Online selection of a physician by patients: Empirical study from elaboration likelihood perspective

Accepted Manuscript Online selection of a physician by patients: Empirical study from elaboration likelihood perspective Xianye Cao, Yongmei Liu, Zhangxiang Zhu, Junhua Hu, Xiaohong Chen PII:

S0747-5632(17)30220-0

DOI:

10.1016/j.chb.2017.03.060

Reference:

CHB 4885

To appear in:

Computers in Human Behavior

Received Date: 9 November 2016 Revised Date:

25 March 2017

Accepted Date: 27 March 2017

Please cite this article as: Cao X., Liu Y., Zhu Z., Hu J. & Chen X., Online selection of a physician by patients: Empirical study from elaboration likelihood perspective, Computers in Human Behavior (2017), doi: 10.1016/j.chb.2017.03.060. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Online Selection of a Physician by Patients: Empirical Study from Elaboration Likelihood Perspective Cao Xianye, Liu Yongmei*, Zhu Zhangxiang, Hu Junhua, Chen Xiaohong

RI PT

(School of Business, Central South University, Changsha 410083, China)

EP

TE D

M AN U

SC

Abstract: With the rapid development of Web 2.0 technologies, an increasing number of physicians are providing services through websites that enable patients to consult with them online. Patients can find a wealth of information about the healthcare community, but research has not explored how patients process this information, or how this processing might influence their decisions to consult a physician online. To fill this gap, we used the Elaboration Likelihood Model and the Service Quality theory to investigate patients’ selection decisions. We considered service quality as the central route, and electronic word-of-mouth (eWOM) as the peripheral cue, and explored their importance. We also examined the moderating effects of disease risk and disease knowledge on patients’ consulting intention. We developed an empirical econometric model to evaluate our hypotheses. Using data from an online healthcare site in China, our results revealed that service quality and eWOM both had positive effects on patients’ selection decisions. Disease knowledge increased the importance of service quality on patients’ choices. Furthermore, disease risk and disease knowledge decreased the influence of eWOM on patients’ choices. We conclude that our research into the impact of information processing on how patients select their physicians has strong theoretical and practical implications.

AC C

Key words: disease type; Elaboration Likelihood Model; online healthcare community; Service Quality; electronic word-of-mouth

*

Corresponding author at: Address: School of business, Central South University, 932 Lushan South Road, Yuelu District, Changsha, Hunan Province 410083, China. Tel.:+86 73188879881; fax. : +86 731-88710006. E-mail address: [email protected]

ACCEPTED MANUSCRIPT

Disease knowledge

H4A(+)

H3A(+)

H4B(-)

H3B(-) The number of current patients who repeatedly interact with physician

Consulting intention (new patient next week)

H2(+)

M AN U

SC

H 1(+)

peripheral clue Electronic word-of-mouth

RI PT

central route Service quality

Disease risk

Vote heating

TE D

Service star

Fig. 1 Research concept model.

AC C

EP

According to Elaboration Likelihood Model and the Service Quality theory, service quality and eWOM have positive effects on patient consulting intention. Disease risk increases the importance of service quality on intention, and decreases the importance of eWOM on intention. Meanwhile, disease knowledge amplifies the effects of service quality on intention and abates the effects of eWOM on intention. Our hypotheses were shown in the fig.1., H1 to H4. By collecting data from an online healthcare in China, we verified most of the hypotheses, and only H4A was not proved.

ACCEPTED MANUSCRIPT

Online Selection of a Physician by Patients: Empirical Study from Elaboration Likelihood Perspective

TE D

M AN U

SC

RI PT

Abstract: With the rapid development of Web 2.0 technologies, an increasing number of physicians are providing services through websites that enable patients to consult with them online. Patients can find a wealth of information about the healthcare community, but research has not explored how patients process this information, or how this processing might influence their decisions to consult a physician online. To fill this gap, we used the Elaboration Likelihood Model and the Service Quality theory to investigate patients’ selection decisions. We considered service quality as the central route, and electronic word-of-mouth (eWOM) as the peripheral cue, and explored their importance. We also examined the moderating effects of disease risk and disease knowledge on patients’ consulting intention. We developed an empirical econometric model to evaluate our hypotheses. Using data from an online healthcare site in China, our results revealed that service quality and eWOM both had positive effects on patients’ selection decisions. Disease knowledge increased the importance of service quality on patients’ choices. Furthermore, disease risk and disease knowledge decreased the influence of eWOM on patients’ choices. We conclude that our research into the impact of information processing on how patients select their physicians has strong theoretical and practical implications.

AC C

EP

Key words: disease type; Elaboration Likelihood Model; online healthcare community; Service Quality; electronic word-of-mouth

1

ACCEPTED MANUSCRIPT

1. Introduction With the development of online healthcare services, an increasing number of

RI PT

people choose to visit healthcare websites for healthcare information and consultations (Xiao, Sharman, Rao, & Upadhyaya, 2014). Through these websites, a patient can post a text description of his/her symptoms along with any appropriate images. In turn, a physician can use this online information to determine the patient’s disease condition and provide a diagnosis, treatment, and suggestions (Yang, Guo, Wu, & Ju, 2015).

TE D

M AN U

SC

There are two primary features of the physician-patient interaction via a healthcare website that differ from traditional in-person methods of interaction between physicians and patients (Ba & Wang, 2013). First, the website gives patients the opportunity to review the rich amount of information presented about various physicians, and then use this information to choose the doctor they wish to consult. In contrast, in medical settings such as hospitals, patients can obtain only limited information about the facility and the physicians, for example, the hospital’s standing and the physicians’ positions. Second, a healthcare website enables physicians to show information about their knowledge and treatments to everyone who visits the online site. In a medical office or complex, physicians meet face to face with a patient, and they can provide only limited information because of limited time.

AC C

EP

Although website information is very important for patients’ selection of their physicians, only a few researchers have been interested in investigating the effects of online information on patients’ decisions. Yang, Guo, Wu, et al. (2015) demonstrated that both patient-generated and system-generated information can reflect the quality of physicians’ service outcomes and delivery processes. They found that positive information about the quality of physicians’ services, whether patient-generated or system-generated, had a positive influence on both the patients’ visits and website consultations. Lu and Wu (2016) found that word-of-mouth information about physicians reflecting the physicians’ service quality, and it has a positive effect on patients’ office appointments. Disease risk is another factor that can moderate the relationship between service quality and patients’ office appointments. However, there has been very little research exploring how patients’ processing of information affects their online consultation intentions, and no one has considered the effects of disease type on how the patients process information. We employed the Elaboration Likelihood Model (ELM) as the theoretical base by which to understand how patients process information regarding their selections for 2

ACCEPTED MANUSCRIPT

RI PT

online health services. The ELM proposes that the decisions people make are influenced in two ways: by central route processing and by peripheral cues (R. E. Petty, &Cacioppo, J. T., 1981; R. E. Petty & Cacioppo, 1986b). When using the “central route,” an individual processes information related to a task based on careful consideration of the merits. Central route reasoning requires more effort than taking the “peripheral route,” in which the individual makes a decision based on information cues such as reputation (Zhou, 2012).

