Sage Journals: Discover world-class research

Abstract

Background

Mobile health (mHealth) is considered an effective way to manage chronic disease patients’ health. However, patients often do not use or continue using mHealth services due to concerns about service quality. Patient satisfaction may influence their utilization, and this effect may be moderated by platform regulations and differences in physicians’ qualifications and regions.

Objective

This study aims to enhance mHealth service utilization by examining how patient satisfaction affects both use and continuous use behaviors. It also explores the role of platform regulations (recommendation, commenting, and reward systems) and assesses differences due to physician proficiency and regional variations.

Methods

Data were collected from Haodf, a leading mHealth platform in China, from October 2021 to March 2022 at 3-month intervals. Generalized additive models were used to verify nonlinear relationships between patient satisfaction and their utilization behaviors. The role of platform regulations and heterogeneity analysis regarding physician region was also examined.

Results

Efficacy satisfaction positively influenced mHealth service utilization (p < 0.01), while efficacy and attitude satisfaction positively influenced continuous use (p < 0.01). Platform regulations altered these relationships (p < 0.01). Comprehensive recommendations had a greater effect than patient recommendations (p < 0.01). High-quality evaluations played a significant role (p < 0.01), while nonmaterial rewards had no significant effect (p ≥ 0.05). These relationships varied by physician region (p < 0.01).

Conclusions

Efficacy satisfaction drives mHealth service utilization, while efficacy and attitude satisfaction drive continuous use. Recommendation and commenting systems exhibit a positive synergistic effect, while the reward system plays a supplementary role. Professional recommendations and high-quality evaluations enhance the evaluation system's reliability, promoting service utilization. Patients in big cities are more sensitive to physicians’ proficiency and therapeutic effects. A multidimensional, transparent evaluation and feedback system should be integrated into the platform to improve service quality and increase patient utilization and loyalty.

Keywords

mHealth patient satisfaction use behavior generalized additive models chronic disease management

Introduction

Chronic diseases are typified as persistence, association with multimorbidity, and potential for prevention and control. Driven by rapid urbanization, aging demographics, and lifestyle changes, the incidence and prevalence of chronic diseases in China are continuing to rise. According to the Report on Nutrition and Chronic Disease Status of Chinese Residents (2020), over 20% of the population currently lives with chronic diseases, with hypertension, diabetes, hypercholesterolemia, and chronic obstructive pulmonary disease (COPD) being the most prevalent types. An increasing number of chronic disease patients manifest as multimorbidity—approximately 30% of chronic disease patients suffer from two or more chronic conditions simultaneously, a proportion that increases to 70% among seniors aged 60 and above.¹ These diseases place significant strain on the healthcare system, accounting for 80% of deaths and 70% of the national disease burden.² Adequate screening, diagnosis, and treatment, as well as proper rehabilitation management of chronic diseases help for achieving the goal of preventing and controlling chronic diseases.³ Traditional in-person health facility-based models of chronic disease management face challenges in meeting the continuous and intensive needs of patients due to spatial constraints, potentially limiting service accessibility.⁴

Mobile Health is considered more effective than traditional methods in managing chronic diseases due to its ability to overcome spatial barriers, facilitate behavioral changes, and improve patient health outcomes. Thakkar et al. revealed that mHealth interventions significantly enhance medication adherence in chronic disease patients compared to conventional interventions.⁵ Furthermore, Fortmann et al., through a randomized controlled trial, demonstrated that mHealth-based diabetes self-management and support systems more effectively addressed patient needs and markedly improved glycemic control relative to usual care.⁶ However, patients tend to be hesitant to use mHealth services due to a variety of reasons including knowledge literacy, psychological concerns, and economic considerations. Deng et al., through a survey conducted in three large hospitals in China, identified that trust and perceived risks, in addition to technical factors, are critical determinants of patients’ willingness to use mHealth service.⁷ Liu et al. further found that key barriers to users’ willingness to pay for mHealth services include the belief that the government should provide free mHealth, distrust in mHealth platforms, and low digital health literacy.⁸ Thus, even when patients seek mHealth services, they prefer occasional rather than continuous use,^9,10 which does not effectively make use of the potential of mHealth services to help manage chronic diseases.^11,12

The aforementioned reasons can be attributed to the varying service quality of mHealth platforms.¹³ User evaluations serve as the primary means for patients to evaluate the quality of mHealth services.¹⁴ Among various types of patient's evaluations, patient satisfaction reflects the users’ feelings, attitudes, and expectations and can objectively reflect the quality of mHealth services,^15,16 which is an intuitive criterion to help patients make decisions on whether to utilize the mHealth services or not.¹⁷ The quality of modern health care and medical services can be classified from the perspective of technical quality (e.g., safety, effectiveness, efficiency) and functional quality (patient-centered, accessibility, equity).¹⁸ Technical quality reflects the outcomes of health care and medical services and functional quality reflects its processes.¹⁹ Correspondingly, patient satisfaction includes both efficacy satisfaction and attitude satisfaction.²⁰ For patients with chronic disease who require lifelong care, this means they will focus on physician expertise²¹ as well as their attitude when seeking health care and medical services.²² This tells us that it is not only the efficacy satisfaction of patients who have already used the mHealth services is an important factor influencing patients’ use and continuous use of mHealth services but also that patients’ attitude satisfaction may play an equally important role.¹⁴ In practice, mHealth platforms usually also include recommendation system,^23,24 commenting system,^25,26 and reward system^16,21 to maximize the functionality of user evaluations in facilitating patients’ use and continuous use of mHealth services. These three systems can influence physician behavior and motivate physician efforts, which enhances patient satisfaction and increases the possibility of patients’ continuous use of mHealth services.^27,28 Previous studies have also shown that there are differences in the impact of recommendation system, commenting system, and reward system on the relationship between patient satisfaction and patients’ use and continuous use of mHealth services.^25,29,30 In addition, given the reality that China's high-quality medical resources are insufficiently allocated and unevenly distributed among regions, patient satisfaction, and the above three systems may also differ in influencing patients’ decision to use mHealth services due to differences in physician occupation titles and medical development in different regions.³¹

To this end, we aimed to examine how patient satisfaction influences the utilization of mHealth services among patients diagnosed with chronic diseases in China from the following three perspectives. First, we sought to understand the dynamic trajectory of patient satisfaction (efficacy satisfaction and attitude satisfaction) on the use and continuous use of mHealth services by patients with chronic diseases. Second, we aimed to further analyze the role of institutional efforts such as recommendation system, commenting system, and reward system in changing the effect of patient satisfaction on their use and continuous use of mHealth services. Third, we sought to identify the differences among the relationship of patient satisfaction, institutional efforts, and use and continuous use of mHealth services by patients with chronic diseases in terms of differences in physician-level and regional development, to provide guidelines for the construction of mHealth services that meet the needs of both patients with chronic diseases and regional development.

Theory framework and hypotheses

Patient satisfaction encompasses both efficacy and attitude satisfaction and is a comprehensive evaluation of various dimensions of mHealth services, representing a subjective standard derived from patients’ perception, assessment, and comparison.³² It reflects the effectiveness of physician–patient interaction and influences patients’ utilization of mHealth services through both self-perception and external information guidance.³³ In practice, we aim for patients to not only adopt mHealth services but also continue use it over time. According to the Use and Satisfaction Theory, satisfactory diagnostic outcomes and service attitudes can effectively meet patients’ needs for emotional engagement with physicians, thereby fostering continued use of mHealth services.³⁴ Meanwhile, patient satisfaction serves as a dynamic indicator of real-time patient feedback following service receipt. This indicator can evolve over time due to individual differences in disease progression, perceptions of service attitudes, and understanding of diagnosis and treatment outcomes.³⁵ Furthermore, patients’ utilization behavior, influenced by psychological and environmental factors, does not follow a linear pattern but rather exhibits a nonlinear relationship with patient satisfaction.^36,37 Patient satisfaction is more likely to show a nonlinear relationship with the utilization of mHealth services. Thus, we hypothesized the following.

H1a: Patient efficacy satisfaction may have a nonlinear effect on patients’ utilization of mHealth services.

H1b: Patient attitude satisfaction may have a nonlinear effect on patients’ utilization of mHealth services.

