The Impact of Narrative Reviews on Patient E-doctor Choice in Online Health Communities

Abstract

As online health communities (OHCs) continue to proliferate, narrative reviews on doctors have become a vital reference source for patients when choosing online health services. However, the potential value of subjective information reflecting patient experiences in OHCs has not been fully explored. The present study seeks to investigate the impact of narrative reviews on patients’ selection of e-doctors and the extent to which such reviews are moderated by doctors’ specialties. This paper collected data from 747 doctors and 105 032 reviews from WeDoctor, one of China’s most popular OHCs, in 2019. We employed Latent Dirichlet Allocation topic modeling to extract 3 topics and analyzed their effects on patient e-doctor choice using a multiple regression method. Our findings indicate that Topic 1, clinical skills and effects, had a positive impact on patient choice in OHCs (β₁ = .243, P < .001), as did Topic 2, service attitude and trust (β₂ = .130, P < .05). However, the impact of Topic 3, convenience, did not show a significant effect in this study. Moreover, our results suggest that the specialty of Internal Medicine can positively moderate the relationship between Topic 1 (clinical skills and effects) and patient e-doctor choice (β₉ = .087, P < .05). Based on the findings of this study, e-doctors are encouraged to enhance their technical competence to improve treatment effectiveness and adjust their communication methods to increase patient trust and sense of security. OHC platform managers should accurately understand the key factors that influence patient choice and take measures to improve their service quality accordingly.

• What do we already know about this topic?

• The ever-increasing technological advancements and the global outbreak of COVID-19 have led to a surge in demand for online health services. As a result, previous patient reviews have emerged as a crucial factor influencing other patients’ online consultation decision-making.

• How does your research contribute to the field?

• Our study distinguishes itself from previous literature regarding research objectives, variable selection, and the data source. Firstly, we specifically target subjective information by utilizing topics extracted from unstructured narrative reviews as independent variables, as opposed to statistical data of reviews. Secondly, we evaluate patient choice by measuring the 3-month change in consultation volume per doctor, a dependent variable that has yet to be extensively examined in previous research. Finally, to account for patients’ online searching behavior, we introduce the doctor’s specialty/department as a moderator variable instead of disease risk.

• What are your research’s implications toward theory, practice, or policy?

• With the outbreak of the COVID-19 pandemic, health systems worldwide are rapidly adopting online health services to avoid physical contact between patients and healthcare providers. The escalating uptake of telehealth services and a surge in medical appointment cancellations and postponements underscores the need for further research into people’s experiences and satisfaction levels regarding online health services. Our study revealed that e-doctors with strong clinical skills, a compassionate attitude, and a higher level of trustworthiness tend to attract more patients. In contrast, the convenience of an online platform appears to have no significant impact on patient choice. Furthermore, our results indicate that the “clinical skills and effects” topic has a more pronounced effect on Internal Medicine patients. The insights from our research based on China’s experiences contribute theoretically and practically to patient decision-making, online doctor performance, and platform mechanism design. These results could help to sustain online health services beyond the COVID-19 pandemic.

Introduction

The proliferation of Web 2.0 technologies and the widespread adoption of information technology have led to the emergence of online health communities (OHCs). ¹ These communities typically fall into patient-to-patient communities and patient-to-doctor communities. ² Patient-to-patient communities offer platforms for patients and their families to exchange health-related information, share experiences, and provide emotional support to one another. ³ Meanwhile, patient-to-doctor communities enable doctors to provide consultation services remotely via laptops, smartphones, tablets, and desktop computers. ⁴ Compared to traditional medical mode, the advent of patient-to-doctor communities does away with time and distance limitations, leading to cost reductions, improved privacy, avoidance of embarrassment, and time-saving.^5,6 In this study, we concentrate primarily on patient-to-doctor communities.

As with consumers on business-to-consumer (B2C) e-commerce platforms, sharing patient experiences about doctors on online forums and medical websites is now a widespread international phenomenon.^7,8 Such prior reviews help patients acquire relevant information about a doctor’s service quality and thus influence their choices.^9,10 Many OHCs, also known as doctor-rating websites, serve as word-of-mouth platforms in several countries. For instance, in Britain, doctors can be evaluated on the National Health System website. ¹¹ Similarly, in the United States, popular doctor-rating websites like RateMDs.com, HealthGrades.com, and Vitals.com facilitate this process. Due to increasing medical demands and the push of government policies, ¹² China’s healthcare industry has significantly improved, and online health services have become increasingly popular. ¹³ Online health platforms offer a wide range of services, including online written consultations, telephone consultations, video consultations, and outpatient registration services. In addition, OHC platforms have adopted a profit-sharing mechanism, enabling doctors to earn income by providing high-quality online services. ¹⁴ Two of China’s notable patient-to-doctor communities are Good Doctor Online and WeDoctor (formerly Guahao.com).

