Willingness to Pay for Mobile Health Live Streaming during the COVID-19 Pandemic: Integrating TPB with Compatibility

Abstract

The COVID-19 pandemic has forced people to limit their physical interactions, which has led to unprecedented growth in mobile health live streaming (MHLS). Such practices have been facilitated by the rapid development of consumers’ willingness to pay for MHLS. However, few efforts appear in the literature to understand this change. This study aims to integrate the theory of planned behavior with compatibility to clarify payment motivations for MHLS during the COVID-19 pandemic in China. Accordingly, the current study used a web-based, self-reported questionnaire to collect data from 506 Chinese consumers. Of the 535 distributed questionnaires, we received 505 valid responses, yielding an effective rate of 94.3%. The valid responses were analyzed using structural equation modeling, and the associated hypotheses were tested using partial least squares regression. The results showed that attitude, self-efficacy, preferred lifestyle, information quality, and interactive immersion (but not subjective norms) significantly predicted consumers’ willngness to pay (WTP). In addition, attitude partially mediated the association between preferred lifestyle, information quality, and WTP, respectively, while the mediating role in the relationship between interactive immersion and WTP was not confirmed. These findings can be used to formulate effective marketing strategies to benefit MHLS services and mobile payment technology in the country.

1. Introduction

Information and communication technologies (ICTs) continue to progress at a phenomenal rate in China, spawning a robust growth of new health promotion methods using ICTs, such as mobile health live streaming (MHLS) services. As a platform for optimizing offline health, MHLS provides services such as synchronous physician-patient consultations and healthcare activities for consumers. Mobile health has exhibited clear advantages, including greater autonomy, fewer operational difficulties, and a reduced time commitment [1,2,3], as patients can easily reach medical professionals for health support without spatial or temporal constraints. Furthermore, mobile health has arguably not only enhanced the public’s ability to self-manage health but also raised their willingness to pay (WTP) for health via healthcare devices [4,5], particularly during the COVID-19 pandemic [6,7]. In China, MHLS is a new field that has been growing rapidly over the past decade, fueled by the increasing availability of data and computing power, as well as by the diversity of consumer health needs. Many mobile health platforms mushroomed during the pandemic, such as “Haodf” and “Guahao” to cope with public health issues, and they have exhibited a great commercial value for MHLS. According to an industry report released in 2022 by iResearch [8], 27.5% of live streamers are from the medical field or medical aesthetic field in China, ranking first in all application scenarios. The trend in market performance highlighted the importance of live streaming in the health services industry.

In response to the COVID-19 pandemic, China’s mobile health platforms use live streaming technology, which is embedded in mobile applications such as “Haodf” and “Guahao”. In addition, professionals also conduct mobile health live streaming on China’s social media and entertainment media, such as “Wechat” and “Tiktok”. This technology aims to proactively prevent and control pandemics, deliver health knowledge synchronously, and strategically promote public health literacy [9]. Viewers in MHLS tend not to consume health services initially unless they have undergone several pre-defined processes, such as free consulting and service-to-need matching [10]. Thus, this phenomenon naturally introduces the question of what factors drive consumers’ WTP for MHLS. WTP measures consumers’ willingness to pay for services, which is a non-persistent personal characteristic influenced by applications, scenarios, reviews, and other social characteristics [11]. Until recently, a small but expanding body of work has begun to understand the factors influencing consumer adoption of new health promotion using ICTs [12]. However, facing the health uncertainties associated with public crises, MHLS received little attention regarding their evolution. Few studies have addressed the behavioral preferences for MHLS among health knowledge seekers and how these preferences affect their willingness to pay for MHLS.

A careful examination of WTP can reveal the mechanisms behind consumers’ payment for MHLS and also serve mobile health service providers to develop marketing strategies, as well as benefit the public by improving health literacy in the post-pandemic era. Therefore, to achieve this goal, the current study aims to propose a research model based on the theory of planned behavior (TPB) and compatibility to investigate the WTP for MHLS. First, TPB has robust predictive power and is thus widely used to predict various human behaviors in different contexts [13]. Although it has been used to validate consumers’ WTP for mobile health [5,14], research on its role as a systematic framework to explore MHLS using ICTs is still vacant [15]. Therefore, this study introduced the TPB model as a grounded theory to understand consumer WTP for MHLS. In addition to being influenced by COVID-19, the success of MHLS may also be explained by various compatibilities. Compatibility related to health technology is fundamental in the physician-patient relationship and plays a central role in consumers’ intentions to seek information [16]. In this sense, compatibility provides an opportunity to develop new factors related to new health promotion using ICTs and has the potential to understand consumers’ WTP for MHLS based on valuable engagement behaviors. Therefore, the current study attempts to address the following questions:

• Is TPB applicable to the MHLS environment?
• Can compatibility be applied to MHLS environments?
• How does compatibility integrate into the TPB when interpreting consumers’ WTP in an MHLS environment?

To address these research questions, the current study used the TPB framework to provide a holistic perspective to examine the WTP for MHLS. In the TPB model, individuals’ behavioral intentions are influenced by three related factors: attitude stemming from behavioral evaluations, subjective norms stemming from social pressures, and perceived behavioral control (PBC) stemming from individuals’ past experiences and expectations [13]. Existing literature on mobile health confirms the significant impact of attitude, subjective norms, and PBC on behavioral intention [14,17]. Since scholars doubt whether TPB, as a general behavioral model, is sufficient to predict specific behaviors, it is necessary to develop TPB further by including structure in specific contexts [18,19]. Based on this, TPB was modified in the current study by replacing PBC with SE, which refers to an individual’s belief that he or she can succeed, i.e., "I can do it" [20]. SE reflects one’s subjective psychological feelings [21]. In contrast to SE, PBC includes both objective environmental and subjective psychological conditions that have an impact on individuals’ beliefs [22], which is not conducive to a detailed explanation of the mechanisms that determine WTP for MHLS. To further extend TPB, we also integrated compatibility with it.