M AN U

SC

The ELM indicates that the ability and motivation of the user are two important considerations that moderate the effects of information on their decisions (R. E. Petty & Cacioppo, 1986a, 1986c; R. E. Petty, Cacioppo, & Goldman, 1981). When individuals who receive information have a higher level of prior knowledge and comprehension about the message topic, they are likely to have a greater number of issue-relevant thoughts and ideas. Their ability to understand the issues and related questions will be greater as well, which in turn will increase the likelihood of elaboration and decrease reliance on peripheral cues (Cheung, 2008; Sussman & Siegal, 2003). To reflect these considerations, we chose disease knowledge as a proxy for this ability. Our main research questions were the following:

TE D

1. Could information about the service quality of the physician and peripheral cues affect the patients’ decisions about which physician to consult? 2. Could the patients’ disease knowledge and their motivation moderate the relationship between service quality or peripheral cues and the patients’ selection decisions?

AC C

EP

For this research, we collected data for 15,617 physicians from the Good Physician (also translated as Good Doctor) Online website (www.haodf.com), one of the largest online healthcare communities in China. We established an empirical model to test our hypotheses. The contributions of this paper are threefold. First, our study adds to the literature about online health consulting by testing the effects of service quality and electronic word-of-mouth (eWOM) on patients’ selection decisions. The results supported the effects of them. Second, this paper contributes to the body of research about online services by expanding the Elaboration Likelihood Model to encompass online health consulting. Last, this paper investigates the moderating effects of disease risk and disease knowledge on the relationship between peripheral cues and patients’ selection of their physicians. In this regard, we verified the moderating effects of disease risk and disease knowledge on them. The remainder of this paper is organized in the following manner. In Section 2, we 3

ACCEPTED MANUSCRIPT review the online health community, online service quality, and online healthcare information, along with patients’ decision processing and the ELM theory. In Section 3, we present our research model and hypotheses. Our method, data, and results are presented in Section 4. In Section 5, we discuss our results. The conclusion is presented in Section 6.

RI PT

2 Theoretical backgrounds 2.1 Online health community

TE D

M AN U

SC

Through an online health community, a patient can search for information about diseases and consult physicians to get treatment prescriptions without needing face-to-face communication. This type of healthcare website provides a convenient, time-saving platform for patients while offering lower cost. The rapid development of online healthcare has also received the attention of the Information System (IS) academic community. Most existing IS studies, however, have examined topics relating to the patients’ adoption and post-adoption of online healthcare (Guo, Zhang, & Sun, 2016; Okazaki, Blas, & Castaneda, 2015; Ozdemir, Barron, & Bandyopadhyay, 2011). Some researchers have focused on the benefits or motivation for choosing online healthcare from different perspectives. Galehbakhtiari and Pouryasouri (2015) identified the strongest motivational paths toward participation in online healthcare communities. Ba and Wang (2013) studied the effectiveness of the motivation mechanisms provided by social networks.

EP

However, exploration of the factors that influence a patient’s decision to consult a physician online has been limited, and there are no research studies about the moderating effects of disease risk and disease knowledge on the selection process. To better understand these influencing factors and moderating effects, more studies are needed.

AC C

2.2 Online health service quality Service quality is defined in terms of a user’s judgment about the overall excellence or superiority of service (Zeithaml & Institute, 1987). Service quality has also been defined as measuring performance against expectations, or in terms of the gap between expected and perceived service, or as performance-only measures (Akter, D’Ambra, & Ray, 2013). Service quality is generally specified as a multidimensional and hierarchical concept (Akter et al., 2013). In healthcare, most service quality research has focused on either a two-dimensional model (Grönroos, 1993) (i.e., functional quality and technical quality) or on the SERVQUAL model for measuring a patient’s perception of service (Parasuraman, Zeithaml, & Berry, 1988). Donabedian (2005) measured medical service quality under two dimensions: 4

ACCEPTED MANUSCRIPT technical and interpersonal quality. Technical quality referred to the application of medical science and technology to healthcare, and interpersonal quality referred to the interaction that occurs between the service provider and the consumer. Akter et al. (2013) developed a scale of service quality for health which contained three dimensions: system quality, information quality, and interaction quality.

M AN U

SC

RI PT

Patient-physician interaction contains information delivery that can assist patients in understanding their conditions and treatments (Xiao et al., 2014). A patient’s intention to consult a physician can be enhanced by previous positive interaction with that physician. According to Delone and Mclean (2004), perceived service quality can lead to use intention and usage. Lu and Wu (2016) studied and confirmed that service quality evaluations by other patients (eWOM) online can have a positive effect on a patient’s choice to make an office visit. However, eWOM cannot fully reflect the quality of products – in this case, physician services – because of two sources of bias in product review systems: purchasing and under-reporting. (Hu, Zhang, & Pavlou, 2009). We need a better measurement of service quality in the online health context.

2.3 Online healthcare information and patients’ decision process

AC C

EP

TE D

Healthcare websites provide patients with rich information, including reviews by former patients, data about physicians’ interactions with patients, and even emotional support (Yan & Tan, 2014). Online health information could empower patients by assisting them in managing their illnesses, helping them cope with the associated pressure and stress, and facilitating their medical decision-making (Xiao et al., 2014). Some researchers have investigated the factors that influence users’ information adoption or search behaviors in online healthcare (Lin et al., 2015; Mano, 2014; Xiao et al., 2014). Jin, Yan, Li, and Li (2015) applied the ELM to investigate and confirmed the positive impacts of information quality, emotional support, and source credibility on the likelihood of healthcare information adoptions. The use of social media and users’ health conditions/illnesses are associated positively with searching for online health information (Mano, 2014). Xiao et al. (2014) found that perceived health status had a negative impact on the frequency and diversity of information search behavior. In the online healthcare context, patients make their choice of physicians based on information on the website. As mentioned above, Yang, Guo, Wu, et al. (2015) confirmed that patient-generated and system-generated information can reflect the quality of physicians’ service outcomes and delivery processes in a manner that helps patients select doctors. By viewing a physician’s homepage, patients can learn about the physician’s training, specialties, and the effectiveness of the treatments 5

ACCEPTED MANUSCRIPT offered. In addition, new or existing patients might make inquiries via the website. The physician can provide prompt, high quality responses to patients’ questions and concerns that could persuade patients to consult her/him further.

