Introduction

With continued advancements in information and communication technology (ICT), online health consultation (OHC) platforms have become a popular source for patients to receive healthcare information and treatment, especially during the COVID-19 pandemic (Kumar et al., 2023; Nasseef et al., 2022; Zhang et al., 2020). The process of OHC provides patients with a means to acquire social support, including information and emotional support from doctors (Mirzaei & Esmaeilzadeh, 2021). The “online” setting of OHC offers many advantages compared to traditional face-to-face medical settings, such as time and location-independent access to support (Chen et al., 2022) and the opportunity to communicate anonymously, which makes it easier for patients to disclose sensitive or “taboo” information with less risk than in traditional offline healthcare consultation (Liu et al., 2022b; Sun et al., 2022). In fact, a nationwide survey in 2021 indicated that 28.9% of Chinese residents intended to obtain diagnosis and treatment through OHC.Footnote 1 Thus, OHC appears to be a popular doctor-patient communication pattern, ensuring an integrative social support environment where patients can obtain professional guidance and access help to manage illnesses (Wang et al., 2021a).

Receiving social support from doctors is essential in helping patients to deal with health challenges and improving overall well-being during OHC (Tan and Yan, 2020). Most previous studies have highlighted factors that influence patients’ social support acquisition from others in online health communities from the perspective of structural social capital (Chen et al., 2019), community engagement (Tseng et al., 2022), and social relationships (Lin and Kishore, 2021). Recently, from a patient-centered perspective, researchers have found that the amount of information patients self-discloses positively affects the social support received from doctors (Chen et al., 2020b; Liu et al., 2022a). However, such studies do not provide evidence on how patients write self-disclosure to elicit doctors’ social support from the perspective of linguistic features. Linguistic cues come from texts and include features such as the emotions expressed or the diversity of the vocabulary used (Siering et al., 2016). In computer-mediated communication, interaction is text-based and lacks some of the social cues that typically occur in face-to-face interactions, allowing linguistic features to play an increasingly important role (Jiang et al., 2022; Rain, 2016). Linguistic features do not necessarily change the content of text but can affect the willingness of the recipient to spend time and effort processing the information, which in turn affects reader behavior (Chen et al., 2020a; Wang et al., 2021c; Wu et al., 2024). Scholars have also suggested that linguistic signals, embedded in the text written by patients, could also affect the social support they receive from doctors (Chen et al., 2020a; Jiang et al., 2022). The influence of the language used by patients differs considerably from other contexts, with most patients tending to use plain language to describe complicated disease symptoms (Vasilevsky et al., 2018). Patient self-disclosure usually includes information about the illness, disease type, duration of the symptoms, clinical experience, current medication, past medical history, and expectation of online medical treatment. It is challenging, therefore, for doctors to specify the details of diseases, which largely determine the level of social support provided by the doctor (Hu et al., 2020). For example, the discrepancy between plain healthcare language (i.e., what patients actually use) and professional medical terms (i.e., what doctors try to understand) may change the ways in which linguistic features affect the process of doctor-patient interaction. While recent studies offer evidence for the implications of description design for acquiring social support from others in online health communities (Chen et al., 2020a; Jiang et al., 2022), few has provided evidence of how patient self-disclosure strategies influence the social support received from doctors from the perspective of linguistic cues.

At the same time, we identify significant gaps in our knowledge of the links between doctors’ social support in the process of the first OHC and patients’ follow-up online consultation behavior. Several studies provide evidence for patients’ continued online consultation behavior, which could enhance treatment efficiency, especially for patients suffering from chronic disease (Li et al., 2021; Yang et al., 2019). In addition, patients’ follow-up consultation behavior could also produce positive effects in increasing their trust towards the online health community (Wang et al., 2021b). Many studies reveal that the greatest benefit of doctors’ social support in online health communities is the strengthening of patient-community relationships, including patient satisfaction (Chen et al., 2020a), community commitment (Zhang & Liu, 2022), and community reputation (Chen et al., 2022). The primary issue is that extant research usually treats doctors’ social support behaviors as integrated behavior (e.g., relies on a Q&A template and free knowledge contribution behavior) in online health communities or chooses one specific activity (e.g., prosocial behavior) as the substitute for doctors’ social support (Wang et al., 2023a; Xia et al., 2022; Zhang et al., 2020), whereas social support behavior from doctors can be implemented in multiple online medical processes. The multiple types of doctors’ social support behaviors play different roles in shaping doctor-patient relationships (Chen et al., 2019; Liu et al., 2020b). However, how the various specific social support from doctors in the process of first OHC construct the doctor-patient relationships, which in turn promotes patients’ follow-up online consultation behavior, remains unclear. Considering these focuses, there is a need for research to understand the relationships between various social support mechanisms from doctors in the process of OHC and patients’ continued online consultation behavior.

Furthermore, drawing on signaling theory, we demonstrate how different types of patients’ self-disclosure content has diverse impacts on doctors’ social support behavior in OHC under different boundary conditions by introducing one moderating variable—patients’ offline visit experience (Ye & Wu, 2023). Compared with patients without offline visit experience, patients with previous offline medical treatment can upload their medical records, including previous health records containing tests conducted in hospitals (Chen et al., 2020b). As a result, doctors in the online environment can provide supplemental social support based on the offline medical records of patients, while reducing any repetitive statements (Xiang & Stanley, 2017). Accordingly, this study examines whether the impacts of linguistic signals in patients’ self-disclosure on doctors’ social support behavior is based on patients’ offline visit experience.

In summary, we design our study to answer the following three questions: (1) During OHC, how do the linguistic features of patients’ self-disclosure content affect the amount of social support provided by doctors? (2) Does the social support provided by doctors during the first online consultation affect patients’ continued consultation behavior? If so, how does this happen? (3) Does patients’ offline visit experience moderate the effects of the linguistic features of patients’ self-disclosure content on the social support received from doctors?