Notably, the other three ways to provide feedback and obtain information from patients in existing mHealth platforms, recommendation system, commenting system, and reward system^24,25,27 may influence patients’ utilization of mHealth services by providing more reliable and comprehensive information than patient satisfaction. This because these three systems require patients to bear more costs of reputation, time, and money. A high recommendation rating reflects the patients’ recognition of a physicians’ professional and technical level to some extent, which may induce other patients to develop a positive medical care-seeking attitude.³⁸ The details and examples in patient comments may deeply influence a physicians’ reputation and regulate their online diagnosis and treatment services, and rewards can help to enhance physicians’ online contributions.³⁸ Patients’ positive medical care-seeking attitude and physicians’ positive online contribution, good reputation, and standardized diagnosis and treatment behavior can positively affect patient satisfaction and promote their use and continuous use of mHealth services.³⁹ Thus, we hypothesized the following.

H2a: The recommendation system may change the effect of patient satisfaction on patients’ utilization of mHealth services.

H2b: The commenting system may change the effect of patient satisfaction on patients’ utilization of mHealth services.

H2c: The reward system may change the effect of patient satisfaction on patients’ utilization of mHealth services.

Physician occupation titles and region to which a physician belongs have a significant impact on patients’ utilization of mHealth services.^38–40 Physicians’ occupation title represents their professional level and authority. Physicians with a higher-level title mean they may have more medical practice experience or a higher professional level of specialization. However, the siphon effect in more economically developed areas provides physicians with more sophisticated equipment and higher professional and technical levels. Therefore, patients may be more inclined to choose physicians with higher-level title and those working in economically developed regions when choosing mHealth services, thereby ignoring patient satisfaction and the influence of the three institutional efforts (recommendation system, commenting system, and reward system). For this reason, we proposed the following hypotheses.

H3a: The effect of patient satisfaction and institutional efforts on patients’ utilization choices regarding mHealth services may vary with physician occupation titles.

H3b: The effect of patient satisfaction and institutional efforts on patients’ utilization choices regarding mHealth services may vary with the region to which a physician belongs.

Methods

Study design

This is an observational, longitudinal study, which employs a retrospective panel dataset collected from the Haodf mHealth platform in China, spanning October 2021 to March 2022 at 3-month intervals for a total of four times. Haodf website is one of the leading mHealth platforms in China. There are more than 13,000 hospitals and over 890,303 physicians on this platform, along with 680 million physician–patient communications and 5,269,000 online evaluations from patients after mHealth platform visits.¹⁴

Data collection

Data from the homepage of the physicians who diagnose eight chronic diseases (lung cancer, liver cancer, hypertension, coronary heart disease, COPD, diabetes, gastric cancer, and asthma) with a prevalence rate of more than 80% of the whole population as the respondents in China were obtained from the Haodf website through Python. We collected data at 3-month intervals from October 2021 to March 2022, resulting in four collection points. Each physician's data and the patients they treated were matched into subdatasets, forming an initial longitudinal panel dataset comprising 14,269 subdatasets (each containing medical consultation data for a physician and their patients across four time points). To ensure data quality, we performed data cleaning by removing subdatasets with missing values and duplicate entries. Specifically, to maintain temporal consistency, we eliminated subdatasets with missing values in either independent or dependent variables at any time point, reducing the dataset to 3948 subdatasets. Additionally, we removed duplicate data for eight types of chronic diseases, further refining the dataset to 1370 subdatasets. The final dataset comprised 2578 subdatasets, of which 991 subdatasets included physicians who had opened follow-up channels. In this study, we examined the influence of prior patient satisfaction on subsequent alterations in patients’ utilization patterns of mHealth services. Therefore, the final dataset included data from 2578 physicians and corresponding information on 10,312 instances of mHealth service usage and 3964 instances of continuous mHealth service usage. Ethical permission of the study was approved by the Ethics Committee of Xi’an Jiaotong University (LLSBPJ-2024-WT-020). Data cleaning and processing were performed using Microsoft Excel (2019) and R (version 4.1.2) to guarantee the authenticity and objectivity of the data.

Variables

Dependent variables

Patients’ utilization of mHealth services refers to the patients’ use⁴¹ and continuous use⁴² of mHealth services. Patients’ use of mHealth services was assessed according to the change in the number of diagnosed patients (△NPD_t), which reflects the change in the number of patients consulting a physician one time. Patients’ continuous use of mHealth services was evaluated according to the change in the number of followed-up patients (△NPF_t), which indicated the change in the number of patients followed up by physicians.

Independent variables

Patient satisfaction was measured by patients’ efficacy satisfaction (ESt-1)^34,39 and attitude satisfaction (ASt-1)⁴³ with mHealth services. ESt-1 was calculated as each patient's satisfaction score on the technical and inquiry accuracy as well as the treatment effect for the physician. ASt-1 was measured as the score for the patient's affinity for the physician's service attitude during treatment, including whether the length of the consultation was appropriate, and whether the patient's questions were answered in a timely manner.

To analyze the effects of the recommendation system, commenting system, and reward system, we included three types of variables, based on previous studies.^16,44–48 The first category involved indicators that reflect the recommendation system, that is, the degree of comprehensive recommendation (DCRt-1) and the degree of patient recommendation (DPRt-1).^23,24 The score for DCRt-1 ranged from 0 to 5, which is calculated by the platform combined with the information integrity of the physician, whether the physician opens the follow-up channel, the physician's online efforts, and other information according to the internal evaluation criteria of the platform. DPRt-1 is a continuous variable, calculated as the number of patients who are willing to recommend a physician after the physician's consultation or diagnosis and treatment. The second category involved indicators that reflect the commenting system, that is, the number of comments (NCRt-1), the praise rate in comments (PCRt-1), and the presence of high-quality comments (HQRt-1).^25,26 NCRt-1 refers to the total number of comments received by a physician, and PCRt-1 is the number of comments containing favorable ratings as a percentage of the total number of comments for that physician. HQRt-1 are judged using an indicator construction methodology proposed by Park,⁴⁹ which measures the presence of high-quality comments for the physician in terms of four dimensions (relevance, reliability, accessibility, and adequacy). We used a dummy variable that was coded as 1 if the physician's comments contained a high-quality review and 0 otherwise. The third category was indicators reflecting the reward system, that is, the number of thank you letters (TLt-1) or virtual gifts (HGt-1).⁴⁰ Patients can select and purchase a variety of virtual gifts on mHealth platform to reward their physicians. These gifts are showcased on the physician's profile page, and the proceeds from these purchases are transferred in full to the respective physician through the mHealth platform. TLt-1 is the number of letters from patients who thanked their physician. HGt-1 is the number of additional virtual gifts from patients to their physicians, both are continuous variables as counted by the platform.

Control variables

We included physician occupation titles (PT_t−1) and region to which a physician belongs (CITY_t−1) as the control variables.^15,33 Given the political and economic centrality of Beijing and Shanghai, we used dummy variables coded as 1 if the physician belonged to a hospital located in Beijing or Shanghai and 0 otherwise. Physician titles of chief physician or associate chief physician were coded as 1; other titles were coded as 0 (Table 1).

Table 1.

Variable definitions and measurements.

Variables	Definitions	Measurements
Dependent variables
△NPD_t	Patients’ use of mHealth services	Continuous variable, the change in the number of diagnosed patients
△NPF_t	Patients’ continuous use of mHealth services	Continuous variable, the change in the number of followed-up patients
Independent variables
ES_t−1	Efficacy satisfaction	Continuous variable, each patient's satisfaction score on the technical and inquiry accuracy as well as the treatment effect for the physician
AS_t−1	Attitude satisfaction	Continuous variable, the score for the patient's affinity for the physician's service attitude during treatment
DCR_t−1	The degree of comprehensive recommendation	Continuous variable, calculated by the platform with physician's all information according to the internal evaluation criteria of the platform
DPR_t−1	The degree of patient recommendation	Continuous variable, the number of patients who are willing to recommend a physician after receiving mHealth services
NCR_t−1	The number of comments	Continuous variable, the total number of comments received by a physician
PCR_t−1	The praise rate of comments	Continuous variable, the number of comments containing favorable ratings as a percentage of the total number of comments for that physician
HQR_t−1	The presence of high-quality comments	A dummy variable, coded 1 if the physician's comments contained a high-quality review and 0 otherwise
TL_t−1	The number of thank you letters	Continuous variable, the number of thank-you letters delivered to the physician by patients
HG_t−1	The number of virtual gifts	Continuous variable, the number of additional virtual gifts from patients to their physicians
Control variables
PT_t−1	Physician occupation title	A dummy variable, coded 1 if physician title is chief physician or associate chief physician, and 0 otherwise
CITY_t−1	Region to which a physician belongs	A dummy variable, coded 1 hospitals located in Beijing and Shanghai and 0 otherwise

Statistical analysis

Data analysis was performed using R software (Version 4.1.2). Descriptive statistics are given as average value, median, and standard deviation. In order to test the hypotheses proposed in this study, we used generalized additive models (GAMs) to verify the nonlinear relationship between two types of patient satisfaction and patients’ utilization of mHealth services. Hastie and Tibshirani⁵⁰ first proposed the GAM in 1986. In the analysis, continuous variables were first treated in a nonparametric form, and meaningful variables from the single-factor analysis were included in the constructed logistic GAM in multivariate analysis, including model fitting, co-curve diagnosis, and interaction analysis.