M AN U

SC

RI PT

The electronic word-of-mouth shown on a website is an important source of information relied upon by users when making a choice. Many researchers have explored the effects of eWOM in the context of consumers’ purchase intentions and actual sales (Chevalier & Mayzlin, 2004; Iii, 2001; Prendergast, Ko, & Yin, 2010). The online health community also shows eWOM about physicians, such as the voting heat about that doctor and whether s/he has received Service Star recognition. There are a few researchers who have explored the effects of eWOM on patients’ choices concerning online consulting services in a health context (Yang, Guo, Wu, et al., 2015). In addition to eWOM information, healthcare websites also can show the interaction between the patients and the physician. To ensure the patients’ privacy rights, all patients are anonymous to others. Users who view a website can infer the service quality of the physicians by scrutinizing the interaction information. Despite the availability of extensive health-related information on the Internet, limited attention has been given to the influence of that information on a patient’s selection decision process in an online healthcare domain. 2.4 Elaboration likelihood model

TE D

The Elaboration Likelihood Model (ELM) is a psychological theory that addresses the process of persuasion (Marquart & Naderer, 1986; R. E. Petty et al., 1981). The ELM is a “dual-process” theory that posits two routes through which persuasion

EP

takes place: the central route and the peripheral route (Angst & Agarwal, 2009; R. E. Petty & Cacioppo, 1986b). The central route involves carefully scrutinizing the issue-related information with extensive cognitive effort, whereas the peripheral route often relies on the environmental characteristics associated with the

AC C

information, without giving the matter any extensive thought (Bi, Liu, & Usman, 2017; Lee, Park, & Han, 2008). Researchers have explored the influence of the factors related to these two routes on consumers' attitudes toward products and services. Generally, researchers have considered that users process information about the quality of products and services using the central route, while eWOM cues are processed using the peripheral route (Bi et al., 2017; Cheung, 2008; Lee et al., 2008; Park, Lee, & Han, 2007). For reasons explained above in this paper, while a physician’s home page may show the eWOM about him/her, patients should not rely solely on the eWOM when making a decision. There is other information requiring more cognitive processing effort that can serve as a better signal of the service quality of a physician, such as the number 6

ACCEPTED MANUSCRIPT of repeat patients who have returned for further consultation. Therefore, based on the literature, our study considers signal of service quality to be information that is processed utilizing the central route, and eWOM as processed using the peripheral route.

RI PT

According to ELM, information recipients can vary widely in their ability and their motivation to process the information. When the recipients have more motivation, more knowledge, and/or more cognitive ability to examine the message, they engage in the central route. When the recipients have less ability or motivation to elaborate on the message, they take the peripheral route. Many research studies explore the empirical validity of the ELM in various IS

SC

contexts (Angst & Agarwal, 2009; Bhattacherjee & Sanford, 2006; Fadel, Meservy, & Jensen, 2015; Ho & Bodoff, 2014; Meservy, Jensen, & Fadel, 2014; Tam & Ho,

M AN U

2005; Zhang & Watts, 2008). Although the above studies testify to the validity of the ELM in IT adoption, knowledge adoption, and set reduction, none were concerned with the process of selecting a physician in the context of online healthcare. 3. Research Hypotheses

3.1 The direct impact of central route and peripheral route processing on patients’ decisions

EP

TE D

To form our hypotheses, we first addressed the central route of online healthcare evaluation. Central route information (signals of project quality) has have significant positive effects on user’s choice(Bi et al., 2017). And service quality are important central route factors that persuade users (Zhou, 2012) . In our online healthcare context, patients believe physicians who have higher service quality can better diagnose and treat their disease, so they are most concerned with a physician’s

AC C

service quality. Information about service quality is processed using the central route because it is a core criterion when a patient choses a physician, and patients need to spend sufficient time to review the interaction information between the physician and his/her current patients, and assess the service quality of him/her. These inspections require an investment of effort from users (Zhou, 2012). Since service quality has a positive effect on patients’ choice of their physicians (Lu & Wu, 2016). We hypothesize the following: H1: The service quality provided by a physician to current patients has a positive effect on a patient’s consulting intention. The eWOM are considered as peripheral cues (Bi et al., 2017), and peripheral cues also have a positive effect on users’ choices (Bhattacherjee & Sanford, 2006; Ho & Bodoff, 2014; Richard E. Petty et al., 1981). Patients rely on eWOM, such as voting 7

ACCEPTED MANUSCRIPT heat, to judge the service quality of a physician. Patients prefer to choose a physician who has a higher voting heat. H2: The voting heat for a physician has a positive effect on a patient’s consulting intention.

RI PT

3.2 The moderating effect of central route and peripheral route on patients’ decisions

The following hypotheses concern the moderating effects of motivation and disease knowledge on the relationship between the central route or the peripheral route and the selection of patients. Meservy et al. (2014) argued that elaboration

SC

positively moderates the influence of central route reasoning on choice set retention, and negatively moderates the influence of validation on set retention. Motivation and

M AN U

ability are two sources of elaboration. According to ELM, the use of central route and peripheral route processing for decision-making is moderated by the potential user’s ability and motivation to elaborate on informational messages (Bhattacherjee & Sanford, 2006; R. E. Petty & Cacioppo, 1986b). As explained above, this ability is related to prior knowledge, and we have chosen to use disease knowledge as a proxy for this ability. Patients who have a higher level of disease knowledge, are more likely to rely on service quality and exhibit a decreased reliance on peripheral cues. Therefore, we can hypothesize the following:

TE D

H3a: Disease knowledge has a positive moderating effect on the association between service quality and a patient’s consulting intention.

EP

H3b: Disease knowledge has a negative moderating effect on the association between voting heat and a patient’s consulting intention.

AC C

Motivation levels also change the likelihood of elaboration by a user（Petty & Cacioppo, 1986b; Petty & Cacioppo, 1990). Patients who have suffered from high-risk disease may have poorer health status than those with lower disease risk. High-risk patients will be more worried, and will be more motivated by the hope of finding a higher quality physician (Lu & Wu, 2016). Moreover, patients with high-risk disease need higher quality service than patients with low-risk disease (Yang, Guo, & Wu, 2015). Because of its association with mortality, patients with high-risk disease may have more motivation to undertake more cognitive effort to attain a better physician. These patients are likely to be drawn to a physician who provides higher service quality, rather than a physician who solely provides higher voting heat. Consequently, we can hypothesize the following: H4a: Disease risk has a positive moderating effect on the association between 8

ACCEPTED MANUSCRIPT service quality and a patient’s consulting intention. H4b: Disease risk has a negative moderating effect on the association between voting heat and a patient’s consulting intention.

SC

Fig. 1 Research concept model.

M AN U

4. Method, Data, Model, and Analysis 4.1 The research context

RI PT

In summary, we developed our research model according to both the ELM and the service quality theory, as depicted in Fig. 1. Both service quality and eWOM have positive effects on a patient’s consulting intention. Disease risk increases the importance of service quality on intention, and decreases the importance of eWOM on intention. Meanwhile, disease knowledge amplifies the effects of service quality on intention, and abates the effects of eWOM on intention.

TE D

Our research context was the Good Physician Online website (www.haodf.com), one of the most popular online health communities and the largest in China. The Good Physician Online website was founded in 2006, and it is a physician-patient interaction platform (Yang, Guo, & Wu, 2015). Currently, over 5,000 hospitals and more than 400,000 physicians are listed on the Good Physician Online. A physician can open his/her homepage on the Good Physician Online website and provide service though the homepage. More than 110,000 physicians are providing online website consulting.