The success of MHLS services may be explained by their various compatibilities. Compatibility with MHLS technology is fundamental in physician-patient relationships, and it plays a central role in viewers’ intentions to seek information [16]. Compatibility is manifested in two aspects: value compatibility and practical compatibility [23]. As the name suggests, value compatibility reflects the extent to which technology consumers’ existing values (e.g., preferred lifestyles) align with the perception of innovation [14,20,24]. Efforts were also made to understand practical compatibility, which describes the suitability of innovation to one’s behavior [25,26]. In the mHealth context, practical compatibility is related to the quality of the information, such as accurate health guidance [27] and the interactive immersion brought by images, voice, and video [28,29,30]. Not surprisingly, live streaming platforms perfectly meet individuals’ health needs and increase individuals’ willingness to try or consume health services [31]. Furthermore, when consumers are satisfied with IT compatibility, they are expected to develop a more committed attitude toward it; when consumers find it too much effort to use IT, they consider it not worth using and thus develop a negative attitude [32,33]. Given this, attitudes are considered a mediator of the relationship between compatibility and attitudes toward purchasing MHLS.

Specifically, the objectives of this study were to:

• incorporate compatibility into TPB to form an integrated framework for explaining the formation of WTP for MHLS.
• examine the structures that influence the process of attitude formation.
• examine the mediating role of attitude in the mechanism of WTP for MHLS.

2. Literature Review

This section presents a review of TPB and compatibility in new health promotion using ICTs. Specifically, the literature on TPB is divided into three subsections: attitude, subjective norms, and self-efficacy; the literature on compatibility is divided into three subsections: preferred lifestyle, information quality, and interactive immersion.

2.1. TPB in MHLS

TPB, the abbreviation of the theory of planned behavior, originally developed in the field of social psychology, has been a popular theoretical framework to explain behavioral intentions by attitude, subjective norms, and perceived behavioral control [13]. The theory has been applied to ICTs such as e-commerce [34,35], social media [36,37], and mobile payment [38,39]. The framework is also found to be applicable to predicting live streaming sales and shopping intentions [40,41]. In particular, TPB has been used to verify consumers’ WTP for mobile health [5,14]. Some studies have examined the utility of TPB in live streaming [40,42,43,44]. Therefore, TPB was used as a foundation for understanding the consumer’s willingness to pay for MHLS in our study [15].

Despite the widespread use of the TPB in mobile health and other health-related services [45], existing criticisms of this model point out that there is a need to include additional variables to improve its predictive and explanatory power further. Some researchers questioned the lack of value considerations and thus started to add “materialism” to the TPB [46]. Several studies suggested replacing the perceived behavioral control construct with an internal motivation variable, self-efficacy, to improve the predictive ability of the TPB [47]. Collectively, these mixed results suggest that research should assess the joint effects of other additional motivation variables, instead of raw variables, on health behavior [13,15]. Similarly, perceived behavioral control is substituted by self-efficacy in this study, representing individuals’ internal motivation to purchase MHLS services.

2.1.1. Attitude (AT)

In the initial TPB model, AT is the degree to which an individual holds positive or negative beliefs about the consequences of behavior [48]. It is an implicit response that occurs within the individual and affects subsequent overt responses. In this study, AT refers to individuals’ feelings or emotions that influence their intention to use MHLS technology. It measures the utility of MHLS technology from the perspective of individual cognition. AT has been found to be a significant determinant of consumers’ WTP for new health promotion using ICTs. For example, by analyzing an international message from a developing country, Jaber et al. [49] found that patients’ AT has a strong impact on WTP for pharmaceutical care. During a lockdown, MHLS can strengthen the promotion of health knowledge by initiating live medical interactions and enabling the public to exercise simultaneously at home with intelligent applications [9,50]. This encourages the public to use MHLS technology. Traditional clinical practices are time-consuming and geography-restricted, but MHLS can reduce the redundant work involved in offline medical interaction by communicating health information synchronously online. Most importantly, consumers can be exposed to easy-to-understand health knowledge by engaging in live physician-patient interactions. This can create a sense of attraction for consumers, which might motivate them to develop a more positive AT toward MHLS and, in turn, lead to a greater WTP.

2.1.2. Subjective Norms (SN)

According to the initial TPB model, SN are one’s perceptions of how particular referents explain the behavior and extrinsic inspiration gained from the expectations and beliefs of these referents [48]. In this study, SN refers to responses to the MHLS payment behaviors of important referents (acquaintances or social groups) and the extent to which individuals will consider these responses. SNs reflect the social pressure that consumers face in making payment decisions for MHLS services. SNs are also an indispensable factor affecting willingness to pay. For example, Bettiga et al. [5] conducted an empirical study on individuals’ adoption of smart technologies for preventive health care, and the result shows a direct effect of SN on willingness to pay. In the MHLS context, SN usually affect an individual’s willingness to pay for services through two mechanisms: online viewer review [51] and streamer recommendations [52]. The online review mechanism reflects “after-sales feedback” from online patients rating the experience and effectiveness of the health program as well as the attitude and professionalism of its providers, which in turn creates a social impetus to change potential consumers’ willingness to pay [53]. And to reduce the information asymmetry between patients and services, providers typically push a ranking of streamer recommendations to the patients, which are calculated by a big data algorithm based on keywords in a search engine input by the patients. With the ease of free access to streamer ranking mechanisms, the public gradually internalized and embraced the normative social pressure of new health promotion using ICTs [54].

2.1.3. Self-efficacy (SE)

SE reflects the degree to which consumers perceive their ability and motivation to perform the desired behavior [21]. In this study, SE is defined as the extent to which people perceive ease or difficulty in performing a behavior when using MHLS. It reflects the perceived beliefs that accumulate inside consumers during their use of MHLS, such as confidence, readiness, reluctance, and rejection. SE also directly affects a consumer’s WTP. For example, in the live streaming scenario, Gong et al. [2] found that SE positively affected student participants’ sustainable impulse buying. The self-efficacy perceived by online patients in ICT-based remote health promotion technologies consists of two main components: the efficacy expectancy of physician-patient matching and the outcome expectancy of medical guidance [26,55]. First, the efficacy expectancy of physician-patient matching, which is the personal conviction that one succeeds in getting the medical help he/she needs; and second, the outcome expectancy of medical guidance, which is the belief that medical guidance will lead to a patient expectation for mobile health outcomes [21]. The aforementioned mechanism reflects a consumer’s expectation that health care will meet their needs. Especially in today’s COVID pandemic, medical resources are being extended online, and healthcare programs arising from technology applications have an important impact on individuals’ purchase intentions for health. Based on this, our study suggests that consumers’ SE might psychologically affect their WTP for MHLS.