The GAM is formulated as: $E (Y | X_{1}, X_{2}, \dots, X_{n}) = α + f_{1} (X_{1}) + f (X_{2}) + \dots + f_{n} (X_{n})$ (1)

where Y represents the dependent variable; the two dependent variables in this study satisfied the Gaussian distribution. $X_{1}, X_{2} \dots X_{12}$ represent 11 independent variables.

The model can be constructed as: $△ P R V_{t} = α + f_{1} (X_{1}) + f_{2} (X_{2}) + \dots + β_{11} X_{11} + β_{12} X_{12}$ (2) $△ P F U_{t} = α + f_{1} (X_{1}) + (X_{2}) + \dots + β_{11} X_{11} + β_{12} X_{12}$ (3)

Generalized additive models can be adjusted to the new model, partial $f_{j}$ can be a linear relationship function. $β_{j}$ represents the parametric part; $f_{j}$ represents the smoothing (nonparametric) function. The parametric part is equivalent to the linear term or generalized linear term, and the nonparametric smoothing part is the key part of the generalized summation model. Statistical significance was defined as p < 0.05.

Results

Descriptive statistics and single-factor analysis

As shown in Table 2, the descriptive statistics of our main variables included patients’ use and continuous use of mHealth services, two types of patient satisfaction (efficacy satisfaction and attitude satisfaction), and information on the three systems. Table 3 showed that the independent variables and covariates were significant (p < 0.01) for both patients’ use and continuous use of mHealth services, with estimated degrees of freedom for all variables greater than 1, indicating that efficacy satisfaction (p < 0.01; Figures 1Aa and 1Ba), attitude satisfaction (p < 0.01; Figures 1Ab and 1Bb), and each covariate had a nonlinear relationship with patients’ utilization of mHealth services. Thus, H1a and H1b were supported.

Table 2.

Summary statistics.

Variable name	Patients’ use of mHealth services (n = 10312)			Patients’ continuous use of mHealth services (n = 3964)
Variable name	Average value	Median	Standard deviation	Average value	Median	Standard deviation
△NPD_t	886.6	377	1520.097	-	-	-
△NPF_t	-	-	-	1287.263	731.5	1811. 11
ES_t−1	0.994	1	0.024	0.995	1	0.013
AS_t−1	0.995	1	0.023	0.996	1	0.011
DCR_t−1	3.783	3.7	0.485	4. 115	4.1	0.470
DPR_t−1	37.849	26	39.495	38.047	27	38.938
NCR_t−1	39.981	32	26.319	40.566	33	25.891
PCR_t−1	0.913	0.92	0.074	0.908	0.917	0.082
TL_t−1	78.278	38	124.836	128.233	71	172.438
HG_t−1	161.837	54	350.406	281. 139	117	495.795
HQR_t−1	0.740	1	0.436	0.730	1	0.443
PT_{t −1}	0.800	1	0.397	0.850	1	0.352
CITY_{t −1}	0.390	0	0.489	0.390	0	0.487

Table 3.

Main results of single-factor analysis.

Category	Patients’ use of mHealth services				Patients’ continuous use of mHealth services
Parameters	Estimate std.	Error	t-value	p value	Estimate std.	Error	t-value	p value
PT_t−1	532.910	37.320	14.280	0.000**	691.260	80.890	8.545	0.000**
CITY_t−1	−283.380	30.520	−9.286	0.000**	−197.540	58.970	−3.350	0.001**
HQR_t−1	17.540	34.330	0.511	0.610	127.900	64.970	1.969	0.0490*
Smoothing effect part	Edf	Ref.df	F-value	p value	Edf	Ref.df	F-value	p value
s(ES_t−1)	8.917	8.997	149.200	0.000**	7.755	8.432	93.250	0.000**
s(AS_t−1)	8.905	8.997	98.440	0.000**	8.017	8.677	89.820	0.000**
s(DCR_t−1)	7.356	7.930	600.700	0.000**	7.807	8.570	187.600	0.000**
s(DPR_t−1)	8.218	8.813	8.518	0.000**	8.395	8.881	7.740	0.000**
s(NCR_t−1)	2.902	3.700	9.053	0.000**	8.825	8.982	5.521	0.000**
s(PCR_t−1)	1.904	1.991	7.516	0.000**	7.347	7.747	9.936	0.000**
s(TL_t−1)	8.967	9.000	929.800	0.000**	8.936	8.999	386.600	0.000**
s(HG_t−1)	8.748	8.977	1099.000	0.000**	8.502	8.915	493.300	0.000**

Note: *<0.05; **<0.01.

Edf: estimated degrees of freedom; Ref.df: reference degree of freedom.

Figures 1A and 1B showed the estimated smoothing components for patients’ use and continuous use of mHealth services with their corresponding 95% confidence intervals, respectively. In addition to two types of patient satisfaction (p < 0.01), the degree of comprehensive recommendation (p < 0.01), the degree of patient recommendation (p < 0.01), the number of comments (p < 0.01), the praise rate of comments (p < 0.01), the number of thank you letters (p < 0.01), and the number of virtual gifts (p < 0.01) were statistically significant and nonlinearly associated with patient use behavior (Figures 1Ac, 1Ae–1Ah) and patient continuous use behavior (Figures 1Bc, 1Be–1Bh), and the overall impact was significantly positive. However, there were significant differences in patient use behavior, which varied with physicians’ occupation title (p < 0.01) and region to which the physician belonged (p < 0.01). There were also significant differences in patient continuous use behavior, which varied with physicians’ occupation title (p < 0.01), region to which the physician belongs (p < 0.01), and the presence of high-quality comments (p = 0.049).

Figure 1.

The effect diagram of single-factor analysis. (A) Estimated smoothness of single-factor (patients’ efficacy satisfaction (ESt-1), patients’ attitude satisfaction (ASt-1), the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), the number of comments (NCRt-1), the praise rate in comments (PCRt-1), the number of thank you letters (TLt-1), and the number of virtual gifts (HGt-1)) on patient use behavior. (B) Estimated smoothness of single-factor (patients’ efficacy satisfaction (ESt-1), patients’ attitude satisfaction (ASt-1), the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), the number of comments (NCRt-1), the praise rate in comments (PCRt-1), the number of thank you letters (TLt-1), and the number of virtual gifts (HGt-1)) on patient continuous use behavior. The ordinate represents the smoothing function value, the number in the bracket represents the estimated degree of freedom EDF, and the dotted line represents the upper and lower limits of the confidence interval.