AC C

EP

We chose the Good Physician Online website to test our hypotheses because this site possesses rich information about each physician, the number of patients interacting with him/her and his/her interaction information with each patient, especially the interaction times, the physician’s position (director physician, associate director physician, chief physician or else), and peripheral cues such as voting heat and awards or recognitions, such as receipt of a Service Star. On this online website, a physician’s homepage presents the disease conditions for which the physician has specific skills or interests. We can consider the diseases from two dimensions: disease risk and disease knowledge. For these reasons, it is suitable for us to collect the data from this website to test our proposed hypotheses. 4.2 Method and Data To collect data from the Good Physician Online website (www.haodf.com), we chose 32 types of diseases. Then we developed a C# program to download 9

ACCEPTED MANUSCRIPT information automatically about each physician’s listing of diseases on this site, along with the disease “voting heat” for the physician. Voting heat was calculated by the platform based on the vote count by patients per disease. Because physicians received votes from their patients who suffered from various diseases, for the purposes of this study, we set the physician’s “disease category” as the disease for

AC C

EP

TE D

M AN U

SC

RI PT

which s/he received the highest number of votes, i.e., the highest voting heat. Considering the types of diseases, we chose 20 types of disease. Next, we collected physicians’ homepages and consult list pages based on two norms: First, does the physician treat one (or more) of the 20 kinds of disease on our list, and is the physician’s voting heat the highest for his/her treatment of this disease? Second, does the physician provide online consultation services via the website? We collected 20 disease categories with 15,617 physicians’ homepages and consult list pages from July 1, 2016 to July 2, 2016. These 20 kinds of diseases included anxiety disorders, depression, bipolar disorder, menstrual disorder, obsessive-compulsive disorder, nasopharyngeal carcinoma, leukemia, breast cancer, gastric cancer, liver cancer, rheumatoid arthritis, gastritis, hepatitis B, hypertension, anemia, gastric ulcers, diabetes, liver cirrhosis, cerebral infarction and coronary heart disease. After a week, we collected the number of patients that interacted from the physicians’ homepages, and calculated the number of new patients for that week. The dependent variable of this research was the consulting intention of patients. We used the number of new patients each week as a proxy for patient consulting intention in accordance with the research of Yang, Guo, Wu, et al. (2015). Consulting intention is a psychological indicator that can predict consult behavior. Therefore, the number of patients was a useful method for expressing patients’ consulting intentions. The independent variables were service quality and voting heat. If a patient considered the initial service provided by the physician to be good enough, s/he would interact with the physician more than once to gain more information. We considered that the more the patient interacted, the better the service quality. “Interaction” was understood to mean an information exchange between patient and physician. We considered that the number of patients who repeatedly interacted with a physician may be a measurement of service quality. To secure the best possible comparison and expression of a physician’s current service quality, we selected the number of current patients who interacted repeatedly with the physician. Current patients were the patients whose latest update time was within a month before we collected the data. The voting heat of a physician was shown in terms of the homage for him/her. The voting heat was a number between 0 to 5, with higher numbers indicating higher eWOM for a physician. The moderation variables were disease knowledge and disease risk. Since there is 10

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

more information about common diseases on the Internet than about uncommon diseases, patients can get more disease knowledge about the common illnesses. Consequently, the patients’ knowledge about common diseases was higher than about uncommon diseases, and there was less information asymmetry between physicians and patients. We classified these diseases according to the work by Yang, Guo, and Wu (2015) and the mortality and morbidity data from China Health Statistics Yearbook 2015. The classification basis and results are shown in Appendix Table A.1. High-risk diseases and low-risk diseases were expressed as 1 and 0, respectively. Common diseases were expressed as 1, with other diseases expressed as 0. In our model, we also included other variables that would impact patients’ selections: the physician’s position (Director Physician, Associate Physician, Chief Physician, and Physician), as well as the number of patient visits to the homepage, the rating and grade of the physical hospital where a physician worked, the city rank where the hospital was located, and other service categories such as telephone services, transfer treatment, the volume of digital gifts, and the number of months the physician’s home page had existed. We used these variables in our research model to control the effects of a physician’s status on patients’ selection. All variables and their descriptions are listed in Table 1.

EP

TE D

Table 1 Variable description. Descriptive statistics and correlations of variables are shown in Table 2. Table 2 lists types of variables, variable name, the proxy, description, and abbreviation used in models. Note that position and city rank are categorical variables. As we can see in Table 2, the main independent variables (Current, Heat, Star) were correlated significantly with the dependent variable (New), and consistent with the hypothesis.

AC C

Table 2 Descriptive statistics and correlations of variables.

4.3 Model estimation

To test our hypotheses about the direct and moderating effects on the consulting intention, we created an empirical model as follows:

Newi = a0 + a1Currenti + a2Currenti * Riski + a3Currenti * Knowledgei +a4 Heati + a5 Heati * Riski + a6 Heat * Knowledgei +α10Teli + α11Transferi + α12Visitsi + α13Monthsi + α14 Lettersi + α15Giftsi +α16 HospitalRatingi + α17 HospitalGradei + α18 Positioni + α19Cityi + ui 11

ACCEPTED MANUSCRIPT Let i = 1,..., N index the physician. Where estimated,

a1

to

a6

are the focus parameters to be

a0 is the intercept, and a10 to a17 represent the coefficients of the control

variables, and they are not of interest in the current work. Position and city are category

a18 to a19 are the vector coefficients of these category control

variables. ui is the error term associated with observation i . 4.4 Results

RI PT

control variables, and

SC

Table 3 Parameter estimates of the consulting intention. Table 3 presents the results of our model estimated by ordinary least squares. We present this equation hierarchically, first showing a model with control variables in Column 1, and then introducing independent variables and interaction terms in

AC C

EP

TE D

M AN U

Column 2 and Column 3. The Adjusted R-squared and F values were reasonable and significant, since the Adjusted R-squared was not low (more than 0.25) and Prob > F is 0.000. None of the variance inflation factor (VIF) statistics for the variables were greater than 6.0, which indicated the absence of multicollinearity. The central route Hypothesis (H1) predicted that a physician’s service quality would affect patients’ selections. Per Column 2 of Table 3, this hypothesis was supported because the coefficient of Current (B = 0.066, T = 61.06, P < 0.01 in Table 3) was positive and statistically significant. The peripheral cue hypothesis (H2) predicted that a physician’s voting heat would affect patients’ selections. Per Column 2 of Table 3 , this hypothesis was supported because the coefficient of Heat (B =1.716, T = 14.29, P < 0.01 in Table 3) was positive and statistically significant. The ELM suggests that as patients select their physicians, disease knowledge will increase the relative importance of central route processing such as service quality (H3a), and decrease the importance of peripheral cues, such as voting heat (H3b). Our results provided support for these hypotheses. First, as shown in Table 3, our analysis revealed a significant increase in the importance of service quality (H3a), as evidenced by a significant coefficient of the interaction item (B=0.016, T=7.06, P<0.01). Meanwhile, disease knowledge did decrease the importance of voting heat (H3b), as evidenced by a significant coefficient of the interaction item (B = -1.529, T = -7.37, P < 0.01 in Table 3). According to the ELM, when patients select a physician online, disease risk will increase the relative importance of central route factors, such as service quality (H4a), and decrease the important of peripheral cues, such as voting heat (H4b). Our results provided mixed support for these hypotheses. First, as shown in Table 3, our 12

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

analysis did not show a significant increase in the importance of service quality; therefore, H4a was not supported. In contrast, disease risk did decrease the importance of voting heat (H4b), as evidenced by a significant coefficient of interaction item (B = -1.327, T = -7.37, P < 0.01 in Table 3). Overall, we can see that most of our hypotheses were supported except H4a. Disease risk did not amplify the effects of service quality on a patient’s consulting intention. A possible reason is that patients suffering from high-risk disease are less likely to rely upon online consultation for accessing health-related services, as shown by the results of Mano (2014). Although online health sites can help patients gain access to health services conveniently at a lower cost, most online health consultations are not done in real time, and they cannot calm the patient's immediate anxiety and worry. For these reasons, we propose that the service quality was still unable to meet the needs of patient suffering from high-risk disease who might prefer face to face communication.