2.2. Compatibility

Compatibility is perceived as the intensity of innovation that fits the current needs, values, and prior experiences of its potential adopters [56]. In the live streaming field, compatibility has been recognized as one of the basic prerequisites for consumers to adopt new technologies or applications [57]. Likewise, Clement et al. [58] revealed that it is an influential predictor of what attitudes people hold toward live streaming. Nevertheless, most existing ICT-based literature treats compatibility as a single-dimensional variant, and no sub-category related to a specific technology (e.g., mHealth) is established. In this study, compatibility is defined as the intensity with which MHLS technologies adapt to consumers’ values (preferred lifestyles) and practical use (information quality and interactive immersion). Specifically, from a value compatibility perspective, Belanche et al. [59] discovered that preferred lifestyle is the typical trait of value compatibility, and Wu et al. [60] further proposed that preferred lifestyle is an important reason why MHLS is superior to face-to-face health interventions. From a practical compatibility perspective, Wood et al. [61] found that information quality compatibility significantly facilitates consumers’ actual use of mobile health technologies. Xiong et al. [62] and Rajak et al. [63] also indicated that interactive system-enabled phones could effectively improve the effectiveness of MHLS interventions. As the above argumentation is generally applicable to ICTs, they facilitate our study to identify the driving role of compatibility in the context of MHLS.

2.2.1. Preferred Lifestyle (PL)

“Lifestyle” is a psychological term that describes the ways in which individuals behave in their everyday lives [64]. Lifestyle varies between individuals, and a unique lifestyle type is affected by several factors, such as ethnic affiliation, religious status, family background, temperament, and moral perception [65]. In the context of MHLS, PL is a consumer’s interests and opinions related to the purchase intention of MHLS products/services. Researchers indicate that PL is an important characteristic of consumers [66] and can be considered compatibility because PL allows for the rapid evolution of technology without detracting from the value that people hold for determining adoption intention. For example, Chen et al. [67] identified that the live streaming purchase intention of experience products is significantly affected by lifestyle indicators. In a study of mobile social media advertising, Zhang et al. [68] conducted an online questionnaire survey of WeChat consumers and found that presenting consumers’ mobile lifestyles increased their purchasing intentions. Therefore, we expect MHLS consumers to make payment decisions based on their dietary lifestyle or work pace. In addition, several studies suggested, however, that in the context of consumer behavior, this effect is mediated by AT. Specifically, Lee [69] investigated various types of anti-consumption lifestyles and found the relationship between lifestyle and purchase intention was significantly mediated by the AT toward commercial sharing systems. A study of the farming system in the UK also found that lifestyle’s influence on WTP is mediated by AT [70].

2.2.2. Information Quality (IQ)

IQ has become a critical concern in live streaming research, which refers primarily to the usefulness of information [71]. IQ’s significant influence on AT has been clearly illustrated by studies in the context of live streaming [72,73]. As Handayani et al. [74] put it, IQ in mobile health is characterized by sufficient and relevant information, especially those that can be matched to the patient’s physical file. To meet the needs of each individual’s health accurately, MHLS technology requires patients to fill in basic information, such as their disease symptoms, service type, price range, and preferred physician title. With the help of big data, MHLS could screen out qualified doctors in the database based on the information provided by the patient. Therefore, the subsequent matching results (e.g., medical guidance) will be a crucial indicator for the patient to evaluate AT toward MHLS. Previous research on live streaming commerce also pointed out that the more informative or practically useful the IQ is perceived to be, the more positive the WTP [75]. Considering the similarity between MHLS and live streaming commerce (both platforms feature a synchronous virtual consumer technology orientation), we expect the influence of IQ on WTP is consistent with prior research. If IQ in MHLS is compatible with the consumer’s health needs, the consumer’s purchase intention will increase as they will receive more accurate responses to illnesses and access to remedies similar to those ones they would receive in an offline consultation. As Bonnie [76] noted, the pandemic created a high demand for quality information about healthcare resources, which in turn provides unparalleled opportunities for physicians and policymakers to promote health technologies.

2.2.3. Interactive Immersion (II)

Immersion is conceptualized as the cognitive state of being enveloped by an interactive environment, such as a continuous stream of stimuli, activities, or events [77]. Immersion has been understood to be an essential antecedent factor affecting WTP in the context of live streaming [78] and online tourists [79]. The degree of perceived immersion of consumers reflects the breadth and depth of information sharing [80], which helps identify the values and benefits they can derive from a visual scene [81,82]. In addition, -values and benefits can facilitate consumer immersion in the usage environment created by mobile technology, which in turn will increase the consumer desire to purchase [83]. In MHLS, immersion is often followed by patient-physician interaction (e.g., audio, chats, and video) [84]. Concretely, when a patient feels absorbed by mHealth, he/she may think that offline care is being delivered to his/her location. This mental immersion creates an illusory interactive experience in which the patient feels that the imagined “seeing public doctors in outpatient clinics” is present in the physical environment. Accordingly, we argue that II will enhance consumers’ AT toward MHLS, similar to previous studies in e-commerce [85] and social media [86]. The COVID-19 pandemic disrupted the traditional health preferences of health seekers, and emerging technologies such as artificial intelligence, big data, and virtual reality brought about a new interactive experience. Collectively, these factors are driving immersion out of many health preferences [87].

3. Methodology

To effectively obtain the data for testing, this section focuses on clarifying the research model and hypotheses, construct measures, data collection and analysis methods.

3.1. Research Model and Hypotheses

A hypothesis model was constructed using exogenous variables such as preferred life-style (PL), information quality (IQ), interactive immersion (II), self-efficacy (SE), sub-jective norm (SN); endogenous variables of willngness to pay (WTP); and mediating variables of attitude (AT) (see Figure 1). The hypotheses are coded as follows:

Hypothesis 1 (H1). 

AT positively affects consumers’ WTP for MHLS.

Hypothesis 2 (H2). 

SN positively affects consumers’ WTP for MHLS.

Hypothesis 3 (H3). 

SE positively affects consumers’ WTP for MHLS.

Hypothesis 4 (H4). 

PL positively affects consumers’ WTP for MHLS.

Hypothesis 4a (H4a). 

The influence of PL on WTP for MHLS is mediated by AT.