Multifactor analysis

The rates of the model for patient's use of mHealth services and the model for patient's continuous use of mHealth services in this study reached 60% and 63%. Table 4 (Figure 2) showed that efficacy satisfaction (p < 0.01; Figure 2Aa) had a significant fluctuating influence on patient use behavior, whereas efficacy satisfaction (p < 0.01, Figure 2Ba) and attitude satisfaction (p < 0.01, Figure 2Bb) had a significant and positive influence on patient continuous use behavior. Especially, when attitude satisfaction was at a high level (0.96–0.99), it had the greatest effect on patient continuous use behavior. The results showed that the degree of comprehensive recommendation (p < 0.01; Figure 2Ab), the degree of patient recommendation (p < 0.01; Figure 2Ac), the praise rate of comments (p < 0.01; Figure 2Ad), and virtual gifts (p < 0.01; Figure 2Ae) had a significant positive influence on patient use behavior. Similarly, the degree of comprehensive recommendation (p < 0.01; Figure 2Bc), the degree of patient recommendation (p < 0.01; Figure 2Bd), the number of comments (p < 0.01; Figure 2Be), the praise rate of comments (p < 0.01; Figure 2Bf), and virtual gifts (p < 0.01; Figure 2Bg) had a significant positive influence on patient continuous use behavior. Among them, the praise rate of comments showed a U-shaped influence on patient use behavior and continuous use behavior; when it reached 50%, it had a positive influence. In addition, patient use behavior and continuous use behavior were only significantly different due to the difference between offline hospital locations (p < 0.01). Thus, H3b was supported.

Figure 2.

The effect diagram of each factor with patient use behavior and continuous use behavior. (A) Estimated smoothness of five variables (patients’ efficacy satisfaction (ESt-1), the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), the number of comments (NCRt-1), the praise rate in comments (PCRt-1), and the number of virtual gifts (HGt-1)) on patient use behavior. (B) Estimated smoothness of seven variables (patients’ efficacy satisfaction (ESt-1), patients’ attitude satisfaction (ASt-1), the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), the number of comments (NCRt-1), the praise rate in comments (PCRt-1), and the number of virtual gifts (HGt-1)) on patient continuous use behavior. The ordinate represents the smoothing function value, the number in the bracket represents the estimated degree of freedom EDF, and the dotted line represents the upper and lower limits of the confidence interval.

Table 4.

Results for the model refitting hypothesis test.

Category	Patients’ use of mHealth services		Patients’ continuous use of mHealth services
Parameters	Estimate std.	p value	Estimate std.	p value
Intercept	1029.180	0.000**	1516.940	0.000**
PT_t−1	−25.970	0.311	−93.780	0.318
CITY_t−1	−281.560	0.000**	−335.240	0.000**
HQR_t−1	-	-	−53.510	0.086
Smoothing effect part	Edf	p value	Edf	p value
s(ES_t−1)	8.651	0.000**	2.801	0.000**
s(AS_t−1)	1.000	0.660	5.877	0.000**
s(DCR_t−1)	6.932	0.000**	4.517	0.000**
s(DPR_t−1)	8.973	0.000**	8.299	0.000**
s(NCR_t−1)	3.702	0.109	8.679	0.000**
s(PCR_t−1)	1.933	0.000**	1.616	0.027**
s(HG_t−1)^a	8.725	0.000**	8.434	0.000**

Note: *<0.05; **<0.01.

Edf: estimated degrees of freedom.

^aDue to the high concavity of thank you letter and hearty gift, in view of the fact that hearty gift not only express gratitude but also reflect additional effort, this study chooses to eliminate thank you letter when re-fitting the model.

Interaction analysis

Table 5 summarizes the results of the interaction analysis. The degree of comprehensive recommendation (p < 0.01), the degree of patient recommendation (p < 0.01), and virtual gifts (p < 0.01) significantly enhanced the positive effect of efficacy satisfaction in two models of patients’ use and continuous use of mHealth services. Thus, H2a and H2c were supported. The degree of comprehensive recommendation (p < 0.01), the degree of patient recommendation (p < 0.01), the number of comments (p < 0.01), the praise rate of comments (p < 0.01), and virtual gifts (p < 0.01) had a significant fluctuating interaction with attitude satisfaction in the model of patients’ continuous use of mHealth services. Thus, H2b was supported.

Table 5.

Results of interaction analysis to test the hypothesis.

Category	Patients’ use of mHealth services		Patients’ continuous use of mHealth services
Parameters	Estimate std.	p value	Estimate std.	p value
Intercept	976.50	0.000**	1472.21	0.000**
PT_t−1	−52.15	0.019*	−41.160	0.442
CITY_t−1	−249.6	0.000**	−354.010	0.000**
HQR_t−1	-	-	−105.410	0.009*

Smoothing effect part	Edf	p value	Edf	p value
ti (ES_t−1)	2.525	0.000**	1.339	0.044*
ti(AS_t−1)	3.771	0.362	3.373	0.029*
ti(DCR_t−1)	1.000	0.000**	2.631	0.000**
ti(DPR_t−1)	4.000	0.000**	3.805	0.000**
ti(NCR_t−1)	3.627	0.319	3.817	0.000**
ti(PCR_t−1)	1.925	0.000**	1.000	0.137
ti(HG_t−1)	1.000	0.000**	3.942	0.000**
ti(ES_t−1, AS_t−1)	1.000	0.376	1.000	0.483
ti(ES_t−1, DCR_t−1)	5.633	0.000**	7.927	0.004**
ti(ES_t−1, DPR_t−1)	11.996	0.000**	4.254	0.000**
ti(ES_t−1, NCR_t−1)	1.718	0.102	6.215	0.866
ti(ES_t−1, PCR_t−1)	1.529	0.107	1.681	0.258
ti(ES_t−1, HG_t−1)	9.497	0.000**	4.003	0.000**
ti(AS_t−1, DCR_t−1)	10.935	0.000**	7.034	0.000**
ti(AS_t−1, DPR_t−1)	3.315	0.006**	7.982	0.000**
ti(AS_t−1, NCR_t−1)	0.831	0.137	3.696	0.000**
ti(AS_t−1, PCR_t−1)	1.000	0.226	4.000	0.000**
ti(AS_t−1, HG_t−1)	14.343	0.000**	1.116	0.000**

Note: *<0.05; **<0.01.

Edf: estimated degrees of freedom.

To illustrate the interaction effects more clearly, an interaction diagram is shown in Figure 3. The degree of comprehensive recommendation (Figure 3Aa) and virtual gifts (Figure 3Ac) significantly enhanced the positive effect of efficacy satisfaction on patients’ use of mHealth services. When the efficacy satisfaction reached 0.7–0.9, the degree of its influence increased with an increased degree of comprehensive recommendation. An extremely high degree of patient recommendation (Figure 3Ab) increased the negative effect on patients’ use of mHealth services when efficacy satisfaction was low. The degree of comprehensive recommendation (Figure 3Ba) and virtual gifts (Figure 3Bc) significantly enhanced the positive effect of efficacy satisfaction on patients’ continuous use of mHealth services. The degree of patient recommendation (Figure 3Bb) showed a U-shaped influence of first, inhibition and then promotion. Even if the efficacy satisfaction was close to very satisfactory, as long as the degree of patient recommendation was close to 0, it showed a strong inhibitory effect. The degree of comprehensive recommendation (Figure 3Bd), the degree of patient recommendation (Figure 3Be), and the number of comments (Figure 3Bf) significantly enhanced the positive effect of attitude satisfaction on patients’ continuous use of mHealth services. The degree of influence increased with increases in those variants. The praise rate of comments (Figure 3Bg) showed a U-shaped effect of first, inhibition and then promotion. Even if the attitude satisfaction was close to very satisfactory, the praise rate of comments showed a strong inhibitory influence when it was lower than 50%, but the degree of inhibition was lower than that of promotion. In addition, the promotion influence of virtual gifts (Figure 3Bh) was only significant when the patient had a higher attitude satisfaction (0.85–0.90).

Figure 3.

Interaction effect diagram of efficacy satisfaction and attitude satisfaction with each factor. (A) Three-dimensional effect graph of patients’ efficacy satisfaction (ESt-1) interacting with the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), and the number of virtual gifts (HGt-1) influencing on patient use behavior, respectively. (B) Figure3Ba-Figure3Bc: Three-dimensional effect graph of patients’ efficacy satisfaction (ESt-1) interacting with the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), and the number of virtual gifts (HGt-1) influencing on patient continuous use behavior, respectively. Figure3Ba-Figure3Bc: Three-dimensional effect graph of patients’ attitude satisfaction (ASt-1) interacting with the degree of comprehensive recommendation (DCRt-1), the degree of patient recommendation (DPRt-1), the number of comments (NCRt-1), the praise rate in comments (PCRt-1), and the number of virtual gifts (HGt-1) influencing on patient continuous use behavior, respectively.