To empirically answer these questions, an integrated model, based on signaling theory and social support theory, is proposed. In total, 46,012 patient consultation records were collected from a prominent online healthcare platform in China. Then, data crawling and empirical analysis were completed to validate the hypotheses. Overall, this study contributes to the extant literature on online health consultation in three main ways. First, based on signaling theory, it provides a nuanced investigation of the effects of different types of linguistic features of patient disclosure content on doctors’ social support provided during the process of OHC. To the best of our knowledge, this study is among the first to compare quantitively patient disclosure effects under different signal types in the context of OHC. Second, it extends the social support literature by revealing the underlying mechanism of the effects of doctors’ social support on patients’ follow-up consultation behavior in the first stage of OHC. Finally, based on online-offline integration theory (Huang et al., 2021), this study expounds theoretically and demonstrates empirically the moderating role of patients’ offline medical treatment on the effects of patient self-disclosure and doctors’ social support behavior in the process of OHC. Overall, the findings of our study could serve as guidelines on how to write self-disclosure content to acquire more social support from doctors in the process of OHC.

The remainder of this research is structured as follows. The “Related literature and theoretical foundations” section reviews relevant literature on social support theory, signaling theory, and online-offline channel integration. The “Research hypotheses and model” section presents the proposed research model and hypotheses. Then, we describe the data source and variables in this study that are presented in the “Methods” section. The analysis in the “Results” section details the empirical results. The “Discussion and conclusion” section discusses the theoretical implications, practical implications, limitations, and future research.

Related literature and theoretical foundations

Social support from doctors in OHC

Social support refers to a process involving the exchange of resources whereby providers and recipients convey information and emotions through the exchange of verbal and non-verbal messages, helping to reduce the recipient’s uncertainty or stress (Shumaker & Brownell, 1984; Wang et al., 2021d). Researchers view social support as a multidimensional construct informational support and emotional support (Chen et al., 2020b; Hur et al., 2019; Tan & Yan, 2020; Tan et al., 2023). In the context of online health communities, patients’ informational support relates to the helpful messages conveyed by doctors, such as medical knowledge, experience, advice, treatments, and guidance to assist patients in addressing patient issues. In addition, emotional support mainly includes medical care, trust, encouragement, empathy, understanding, and compassion from doctors during online health consultations (Liu et al., 2020b; Tan et al., 2023; Yan & Tan, 2014).

Recently, the importance of social support has gained increasing attention in the fields of online health communities (Chen et al., 2019; Connolly et al., 2023; Goh et al., 2016; Oh & Lee, 2012; Park et al., 2020; Tan & Yan, 2020; Tseng et al., 2022; Wang et al., 2015).Footnote 2 On the one hand, the significant role of social support from doctors has been widely emphasized (Erfani et al., 2017; Oh et al., 2013; Yan & Tan, 2014). For instance, by employing a 5-year longitudinal data set, Song and Xu (2023) found that both emotional support and informational support received by patients are positively associated with the generation of self-reflective health updates in journals. Emotional support from doctors is also found to positively affect patient mental well-being and optimism (Gu et al., 2023; Liu & Jia 2023). Furthermore, several scholars also highlighted the significance of social support provided by doctors in shaping close relations with patients in online health communities. Chen et al. (2020b) indicated that social support from doctors has a positive influence on patient satisfaction in online health communities. Tan and Yan (2020) demonstrated that doctors’ informational and emotional support are essential factors to improve users’ perceived service quality during online healthcare consultation. Overall, there is existing evidence linking social support from doctors to patients’ decision-making in online health communities; current literature is yet to uncover the distinct mechanisms through which social support provided by doctors affects patients’ continued consultation behavior during online health consultation. However, patients’ continued consultation behavior could strengthen the relation between patient and online health communities. Thus, our research advances the current studies by investigating the impact of social support from doctors during online health consultation on patients’ continued consultation behavior.

Furthermore, among the reviewed studies, scholars in the field of online health communities also have focused on the antecedents of social support received by patients during online health consultation from various perspectives (Chen et al., 2019, 2020a, 2020b; Liu et al., 2022a; Marco Leimeister et al., 2008; Yan & Tan, 2014). Chen et al. (2019) have identified that individuals’ network position is an essential driver of the social support received from other agents within an online health community. Marco Leimeister et al. (2008) obtained similar findings. Several studies have also claimed that the amount of patient self-disclosure in online health communities might affect social support received from doctors. For instance, Chen et al. (2020b) revealed that offline hospital records, disclosed by patients, would shape the social support from online health community participants. Further, Chen et al. (2020a) and Jiang et al. (2022) highlighted the influence of linguistic signals embedded in patients’ self-disclosure on their social support from doctors. However, less attention has been paid to how patients can disclose disease information effectively to gain more social support from doctors during OHC. Therefore, our study was conducted to explore the characteristics of patients’ self-disclosure in disease description on their social support received from doctors.

Linguistic features in online communities

Linguistic features are derived from the text and encompass abstract aspects such as appropriate amount of information, readability, timeliness, relevancy, completeness, and sentiment (Chen et al., 2020a; Jiang et al., 2022; Ma et al., 2018; Wu et al., 2024; Xiang et al., 2017). Abundant research attention has been devoted to exploring the relationships between linguistic features and outcomes in online communities. For example, readability and comprehensibility of narratives in crowdfunding projects can affect financing outcomes (Wang et al., 2022). Chatterjee (2020) found that the sentiment of online customer reviews was related to the helpfulness of the reviews. In contrast to content-based features that focus on what is conveyed, linguistic features focus on how information is conveyed, indicating people’s attentional focus, emotions, social relationships, thinking styles, and personal characteristics (Wu et al., 2024; Zhang et al., 2023). The selection of appropriate linguistic features in a text could enhance the observation and understanding of the signal by the signal recipient, which is essential for improving the quality of communication (Wang et al., 2022). The sender’s message (using a certain language) affects the receiver’s perception and influences their behavior (Krishnamoorthy, 2015).

Informational signals and affective signals are the two main attributes embedded in text content (Chen et al., 2020a). Informational attributes can disclose highly accurate information to minimize information asymmetry between the communicating parties, while affective attributes express emotions or feelings (Wu et al., 2024). Text length and complexity (also referred to as readability) (Liang et al., 2019) have been applied to characterize information signals embedded in text in healthcare settings (Chen et al., 2023; Shah et al., 2021). With regard to emotional signals, affective expression includes both positive and negative emotional discourses that can trigger or evoke an emotional response in recipients, thereby influencing their engagement (Wu et al., 2024). Patients may intentionally include some form of emotional expression in messages to add meaning or connect with others within online healthcare communities (Myrick et al., 2016). Prior studies suggest that linguistic features embedded in online healthcare community posts affect the informational support and emotional support provided by others (Chen et al., 2023; Jiang et al., 2022). However, it is not clear how patients can use different linguistic features effectively in their self-disclosure to receive more social support from doctors during online healthcare consultation process. Therefore, we propose that when three types of linguistic features—complexity, text length, and affective expression—are embedded in patient self-disclosure in disease description, they may also affect doctors’ social support in OHC.