Hypothesis 5 (H5). 

IQ positively affects consumers’ WTP for MHLS.

Hypothesis 5a (H5a). 

The influence of IQ on WTP for MHLS is mediated by AT.

Hypothesis 6 (H6). 

II positively affects consumers’ WTP for MHLS.

Hypothesis 6a (H6a). 

The influence of II on WTP for MHLS is mediated by AT.

3.2. Construct Measures

To investigate Chinese consumers’ WTP for mHealth, we designed a questionnaire with three sections: The first section collected participant demographic information, such as gender, age, educational level, and monthly household income; the second section described mobile live streaming payment background, including mobile live streaming experience and disease type. The third section reported participants’ current mHealth payment usage regarding AT, SE, SN, WTP, PL, IQ, and II. All items in this study were adapted from the construct validated in extant research and measured on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree). The questionnaire was also modified based on participants’ recommendations to correspond with the mHealth context. To maintain consistency with the main references, the questionnaire was initially developed in English. A professional translator translated it into the Chinese version, and then another expert translated the Chinese version back into English. We compared the two English versions to ensure that the Chinese questionnaire is in line with the original English questionnaire. A pilot test was also conducted among a group of 50 mHealth payment consumers. The final questionnaire items are indicated in Appendix A 

Table A1. Questionnaire items.

Variables	Items	Sources
Preferred lifestyle (PL)	PL1: MHLS is in line with my own personal health value PL2: MHLS fits the way I view health PL3: MHLS is consistent with the way I to keep healthy	Dhir et al. [122]
Information quality (IQ)	IQ1: MHLS provides appropriate information IQ2: MHLS provides important information IQ3: MHLS provides effective information	Kim & Lee [123]
Interactive immersion (II)	II1: I feel absorbed when using MHLS II2: I feel focused when using MHLS II3: I get a good interactive experience when using MHLS	Song et al. [124]
Self-efficacy (SE)	SE1: I can use MHLS if there is no one around to assist me SE2: I can use MHLS if the technology has never been similar to me before SE3: I believe I can use mobile technology effectively	Reychav et al. [125]
Subjective norms (SN)	SN1: people who matter to me prefer that I use MHLS SN2: People who affect me prefer that I use MHLS SN3: People who value me prefer that I use MHLS	Ajzen & Fishbein [126]
Attitude (AT)	AT1: I think it is wise to use MHLS AT2: I think it is a pleasant experience to use MHLS AT3: I am satisfied with MHLS	Deng et al. [127]
Willingness to pay (WTP)	WTP1: Assuming I have access to MHLS, I intend to pay more for it WTP2: Given that I have access to MHLS, I predict that I will pay more for it WTP3: If I have access to pay for MHLS, I want to pay as much as possible	Bettiga et al. [5]

.

3.3. Data Collection

Due to the pandemic, the questionnaire was released through Wenjuanxing, a web-based survey platform in China. People with mobile health living streaming payment experiences between January and April 2022 were recruited for this study. Two authors of our study are trained investigators who sent a website link to the questionnaire to viewers of mobile health live streaming. Initially, the questionnaire was sent to 50 respondents, who had to complete it before they would receive a gift worth $1. However, before the investigators formally sent the questionnaire link to potential respondents, respondents were asked to provide the name of the mobile health and living application used in their most recent purchase and to report the specific usage year with their payment channel. We excluded respondents whose recent usage year was not 2022 and who were unable to confirm payment details. The data were collected over the course of one month in May 2022. A total of 535 valid responses to the questionnaire were received. After examining the data, a final sample of 506 valid responses was yielded by the questionnaire, with an effective rate of 94.3%. To examine our model, we conducted the two-step approach introduced by Anderson and Gerbing [88], which first tests the measurement model and then tests the structural model.

3.4. Data Analysis Methods

First, structural equation modeling (SEM) analysis was performed to assess the impact of various drivers on WTP for MHLS. The SEM approach was chosen because it is considered a sophisticated next-generation analysis method that allows researchers to model the relationships between multiple independent and dependent variables simultaneously [89]. SEM is appropriate for this study due to the presence of complex mediating mechanisms in our model. Second, our analysis used the PLS-SEM method to create the research model and determine the relationship between factors and WTP. PLS-SEM is a well-known analytical technology and a widely used analytical method in the social sciences [90]. PLS-SEM is also a validated non-experimental research technique with a wide statistical range for evaluating measurement and structural models [91]. Therefore, this method was used in this study to analyze the data obtained from the MHLS industry. The computational software and its version number accompanying the method used in this paper are SmartPLS 3.0 (SmartPLS v3.3.3. https://www.smartpls.com/, accessed on 3 September 2021). SmartPLS 3.0 software was used in this study for its ability to process complex model analysis without strict data assumptions, such as residual distribution and a large sample [89].

4. Results

To examine our model, we conducted the two-step approach introduced by Anderson and Gerbing [88], which first tests the measurement model and then tests the structural model. SmartPLS 3.0 software was used in this study for its ability to process complex model analysis without strict data assumptions, such as residual distribution and a large sample.

4.1. Respondents’ Background

Of the 506 respondents, 53.2% of consumers were male and 46.8% were female. Regarding age groups, respondents between 31 and 40 years old (39.6%) dominate the largest group, followed by respondents between 41 and under 50 years old (32.0%), over 50 years old (11.9%), and under 20 years old (12.5%). Among all respondents, 55.1% were more concerned about chronic diseases, and 44.9% were more concerned about acute diseases. Most individuals reported monthly household income between $1200 and $1700 (40.7%), followed by $700 to $1200 per month (25.7%), more than $1700 per month (15.8%), and less than $700 per month (6.3%). Regarding the education level, the majority had an undergraduate degree (52.8%), followed by the technical school (19.2%), a high school certificate or below (15.6%), and a master’s or higher degree (12.4%).

Table 1. Demographics of respondents (n = 506).

Characteristics	Frequency	Percent (%)
Gender
Male	269	53.2%
Female	237	46.8%
Age
20s	63	12.5%
30s	150	39.6%
40s	162	32.0%
50s	131	15.9%
Mean age = 41.49 years old
Education level
High school certificate or below	79	15.6%
Technical school	97	19.2%
Undergraduate degree	267	52.8%
Master or higher degree	63	12.4%
Monthly household income, $
<700	32	6.3%
(700, 1200]	188	37.2%
(1200, 1700]	206	40.7%
>1700	80	15.8%
Mobile live streaming experience
Under 1 year	56	11.0%
1–3 years	225	44.5%
Over 3 years	225	44.5%
Disease type
Chronic disease	279	55.1%
Acute disease	227	44.9%

demonstrates the demographic characteristics of the respondents.