Heterogeneity analysis

Table 6 summarizes the results of a heterogeneity analysis regarding the region to which physicians belong. In Beijing and Shanghai, the effects of efficacy satisfaction and attitude satisfaction on patients’ utilization of mHealth services were consistent with the main analysis. However, patient use behavior in other regions was significantly influenced by both efficacy satisfaction (p < 0.001) and attitude satisfaction (p < 0.001). Unlike the whole country, the degree of patient recommendation in Beijing and Shanghai promoted the positive effect of efficacy satisfaction on patient use behavior, but the degree of influence was gradually weakened. Also, the positive effect of attitude satisfaction on patient continuous use behavior was more affected by the degree of patient recommendation and the praise rate of comments. The higher the degree of patient recommendation, the higher the number of comments, and the higher the praise rate of comments, the greater the influence. In other regions, the praise rate of comments also significantly improved the positive influence of efficacy satisfaction and attitude satisfaction on patient continuous use behavior rather than showing a U-shaped relationship. Other results were consistent with the main analysis.

Table 6.

Results of heterogeneity analysis (region to which a physician belongs).

Category	Beijing and Shanghai								Other cities
Category	Patients’ use of mHealth services				Patients’ continuous use of mHealth services				Patients’ use of mHealth services				Patients’ continuous use of mHealth services
Parameters	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value	Estimate std.	p value
Intercept	714.740	0.000**	591.170	0.000**	1166.270	0.000**	1096.700	0.000**	996.900	0.000**	954.810	0.000**	1365.100	0.000**	1308.480	0.000**

Smoothing effect part	Edf	p value	Edf	p value	Edf	p value	Edf	p value	Edf	p value	Edf	p value	Edf	p value	Edf	p value
s/ ti(ES_t−1)	7.993	0.000**	3.982	0.000**	3.476	0.00267	3.198	0.001**	6.870	0.000**	1.907	0.663	2.211	0.000**	1.000	0.000**
s/ti(AS_t−1)	1.000	0.250	2.815	0.065	3.737	0.00697	1.000	0.746	8.598	0.000**	1.000	0.000**	6.413	0.000**	1.000	0.026
s/ti(DCR_t−1)	1.917	0.000**	1.000	0.000**	4.401	0.000**	1.000	0.000**	1.984	0.000**	3.762	0.011	5.484	0.000**	1.000	0.050
s/ti(DPR_t−1)	8.473	0.000**	3.080	0.000**	8.177	0.000**	3.775	0.000**	7.739	0.000**	1.000	0.000**	8.634	0.000**	1.000	0.496
s/ti(NCR_t−1)	4.591	0.022	2.608	0.177	4.443	0.000**	3.748	0.003**	4.684	0.045	1.001	0.448	8.586	0.000**	3.675	0.013*
s/ti(PCR_t−1)	1.457	0.000**	3.884	0.007**	3.812	0.0113	3.706	0.014*	1.707	0.000**	4.000	0.000**	7.976	0.000**	4.000	0.000**
s/ti (HG_t−1)	8.861	0.000**	1.000	0.000**	8.77	0.000**	2.408	0.000**	8.921	0.000**	4.000	0.000**	8.967	0.000**	3.319	0.000**
ti(ES_t−1, AS_t−1)	-	-	1.177	0.774	-	-	2.522	0.361	-	-	2.245	0.120	-	-	1.000	0.627
ti(ES_t−1, DCR_t−1)	-	-	11.488	0.000**	-	-	4.285	0.577	-	-	13.330	0.000**	-	-	14.282	0.000**
ti(ES_t−1, DPR_t−1)	-	-	5.317	0.007**	-	-	1.562	0.979	-	-	8.208	0.094	-	-	9.741	0.000**
ti(ES_t−1, NCR_t−1)	-	-	5.272	0.230	-	-	2.634	0.000**	-	-	2.889	0.097	-	-	5.347	0.184
ti(ES_t−1, PCR_t−1)	-	-	4.616	0.000*	-	-	1.071	0.578	-	-	12.638	0.000**	-	-	5.381	0.000**
ti(ES_t−1, HG_t−1)	-	-	9.224	0.000**	-	-	3.452	0.138	-	-	15.001	0.000**	-	-	7.272	0.000**
ti(AS_t−1, DCR_t−1)	-	-	1.3266	0.201	-	-	0.601	0.102	-	-	6.140	0.000**	-	-	8.174	0.021*
ti(AS_t−1, DPR_t−1)	-	-	16.0000	0.000**	-	-	1.693	0.017*	-	-	14.229	0.001**	-	-	0.819	0.000**
ti(AS_t−1, NCR_t−1)	-	-	0.6341	0.173	-	-	8.363	0.000**	-	-	4.509	0.051*	-	-	10.029	0.000**
ti(AS_t−1, PCR_t−1)	-	-	0.6341	0.004**	-	-	8.603	0.000**	-	-	1.000	0.015*	-	-	4.671	0.008**
ti(AS_t−1, HG_t−1)	-	-	9.9173	0.000**	-	-	1.561	0.946	-	-	10.247	0.001**	-	-	4.728	0.000**

Note: *<0.05; **<0.01.

Edf: estimated degrees of freedom.

Discussion

It is known from the Information Asymmetry Theory and Social Information Processing Theory that knowledge-intensive medical information is likely to produce “information asymmetry” between physicians and patients,^42,51 which triggers behavior biased toward misleading patients seeking consultation, treatment, and medication service. Coincidentally, patient satisfaction can be used as a reference for patients’ judgment when the available information is insufficient.⁵² Thus, previous studies have shown that both efficacy satisfaction and attitude satisfaction positively influence patients’ utilization of mHealth services.^{14–16,21,31,39,40} However, it is known from Self-Determination Theory and Use and Satisfaction Theory that patient satisfaction with mHealth services reflects the degree of satisfaction according to patients’ own psychological needs after receiving mHealth services, which is one of the key factors contributing to their autonomy in pursuing better health outcomes.⁵³ This autonomy may determine patients’ utilization of mHealth services.³³ This suggests that the impact of patient satisfaction on decision-making behavior varies according to the distinct levels of psychological fulfillment patients experience following each medical encounter. Our study also revealed a wavy curve correlation between patient efficacy satisfaction and patients’ use of mHealth services, whereas there was no significant correlation between patient attitude satisfaction and patients’ use of mHealth services. This aspect reflects that the psychological satisfaction derived from the therapeutic outcomes is more significant for patients than the attitude of physician during a single medical visit. Also, because there are individual differences in the conditions (etiology and course of disease) of patients with chronic diseases, the treatment effect and psychological satisfaction on patients in a single visit may also be different, which accounts for the wave-shaped correlation between efficacy satisfaction and patients’ use of mHealth services.¹³

Meanwhile, we found that both efficacy satisfaction and attitude satisfaction had a significant and positive effect on patient continuous use behavior, with a greater effect when patient attitude satisfaction was at a higher level. This is because, it is an observable and reliable signal that individuals can obtain from the social environment in evaluating the quality of services, in the case of protecting patients’ privacy without access to their personal information, patients who have received mHealth services in the current period may be more objective in evaluating their satisfaction with mHealth services because they have no fears^33,54; further, patients may be more inclined to trust such evaluation information as the main basis for their own medical decision-making in the next period. This indicates that both physicians’ technical level and service attitude influence patients’ continuous use of mHealth services.⁵⁵ Patients are more likely to choose physicians with a good service attitude, high diagnostic accuracy, and adequate service.^39,56 Therefore, mHealth platforms should encourage physicians with a higher technical level to participate in mHealth services. Therefore, mHealth platforms should encourage physicians with a higher technical level to participate in mHealth services. Existing research demonstrates that while physicians generally exhibit favorable perceptions toward mHealth engagement, their effective participation remains constrained by limited awareness of mHealth and inadequate digital skill.⁵⁷ Existing research demonstrates that while physicians generally exhibit favorable perceptions toward mHealth engagement, their effective participation remains constrained by limited awareness of mHealth and inadequate digital skill. Government and medical institutions can provide physicians with modular online courses (such as the medical artificial intelligence [AI] course offered by Coursera), virtual simulation training (such as surgical simulators), and interactive training based on real cases (such as DiagnosUs) via low-bandwidth optimized MOOC platforms (such as WHO Academy) and AR remote guidance tools (such as Proximie). These digital education or training initiatives can be implemented to enhance physicians’ digital skill and improve physicians’ knowledge about mHealth services, thereby supporting broader participation by health professionals. Thank you again for your valuable suggestions.⁵⁸