4.5 Robustness check

AC C

EP

TE D

In order to check the robustness of our results, we used “Service Star” as another measure of peripheral cues. Table 4 presents the results of our model estimated by ordinary least squares. We present this equation hierarchically, first showing a model with control variables in Columns 1, then introducing independent variables and interaction terms in Column 4 and Column 5, respectively. The Adjusted R-squared and F values were reasonable. Adjusted R-squared values were higher than 0.2, and F values were significant. No variance inflation factor (VIF) statistics for the variables were greater than 7, which indicates the absence of multicollinearity. The results are consistent with the results of the previous model. Therefore, the results are robust.

Table 4 Parameter estimates of the consulting intention (robust check).

5. Discussion and implications 5.1 Discussion In this paper, we investigated the effects of the ELM central route and peripheral cues on patients’ selection of a physician in the online health community. We hypothesized that service quality is positively associated with a patient’s consulting intention. We also hypothesized that if a physician receives a higher eWOM, s/he would be more likely to be chosen by patients. In addition, we hypothesized that disease risk and disease knowledge moderates the relationship between the service 13

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

quality or eWOM and a patient’s consulting intention. Using data about physicians collected from the largest online healthcare community in China, we created an empirical model to test our hypotheses. We chose control variables such as visits, telephone services, treatment transfer, opening months (how many months the home page had existed), and the position of the physician, along with the hospital rating, hospital grade, city rank of the hospital, and the volume of digital gifts and letters, to address potential endogeneity issues. The results demonstrated that patients’ consulting intention was related to a physician's service quality. Online consultation can provide patients with convenient access to physicians at low cost. The service provided by physicians online can influence the health condition and life quality of patients, and health is a matter of great concern to patients. Every patient wants to receive the best service, so the better the service quality provided by a physician, the more patients will want to consult him/her. The results relating to the eWOM about physicians indicated that this variable was closely related to a patient’s consulting intention. The eWOM comes from the evaluations of physicians’ existing patients, so it is a simple and well-understood index for use by potential new patients when choosing a physician. Consequently, when a physician has a higher eWOM, s/he is more likely to be chosen by new patients. We also found that disease knowledge plays a role in moderating the relationship between service quality or eWOM and patient satisfaction. Since there are so many disease types, we could not measure knowledge of these diseases one by one. Instead, we used the morbidity rate to classify these diseases into two types: common diseases and uncommon diseases. When a patient suffers from a common disease, s/he can gain a substantial amount of knowledge from websites and from other people, so s/he can focus on the service quality of the physicians, relying less on eWOM. The results relating to the moderating influence of disease risk showed mixed support for our hypotheses. Patients suffering from high-risk disease do rely less on eWOM. However, we failed to verify the moderating effect of disease risk on the association between service quality and a patient’s consulting intention. A possible explanation is that patients suffering from high-risk disease may be more concerned about service quality, but they may prefer office visits. Office visits can communicate rich information in a short period, and patients may be more accustomed to and place more trust in the face to face approach, especially when they are confronting high-risk disease.

14

ACCEPTED MANUSCRIPT 5.2 Theoretical contributions

AC C

EP

TE D

M AN U

SC

RI PT

Our research offers theoretical contributions in the following ways. First, although extensive studies have investigated the effects of user-generated information (such as consumer reviews and votes) and other types information (such as recommendations and ranking), generally these previous studies have focused on consumer decisions about products such as hotel reservations, mobile applications (APPs), and books (Anindya Ghose, 2012; Ghose, Ipeirotis, & Li, 2014; Nikolay Archak, 2011; Yin, Mitra, & Zhang, 2016). In contrast, there is a lack of adequate literature about the effects of various types of information on patient decisions regarding online healthcare services. Since health services are intangible, heterogeneous, and inseparable, service quality is more difficult to evaluate than product quality (Parasuraman & Berry, 1985). To evaluate a physician’s service quality, patients can depend not only on the eWOM about the physicians and feedback of other patients, but also on more specific information about the service provided to previous patients as shown on the physician’s website. Through these means, the asymmetry of information can be reduced. Second, although extensive studies have examined the validation of the ELM, those studies focused mainly on information adoption, IT adoption, customer satisfaction, and similar areas. A key contribution of our research is that we extended the application of the ELM to the online health community. Based on the ELM and the service quality theory, our study added to this research by investigating the effects of service quality and eWOM on patients’ consulting intention. Moreover, compared to previous research, this paper used disease knowledge and disease risk as detailed proxies for expertise and motivation in the health context, receptively. Third, we investigated the moderating effects of patient disease risk and disease knowledge on the relationship between the service quality or eWOM and a patient’s consulting intention. Although previous studies have investigated the moderation effect of motivation and knowledge between the ELM central route and peripheral cues for user’s decisions, the characteristics of patients are different from those of retail customers. We explored the effects of disease risk and disease knowledge on their consulting intention, and most of our hypotheses were proven. 5.3 Implications for practice Our research has three major implications for practice. First, the results showed that eWOM and service quality affect patients’ selection decisions. The eWOM, such as voting heat or the receipt of awards such as a Service Star, reflects the outcome and satisfaction of previous patients. eWOM provides a simple reference index that does not require potential new patients to undertake much cognitive effort. Since information about service delivery is shown to the online community, patients can 15

ACCEPTED MANUSCRIPT

TE D

M AN U

SC

RI PT

look over the records and evaluate the service quality of physicians. The number of current repeat patients is a useful variable that is reflective of service quality in the online health context, and our results demonstrated that it had positive effects on patients’ decisions. Second, for the designers of platforms such as wwww.haodf.com, our findings can help them identify strategic elements to include in the website. They should provide more information or statistics about the physicians’ service quality – such as the number of current repeat patients consulting the physician, interaction frequency, and physician response time – that can reduce the cognitive effort of patients when choosing a physician, and help patients process relevant information in a better way. Third, for physicians who provide online consultation services, our findings can help them manage their information to enhance their online reputations and acquire more patients. Although physicians cannot change the eWOM directly, they can encourage their patients in a physical hospital to use the platform to manage their disease, thereby providing better service to these patients. Physicians can make a better impression when potential patients look over the interactions between the doctors and their current or past patients. Fourth, high-risk disease did not amplify the effects of service quality on patients’ decisions, which means that platform and health providers should try to gain a clearer understanding of what motivates high risk disease patients, and address their needs in new ways. The rapid growth of online health services is a relatively recent phenomenon that will continue to grow and address the ways in which high-risk disease amplifies the effects of service quality on patients’ choices. 5.4 Limitations and suggestions for future research