4.2. Measurement Model Analysis

The measurement model results include the analysis of internal reliability, convergent validity, and discriminant validity. First, internal reliability was evaluated by assessing Cronbach’s α and composite reliability. After the calculation, all Cronbach’s α and composite reliability values were well above the minimum threshold of 0.70, which ensures the internal consistency of the constructs [92].

We then tested for validity by assessing the average variance extracted (AVE), the square roots of AVE, item loadings, and cross-loadings. AVE directly reveals the fraction of the variance captured by the construct that comes from the measurement error, and its ideal value should be greater than 0.5 [93]. The square roots of AVE are larger than the correlation coefficients, implying good discriminant validity between the constructs. In addition, we found that factor loadings that loaded well on their construct were significantly larger than the cross-loadings that loaded poorly on other constructs, indicating that all variables in this study had good convergent and discriminant validity. See

Table 2. Assessment of the reliability and validity.

Variables	Items	Ladings	Cronbach’s α	CR	Rho_A	AVE
Preferred lifestyle (PL)	PL1	0.875	0.844	0.906	0.846	0.763
	PL2	0.838
	PL3	0.858
Information quality (IQ)	IQ1	0.867	0.838	0.903	0.849	0.755
	IQ2	0.844
	IQ3	0.896
Interactive immersion (II)	II1	0.879	0.904	0.940	0.910	0.839
	II2	0.858
	II3	0.856
Self-efficacy (SE)	SE1	0.879	0.817	0.891	0.831	0.731
	SE2	0.873
	SE3	0.868
Subjective norms (SN)	SN1	0.858	0.832	0.899	0.838	0.747
	SN2	0.823
	SN3	0.883
Attitude (AT)	AT1	0.921	0.854	0.911	0.854	0.774
	AT2	0.932
	AT3	0.895
Willingness to pay (WTP)	WTP1	0.884	0.820	0.893	0.822	0.735
	WTP2	0.889
	WTP3	0.865

Note: composite reliability, CR; average variance extracted, AVE.

and

Table 3. Discriminant validity: Fornell-Larcker criterion.

Items	(1) PL	(2) IQ	(3) II	(4) SE	(5) SN	(6) AT	(7) WTP
(1) PL	0.873
(2) IQ	0.465	0.869
(3) II	0.537	0.495	0.916
(4) SE	0.526	0.456	0.351	0.855
(5) SN	0.468	0.350	0.475	0.482	0.865
(6) AT	0.530	0.579	0.438	0.540	0.528	0.880
(7) WTP	0.607	0.534	0.545	0.519	0.476	0.590	0.857

Note: the square roots of AVE are along the diagonal; correlations are below the diagonal.

for detailed measurement information.

In addition, the cross-loading criterion was also applied to assess discriminant validity [94]. It aims to evaluate the loading on the constructs and confirm that it performs better in each row, which implies that the loading value of the construct in the item must be greater than the loading of the other items. By calculation, we found that factor loadings that loaded well on their structures were significantly larger than cross-loadings that loaded poorly on other structures. As shown in

Table 4. Item loadings and cross-loadings.

Items	(1) PL	(2) IQ	(3) II	(4) SE	(5) SN	(6) AT	(7) WTP
PL1	0.875	0.498	0.424	0.576	0.473	0.451	0.519
PL2	0.838	0.385	0.405	0.486	0.427	0.468	0.498
PL3	0.858	0.485	0.394	0.496	0.432	0.484	0.501
IQ1	0.408	0.867	0.259	0.391	0.416	0.384	0.490
IQ2	0.473	0.844	0.226	0.353	0.347	0.415	0.430
IQ3	0.505	0.896	0.409	0.459	0.423	0.484	0.576
II1	0.444	0.290	0.879	0.402	0.412	0.478	0.440
II2	0.364	0.300	0.858	0.404	0.395	0.364	0.465
II3	0.418	0.319	0.856	0.409	0.441	0.379	0.466
SE1	0.568	0.416	0.426	0.879	0.464	0.449	0.454
SE2	0.491	0.431	0.357	0.873	0.472	0.458	0.443
SE3	0.528	0.372	0.438	0.868	0.442	0.499	0.489
SN1	0.445	0.400	0.391	0.474	0.858	0.323	0.429
SN2	0.378	0.396	0.447	0.463	0.823	0.296	0.477
SN3	0.496	0.380	0.407	0.420	0.883	0.286	0.482
AT1	0.489	0.438	0.440	0.498	0.336	0.921	0.386
AT2	0.542	0.468	0.449	0.507	0.370	0.932	0.431
AT3	0.462	0.455	0.414	0.469	0.253	0.895	0.384
WTP1	0.520	0.491	0.455	0.456	0.504	0.364	0.884
WTP2	0.511	0.535	0.451	0.491	0.458	0.427	0.889
WTP3	0.527	0.500	0.486	0.450	0.463	0.364	0.865

, the outcomes indicate that all constructs loaded higher than the cross-loadings of the other constructs in the row, which implies a large amount of one-dimensionality for each construct. The results exhibit good discriminant validity for all of the constructs in our study.

We then tested for common method bias. Generally, self-reported data are considered potentially subject to common method bias (CMB). Here we took a series of measures to reduce the risk of CMB (see Appendix A 

Table A2. Common method bias measurement.