We also revealed that recommendation system, commentary system, and reward system played different roles in influencing the two types of patient satisfaction in the use and continuous use of mHealth services by patients with chronic diseases. When patients decide to use mHealth services and choose a physician, they not only pay attention to past patients’ efficacy satisfaction with the physician's treatment but also take the degree of comprehensive recommendation, the praise rate in comments by patients who have already visited the physician, and virtual gifts received by the physician from patients as their basis for judgment.⁵⁹ Also, when patients decide to continue using mHealth services, they are concerned with past patients’ efficacy satisfaction and attitude satisfaction of the physician, as well as the degree of comprehensive recommendation, the degree of patient recommendation, the praise rate of comments by patients who have already visited the physician, and virtual gifts received by the physician from patients.²⁶ Among them, the degree of comprehensive recommendation and degree of patient recommendation are important elements in the recommendation system of mHealth platforms.^47,59 Our results showed that the degree of comprehensive recommendation has a high promoting effect on patients’ use of mHealth services when patients’ efficacy satisfaction was at a low to medium level, and it could continuously enhance the positive effects of efficacy satisfaction and attitude satisfaction on patient continuous use behavior. However, the degree of patient recommendation has a U-shaped effect on changing the correlation between efficacy satisfaction and patients’ use and continuous use of mHealth services. When both patient efficacy satisfaction and patient recommendation levels were at medium–high, it was beneficial for patient continuous use behavior. Otherwise, the patient recommendation level would act as a disincentive. The degree of comprehensive recommendation is considered objective information reflecting a physician's professional competence, and its objectivity and the amount of information contained in it can influence a patient's trust, thereby prompting the patient to try a new service even when they are less satisfied.^47,60 The degree of patient recommendation reflects the role of conformity in promoting patient visits, as the same disease experience promotes identity among patients who have not visited a physician and classifies them into the same group of patients who have previously visited the physician.⁴⁰ This cohort effect leads patients to be more willing to believe in and choose a physician with a high degree of patient recommendation, which has an important role in changing the patients’ efficacy satisfaction and affecting their utilization of mHealth services.⁶¹ Similarly, this role of the degree of patient recommendation was reflected in the influence of the praise rate of comments in altering the effect of efficacy satisfaction on patient use behavior and continuous use behavior with chronic diseases.^25,62 However, the influence of the number of comments was not significant. Compared with the number of comments, the role of the praise rate of comments is more important, which reflects improvement in the health literacy of patients, and also requires that commenting system no longer be limited to quantitative evaluation but these should further penetrate the evaluation content.^46,51 Some scholars have found the content of comments has an influence on patient perceptions, which in turn affects the utilization of mHealth services by patients.^24,48 The more information on diagnosis and treatment and services contained in a comment, the greater the effect on the utilization of mHealth services by patients with chronic diseases.⁶³ However, the high-quality comments in this study did not have a significant impact on patients’ utilization of mHealth services, suggesting that it is difficult for patients to generate professional perceptions of technology-intensive and knowledge-intensive medical services; more popular content comment methods are still needed to facilitate patient decision-making reference.²⁵ In addition, in the reward system, the number of gifts positively changed the effect of efficacy satisfaction and attitude satisfaction on patients’ use and continuous use of mHealth services. Research has demonstrated that reward system can increase a physician's online contribution, which can improve patient satisfaction and promote patients’ utilization of mHealth services.²³ In contrast, gifts are a kind of material reward that requires actual payment, whereas a thank you letter does not require payment and is more similar to a spiritual reward.⁴⁰ Thus, virtual gifts can more accurately reflect the patient's evaluation of a physician's diagnosis and treatment level.³⁹ We also found that the degree of comprehensive recommendation, the degree of patient recommendation, the praise rate of comments, and virtual gifts can continuously and incrementally improve the promotional influence of attitude satisfaction on patients’ continuous use of mHealth services. The service process is only an influential factor, not a major one, in the patients’ use of mHealth services, as reflected by this.

This study found that when patients used mHealth services from physicians in Beijing and Shanghai, they only focused on physicians’ treatment level rather than service attitude and paid more attention to the degree of comprehensive recommendation and degree of patient recommendation. The evaluation of past patients also played a more important role. When patients decided to use and continue to use mHealth services from physicians in other regions, they focused not only on the level of treatment but also on the attitude of physicians providing services. The recommendation system, commenting system, and reward system were⁶⁴ only used as auxiliary references for patient medical decision-making. The unbalanced allocation of medical resources in China, where high-quality medical resources are concentrated in large cities, is a contributing factor to this difference.³⁹ Patients often choose mHealth services provided by physicians in large cities with a certain purpose and are more concerned about whether the disease can be cured rather than health care experience.^31,65 Therefore, patients who use mHealth services provided by physicians in large cities tend to pay greater attention to the level of diagnosis and treatment rather than other service qualities, and they will pay more attention to evaluations of a physician's technical level by patients who have already been treated.

There are three strengths in this study. First, refine the research variables. Previous studies have predominantly treated patient satisfaction as a single variable to evaluate “service quality”^33–35 when determining its influence on patients’ tendency to use mHealth services.³⁶ These studies have overlooked patients’ evaluations of physicians’ professional skills and service attitudes, as well as the distinction between initial use and continuous use of mHealth services.³⁷ In contrast, our study differentiates patient utilization of mHealth services into initial use behavior and continuous use behavior and further categorizes patient satisfaction into efficacy satisfaction and attitude satisfaction. This approach reveals the heterogeneity in satisfaction-driven mechanisms across different decision-making stages, thereby providing a more nuanced analytical framework for understanding the utilization of mHealth services among patients diagnosed with chronic disease in China. Second, employ innovative methodology. Previous research has primarily identified a linear correlation between patient satisfaction and medical decision-making.³⁸ However, patient satisfaction is dynamic and can be easily influenced by daily encounters, suggesting a potential nonlinear relationship between patient satisfaction and medical decision-making. To address this, we employed GAMs,³⁹ which can accurately capture both linear and nonlinear relationships between input variables and response variables under relaxed assumptions, thereby better reflecting the dynamic nature of patient satisfaction and its impact on medical decision-making. Third, expand and deepen the research content. Prior studies have not adequately considered the institutional efforts of service providers in promoting patient stickiness or the differences in physician-level and regional development. Our study addresses these gaps by focusing on the needs of patients with chronic diseases and regional development requirements, providing a more comprehensive contribution to the advancement of mHealth services for chronic disease management.

We acknowledge that this study not only had some limitations but it also identifies key directions for future research. First, while the Haodf website is one of the leading mHealth platforms in China and its findings are representative to an extent, the generalizability of the results to broader populations requires further validation through data collection from other mHealth platform. Second, in our dataset, 73% of physicians registered on the Haodf platform were from tertiary hospitals, whereas 93% of all physicians in tertiary hospitals were represented. This discrepancy suggests a need for further analysis of how hospital-level heterogeneity affects the use and continuous use of mHealth services by patients with chronic diseases. Finally, we utilized GAMs, which are nonparametric models, to optimize the number of nodes for each variable using spline functions, thereby adjusting for the influence of various factors on mHealth service utilization among patients with chronic diseases. However, due to platform privacy policies, several potential influencing factors such as sociodemographic/economic characteristics and digital health literacy were not included in the analysis, potentially biasing the results. Future research should incorporate these variables to provide a more comprehensive understanding of factors influencing mHealth usage of patients.