Signaling theory in OHC

Signaling theory has been developed to explain how one party uses signals to convey information to another party to facilitate an exchange in the context of information asymmetry (Shah et al., 2021; Zhang et al., 2019a, 2019b). Scholars have applied signaling theory to a wide range of contexts, such as corporate management (Miller and del Carmen Triana, 2009), marketing (Connelly et al., 2011; Mavlanova and Benbunan-Fich, 2010), and information systems. For example, Piazza et al. (2022) demonstrated that, in online crowdsourcing platforms, solvers send signals of their past experiences which highlight that they have, e.g., excellent skills in developing innovative and creative solutions.

In recent years, signaling theory has been applied to OHC to solve information asymmetry between doctors and patients (Ouyang & Wang, 2022). Although information asymmetry in OHC exists between both sides (i.e., the doctors and the patients), in the literature, the application of signaling theory has often reflected how the doctors serve as a signaler sending signals to patients (Zhang et al., 2019a, 2019b). In general, doctors will send signals to patients in order to help them distinguish between doctors with high- and low-quality medical care, such as their reputation, professional ranking, and online efforts (Liu et al., 2016; Shah et al., 2021; Zhang et al., 2019a, 2019b). For instance, Zhang et al., (2019a, 2019b) stressed that free services are signals about the quality of the doctor, which help the patient to make suitable choices. There is a shortage of literature that systematically discusses how patients clearly articulate their disease conditions to reduce information asymmetry. This is particularly important considering that the primary communication signal between doctors and patients is textual, in which the efficiency of signaling relies more on signals that are accurately sent, observed, and understood (Wang et al., 2022).

Traditional signaling theories have always ignored the impact of rhetorical signals (language-based messages). Recent research in signaling theory, however, has demonstrated the importance of rhetoric as a signal, in which the signal sender indicates qualities or attracts attention through the linguistic features, and reveals that it has been widely applied in online communities (Moradi et al., 2023; Steigenberger & Wilhelm, 2018; Wu et al., 2024). This literature provides a theoretical foundation for using signaling theory to analyze the role of linguistic features in healthcare settings. Patients can choose appropriate linguistic features in their self-disclosure to enhance the observability and comprehensibility of the signals, allowing doctors to better understand their disease condition and patients to receive effective treatment options from their doctors. Accordingly, based on signaling theory, this study investigates how linguistic signals in patient self-disclosure induce social support provided by doctors in OHC.

Online-offline channel integration

Online-offline channel integration refers to a phenomenon where platforms integrate existing offline operations into online channels or vice versa (Huang et al., 2021). In the context of medical care, repeated interactions and continuous responses between service providers and patients are required across the two channels (Ayabakan et al., 2017). The integration of online-offline services plays an important role in enabling the continuous provision of medical care to patients and the synchronization of electronic medical records (Goh et al., 2016).

Extant research on online-offline channel integration in online healthcare is analyzed mainly from the perspective of doctors and patients. For doctors, online-offline channel integration benefits them through the fact that doctors’ service integration will increase the total number of consultations and the number of online consultations they receive (Huang et al., 2021; Shah et al., 2021). Specifically, regarding the interplay between online and offline services, studies have shown that free online consultations are complementary to offline consultations. Paid online consultations act as substitutes for offline consultations (Wu & Lu, 2017; Yu et al., 2016), while an increase in offline consultations can lead to a decrease in online consultations (Wang et al., 2020). For patients, e-health consumption will increase the chances of patients’ hospital consultations (Xiang & Stanley, 2017). However, previous research has focused mainly on information and service integration related to healthcare providers. The existing literature is largely silent on how the integration of patients’ offline behaviors and online activities affects the quality of doctor services, and how the integration of patients’ offline behaviors with doctors’ online behaviors predicts subsequent patient behaviors.

Research hypotheses and model

Based on signaling theory and social support theory, this study aims to examine how sentence complexity, text length, and affective expression in patient self-disclosure as forms of linguistic features explain variations in doctors’ social support. This study also explores the moderating role of patients’ offline visit experience. Further, we analyzed whether there are differences in the impact of social support received from doctors on patients’ continued consultation behavior between patients with and without offline visit experience. The research model adopted in this study is shown in Fig. 1.

Fig. 1
figure 1

Research model

Full size image

The effect of linguistic features of patient self-disclosure

The effect of sentence complexity

Sentence complexity is used to indicate the degree of comprehension of texts, which predicts readability and has been shown to have an impact on the processing needs of information receivers (Lee, 2022; Lu et al., 2022). During OHC, sentence complexity is used to indicate the writing style of patient self-disclosure, making the meaning of the self-disclosure content presented easier to understand or more comprehensive. This section explains that the sentence complexity found in patient self-disclosure presents both benefits and pitfalls for obtaining doctors’ social support. We contend that sentence complexity can be a double-edged sword.

When sentence complexity is low to moderate, it can significantly increase the amount of social support provided by doctors. First, complex sentences help readers both temporally and causally to sequence events and improve narrative coherence (Kuan et al., 2015). Complex sentences may enhance the perceived reasoning of the text, such that readers’ trust and recognition of the text would also be increased (Lu et al., 2022). Furthermore, when sentence complexity is kept at a moderate level, it is the reader’s motivation rather than the ability to process the information (Bradley & Meeds, 2002). Text that is easy or difficult to comprehend is influenced by the intrinsic motivation of the readers (Hu et al., 2020). Doctors are intrinsically motivated to engage in challenging activities, due to the fact that they are highly educated and high-skilled readers. Doctors who are highly skilled readers are skilled at deftly unraveling complex patient disease descriptions and obtaining the required information (Lu et al., 2019). As a result, they can process patients’ information more efficiently, integrate more comprehensive information and patients’ psychological reactions, and instigate more treatment and appeasing behaviors (Chen et al., 2020a, 2020b). Accordingly, higher sentence complexity can stimulate doctors to provide more informational support and emotional support.