Construct	Indicator	Substantive Factor Loading	Construct	Indicator	Substantive Factor Loading
Preferred lifestyle (PL)	PL1	0.922	0.850	0.073	0.005
	PL2	0.928	0.861	0.077	0.006
	PL3	0.898	0.806	0.07	0.005
Information quality (IQ)	IQ1	0.872	0.760	0.066	0.004
	IQ2	0.848	0.719	0.064	0.004
	IQ3	0.888	0.788	0.077	0.006
Interactive immersion (II)	II1	0.877	0.769	0.076	0.006
	II2	0.878	0.770	0.073	0.006
	II3	0.865	0.748	0.076	0.006
Self-efficacy (SE)	SE1	0.857	0.734	0.068	0.005
	SE2	0.839	0.703	0.067	0.004
	SE3	0.871	0.759	0.068	0.005
Subjective norms (SN)	SN1	0.871	0.759	0.068	0.005
	SN2	0.871	0.759	0.064	0.004
	SN3	0.852	0.726	0.067	0.004
Attitude (AT)	AT1	0.887	0.787	0.076	0.006
	AT2	0.886	0.785	0.078	0.006
	AT3	0.865	0.748	0.076	0.006
Willingness to pay (WTP)	WTP1	0.87	0.757	0.079	0.006
	WTP2	0.843	0.711	0.073	0.005
	WTP3	0.859	0.738	0.076	0.006
Average	N. A	0.874	0.763	0.072	0.005

). First, the questionnaires were distributed over four non-consecutive weeks, with three to five days spaced between each period, and the response screens for each participant were arranged in a randomized pattern. Next, we created a PLS model to assess CMB following the Podsakoff et al. [95] method, which compares the variance of each observation explained by its substantive factors and the method factors. The statistical results showed that the average substantive variances (0.763) are substantially greater than their method variances (0.005), which indicates that CMB is unlikely to be a serious concern in this study. Finally, we conduct a full collinearity test to examine the level of variance expansion factor (VIF) of all factors [96]. All the internal VIF values for each construct were not greater than the threshold (3.3) (see

Table 5. Variance inflation factor (VIF).

Construct	VIF	Construct	VIF
PL1	2.047	SN1	1.961
PL2	2.113	SN2	2.008
PL3	1.926	SN3	1.806
IQ1	2.049	AT1	2.218
IQ2	1.819	AT2	2.216
IQ3	2.105	AT3	1.951
II1	3.148	WTP1	1.933
II2	3.252	WTP2	1.734
II3	2.539	WTP3	1.852
SE1	1.821
SE2	1.739
SE3	1.898

Note: PL, preferred lifestyle; IQ, information quality; II, interactive immersion; SE, self-efficacy; SN, subjective norm; AT, attitude; WTP, Willingness to pay.

). These three tests show that the common method bias is not an issue in our study.

4.3. Structural Model Analysis

After the reliability and validity assessment was performed, we tested the research hypotheses using an SEM with the understanding that an SEM can reveal causal relationships among multiple latent variables effectively and simultaneously [97]. In addition, SEM has a statistical advantage over regression analysis with a more elaborate measurement [98]. In general, measures of variables with multiple items in regression analysis are usually condensed to mean scores, whereas SEM independently allows for the measurement of multiple latent variables with errors.

To evaluate the structural model, quality indicators were first measured using R squared ($R^2$) effect size ($f^2$) and prediction relevance ($Q^2$). $R^2$ ≥ 0.25 was considered acceptable [92]. $f^2$ was interpreted as follows: 0.02 corresponds to a small effect, 0.15 indicates a medium effect, and 0.35 for a large effect [99]. An acceptable $Q^2$ value was defined ≥ 0 [89]. According to the statistical analysis, our model accounted for 53.4% ($R_1^2$ = 0.534) of the variance in WTP and 42.6% ($R_2^2$ = 0.426) of the variance in AT. The $Q^2$ values were 0.381 for WTP and 0.321 for AT, which indices confirmed the good predictive performance of the structural model. Effect sizes were presented in

Table 6. Direct hypotheses test.

Hypotheses		Path Coefficient	Standard Deviation	Lower CI 2.5%	Upper CI 97.5%	p-Value	Supported	f2
H1	AT→WTP	0.203	0.065	0.074	0.328	0.002	Yes
H2	SN→WTP	0.058	0.052	−0.042	0.161	0.266	No	0.004
H3	SE→WTP	0.127	0.049	0.033	0.227	0.010	Yes	0.020
H4	PL→AT	0.302	0.048	0.209	0.394	0.000	Yes	0.105
H5	IQ→AT	0.399	0.050	0.297	0.494	0.000	Yes	0.194
H6	II→AT	0.078	0.046	−0.012	0.168	0.092	No	0.007

Note: preferred lifestyle, PL; information quality, IQ; interactive immersion, II; self-efficacy, SE; subjective norms, SN; attitude, AT; willingness to pay, WTP; →, pathway between variables; Effect size, f².

.

Then, we performed 5000 bootstrapping iterations to assess the statistical significance of the path coefficients. All hypotheses were supported except for H6a (b = 0.078, p = 0.093) and H2 (b = 0.058, p = 0.273). In particular, AT had a significant impact on WTP (b = 0.203, p = 0.002), which supported H1. PL (b = 0.399, p = 0.000) and IQ (b = 0.302, p = 0.000) were associated with AT. Hence, H4 and H5 were supported. More structural model results are illustrated in Table 6 and Figure 2.

4.4. Mediation Effects

To test whether the influence of PL, IQ, and II on WTP is mediated by AT, we adopted a total of 5000 bootstrap samples within a 97.5% confidence interval. When the interval does not contain zero, mediation is confirmed [100]. Typically, the mediation effects are captured by direct, indirect, and total effects. In our study, one of the indirect effects was insignificant, which is for the path “II→AT→WTP” and the 97.5% confidence interval includes zero (−0.001, 0.044). The remaining two indirect effects are significant with a 97.5% confidence interval excluding zero. Hence, we can deduce that the effects of PL and IQ on WTP are mediated by AT.

The magnitude of mediation was evaluated immediately in the next step. Thus, we performed Variance Accounted For (VAF) as recommended by Helm et al. [101], with values falling between 0 and 100%. Full mediation occurs when VAF ≥ 80%, partial mediation occurs when 80% ≥ VAF ≥ 20%, and there is no mediation when 20% ≥ VAF [102]. To rephrase, VAF is instrumental in measuring the relationship between indirect effects and total effects [103]. In the current study, VAF values range between 20% and 80%. We can conclude that AT partially mediates the effect of PL and IQ on WTP. All mediation pathways are listed in

Table 7. Mediation test results.