Implications

This research holds significant policy and practical implications for the future development of mHealth systems. First, this study highlights patients’ emphasis on therapeutic efficacy when utilizing mHealth services, which underscores the need for policymakers and platform developers to prioritize treatment outcomes. Specifically, it can guide policymakers in designing an online therapeutic efficacy-centered service evaluation system, mandating platforms to disclose efficacy metrics (e.g., symptom relief rates, long-term health improvements) and integrate these into performance assessments. Additionally, it can motivate platforms to optimize their service design via chronic disease management modules, personalized tracking tools (e.g., medication reminders and health index monitoring), and feedback loops to incentivize quality care. Second, the findings of this study regarding the differential impacts of recommendation system, commenting system, and reward system on patient satisfaction and mHealth service utilization provide valuable evidence for enhancing platform operations mechanism. To optimize the design of the recommendation system, policies should enhance credibility through third-party verification of physicians’ qualifications (e.g., unifying the certification of physicians’ academic qualifications, and clinical experience) /treatment success rates, reducing overreliance on patient-provided labels. Standardizing the evaluation system to encompass multidimensional assessment criteria (e.g., patient recovery cases, physicians’ professional backgrounds, and treatment plan compatibility) to enhance the comprehensiveness and fairness of evaluation system. Implementing dynamic incentives (e.g., offering traffic recommendations or honorary certifications) to physicians who demonstrate significant therapeutic outcomes and high patient follow-up rates. Third, this research concludes that the inequality in medical resource distribution leads to patients’ tendency to blindly choose physicians with high professional titles or from economically developed regions. This finding is instrumental for policymakers in addressing this negative influence of uneven distribution of high-quality medical resources by fostering cross-regional collaborations (e.g., tertiary-grassroots hospital partnerships) and upskilling rural physicians through subsidies and training. Also, platforms should redesign physician profiles to highlight expertise (e.g., diabetic complication management) and case evidence, reducing patients’ overreliance on titles/regions.

Conclusion

mHealth provides access to quality health care services for patients with chronic diseases by breaking through constraints of time and location. Our study reveals that, first, efficacy satisfaction is the primary determinant of patients’ use of mHealth services. Both efficacy satisfaction and attitude satisfaction jointly drive patients’ continuous use behavior. Second, the recommendation system and commenting system can synergistically influence the impact of treatment efficacy assessment on patients’ initial and continued use of mHealth services. The reward system plays an auxiliary role in patient decision-making. Notably, a high comprehensive recommendation score can more effectively enhance patients’ initial experience with the service, while only higher patient recommendations will prompt individuals to continue using the service. Patients prioritize high-quality information as a criterion for evaluating service utilization. Material rewards have a significantly greater impact on both the initial and continued use of services compared to nonmaterial rewards. Third, in China, high-quality medical resources are concentrated in major cities, leading patients to prioritize disease cure over service experience when choosing physicians in these areas. In contrast, when selecting physicians in other regions, patients consider both technical ability and service attitude. In the future, it is imperative to integrate patient feedback to enhance the quality of mHealth services. By leveraging AI and big data analytics, we can continuously refine and optimize both the recommendation and evaluation systems. This will facilitate the development of a multidimensional evaluation framework focused on efficacy assessment and cumulative assessments, thereby encouraging more robust patient feedback on the platform. Additionally, by constructing a comprehensive patient feedback system, we can provide patients with more quantifiable and verifiable information, ultimately improving mHealth service quality and enhancing patient engagement and retention.

Footnotes

Acknowledgements

The authors would like to thank the Haodf website for providing the data.

ORCID iDs

Jiao Lu

Heng Zhao

Zhilin Ji

Yanan Dong

Ethical considerations

Ethical permission of the study was approved by the Ethics Committee of Xi’an Jiaotong University (LLSBPJ-2024-WT-020). The data were obtained from the Haodf website,which is one of the leading mHealth platforms in China.

Author contributions/CRediT

Jiao Lu,Heng Zhao,and Zhilin Ji contributed to the conceptualization and design of the study. Jiao Lu and Heng Zhao involved in collecting the data. Jiao Lu,Heng Zhao,and Zhilin Ji analyzed the data. All authors were involved in the interpretation of results. Jiao Lu and Zhilin Ji wrote the first manuscript draft. Jiao Lu and Heng Zhao engaged in reviewing and editing the manuscript. All authors have read the final version of the manuscript and approved its submission.

Funding

The authors disclosed receipt of the following financial support for the research,authorship,and/or publication of this article: This research was funded by National Natural Science Foundation of China (No.72474174),the Humanity and Social Science Foundation of Ministry of Education of China (No. 22YJA630059),and China Postdoctoral Science Foundation (No. 2022T150514). The funders had no role in the design,data collection,analysis,interpretation of the data,and writing of this work.

Conflicting interests

The authors declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

References

Xiang

Wang

. Comorbidity risk and distribution characteristics of chronic diseases in the elderly population in China. BMC Public Health 2024; 24: 60.

Shengli

Kaiyi

. Research on the status quo, problems and countermeasures of Chinese people’s health Popul Health 2021; 290: 46–48. in Chinese.

Zhao

, et al. Elements of chronic disease management service system: an empirical study from large hospitals in China. Sci Rep 2022; 12: 5693.

Negrini

de Sire

Andrenelli

, et al. International Multiprofessional Steering Committee of Cochrane Rehabilitation REH-COVER action. Rehabilitation and COVID-19: the Cochrane Rehabilitation 2020 rapid living systematic review. Update as of July 31st, 2020. Eur J Phys Rehabil Med 2020; 56: 652–657.

Thakkar

Kurup

Laba

, et al. Mobile telephone text messaging for medication adherence in chronic disease: a meta-analysis. JAMA Intern Med 2016; 176: 340–349.

Fortmann

Gallo

Garcia

, et al. Dulce digital: an mHealth SMS-based intervention improves glycemic control in Hispanics with type 2 diabetes. Diabetes Care 2017; 40: 1349–1355.

Deng

Hong

Ren

, et al. What predicts patients’ adoption intention toward mHealth services in China: empirical study. JMIR Mhealth Uhealth 2018; 6: e172.

Liu

Xie

. Willingness to pay for health apps, its sociodemographic correlates, and reasons for being unwilling to pay. Digit Health 2024; 10: 1–10.

Jian

Wen

, et al. Ping SQ, status quo and influence factors of patients’ presupposed mistrust to doctors. Guangdong Med J 2018; 39: 3520–3524.

10.

Liu

Lee

Sun

, et al. Hospital-Owned apps in Taiwan: nationwide survey. JMIR Mhealth Uhealth 2018; 6: 22.

11.

Sun

Wang

, et al. A Mobile-based intervention for glycemic control in patients with type 2 diabetes: retrospective, propensity score-matched cohort study. JMIR Mhealth Uhealth 2020; 8: e15390.

12.

Poowuttikul

Seth

. New concepts and technological resources in patient education and asthma self-management. Clin Rev Allergy Immunol 2020; 59: 19–37.

13.

Yang

Shu

Zhang

. Understanding “internet plus healthcare” in China: policy text analysis. J Med Internet Res 2021; 23: e23779.

14.

Yang

Song

. Exploring the impact of physicians’ word of mouth on patients’ selection in online health community——taking the website of www.haodf.com as an example. J Mod Inf 2019; 39: 20–28.

15.

Fang

Liu

Fang

. What is the most important factor affecting patient satisfaction - a study based on gamma coefficient. Patient Prefer Adherence 2019; 13: 515–525.

16.

Larrabee

Ostrow

Withrow

, et al. Predictors of patient satisfaction with inpatient hospital nursing care. Res Nurs Health 2004; 27: 254–268.

17.

Han

. Understanding the mobile healthcare applications continuance: the regulatory focus perspective. Int J Med Inform 2023; 177: 105161.

18.

Fiala

. What do patients want? Technical quality versus functional quality: a literature review for plastic surgeons. Aesthet Surg J 2012; 32: 751–759.

19.

Morisada

Hwang

Gill

, et al. Telemedicine, patient satisfaction, and chronic rhinosinusitis care in the era of COVID-19. Am J Rhinol Allergy 2021; 35: 494–499.

20.

JHY

Luk

BHK

. Patient satisfaction: concept analysis in the healthcare context. Patient Educ Couns 2019; 102: 790–796.

21.

Qiu

Zhang

Wang

, et al. Trust-based research: influencing factors of patients’ medical choice behavior in the online medical community. Healthcare (Basel) 2022; 10: 38.

22.

Huang

Wang

, et al. The present situation of chronic disease follow-up in medical institutions and the analysis of key problems. Chin Gen Pract 2020; 23: 3522–3526.

23.

Lai

Young

Kohrman

, et al. Tilting the scale: current provider perspectives and practices on breastfeeding with HIV in the United States. AIDS Patient Care STDS 2023; 37: 84–94.

24.

Yang

Qiao

, et al. Analyze to mediating effects of recommended popularity in online medical platform. J Capital Med Univ 2022; 43: 456–462.