Nevertheless, when the sentence complexity reaches a certain level, too complex sentences may lead to negative effects and incur highly impenetrable text (Stone & Lodhia, 2019). Human attention in a specific environment is limited. When people are presented with a large amount of information, they must filter it for selective reading (Wang et al., 2022). When encountering texts with unnecessary and unfamiliar long sentences, readers will spend more time understanding and may impede readers’ understanding of the intended message (Courtis, 2004). When sentences are too complex, doctors must spend more time understanding the meaning of the text; thus, considering the limited time capacity of doctors, they need to expend more effort in processing the information when sentences are too complex, resulting in texts being less attractive to support providers. This ultimately results in doctors providing less informational and emotional support (Chen et al., 2020a). Therefore, this study proposes the following hypotheses:

  • H1: The sentence complexity of patient self-disclosure has an inverted-U relationship with the amount of informational support provided by doctors.

  • H2: The sentence complexity of patient self-disclosure has an inverted-U relationship with the amount of emotional support provided by doctors.

The effect of text length

In addition to the complexity of texts, another linguistic feature that can be identified is the length of the text. In online health communities, the length of posts is a direct indication of information completeness (Chen et al., 2020a). The longer text implies a greater density of information cues disclosed by patients (Liu et al., 2022a). The more information the patients disclose, the more feedback they will receive from doctors (Lutz et al., 2022). Therefore, this study assumes that text length increases the social support provided by doctors.

Long texts contain a greater density of information clues (Hu et al., 2020), allowing doctors to provide higher-quality responses to patient needs. In the context of this study, the length of patient self-disclosure increases the informational and emotional support received from doctors. As patients communicate with doctors, disclosing more information denotes that they want the doctor to comprehend their symptoms and feelings (Liu et al., 2022a). Longer texts are more detailed and are also likely to contain persuasive arguments which may decrease readers’ uncertainty (Mousavizadeh et al., 2022). As a result, doctors can read longer content to integrate more comprehensive information and make more accurate decisions (Osei-Frimpong et al., 2018), providing better informational support (i.e., disease-specific knowledge, advice, and referrals) and emotional support (i.e., caring and concern) (Yan & Tan, 2014). Thus, for patients, the text reported in the self-disclosure content should involve a complete representation of their condition, the treatment, and what help is required, which will allow more informational or emotional support to be provided by the doctor (Liu et al., 2022a). This study, therefore, proposes the following hypotheses:

  • H3: The text length of patient self-disclosure positively affects the informational support provided by doctors.

  • H4: The text length of patient self-disclosure positively affects the emotional support provided by doctors.

The effect of affective expression

This study proposes that patients’ affective expression positively influences the informational and emotional support provided by doctors. Researchers have found that the social sharing of emotions is an interpersonal process by which people share their emotions with others as a way to gain support, evoke empathy, gain attention, and enhance a link with society (Chen & Farn, 2020). People will extract social information from others’ affective expressions and use this information to guide their behaviors (Chen & Zhang, 2022). In OHC contexts, patients that embed emotional words in their self-disclosure will motivate doctors to respond, especially in a primarily text-based communication environment (Kumar et al., 2022). Affective expression provides illuminating cues that stimulate greater social support exchange between patients and doctors through emotion-invoking linguistic mechanisms (Chen et al., 2020a). In the process of doctor-patient communication, when patients initiate an empathic opportunity, doctors respond in two ways. Firstly, medical knowledge is provided by the doctor in response to the patient’s signals of an illness-related concern. Secondly, the doctor expresses care and concern in response to the patient’s emotional cues (Wang et al., 2020). Thus, affective expression, an important form of the purpose of medical communication, can foster supportive relationships between patients and doctors and should be included in the theoretical construct as an integral part of the exchange of social support. Therefore, this study proposes the following hypotheses:

  • H5: Affective expression contained in patient self-disclosure positively affects the informational support provided by doctors.

  • H6: Affective expression contained in patient self-disclosure positively affects the emotional support provided by doctors.

The moderating role of patients’ offline visit experience (OVE)

The moderating role of OVE on the relationship between patients’ linguistic features of patient self-disclosure and doctors’ social support

This section explains the moderating role of OVE in the relationship between linguistic signals provided by patients in their self-disclosure and the doctors’ social support. This study hypothesized that patients with OVE received less informational support compared to patients without OVE in online healthcare consultation, while patients with OVE received more emotional support compared to patients without OVE.

Regarding informational support, for patients with OVE, the doctor will have already performed a basic examination, providing information about treatment options during previous offline consultations (Lee & Hawkins, 2010). Furthermore, patients often provide online doctors with electronic medical records (EMC) from their offline visits, which may substitute for linguistic features in online consultations (Li et al., 2021). When confronted with both patient self-disclosure and uploaded EMC, increased cognitive and time demands for doctors to focus on EMC, which leads the doctor to reduce the interpretation and perception of the signals provided in the patients’ self-disclosure (EIKefi and Asan, 2021). Therefore, the effects of linguistic features of patient self-disclosure will become less important.

However, with regard to emotional support, the effects of linguistic features of patient self-disclosure on emotional support were stronger for patients with OVE. Patients’ needs for emotional support are long term (Wang et al., 2015). For patients who have OVE, they have more knowledge or information about the course, treatment, and prognosis of their disease. In this situation, they possess high levels of health literacy and experience negative emotions, including fear, anxiety, worry, anger, and worthlessness (Chen et al., 2019; Lee & Hawkins, 2010). Thus, patient self-disclosure may contain more heath care language and negative emotions, which will enhance the doctors’ comprehension of their feelings. Accordingly, the impact of the linguistic features of patient self-disclosure on doctors’ emotional support will be strengthened (Wang et al., 2015). Therefore, this study proposes the following hypotheses:

  • H7: The impact of the linguistic features of patient self-disclosure on doctors’ informational support is weakened by patients’ OVE.

  • H8: The impact of the linguistic features of patient self-disclosure on doctors’ emotional support is strengthened by patients’ OVE.