AC C

EP

There are some limitations in our research. First, we used data from only one online healthcare platform. Although it is the most popular health community in China, there is a possibility that the findings apply only to this platform. It is necessary to collect data from several platforms to test our hypotheses fully. Further, we used data from two time points to test the hypotheses. We lacked information about how any changes to the physicians’ information might affect patients' choices. Future studies could use panel data to investigate the dynamic effects of physicians’ information on patients’ choices, as well as the dynamic effects of patients’ choices on the physicians’ service quality. 6. Conclusion This research explored the effects of service quality, as well as eWOM factors such as voting heat and Service Star, on a patient’s selection of a physician. We also investigated the roles played by disease risk and disease knowledge in these 16

ACCEPTED MANUSCRIPT

RI PT

relationships. We developed an mathematical model to test our hypotheses. The empirical results supported most of our hypotheses, except the moderating effects of disease risk on service quality and patients’ decisions. More research is needed to study the effects of other service qualities (such as information quality) and the influence of these qualities on a patient’s consulting intention. This paper can help academics understand better the evaluation and decision processes used by patients when considering online physicians, thereby contributing to online healthcare research. In addition, this work can provide some implications for the practice of online healthcare.

SC

REFERENCES

AC C

EP

TE D

M AN U

Akter, S., D’Ambra, J., & Ray, P. (2013). Development and validation of an instrument to measure user perceived service quality of mHealth. Information & Management, 50(4), 181-195. Angst, C. M., & Agarwal, R. (2009). Adoption of Electronic Health Records in the Presence of Privacy Concerns: The Elaboration Likelihood Model and Individual Persuasion. Mis Quarterly, 33(2), 339-370. Anindya Ghose, P. G. I., Beibei Li. (2012). Designing Ranking Systems for Hotels on Travel Search Engines by Mining User-Generated and Crowdsourced Content. Marketing Science, 31(3), 493-520. Ba, S., & Wang, L. (2013). Digital health communities: The effect of their motivation mechanisms. Decision Support Systems, 55, 941–947. Bhattacherjee, A., & Sanford, C. (2006). Influence Processes for Information Technology Acceptance: An Elaboration Likelihood Model. Mis Quarterly, 30(4), 805-825. Bi, S., Liu, Z., & Usman, K. (2017). The influence of online information on investing decisions of reward-based crowdfunding. Journal of Business Research, 71, 10-18. doi: http://dx.doi.org/10.1016/j.jbusres.2016.10.001 Cheung, C. M. K. (2008). The impact of electronic word-of-mouth: The adoption of online opinions in online customer communities. Internet Research, 18(3), 229-247. Chevalier, J. A., & Mayzlin, D. (2004). The Effect of Word of Mouth on Sales: Online Book Reviews. Social Science Electronic Publishing, 43(3), 345-354. Delone, W. H., & Mclean, E. R. (2004). Measuring e-Commerce Success: Applying the DeLone & McLean Information Systems Success Model. International Journal of Electronic Commerce, 9(1), 31-47. Donabedian, A. (2005). Evaluating the Quality of Medical Care. Milbank Quarterly, 83(4), 691–729. Fadel, K. J., Meservy, T. O., & Jensen, M. L. (2015). Exploring Knowledge Filtering Processes in Electronic Networks of Practice. Journal of Management 17

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Information Systems, 31(4), 158-181. Galehbakhtiari, S., & Pouryasouri, T. H. (2015). A hermeneutic phenomenological study of online community participation Applications of Fuzzy Cognitive Maps. Computers in Human Behavior, 48, 637-643. doi: 10.1016/j.chb.2015.02.005 Ghose, A., Ipeirotis, P. G., & Li, B. (2014). Examining the Impact of Ranking on Consumer Behavior and Search Engine Revenue. Management Science, 60(7), 1632-1654. Grönroos, C. (1993). A Service Quality Model and Its Marketing Implications. European Journal of Marketing, 18(4), 36-44. Guo, X. T., Zhang, X. F., & Sun, Y. Q. (2016). The privacy-personalization paradox in mHealth services acceptance of different age groups. Electronic Commerce Research and Applications, 16, 55-65. doi: 10.1016/j.elerap.2015.11.001 Ho, S. Y., & Bodoff, D. (2014). The effects of web personalization on user attitude and behavior: an integration of the elaboration likelihood model and consumer search theory. Mis Quarterly, 38(2), 497-520. Hu, N., Zhang, J., & Pavlou, P. A. (2009). Overcoming the J-Shaped Distribution of Product Reviews. Communications of the Acm, 52(10), 144-147. Iii, J. G. M. (2001). Service recovery's influence on consumer satisfaction, positive word-of-mouth, and purchase intentions. Journal of Business Research, 54(1), 11-24. Jin, J., Yan, X., Li, Y., & Li, Y. (2015). How users adopt healthcare information: An empirical study of an online Q&A community. International Journal of Medical Informatics, 86, 91-103. Lee, J., Park, D.-H., & Han, I. (2008). The effect of negative online consumer reviews on product attitude: An information processing view. Electronic Commerce Research and Applications, 7(3), 341-352. doi: http://dx.doi.org/10.1016/j.elerap.2007.05.004 Lin, C. A., Atkin, D. J., Cappotto, C., Davis, C., Dean, J., Eisenbaum, J., . . . Vidican, S. (2015). Ethnicity, digital divides and uses of the Internet for health information. Computers in Human Behavior, 51, 216-223. doi: 10.1016/j.chb.2015.04.054 Lu, N., & Wu, H. (2016). Exploring the impact of word-of-mouth about Physicians' service quality on patient choice based on online health communities. Bmc Medical Informatics & Decision Making, 16(1), 151. Mano, R. S. (2014). Social media and online health services: A health empowerment perspective to online health information. Computers in Human Behavior, 39, 404-412. doi: 10.1016/j.chb.2014.07.032 Marquart, F., & Naderer, B. (1986). Communication and Persuasion: Central and Peripheral Routes to Attitude Change. The American Journal of Psychology, 101(1), 123. Meservy, T. O., Jensen, M. L., & Fadel, K. J. (2014). Evaluation of Competing Candidate Solutions in Electronic Networks of Practice. Information Systems Research, 25(1), 15-34. 18