25.

Guo

Fang

, et al. How doctors gain social and economic returns in online health-care communities: a professional capital perspective. J Manage Inf Syst 2017; 34: 487–519.

26.

Rui

. Can we trust online physician ratings? Evidence from cardiac surgeons in Florida. Manage Sci 2018; 64: 2473–2972.

27.

Moqbel

Hewitt

Nah

, et al. Sustaining patient portal continuous use intention and enhancing deep structure usage: cognitive dissonance effects of health professional encouragement and security concerns. Inf Syst Front 2022; 24: 1483–1496.

28.

Song

Deng

Cui

, et al. Measuring success of Patients’ continuous use of mobile health services for self-management of chronic conditions: model development and validation. J Med Internet Res 2021; 23: e26670.

29.

Xiao

Ting

, et al. Online review analysis-based multi-criteria decision-making for evaluating patient satisfaction: a case study of the Haodf website. J Oper Res Soc 2023; 75: 841–859.

30.

Lang

Yuh

Ajam

, et al. Understanding patient satisfaction ratings for radiology services. AJR Am J Roentgenol 2013; 201: 1190–1195.

31.

Raffaele

Elisabetta

Claudio

. When are extreme ratings more helpful? Empirical evidence on the moderating effects of review characteristics and product type. Comput Hum Behav 2018; 88: 134–142.

32.

Qing

Tong

, et al. Exploring the online doctor-patient interaction on patient satisfaction based on text mining and empirical analysis. Inf Process Manag 2020; 57: 102253.

33.

Sun

Chiu

, et al.

Can online health information sources really improve patient satisfaction?

Front Public Health 2022; 10: 940800.

34.

Liu

North

. Relationship building through reputation and tribalism on companies’ Facebook pages: a uses and gratifications approach. Internet Res 2017; 27: 1149–1169.

35.

Fortin

Cao

Fleury

. A typology of satisfaction with mental health services based on Andersen's behavioral model. Soc Psychiatry Psychiatr Epidemiol 2018; 53: 587–595.

36.

Inmaculada

Francisco

Jose

, et al. Upward voice: participative decision making, trust in leadership and safety climate matter. Sustainability 2020; 12: 3672.

37.

Gelareh

Atish

Mark

, et al. Trust decision-making in online social communities: a network-based model. J Comput Inf Syst 2022; 62: 153–163.

38.

Pan

Song

Zhang

. Dynamic recommendation of physician assortment with patient preference learning. IEEE Trans Autom Sci Eng 2019; 16: 115–126.

39.

Zhao

Luu

Spiegel

, et al. Correlation of online physician rating subscores and association with overall satisfaction: observational study of 212,933 providers. J Med Internet Res 2020; 22: e11258.

40.

Wang

Chiu

, et al. Understanding a nonlinear causal relationship between rewards and Physicians’ contributions in online health care communities: longitudinal study. J Med Internet Res 2017; 19: e427.

41.

Shen

Yang

, et al. Internet-Based medical service use and eudaimonic well-being of urban older adults: a peer support and technology acceptance model. Int J Environ Res Public Health 2021; 18: 12062.

42.

Min

Zhu

Liu

. Factors influencing continued usage behavior on mobile health applications. Healthcare (Basel) 2022; 10: 08.

43.

Gong

Kong

, et al. Factors associated with outpatient satisfaction in tertiary hospitals in China: a systematic review. Int J Environ Res Public Health 2020; 17: 7070.

44.

Heritage

Elliott

Stivers

, et al. Reducing inappropriate antibiotics prescribing: the role of online commentary on physical examination findings. Patient Educ Couns 2010; 81: 119–125.

45.

Knoepke

Allen

Ranney

, et al. Loaded questions: internet Commenters’ opinions on physician-patient firearm safety conversations. West J Emerg Med 2017; 18: 903–912.

46.

Oei

Rombouts

Soeter

, et al. Dopamine modulates reward system activity during subconscious processing of sexual stimuli. Neuropsychopharmacology 2012; 37: 1729–1737.

47.

Hyung

Jumin

Ingoo

. The effect of on-line consumer reviews on consumer purchasing intention: the moderating role of involvement. Int J Electron Commer 2007; 11: 125–148.

48.

Trevor

Tibshirani

. Generalized additive models. In: Balakrishnan

Colton

Everitt

, et al (eds) Wiley statsref: statistics reference online. American: John Wiley & Sons, Inc., 2014, pp.1–11.

49.

Zhou

Wang

Song

. Platform-based mechanisms, institutional trust, and continuous use intention: the moderating role of perceived effectiveness of sharing economy institutional mechanisms. Inf Manag 2021; 58: 103504.

50.

Noorbergen

Adam

MTP

Teubner

, et al. Using co-design in mobile health system development: a qualitative study with experts in co-design and mobile health system development. JMIR Mhealth Uhealth 2021; 9: e27896.

51.

Zhang

. Exploring the effects of health information seeking on e-satisfaction in online health communities: an empirical investigation. BMC Med Inform Decis Mak 2022; 22: 32.

52.

Brereton

Matlock

Fitzgerald

, et al. Content analysis of negative online reviews of hospice agencies in the United States. JAMA Netw Open 2020; 3: e1921130.

53.

Zhang

Chen

Wang

WYC

, et al. Understanding Physicians’ motivation to provide healthcare service online in the digital age. Int J Environ Res Public Health 2022; 19: 15135.

54.

Touati

Sailer

Marchand

, et al. Communication tools and patient satisfaction: a scoping review. J Esthet Restor Dent 2022; 34: 104–116.

55.

Muzammil

Ahmed

Zaidan

, et al. A security framework for mHealth apps on android platform. Comput Secur 2018; 75: 191–217.

56.

Gupta

Rodeghier

Lis

. Patient satisfaction with service quality as a predictor of survival outcomes in breast cancer. Support Care Cancer 2014; 22: 129–134.

57.

Sarradon-Eck

Bouchez

Auroy

, et al. Attitudes of general practitioners toward prescription of mobile health apps: qualitative study. JMIR Mhealth Uhealth 2021; 9: e21795.

58.

Bashir

Lalithabai

AlOtaiby

, et al. Health care professionals’ knowledge and attitudes toward telemedicine. Front Public Health 2023; 11: 957681.

59.

Ming

Wei

Hao

, et al. Consumers’ continuous use intention of O2O e-commerce platform on community: a value co-creation perspective. Sustainability 2022; 14: 1666.

60.

Mahyari

Pirolli

LeBlanc

. Real-time learning from an expert in deep recommendation systems with application to mHealth for physical exercises. IEEE J Biomed Health Inform 2022; 26: 4281–4290.

61.

Zhai

Song

Chen

, et al. A study of mobile medical app user satisfaction incorporating theme analysis and review sentiment tendencies. Int J Environ Res Public Health 2022; 19: 7466.

62.

Lee

Uhm

Cheong

, et al. Patient satisfaction with mobile health (mHealth) application for exercise intervention in breast cancer survivors. J Med Syst 2018; 42: 54.

63.

Mitchell

White

, et al. Uptake of an incentive-based mHealth app: process evaluation of the carrot rewards app. JMIR Mhealth Uhealth 2017; 5: 70.

64.

Hamberger

Ikonomi

Schwab

, et al. Interaction empowerment in mobile health: concepts, challenges, and perspectives. JMIR Mhealth Uhealth 2022; 10: e32696.

65.

Green

Huberty

Puzia

, et al. The effect of meditation and physical activity on the mental health impact of COVID-19-related stress and attention to news among Mobile app users in the United States: cross-sectional survey. JMIR Ment Health 2021; 8: e28479.

Effect of patient satisfaction on the utilization of mHealth services by patients with chronic disease

Abstract

Background

Objective

Methods

Results

Conclusions

Keywords

Introduction

Theory framework and hypotheses

Methods

Study design

Data collection

Variables

Dependent variables

Independent variables

Control variables

Statistical analysis

Results

Descriptive statistics and single-factor analysis

Multifactor analysis

Interaction analysis

Heterogeneity analysis

Discussion

Implications

Conclusion

Footnotes

Acknowledgements

ORCID iDs

Ethical considerations

Author contributions/CRediT

Funding

Conflicting interests

References