The moderating role of OVE on the relationship between doctors’ social support and patients’ continued consultation behavior

This study hypothesized that the impact of doctors’ social support during the first consultation on patients’ subsequent consultation behaviors will vary depending on the patients’ OVE.

Informational support

We conjectured that the effects of the informational support provided by doctors during the first consultation on patients’ continued consultation behaviors will differ for patients with and without OVE. For patients with OVE, doctors’ informational support will encourage the patient to choose the same doctor for follow-up consultations. For such patients, online consultation is an important supplement to offline consultation (Li et al., 2021). When the patients finish the offline diagnosis and treatment, they will develop quasi-professional knowledge of their health condition, which could enhance the perception of the quality of doctor services during OHC (Osei-Frimpong et al., 2018; Yan & Tan, 2014). Then, at the post-diagnosis step, patients will have a better understanding of the symptoms and possible solutions from the doctor’s response during OHC, and they will continue to consult this doctor in the future (Li et al., 2021). Further, understanding the patient’s condition reduces information asymmetry between doctors and patients and helps to improve the efficiency of doctor-patient communication (Wu et al., 2021). Thus, the positive effect of doctors’ information support is strengthened for patients with OVE.

However, for patients without OVE, doctors’ informational support provided during the first consultation may reduce the chances of patients selecting the same doctor for future consultations. Healthcare consultation services still rely on offline services at hospitals; online consultation is usually used as a screening step for patients without OVE (Zhang & Liu, 2022). Since OHC is internet-based, all contact is virtual; therefore, OHC cannot provide tangible diagnoses, such as an examination involving doctor appraisals (e.g., palpation) (Liu et al., 2022a). Two problems may be triggered in this context. On the one hand, the lack of examination leads to limited treatment recommendations being provided by doctors, and on the other hand, it is more difficult for the patient to understand the professional advice being given by the doctor (Virlée et al., 2020; Yan et al., 2020). These will lead to limited trust or confidence in online doctors, and patients switch doctors frequently or switch to offline hospital visits (Yang et al., 2019). Thus, this study proposes the following hypothesis:

  • H9: The effect of doctors’ informational support provided during the first consultation on patients’ continued consultation behavior is contingent on patients’ OVE. The informational support received from doctors positively (or negatively) affects patients’ continued consultation behavior for patients with (or without) OVE.

Emotional support

This study hypothesized that the effect of emotional support provided by doctors during the first consultation on patients’ continued consultation behavior depends on the patients’ OVE. When patients have OVE, it is suggested that doctors’ emotional support will positively affect patients’ continued consultation behaviors. For patients with OVE, OHC may provide an opportunity to access an alternative source of care and comfort, especially if they had a negative experience during offline consultation (Xing et al., 2020). Even if they have had a positive experience with face-to-face interactions, online consultations may provide an additional expression of the doctor’s understanding of patient concerns (Zhang et al., 2018). They are more likely to participate in interactions with doctors during OHC and experience feelings of closeness and belonging, which boosts their self-confidence to solve health problems and substantially affects their decision-making behaviors (Lu, 2023). Therefore, after completing an online consultation, the doctors’ supportive and encouraging responses help develop intimate and trusting relationships, and patients will continue to choose the same doctor if they need to in the future (Ju & Zhang, 2020).

Conversely, for patients without OVE, the emotional support received from doctors during the first consultation will reduce patients’ continued consultation behavior. For patients who have not experienced a traditional offline consultation, their main objectives are to seek information about treatment and advice about their symptoms (Wu & Lu, 2017). During offline consultation, doctors’ emotional support includes an encouraging tone and verbal support (Ong, 1995), while in online consultations, the informational and emotional support provided by doctors is mainly presented in text format (Lee and Zuercher, 2017). Since the cognitive and information processing capacity of humans is limited, patients put the information exchanged in doctor-patient interactions at the center of attention; less attention might be paid to emotions (Van Oerle et al., 2018). Excessive emotional support in the form of text can distract patients’ attention and interfere with the acquisition of informational support due to the patient’s lack of knowledge and limited energy (Tan & Yan, 2020; Wu et al., 2021). Therefore, excessive emotional support during OHC will impede the patients’ continued consultation behavior. Thus, this study proposes the following hypothesis:

  • H10: The effect of doctors’ emotional support provided during the first consultation on patients’ continued consultation behavior is contingent on patients’ OVE. Emotional support received from doctors positively (or negatively) affects patients’ continued consultation behavior for patients with (or without) OVE.

Methods

Research context

To avoid the self-reporting bias of survey data, this study used objective data collected from haodf.com, one of China’s most reputable online healthcare platforms. At present, haodf.com already provides services such as online consultation, telephone consultation, outpatient information inquiry, offline appointments, popularization of disease knowledge, and family doctors. Currently, more than 9000 hospitals and over 880,000 doctors are registered on the platform.

The platform provides a means for patients to consult doctors and integrates online and offline medical resources. Information about doctors can be browsed on the website, including information about the quality of medical care provided by the doctors themselves, and the information generated by patients, such as the number of online visitors and digital gifts. Furthermore, previous doctor-patient communications can be viewed while preserving patient confidentiality, which provides important material for scholars. Therefore, this platform was chosen as a natural research context for empirical analysis.

Data collection

A python-based web crawler was used to collect data from the haodf.com platform with patients being used as analysis units. Figure 2 shows the process of data acquisition and processing. To fit our research questions and ensure validity, the following steps were taken in selecting the final sample. First, three different types of diseases were studied, including serious disease (i.e., lung cancer), chronic disease (i.e., diabetes), and mental disease (i.e., depression). These three diseases are common illnesses and often need regular and repeated consultations (Yang et al., 2019; Zhang et al., 2019b). By focusing on them, we can ensure that we have data from multiperiod consultations.

Fig. 2
figure 2

Process of data acquisition and processing

Full size image

Second, we searched for doctors using the three diseases as keywords, and 300 doctors were randomly selected for each disease separately. We deleted those doctors who had important information missing and who had not logged in for a long period. Then, the major attributes of doctors from their respective homepage (i.e., gender, professional title, hospital level, city, recommendations, grateful letters, service level, and warmth gifts) were obtained (see Fig. 3). After data cleaning, we obtained a sample dataset composed of 686 doctors.