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Nikolay Archak, A. G., Panagiotis G. Ipeirotis,. (2011). Deriving the Pricing Power of Product Features by Mining Consumer Reviews. Management Science, 57(8), 1485-1509. Okazaki, S., Blas, S. S., & Castaneda, J. A. (2015). Physicians' adoption of mobile health monitoring systems in spain: Competing models and impact of prior experience. Journal of Electronic Commerce Research, 16(3), 194-217. Ozdemir, Z., Barron, J., & Bandyopadhyay, S. (2011). An Analysis of the Adoption of Digital Health Records Under Switching Costs. Information Systems Research, 22(3), 491-503. doi: 10.1287/isre.1110.0349 Parasuraman, A., & Berry, L. L. (1985). A Conceptual Model of Service Quality and Its Implications for Future Research. Journal of Marketing, 49(4), 41-50. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). ‘SERVQUAL: a Multiple-item Scale for Measuring Consumer Perceptions of Service Quality’. Journal of Retailing, 64(1), 12-40. Park, D.-H., Lee, J., & Han, I. (2007). The Effect of On-Line Consumer Reviews on Consumer Purchasing Intention: The Moderating Role of Involvement. Int. J. Electron. Commerce, 11(4), 125-148. doi: 10.2753/jec1086-4415110405 Petty, R. E., &Cacioppo, J. T. (1981). Attitudes and persuasion:Classic and contemporary approaches. W.C. Brown, Dubuque, IA. Petty, R. E., & Cacioppo, J. T. (1986a). The Ability to Elaborate in a Relatively Objective Manner Communication and Persuasion:Central and Peripheral Routes to Attitude Change (pp. 61-79). New York: Springer-Verlag. Petty, R. E., & Cacioppo, J. T. (1986b). The Elaboration Likelihood Model of Persuasion. Communication and Persuasion:Central and Peripheral Routes to Attitude Change (pp. 1-24). New York: Springer-Verlag. Petty, R. E., & Cacioppo, J. T. (1986c). The Motivation to Elaborate in a Relatively Objective Manner Communication and Persuasion:Central and Peripheral Routes to Attitude Change (pp. 81-109). New York: Springer-Verlag. Petty, R. E., Cacioppo, J. T., & Goldman, R. (1981). Personal Involvement As a Determinant of Argument-Based Persuasion. Journal of Personality & Social Psychology, 41(5), 847-855. Prendergast, G., Ko, D., & Yin, V. Y. S. (2010). Online word of mouth and consumer purchase intentions. International Journal of Advertising, 29(5), 687-708. Sussman, S. W., & Siegal, W. S. (2003). Informational Influence in Organizations: An Integrated Approach to Knowledge Adoption. Information Systems Research, 14(1), 47-65. Tam, K. Y., & Ho, S. Y. (2005). Web Personalization as a Persuasion Strategy: An Elaboration Likelihood Model Perspective. Information Systems Research, 16(3), 271-291. Xiao, N., Sharman, R., Rao, H. R., & Upadhyaya, S. (2014). Factors influencing online health information search: An empirical analysis of a national cancer-related survey. Decision Support Systems, 57, 417-427. doi: 10.1016/j.dss.2012.10.047 Yan, L., & Tan, Y. (2014). Feeling Blue? Go Online: An Empirical Study of Social 19

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Support Among Patients. Information Systems Research, 25(4), 690-709. doi: 10.1287/isre.2014.0538 Yang, H., Guo, X., & Wu, T. (2015). Exploring the influence of the online physician service delivery process on patient satisfaction. Decision Support Systems, 78, 113-121. Yang, H., Guo, X., Wu, T., & Ju, X. (2015). Exploring the effects of patient-generated and system-generated information on patients’ online search, evaluation and decision. Electronic Commerce Research & Applications, 14(3), 192-203. Yin, D., Mitra, S., & Zhang, H. (2016). When Do Consumers Value Positive vs. Negative Reviews? An Empirical Investigation of Confirmation Bias in Online Word of Mouth. Information Systems Research, 27(1), 131-144. Zeithaml, V. A., & Institute, M. S. (1987). Defining and relating price, perceived quality, and perceived value. Zhang, W., & Watts, S. (2008). Capitalizing on Content: Information Adoption in Two Online communities. Journal of the Association for Information Systems, 9(2), 73-94. Zhou, T. (2012). Understanding users’ initial trust in mobile banking: An elaboration likelihood perspective. Computers in Human Behavior, 28(4), 1518-1525.

20

ACCEPTED MANUSCRIPT

Fig. 1 Research concept model. Table 1 Variable description.

Table 3 Parameter estimates of the consulting intention.

RI PT

Table 2 Descriptive statistics and correlations of variables.

Table 4 Parameter estimates of the consulting intention (robust check).

AC C

EP

TE D

M AN U

SC

Table A. 1 Disease morbidity, mortality and classification of diseases.

21

ACCEPTED MANUSCRIPT

Disease knowledge

H4A(+)

H3A(+)

H4B(-)

H3B(-) The number of current patients who repeatedly interact with physician

Consulting intention (new patient next week)

H2(+)

M AN U

Vote heating

SC

H 1(+)

peripheral clue Electronic word-of-mouth

RI PT

central route Service quality

Disease risk

Service star

Fig. 1 Research concept model.

AC C

EP

TE D

Service quality and eWOM have positive effects on patient consulting intention. Disease risk increases the importance of service quality on intention, and decreases the importance of eWOM on intention. Meanwhile, disease knowledge amplifies the effects of service quality on intention and abates the effects of eWOM on intention.

22

ACCEPTED MANUSCRIPT Table 1 Variable description.

Peripheral route

Service star Voting heat

Showed in the profile homepage Star of the physician directly Heat

Disease risk

Whether a disease is According to the mortality rate of Risk high-risk the disease, we classify the disease into high-risk disease and low-risk disease Whether a disease is According to the morbidity of the Knowledge common disease, we classify the disease into common disease and uncommon disease Whether a physician If provided, then 1; Tel provides telephone If not provided then 0 services

Disease knowledge

Control variable

Telephone services

Visits

Opening months

Gifts

AC C

Hospital rating Hospital grade Position

City rank

If provided, then 1; If not provided then 0

Transfer

The logarithm of number of visit Visits by users How long has the home page Months existed? The number of thank Letters you letters The volume of digital Gifts gifts

EP

Letters

Whether a physician provides transfer treatment The number of visit by users Opening months

TE D

Transfer treatment

The number of patients who Current repeatedly interacted with a physician and whose latest update time was within a month before we collected the data

RI PT

Moderator variable

Description Abbreviation Patients evaluated information New and made a choice of physician

SC

Independent variable

Variable name Proxy Consulting intention The number of new patients who consult the physician next week Service quality The number of current repeat patients

M AN U

Variable type Dependent variable

Dummy variable (Level 3rd 1, else 0) The grade of hospital Dummy variable (Class A 1, else 0) A physician’s position Categorical variable (Director in physician hospital physician 3; Associate director physician; chief physician 1; else 0) The rank of city of the Categorical variable (Rank hospital 0,1,2,3,4,5) The rating of hospital

23

Hospital Rating Hospital Grade Position

City

ACCEPTED MANUSCRIPT

Variable Mean

Std. Dev.

Min

Max

1

RI PT

Table 2 Descriptive statistics and correlations of variables.

2

3

4

5

（

7

8

9

2.979

10.（85

0.000

190

1.000

Current

9.353

（5.042

0.000

45（2

0.542***

1.000

Heat

3.1（（

0.（（8

0.300

5.000

0.32（***

0.19（***

Star

0.244

0.790

0.000

5.000

0.583***

0.382***

0.395***

1.000

Tel

0.241

0.428

0.000

1.000

0.27（***

0.1（0***

0.302***

0.278***

1

Transfer

0.089

0.285

0.000

1.000

0.1（2***

0.094***

0.259***

0.218***

0.24（***

1

Visits

10.438

2.071

2.890

18.29（

0.2（8***

0.170***

0.457***

0.229***

0.3（1***

0.287***

1

Months

49.8（7

30.832

1.000

100.000

-0.05（***

-0.008

0.119***

-0.05（***

0.0（0***

0.1（0***

0.（（8***

Letters

5.439

1（.405

0.000

421.000

0.385***

0.335***

0.422***

0.359***

0.232***

0.282***

0.433*** 0.189***

Gifts

20.248

89.0（4

0.000

2751.000

0.411***

0.350***

0.318***

0.403***

0.220***

0.207***

0.381*** 0.129*** 0.739***

SC

New

AC C

EP

TE D

M AN U

1.000

24

1 1

ACCEPTED MANUSCRIPT Table 3 Parameter estimates of the consulting intention.

