of user web navigation sessions. IEEE Trans Knowl Data Eng 2007 Apr;19(4):441-452. [CrossRef]
    40. Singer P, Helic D, Taraghi B, Strohmaier M. Detecting memory and structure in human navigation patterns using Markov chain models of varying order. PLoS One 2014 Jul 11;9(7):e102070 [FREE Full text] [CrossRef] [Medline]
    41. Puspitasari I, Moriyama K, Fukui K, Numao M. Effects of individual health topic familiarity on activity patterns during health information searches. JMIR Med Inform 2015 Mar 17;3(1):e16 [FREE Full text] [CrossRef] [Medline]
    42. Fackrell M. Modelling healthcare systems with phase-type distributions. Health Care Manag Sci 2009 Mar 7;12(1):11-26. [CrossRef] [Medline]
    43. Carreras M, Ibern P, Coderch J, Sánchez I, Inoriza JM. Estimating lifetime healthcare costs with morbidity data. BMC Health Serv Res 2013 Oct 25;13:440 [FREE Full text] [CrossRef] [Medline]
    44. Duan H, Wang Z, Ji Y, Ma L, Liu F, Chi M, et al. Using goal-directed design to create a mobile health app to improve patient compliance with hypertension self-management: development and deployment. JMIR Mhealth Uhealth 2020 Feb 25;8(2):e14466 [FREE Full text] [CrossRef] [Medline]
    45. Wang Z, Li C, Huang W, Chen Y, Li Y, Huang L, et al. Effectiveness of a pathway-driven eHealth-based integrated care model (PEICM) for community-based hypertension management in China: study protocol for a randomized controlled trial. Trials 2021 Jan 22;22(1):81 [FREE Full text] [CrossRef] [Medline]
    46. Heyse J, Vandewiele G, Ongenae F, De Backere F, De Turck F. How do users interact with mobile health apps? A Markov chain analysis. Proc 14th EAI Int Conf Pervasive Comput Technol Healthc 2020:255-258. [CrossRef]
    47. Stragier J, Vandewiele G, Coppens P, Ongenae F, Van den Broeck W, De Turck F, et al. Data mining in the development of mobile health apps: assessing in-app navigation through Markov chain analysis. J Med Internet Res 2019 Jun 07;21(6):e11934 [FREE Full text] [CrossRef] [Medline]
    48. Tao D, Xie L, Wang T, Wang T. A meta-analysis of the use of electronic reminders for patient adherence to medication in chronic disease care. J Telemed Telecare 2015 Jan;21(1):3-13. [CrossRef] [Medline]
    49. Dell-Kuster S, Sanjuan E, Todorov A, Weber H, Heberer M, Rosenthal R. Designing questionnaires: healthcare survey to compare two different response scales. BMC Med Res Methodol 2014 Aug 03;14:96 [FREE Full text] [CrossRef] [Medline]
    50. Alessa T, Hawley MS, Hock ES, de Witte L. Smartphone apps to support self-management of hypertension: review and content analysis. JMIR Mhealth Uhealth 2019 May 28;7(5):e13645 [FREE Full text] [CrossRef] [Medline]
    51. Hever J, Cronise R. Plant-based nutrition for healthcare professionals: implementing diet as a primary modality in the prevention and treatment of chronic disease. J Geriatr Cardiol 2017 May;14(5):355-368 [FREE Full text] [CrossRef] [Medline]
    52. Mamud S, Bandyopadhyay S, Chatterjee P, Bhandari S, Chakraborty N. A Body Sensor Network for Automatic Dietary Monitoring System. Singapore: Springer; 2021.
    53. Wang Z, Huang H, Cui L, Chen J, An J, Duan H, et al. Using natural language processing techniques to provide personalized educational materials for chronic disease patients in China: development and assessment of a knowledge-based health recommender system. JMIR Med Inform 2020 Apr 23;8(4):e17642 [FREE Full text] [CrossRef] [Medline]
    54. Miller AS, Cafazzo JA, Seto E. A game plan: gamification design principles in mHealth applications for chronic disease management. Health Informatics J 2014 Jul 1;22(2):184-193. [CrossRef] [Medline]
    55. Michie S, Wood CE, Johnston M, Abraham C, Francis JJ, Hardeman W. Behaviour change techniques: the development and evaluation of a taxonomic method for reporting and describing behaviour change interventions (a suite of five studies involving consensus methods, randomised controlled trials and analysis of qualitative data). Health Technol Assess 2015 Nov;19(99):1-188 [FREE Full text] [CrossRef] [Medline]
    56. Zhang X, Guo X, Lai K, Guo F, Li C. Understanding gender differences in m-health adoption: a modified theory of reasoned action model. Telemed J E Health 2014 Jan;20(1):39-46. [CrossRef] [Medline]
    57. Khatun F, Heywood AE, Hanifi SMA, Rahman MS, Ray PK, Liaw S, et al. Gender differentials in readiness and use of mHealth services in a rural area of Bangladesh. BMC Health Serv Res 2017 Aug 18;17(1):573 [FREE Full text] [CrossRef] [Medline]
    58. Al-Ansari JM, Honkala S. Gender differences in oral health knowledge and behavior of the health science college students in Kuwait. J Allied Health 2007;36(1):41-46. [Medline]
    59. Mackenbach JP, Stirbu I, Roskam AR, Schaap MM, Menvielle G, Leinsalu M, et al. Socioeconomic inequalities in health in 22 European countries. N Engl J Med 2008 Jun 5;358(23):2468-2481. [CrossRef] [Medline]
    60. Kunst AE, Bos V, Lahelma E, Bartley M, Lissau I, Regidor E, et al. Trends in socioeconomic inequalities in self-assessed health in 10 European countries. Int J Epidemiol 2005 Apr;34(2):295-305. [CrossRef] [Medline]
    61. van Der Heide I, Wang J, Droomers M, Spreeuwenberg P, Rademakers J, Uiters E. The relationship between health, education, and health literacy: results from the Dutch Adult Literacy and Life Skills Survey. J Health Commun 2013;18 Suppl 1:172-184 [FREE Full text] [CrossRef] [Medline]
    62. Lee SD, Tsai T, Tsai Y, Kuo KN. Health literacy, health status, and healthcare utilization of Taiwanese adults: results from a national survey. BMC Public Health 2010 Oct 16;10:614 [FREE Full text] [CrossRef] [Medline]
    63. Nutbeam D. The evolving concept of health literacy. Soc Sci Med 2008 Dec;67(12):2072-2078. [CrossRef] [Medline]

    ‎
    ANOVA: analysis of variance
    BP: blood pressure
    DBP: diastolic blood pressure
    HR: heart rate
    KL-T: knowledge learning trail
    mHealth: mobile health
    PT-T: perform task trail
    RO-T: result-oriented trail
    SA-T: support acquisition trail
    SBP: systolic blood pressure
    TP value: transition probability value

    Edited by L Buis; submitted 27.08.21; peer-reviewed by L Min, S Nan; comments to author 11.10.21; revised version received 21.10.21; accepted 20.12.21; published 03.02.22

    Copyright

    ©Dan Wu, Xiaoyuan Huyan, Yutong She, Junbin Hu, Huilong Duan, Ning Deng. Originally published in JMIR mHealth and uHealth (https://mhealth.jmir.org), 03.02.2022.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.