Fig. 3
figure 3

Doctor’s homepage on haodf.com

Full size image

Third, we collected detailed information on patient consultation records from January 2017 to December 2021. Our collection regarding the patient consultation process could be divided into three stages. Before doctor-patient interaction, the text of patient self-disclosure (called Disease Description in haodf.com) and previous consultation experience were collected (see Fig. 4). During doctor-patient interaction, a Python program was used to automatically download web pages containing information on doctors responding to patients’ questions. We excluded patient consultations with no more than three doctor-patient interactions as they were ineffective for patient disease management. After doctor-patient interaction, patient follow-up consultations were also tallied. In the end, a total of 46,012 patient consultation records were analyzed, including 16,588 patients with lung cancer, 13,944 patients with diabetes, and 15,480 patients with depression.

Fig. 4
figure 4

Patient description page on haodf.com

Full size image

Variables measurement

Table 1 shows the variables used in this study and describes how they were measured.

Table 1 Variable measurement
Full size table

Continued Consultation Behavior (CCB): This variable refers to whether patients choose to continue with the same doctor for follow-up online consultations, which is denoted as Continued Consultation Behavior (Yang et al., 2019). “Continue to choose the same doctor in subsequent online consultations” was coded as 1, while “stop choosing the same doctor for subsequent online consultations” was coded as 0.

Linguistic Features in patient self-disclosure: The unit of this section is each patient’s Disease Description. The linguistic features provided in the Disease Description consist of sentence complexity, text length, and affective expression (Chen et al., 2020a; Shah et al., 2021; Wu et al., 2024; Xu et al., 2019). The sentence complexity (CO) of the text in this study applies a readability index (Bischof and Senninger, 2018; Liang et al., 2019; Qiao et al., 2022). Readability is seen as an important indicator of how well a person perceives a text to be understandable (Zhuang et al., 2023), which is used to estimate the complexity of the text (Bischof and Senninger, 2018; Liang et al., 2019; Qiao et al., 2022). There are already widely recognized indicators for measuring the readability of English texts, such as ARI, FOG, CLI, and FKGL (Qiao et al., 2022; Zhuang et al., 2023), while for Chinese texts, there are no specific indicators to measure readability. Therefore, we use the indicator developed by Xu et al. (2019) to assess the sentence complexity, which measures the readability of Chinese text in terms of the proportion of adverbs and conjunctions in each sentence. In Chinese contexts, adverbs are used before verbs or adjectives to the effect of modifying the original meaning of the word. The conjunctions can change the complexity of the progressive, transitive relationship of a sentence. Therefore, the proportion of adverbs and conjunctions in a sentence was applied to measure complexity in this study (Xu et al., 2019). The Text Length (TL) is the total number of words contained in the Disease Description provided by the patient (Chen et al., 2020a). Moreover, in this study, the patient’s emotional cues are expressed as Affective Expression (AE), which represents the number of emotional phrases used in the Disease Description (Chen et al., 2020b; Liu et al., 2022b). This variable was analyzed using the TextMind, the automatic Chinese words segmentation developed by the Chinese Academy of Sciences (Liu & Gao, 2021).

Social Support from doctors: In this study, the number information providing behaviors by the doctor in one consultation was used to measure the doctor’s Information Support (IS) (Tan & Yan, 2020). Text categorization methods were applied to identify a doctors’ information providing behaviors. We classified each sentence of a doctor’s response into binary categories to determine whether the response was an act of information support behavior. In this research, information and instructions provided by doctors related to health conditions, treatment plans, and lifestyles were identified as doctor-provided information, such as “There has been a significant increase in the extent of the ground-glass nodule in the right upper lung,” “Increase the dose of insulin to 16 units at bedtime,” and “Exercise, appropriate stress reduction, and hobbies could be taken up.” Firstly, 9605 randomly selected sentences from doctors’ replies were manually labeled as the basic corpus for classification model selection. Second, after removing space returns and other semantically irrelevant information data preprocessing after word splitting, text features were extracted by the term frequency-inverse document frequency (TF-IDF). Third, binary classification models based on naive Bayesian, logistic regression, and support vector machine (SVM) were established. Finally, the optimal classification algorithm SVM was selected. In this model, the precision is 0.8788, the recall is 0.9063, and the F1 value is 0.8923. According to previous research (Chen et al., 2020b; Tan & Yan, 2020), this study measured Emotional Support (ES) by the total positive emotional words used by the doctors in their answers. This variable was analyzed using TextMind, the automatic Chinese words segmentation system developed by the Chinese Academy of Sciences (Liu & Gao, 2021).

Offline Visit Experience (OVE): This study analyzes the moderating role of patients’ offline visit experience prior to the OHC in the effect of the linguistic features in Disease Description on doctors’ social support (Ye & Wu, 2023). “With offline visit experience” was coded as 1, and “without offline visit experience” was coded as 0.

Control variable: Several doctor-level variables were used for control purposes in this study, including the doctor’s gender (GE), doctor’s title (PT), doctor’s hospital level (HL), doctor’s city level (CL), doctor’s recommendations (RE) given by the platform, and doctor’s rating/gifts provided by the patients (denoted as Grateful Letters (GL), Service Level (SL), and Warmth Gifts (WG)).

To reduce the skewness of the variables, we take the natural logarithms of our variables (e.g., the log-transformed variable TL is lnTL). To address the zeros in the data, we take the logarithm of the value of the variable plus one (including GL, WG, AE, IS, ES).

Data analysis

Two regression models were used to test the proposed research model respectively. First, a hierarchical regression model was used to verify the effects of CO, TL, and AE on IS and ES (Meng et al., 2023; Tan & Yan, 2020; Yang et al., 2019). Given that IS and ES are log-transformed and are continuous variable, the models were estimated by ordinary least squares (OLS) (Ye and Wu, 2023). The model was validated in three stages. The first stage contained only control variables, the second stage added control variables and independent variables, and the third stage contained control variables, independent variables, and moderating variables. Second, since CCB is a binary variable, logit regression was used to verify the effects of IS and ES on CCB. Logit regression can test binary dependent variables and can provide unbiased and lower variance estimates (Yang et al., 2019).