New

‘ 1

Current

2

3

0.066*** (61.06)

0.077*** (47.42) 0.016*** (7.06)

Patients*Risk 1.716*** (14.29)

Heat Heat*Knowledge Heat*Risk

-0.041*** (-17.76) 3.549*** (15.32) -1.529*** (-7.37) -1.327*** (-6.43)

RI PT

Patients*Knowledge

Tel

2.784*** (14.55)

2.01*** (11.66)

Transfer

1.582*** (5.74)

1.354*** (5.48)

Visits

1.607*** (28.2)

1.14*** (20.87)

1.067*** (19.77)

Months

-0.105*** (-31.13)

-0.078*** (-24.96)

-0.074*** (-24.06)

Letters

0.092*** (13.42)

0.03*** (4.78)

0.045*** (7.08)

Gifts

0.025*** (20.71)

0.016*** (14.87)

0.015*** (13.48)

-1.31（***

-1.112***

-1.17***

(-3.73)

(-3.52)

(-3.75)

-1.2（8***

-1.508***

-1.518***

(-4.4)

(-5.8)

(-5.93)

SC

M AN U

TE D

Rating

Grade

EP

Category control variables

AC C

Constant N F Prob > F R-squared Adj R-squared

T statistics in parentheses * P<0.1, **P<0.05, P<0.01

-7.767*** (-10.72) 15647 386.960 0.000 0.284 0.283

2.009*** (11.82)

1.339*** (5.48)

Controlled -9.386*** (-13.89) 15547 653.870 0.000 0.431 0.431

-14.794*** (-16.5) 15547 571.630 0.000 0.448 0.447

ACCEPTED MANUSCRIPT

Table 4 Parameter estimates of the consulting intention (robust check).

New

(1)

(2)

(3)

0.054*** (51.75)

2.784*** (14.55) 1.582*** (5.74) 1.607*** (28.2) -0.105*** (-31.13) 0.092*** (13.42) 0.025*** (20.71) -1.31（***

1.176*** (7.33) 0.17 (0.74) 1.064*** (22.21) -0.061*** (-21.42) 0.035*** (6.05) 0.007*** (7.06) -1.05***

0.061*** (37.97) 0.025*** (10.76) -0.041*** (-17.88) 5.273*** (37.53) -1.061*** (-6.37) -0.347** (-2.09) 1.179*** (7.44) 0.218 (0.96) 1.018*** (21.48) -0.059*** (-20.92) 0.049*** (8.45) 0.006*** (6.05) -1.01***

(-3.73)

(-3.58)

(-3.49)

-1.2（8***

-1.282***

-1.299***

(-4.4)

(-5.35)

(-5.49)

Current

Patients*Risk

4.653*** (51.01)

Star Star*Knowledge

Tel

M AN U

Transfer

SC

Star*Risk

RI PT

Patients*Knowledge

Visits Months Letters

TE D

Gifts Rating

EP

Grade Category control variables

AC C

Constant N F Prob > F R-squared Adj R-squared T statistics in parentheses * P<0.1, **P<0.05, P<0.01

-7.767*** (-10.72) 15647 386.96 0.000 0.284 0.283

Controlled

-4.199*** (-6.95) 15647 887.9 0.000 0.505 0.505

-4.244*** (-7.11) 15647 765.77 0.000 0.519 0.518

ACCEPTED MANUSCRIPT

Appendix. Table A. 1 Disease morbidity, mortality and classification of diseases.

Morbidity

Mortality

Lethality

Common rate (1/

Disease name

(1/1,000)

disease

100, 000)

Anxiety disorders

-

0

-

Depression

-

0

-

Bipolar disorder

-

0

-

Menstrual disorder

-

0

-

Obsessive-compulsive disorder

-

0

-

Nasopharyngeal carcinoma

-

0

Leukemia

-

0

Breast cancer

-

0

Gastric cancer

-

Liver cancer

-

Anemia Gastric ulcer Diabetes

EP

Liver cirrhosis Cerebral infarction

Coronary heart disease

0

-

0

-

0

-

0

1.5

1

3.84

1

4.37

1

0

19.19

1

0

2（.2（

1

1

-

0

0

10.3

1

-

0

0

8.3

1

0.03

0

0

2（.2

1

5.42

0.21

0

1.7

1

0.85

0.5

0

2.3

1

2.1（

0.94

1

5.（

1

15.45

2.7（

1

1.2

1

5.28

4.4

1

（.（

1

41.99

（.3（

1

14.3

1

107.5

7.52

1

TE D

Hypertension

-

SC

Hepatitis b

disease

0

M AN U

Gastritis

High-risk

-

8.（

Rheumatoid arthritis

rate (%)

RI PT

rate

AC C

Note: 1. The morbidity rate and mortality rate of disease are derived from the China Health Statistics Yearbook 2015, and the lethality rate equals mortality rate divide by morbidity rate. 2. According to morbidity rate, we divide these diseases into two types: common disease and uncommon disease. According to mortality rate and lethality rate, we divide these diseases into two type: high-risk disease and low-risk disease. 3. The China Health Statistics Yearbook includes the morbidity and relevant data for common diseases, but there are no such relevant data for uncommon diseases. 4. The China Health Statistics Yearbook includes the mortality and relevant data for high-risk diseases, but there are no such relevant data for low-risk diseases. However, some diseases, such as hepatitis B, hypertension, and anemia are common diseases, and their lethality rates are not high, so we classify them as low-risk diseases.

ACCEPTED MANUSCRIPT Acknowledgments and Statements

SC

RI PT

This work was funded in part by “Mobile Health” Ministry of Education - China Mobile Joint Laboratory, the Mobile E-business Collaborative Innovation Center of Hunan Province, and the National Natural Science Foundation of China (Grant Numbers 71210003, 71271219).We certify that we have complied with the ethical guidelines of Committee of Publications Ethics regarding research with human participants and/or care and use of animals in the conduct of the research presented in this manuscript. In addition, we appreciate the academic committee in “Mobile Health” Ministry of Education - China Mobile Joint Laboratory reviewed the project proposal and gave us the ethical approval. We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

AC C

EP

TE D

M AN U

 

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT Highlights Patients’ information processing affects their online health consulting intentions.

• •

The Elaboration Likelihood Model is validated in the online health context. Disease knowledge amplifies the effects of service quality on patients’ intention.

•

Disease knowledge and risk undermines the effects of eWOM on patients’ intention.

AC C

EP

TE D

M AN U

SC

RI PT

•