Hypotheses		Path Coefficient	Standard Deviation	Lower CI 2.5%	Upper CI 97.5%	Bias-Corrected Lower CI 2.5%	Bias-Corrected Upper CI 97.5%	Lower CI 2.5%
Standardized direct effects
H4a	PL→WTP	0.244	0.048	0.150	0.339	0.150	0.339	0.000
H5a	IQ→WTP	0.132	0.047	0.038	0.022	0.044	0.227	0.005
H6a	II→WTP	0.188	0.044	0.100	0.274	0.099	0.274	0.000
Standardized indirect effects
H4a	PL→WTP	0.062	0.021	0.022	0.106	0.026	0.110	0.004
H5a	IQ→WTP	0.081	0.030	0.027	0.143	0.028	0.146	0.007
H6a	II→WTP	0.016	0.011	−0.002	0.040	−0.001	0.044	0.145
Standardized total effects
H4a	PL→WTP	0.305	0.048	0.210	0.397	0.211	0.398	0.000
H5a	IQ→WTP	0.213	0.053	0.104	0.314	0.109	0.317	0.000
H6a	II→WTP	0.203	0.065	0.112	0.294	0.111	0.292	0.000

Note: preferred lifestyle, PL; information quality, IQ; interactive immersion, II; willingness to pay, WTP.

. Table 7 also lists the bias-corrected values of the paths.

4.5. Control Variables

We tested the control variables: gender (b = −0.012, p = 0.688), age (b = 0.034, p = 0.283), monthly household income (b = 0.009, p = 0.777), mobile live streaming experience (b = 0.054, p = 0.100), and disease type (b = 0.024, p = 0.440) and found them to be not significant. Among the control variables, we found that the educational level was significant (b = 0.087, p = 0.005), which is aligned with the findings of Surendran and Sekar [104]. When all control variables were excluded, no differences were observed in the research model. Thus, the major constructs in the research model were not significantly impacted by the control variables.

5. Discussion and Conclusions

5.1. Discussion

This study first developed a theoretical framework, augmenting the TPB model by integrating value compatibility (PL) and practical compatibility (IQ and II) to explain WTP for MHLS. Our findings show evidential support for the mediation effect of AT between compatibility and WTP, except for the non-significant effect of II. Moreover, the effect of SN on WTP was not significant.

• The relationship between PL and WTP is supported and mediated by AT;
  PL’s influence on WTP is supported. Previous studies on social media found that PL has a significant, direct effect on WTP [68], whereas we discovered that its effect is mediated by AT (H4a). Researchers’ attention to lifestyle has exploded in recent years, possibly reflecting consumers’ adaptability to new technologies and health awareness to achieve basic values throughout life. Especially during the COVID-19 pandemic, people are more likely to use MHLS if it allows them to access health information at home and reduces unnecessary travel. Meanwhile, PL is important to maintain the balance between changes in health promotion and individuals’ value systems. Specifically, in MHLS, PL refers to a healthy or mobile lifestyle. In healthy lifestyle terms, many individuals start using health live streaming apps to promote healthy behavior as they desire to maintain a healthy lifestyle, even less helpful health information is available in most apps [105]. A healthy lifestyle requires a sustained attitude and the willingness to pay for MHLS. In mobile lifestyle terms, mobile live streaming services persist because they are often synchronous, authentic, and group-based and provide a means to pay for goods and services in virtually any situation. Each lifestyle may become an incentive that cultivates people to increase their WTP so that the use of a mobile live streaming platform is necessary for consumers. Therefore, the influence of PL remains critical in predicting consumers’ AT and WTP.
• The relationship between IQ and WTP is supported and mediated by AT;
  The study also contributes by integrating IQ into the TPB model. The findings support the role of IQ on WTP (H5), which is consistent with studies relating to live streaming commerce [75] and internet commerce services [106]. This study considers IQ in an MHLS case. Specifically, IQ is a measure of customers’ perceived value of the output generated by MHLS. Pursuing an effective medical consultation is a long and vigorous exploration, not a timely return, which implies that it can be challenging for consumers to achieve substantial positive results from just a few uses of MHLS. Thus, obtaining high-quality health care information is very helpful, as consumers feel the requirements are fulfilled, which can motivate them to pay for their consultations. The effect of IQ on AT towards MHLS is also supported (H5a). Previous studies considered IQ in terms of its impact on health care management. For example, patients perceive that their attitudes toward mobile health services are positively influenced by the quality of mobile health systems [107]. However, prior investigations have not considered the relationship between IQ, AT, and WTP. Our findings illustrate the utility of IQ from a consumer perspective, elaborating on how IQ can motivate WTP via health apps where consumer attitudes are improved.
• The relationship between II and WTP is supported but not mediated by AT;
  Furthermore, it is found that II in MHLS significantly influences the WTP (H6). This relationship follows other studies which established the relationship between II and WTP [78]. Our finding implies that consumers’ active MHLS immersion drives their future purchase intentions. Unlike previous research [108], the results of our study indicated that II in MHLS is not mediated by AT (H6a), which may be due to two reasons. On the one hand, immersive experiences generate positive attitudinal effects in the form of consumers’ emotional and affective responses [108]. In times of pandemic, consumers may feel challenged to identify emotional states accurately because of a negative psychological condition caused by pandemic-induced anxiety [109]. On the other hand, in China, the primary purpose of new health promotion using ICTs may be to make profits rather than disseminate health information [110], which may explain why the II of consumers in MHLS may not lead to a greater AT.
• The relationship between SE and WTP is supported;
  SE’s significant effect on WTP (H3) is based on the work of Chang et al. [111] for health values on functional beverage purchase intentions. As a critical component in the extended TPB model, SE is a strong predictor of WTP in some ICT studies [112,113]. To accomplish health goals, most consumers will actively adopt the MHLS program, thereby making the SE of MHLS an important determinant of WTP. During the COVID-19 pandemic, the public was involved in MHLS with almost minimal technical professional help. This greatly reduces the psychological anxiety associated with the use of technology and strengthens individuals’ self-assurance in their ability to resist public health crises [114].
• The relationship between SN and WTP is not supported;
  Surprisingly, SN had no significant effect on WTP (H2). Our findings are inconsistent with the results of a study on online grocery shopping [115], which found a significant positive effect of SN on WTP. However, the emphasis of online grocery research is on social interaction, allowing consumers to share shopping knowledge or experiences through efficient interplay. The contrast between these two types of research may reveal that the impact of SN on MHLS will be less critical in the post-adoption phase. In the MHLS scenario, app developers rarely design relevant features for social interactions between consumers, which objectively renders it difficult for SN to play a contributing role. To be precise, consumers’ motivation to pay for medical consultations in MHLS is often tied to the characteristics of the private conditions, like disease type. Thus, in the case of MHLS, SN may not be predictive of consumers’ WTP. This is in line with the findings on e-services, where SN showed no significant effect on WTP [116]. Consequently, the impact of SN on consumers’ WTP diminishes before the payment phase, but the experience from the consumer’s own life may be highly influential.