Results

The descriptive statistics and the correlation matrix of the variables are presented in Table 2. A variance inflation factor (VIF) test was conducted to detect any potential multicollinearity problems. The values of VIFs in our estimation models for all the independent variables were found to be below 5 which is less than the threshold of 10 (Choi & Leon, 2020; Zhao et al., 2022). Multicollinearity concerns, therefore, were low in our estimation.

Table 2 Variable statistics and correlations
Full size table

Table 3 aims to quantify the effects of CO, TL, and AE on IS and the moderating effect of OVE. Table 4 shows the effects of CO, TL, and AE on ES and the moderating effect of OVE. Table 5 presents the estimated results of the effects of IS and ES on CCB.

Table 3 Estimates of IS from doctors
Full size table
Table 4 Estimates of ES from doctors
Full size table
Table 5 Estimates of CCB
Full size table

Main effects

In Table 3, model 1 only considers the control variables. Models 2, 3, and 4 were used to verify the effects of CO, TL, and AE on IS, respectively. It is evident that the result of model 2 shows that CO has an inverted-U relationship with IS. Thus, H1 is supported. In model 3, TL has a positive effect on IS (β = 0.0758, p < 0.001) and, thus, H3 is supported. In addition, in support of H5, it can be seen from the findings of model 4 that as the AE increases, the IS increases (β = 0.0160, p < 0.001). As such, H5 is supported.

Model 1 in Table 4 only considers the effect of the control variables on ES. The results of models 2, 3, and 4 show the effects of CO, TL, and AE on ES, respectively. The results of model 2 reveal that CO has a significant inverted-U relationship with ES. Thus, H2 is supported. In model 3, the effect of TL on ES is significantly positive (β = 0.0439, p < 0.001) and, thus, H4 is supported. The results shown in model 4 indicate that AE predicts ES positively (β = 0.0491, p < 0.001) and, therefore, H6 is supported.

The differences in the effect of doctors’ social support on patients’ CCB between patients with and without OVE are shown in Table 5. In the sample, 18,443 patients had OVE, while 27,569 patients had no previous OVE. The results of model 2 suggested that for patients with OVE, IS has a positive and significant effect on CCB (β = 0.0213, p < 0.001), while for patients without OVE, IS negatively affects CCB (β =  − 0.0362, p < 0.05) (model 4). Therefore, H9 is supported. Models 3 and 5 were used to validate the effect of ES on CCB. In model 3, the effect of ES on CCB is significantly positive for patients with OVE (β = 0.2216, p < 0.001). The results shown in model 5 indicate that ES predicts CCB negatively for patients without OVE (β =  − 0.2764, p < 0.001). Therefore, H10 is supported.

Moderating effects

Having estimated the main effects of the patients’ linguistic signals contained in the Disease Description on IS, models 5, 6, and 7 of Table 3 further validate the moderating role of OVE. In model 5, the coefficients of the interaction of OVE with CO^2 were not significant, indicating that OVE has no moderating role in the relationship between CO and IS. In model 6, OVE negatively moderates the relationship between TL and IS (β =  − 0.0351, p < 0.001). In model 7, the result of the moderating effect of OVE indicates that OVE negatively influences the relationship between AE and IS (β =  − 0.0017, p < 0.05). Therefore, H7 is partly supported.

Further, models 5, 6, and 7 of Table 4 further validate the moderating role of OVE in the effect of patients’ linguistic features contained in the Disease Description on ES. The result of model 5 shows that the coefficients of the interaction of OVE with CO^2 were significant (β =  − 2.2798, p < 0.05). The results in models 6 and 7 showed that the moderating effects of OVE were not significant in the effects of TL and AE on ES. Therefore, H8 is partly supported.

Robustness check

A series of robustness checks were performed to ensure that the results were robust and not due to other explanations. The basic and moderating effects in regression models were re-run to test the results and robust linear regression analysis was performed to validate the proposed model. The results are presented in Tables 6 and 7. Probit regression was also performed to test the effect of ES on CCB. The results are shown in Table 8. The main results were consistent with the previous model. Further, we measure CCB with the number of times of patients’ continued consultation. Since this variable was the count variable, Poisson regression methods were adopted to examine the effects of IS and ES on CCB. Relevant results are summarized in Table 9.

Table 6 Robustness check—estimates of IS from doctors
Full size table
Table 7 Robustness checks—estimates of ES from doctors
Full size table
Table 8 Robustness check—estimates of CCB 1
Full size table
Table 9 Robustness check—estimates of CCB 2
Full size table

Discussion and conclusion

Key findings

This study offers several important findings which extend the current understanding of the linguistic features of patient self-disclosure content and their effects on the social support provided by doctors during OHC. Firstly, complexity, text length, and affective expression, three sub-dimensions of linguistic features of patient self-disclosure, have different effects on the level of doctors’ social support. Specifically, the complexity of patient self-disclosure content has an inverted U-shaped relationship with the social support received from doctors. Essentially, if the complexity of patients’ Disease Description is too high or too low, the social support received from doctors will reduce. However, this study also reveals that the text length of patient self-disclosure exerts a significant positive effect on the social support provided by doctors. Consistent with the findings of Chen et al. (2020b), we provide further empirical evidence to prove that the more detailed disease information provided by patients, the more social support is stimulated. Furthermore, this study’s results show that affective expression in patient self-disclosure content is positively associated with the social support received from doctors. This finding confirms the value of emotional dynamics in the social exchange process (Tóth et al., 2022). During OHC, the patient’s expressed emotions are more likely to trigger positive responses from doctors.

Secondly, our findings demonstrate that patients’ offline visit experience moderates the effects of the linguistic features (i.e., complexity, text length, and affective expression) contained in patient self-disclosure content on the social support received from doctors. Specifically, patients’ offline visit experience mitigates the effects of text length and affective expression contained in patient self-disclosure content on the informational support received from doctors. In other words, the patients’ offline visit experience decreases the importance of text length and affective expression of patient self-disclosure for doctors’ informational support. The reason for this can be attributed to the fact that doctors may focus on the hospital medical records uploaded by patients with offline visit experience to provide medical information (Li et al., 2021). Accordingly, it is necessary for patients with offline visit experience to increase their text length and affective expression to get more information support from doctors. Furthermore, for patients with offline visit experiences, the positive relationship between complexity and doctors’ emotional support was enhanced when complexity was low. Hence, within a moderate range of complexity, patients without offline visit experience need to make greater effort to increase complexity in their self-disclosure so that they receive more emotional support from doctors.