5.2. Implications and Future Research

5.2.1. Implications

By validating the theoretical model with survey data from Chinese MHLS consumers, our findings determine the causal forces explaining the WTP of MHLS. These indicated that aside from TPB factors (self-efficacy and attitude toward MHLS), compatibility factors like value compatibility (PL) and practical compatibility (IQ and II) also predict WTP. As far as we know, this study is the first to analyze the factors influencing the public’s adoption of MHLS in response to a public health crisis. Using Chinese health information technology as an empirical context, we provide practical support for the applicability of compatibility theory in public health crises. Additionally, this study evaluates and supports the mediating effect of consumer attitudes on the WTP of MHLS. Our findings yield operational insights for MHLS developers and officials.

PL is critical to promoting the WTP of MHLS to consumers. Consumers typically engage with MHLS for specific lifestyle reasons and expect the app to provide features that will serve as a driver to maintain an active and fit way of life. From our interviewees’ responses, even if a live streaming platform is preferred daily, others with fitted features are registered by consumers for consultation. For example, Haodf, among the most popular MHLS in China, has features like quick consultation, medication shopping, self-diagnosis, and daily health monitoring. However, many Haodf consumers also downloaded Guahao as a substitute for providing medical advice. MHLS service providers should therefore focus on collecting and meeting consumers’ MHLS needs to promote the core products to them. In addition, as awareness of mental and physical health continues to grow and the pace of work increases quickly, more Chinese consumers generate specific health needs such as regular medication, diet management, and sleep testing. Logically, developers should focus more on this expanding market to expand the market share of MHLS.

IQ is a critical factor in predicting consumers’ WTP for MHLS. Previous research has found that one way to increase the likelihood of achieving long-term health goals and maintaining use habits is to improve the quality of information pushed to consumers [117]. In addition, the decision for individuals to engage in healthy behaviors is often deliberate, but it needs to be simple and effective to act on [118]. During the live process, several mobile health live streamers attempt to publicize medical professional knowledge or market products unrelated to medical consultation, which may lead to a rapid fall in flow because consumers are not yet familiar with the detailed health description. Most streamers can offer their viewers very convenient services through detailed condition analysis and a rich case for diagnosis. Applying big data analytics and artificial intelligence in the IoT environment may bring healthcare streamers based on consumer data. For example, several MHLS apps, such as Haodf, use AI to provide immediate and personalized streamer recommendations. With the rapid development of artificial intelligence and big data, technology services for life are becoming increasingly important [119]. Especially in public health crises, people inevitably face a conflict between adopting health technologies and the risk of information leakage. The emergence of the COVID-19 pandemic not only increased health providers’ rethinking of old health technologies but also provided opportunities for developing new health technologies using ICTs [120].

The links drawn between II and WTP are valuable to MHLS streamers. Therefore, MHLS streamers who fail to focus on immersive experiences should concentrate properly on this variable to further improve their business performance. For example, they can create an environment that facilitates consumers’ immersion by actively communicating with them about their conditions and patiently instructing them. MHLS streamers already emphasizing II should maximize the benefits of its visual and voice capabilities. For example, streamers can enhance consumer immersion by providing more details about services and responding quickly to demand. Furthermore, several streamers can strengthen their skills in guiding consumers. For example, streamers would be expected to ensure that they have comprehensive knowledge of the MHLS service before guiding consumers. Consumers are more likely to buy their services when streamers show them professionalism through MHLS and endeavor to help them cope with their illnesses. After the public health crisis ends, health technologies may be developed iteratively so that MHLS can help mitigate the spread of public health crises and other potential general hazards.

We also encourage practitioners involved in MHLS to focus more on consumers’ SE. Any health technology will encounter problems on its way to growth, from the technology level, equipment conditions, or public health habits and preferences. This rule makes it essential for new health promotion initiatives using ICTs to consider mobilizing people‘s initiative in their healthcare skills. SE addresses the balance between challenges and skills, reflecting that MHLS streamers should help consumers set challenging but achievable goals. Providing personalized functions and targets is recommended. For example, several mobile health apps, such as Dingxiang Doctor, offer consumers the option to choose the plan that best suits their condition. Many MHLS apps, such as Chunyu Doctor, monitor consumers’ physical and mental states and, in turn, modify their goals accordingly. To establish customers’ WTP and, in turn, modify their engagement, MHLS should integrate SE like dedicated AI doctors, self-diagnosis, self-checking, and device linkage.

5.2.2. Limitations and Future Research Agenda

Our study is limited in several ways. First, we gathered data on consumers’ WTP. Even though WTP can be used as an indicator of purchase intent, actual purchase data is more accurate in measuring such intentions [121]. If conditions allow, transactional data should be used in future studies to examine consumers’ purchase intentions in the MHLS.

Second, this research includes multiple MHLS apps. In our study, MHLS apps were not restricted to a specific one. Instead, based on respondents’ feedback, at least two MHLS apps are covered in this paper (Haodf and Guahao). Although we can get more data due to this, it is not a rigorous measurement because each app has some unique characteristics, which likely leads to inconsistent results. For example, Haodf live streaming allows being booked in advance. Future research should consider the impact of each app’s unique characteristics on customers’ WTP in MHLS.

Third, applying survey methods makes it impossible to measure WTP based on a consumer’s authentic experience. However, measuring WTP from the experience of a given consumer health consultation helps to understand purchase intention in certain circumstances. Therefore, future research could use experimental methods to examine consumer purchase intentions based on specific patient experiences.

Fourth, II and SN are non-significant predictors of consumer attitudes toward MHLS. However, they may still influence the use of MHLS apps. Even though this paper has presented explanations for these two untested mechanisms, their impact may not always be consciously recognized. Therefore, it may be worthwhile to investigate specific aspects, especially in non-epidemic conditions.