Finally, the results of this study identified mechanisms by which doctors’ behavior during the first-period consultation influenced patients’ decisions to continue their consultations in subsequent periods. Using large-scale patient consultation records collected from a leading and popular OHC platform, we found that the impact of doctors’ social support on patients’ continued consultation behavior can vary depending on the patient’s previous consultation experience. Specifically, for patients with offline visit experience, informational and emotional support received from doctors has a positive effect on their continued consultation behavior. However, for patients without offline visit experience, the informational and emotional support provided by doctors has a negative effect on their continued consultation behavior. Although existing literature has shown that patient’ offline visit experience enhances doctors’ online service quality, this study advances our understanding of how the mechanism of how patients’ offline visit experiences and doctors’ online behaviors are integrated to influence patient follow-up behaviors (Ye and Wu, 2023).

Theoretical implications

This study makes several theoretical contributions to prior literature. First, we enrich signaling theory and social support theory by examining the impacts of the linguistic features contained in patient self-disclosure content on the social support received from doctors. Extant research on patients’ social support antecedents has mainly considered their social capital in online health communities, such as virtual social relationships (Marco Leimeister et al., 2008) and structural social capital (Chen et al., 2019), while little research has examined how to disclose disease information more effectively. Based on signaling theory, this study demonstrates how linguistic features (i.e., complexity, text length, and affective expression) stimulate the social support given by doctors during online consultations. This research, therefore, contributes to extending the application of the existing signaling theory in the context of OHC. In addition, this study sheds light on the formulation of patient self-disclosure content by integrating multiple linguistic features to gain greater social support from doctors.

Second, we contribute to the literature on OHC and patient follow-up behavior, as well as the influencing factors. This research contributes to this literature stream by establishing a two-stage model that empirically verifies the importance of doctors’ social support during first-period consultations, which has significant implications for patients’ follow-up consultations. Prior studies on antecedents of patient continued consultation behavior have mainly focused on patient psychological mechanism, such as trust, perceived justice, satisfaction, and service quality (Hong et al., 2019; Ju & Zhang, 2020; Li et al., 2021; Yang et al., 2021). Further, almost all research use questionnaire survey data and did not use real data, which may lead biased results. Further, almost all research uses questionnaire survey data and did not use real data, which may lead to biased results. However, limited efforts have been made to explore the impacts of doctor-related factors on patient continued consultation behavior through objective data. Broadly speaking, there is insufficient information to explain why doctors’ behaviors affect patient retention. This study fills this gap by empirically examining the impacts of the social support received from doctors during the first consultation on patients’ follow-up consultation behavior, by using secondary data. Moreover, as distinct from previous studies that focus on the online strategy of improving patients’ stickiness, this study enriches the strategy from an online-offline channel integration perspective. Ultimately, we empirically identify that the social support provided by doctors during the first consultation promotes the continued consultation behavior of patients who have had offline visit experience but can inhibit the continued consultation behavior of patients who have no prior consultation experience. This finding provides theoretical evidence that OHC complements offline consultations and confirms the influencing factors of online follow-up service by patients.

Third, this study provides new insights into the literature on online doctor-patient relationship management by considering the moderating role of patients’ offline visit experience. Prior studies have seldom investigated the integrated effects of patients’ offline behaviors and online behaviors (Zhang et al., 2018). However, this study verifies that patients’ self-disclosure interacts with their offline visit experience in influencing the online information support received from doctors during the first online consultation. Overall, these findings provide theoretical evidence that patients’ offline behavior influences the relationship between patients’ online behavior and the response provided by doctors in the context of OHC.

Practical implications

This study also provides several actionable implications for patients, doctors, and the developers of OHC platforms. First, the study’s findings serve as guidelines for how patients write Disease Descriptions to stimulate social support from doctors. Specifically, we provide empirical evidence that patients should avoid excessive use of complex sentences as this may make it more difficult for doctors to comprehend patients’ diseases. We also discovered that longer and more detailed self-disclosure content can help patients in receiving better responses from doctors. Additionally, patients should include more affective expressions in their self-disclosure content to improve the probability of receiving social support from doctors.

Second, this study provides suggestions to doctors to improve patient satisfaction and increase patients continued decision-making during OHC. Our findings reveal that, for patients with offline visit experience, the informational and emotional support provided by doctors during the first consultation has a positive effect on their continued consultation behavior. Therefore, doctors would better provide more unsatisfied informational and emotional support in hospital services, which would increase patients’ intention to select the same doctors in future online consultations. However, for patients without offline visit experience, the informational and emotional support provided by doctors during the first consultation has a negative effect on their continued consultation behavior. This suggests that doctors must provide basic disease-specific information while responding to patients’ problems, to alleviate their anxious emotions or anxiety.

Finally, at a broad level, our findings provide several implications for the design and development of OHC platforms. Platform developers should increase their awareness of the importance of the linguistic features of patients’ self-disclosure content. They also can actively guide patients in writing diagnosis descriptions before conducting their first online consultation. Moreover, the functions of the doctor-patient interaction should be enhanced to provide patients with more informational and emotional support from doctors with offline visit experience during the online consultation. For example, they can encourage doctors to provide more detailed and complete information for patients by introducing different types of reward mechanisms.

Limitations and future research

This study has several limitations that must be taken into consideration. First, our complexity measures are based on morphology complexity, and we focus on specific linguistic features. Future research could use new measures (e.g., lexicography complexity, semantics complexity), as well as additional linguistic signals (e.g., grammar, positive/negative affect, and linguistic style matching) to better represent the role of linguistic features in patient self-disclosure. Second, our research might suffer from data limitations as we do not have detailed information about the patient consultation time. Future studies could, therefore, extend our research by considering the potential impacts of time and frequency of patient consultation. Third, the data reported in this study were collected from a single OHC platform. To ensure broader validation of our findings, future studies can collect data from other online healthcare communities, such as chunyuyisheng.com, zocdoc.com, and dxy.com. Finally, some of the research variables contained in this study only contain information quantities and, therefore, we fail to reflect the content of text data in doctor-patient communication. Future studies are encouraged to apply text mining techniques to capture patient-disclosed disease information and textual content in doctor-patient communication.