When using online health services, patients often seek information about doctors through OHCs. The information available on OHCs can be broadly classified into objective and subjective. ¹⁵ Objective information refers to measurable and observable aspects of a doctor’s service, such as their offline description (eg, academic titles, education, hospital affiliation, and medical specialties) and online information (eg, the number of patients treated online, virtual gifts from patients, and patient ratings). ¹⁶ Several studies have explored the impact of objective information on patient choice in OHCs, such as physician titles, ¹⁷ evaluation scores, ¹⁸ and service stars. ⁵ Subjective information, in contrast, comprises people’s preferences, opinions, or emotional expressions, often expressed through textual reviews and posts. It enables users to express their viewpoints and concerns in a subjective manner. Prior research has examined the effects of various statistical characteristics of subjective information on patient choices, such as the number of online and offline service reviews, ¹⁹ the number of doctors’ posts, ²⁰ and the number of thank-you letters.^12,21

Healthcare services are typically professional, intangible, and heterogeneous, making it challenging to assess their quality. Nevertheless, the rich information from online reviews can help us understand the prior patient’s medical experiences and perceived quality. ²² People tend to read the textual content of reviews more carefully, relying less on summarized statistical information, before making medical decisions. As a result, narrative reviews carry more importance and should be thoroughly examined. ¹⁶ This study diverges from existing OHCs research by focusing on examining the impact of the unstructured narrative content of reviews, rather than strictly focusing on statistical review characteristics, on patient choice. Patients also tend to search for e-doctors by specialty or department. However, it remains unclear whether the impact of reviews differs across doctors with different specialties. This study further explores the moderating effect of a doctor’s specialty in the context of OHCs. Consequently, the following 3 research questions are proposed:

1. What topics can be derived from textual analysis of narrative patient reviews?

2. What topics significantly impact patient e-doctor choice in OHCs?

3. Does the doctor’s specialization exert a moderating effect on the relationship between narrative reviews and patient choice?

Methods

Study Design

The present study follows a methodology similar to that of James et al, ²³ to investigate the impact of narrative reviews in OHCs on patient choice. Our approach comprises 4 fundamental steps: (1) data collection and pre-processing, (2) calculating topic probabilities for narrative reviews of each doctor, (3) identifying dependent, independent, moderator, and control variables, and (4) exploring the influence of different topics on patient choice using a multiple regression model. A graphical representation of our research methodology is presented in Figure 1, and the subsequent sections provide a detailed account of each step.

Figure 1.

Framework for exploring narrative reviews in OHCs.

Variables and Data Collection

A list of the variables utilized in this study is provided in Table 1. The dependent variable for this analysis was the patient choice, which was evaluated based on the 3-month change in the number of consultations per doctor. The probabilities of the 3 topics discovered from the narrative reviews served as the independent variables for this study. Additionally, given the offline appointment habits in hospitals, patients typically search for doctors on OHCs by specialty. Hence, we considered medical specialty as a moderator variable. The control variables encompassed doctors’ titles, hospital level, and the number of previous reviews.

Table 1.

Variables Definition.

Variable type	Variable symbol	Explanation
Dependent variable
Patient choice	ΔLn Selection	The change in the number of consulting for a doctor in 3 months. We used the log value.
Independent variables
Topic probability	Topic_1	The probability of Topic Clinical skills and Effects
	Topic_2	The probability of Topic Service Attitude and Trust
	Topic_3	The probability of Topic Convenience
Moderating variable
Specialties	Spe_dummy1	If a doctor belongs to OBGYN, 1; otherwise, 0
	Spe_dummy2	If a doctor belongs to Chinese Medicine,1; otherwise, 0
	Spe_dummy3	If a doctor belongs to Internal Medicine,1; otherwise, 0
Control variables
Doctors’ title	Title_dummy1	If a doctor’s title is a Chief Doctor, 1; otherwise, 0
	Title_dummy2	If a doctor’s title is an Associate Chief Doctor, 1; otherwise, 0
	Title_dummy3	If a doctor’s title is an Attending Doctor, 1; otherwise, 0
Hospital level	H_ Level	Grade III Level A hospitals are 1; otherwise, 0.
Number of reviews	Ln N_ Reviews	The total number of reviews of a doctor. We used the log value.

The Internet Society of China’s report indicates that the number of online health service users in China rose to 215 million by the end of 2020, accounting for 21.7% of the total number of Internet users. ²⁴ A wealth of information about doctors is available on their homepages in OHCs, including their titles, medical specialties, affiliated hospitals, educational and work experience, the number of online consultations, communication records with patients, service stars, patient feedback, and posts related to health management or diseases. This study collected online information about doctors and patient reviews from WeDoctor (http://www.guahao.com/), one of China’s most popular online health platforms. WeDoctor connects 240 000 doctors across 7200 hospitals in 31 provinces of China across the country. The platform combines offline and online health resources, forming a health maintenance platform with over 200 million users as of March 2020. An interface of the WeDoctor website is outlined in Figure 2.

Figure 2.

The homepage of the WeDoctor website (accessed May 30, 2021).

WeDoctor offers patients multiple healthcare services, including online text and photo consultations, telephone and video consultations, service packages, and offline registration. Online text and photo consultations enable patients to consult with doctors about their medical concerns via text and pictures, and doctors can respond to these inquiries in their spare time. Online telephone and video consultations facilitate real-time communication between patients and doctors during scheduled appointments. The service package provides long-term health management tailored to patients’ needs based on the doctor’s clinical experience. Lastly, offline appointment and registration services allow patients to schedule face-to-face appointments with specific doctors in certain hospitals.

Patients on WeDoctor can search for doctors based on their medical specialty/department, hospital, or disease. They can share their medical experiences, including clinical outcomes, without any time restrictions. This study focused on 4 medical specialties—Internal medicine, Surgery, Obstetrics and Gynecology (OBGYN), and Chinese Medicine—as they have the highest number of doctors registered on WeDoctor. The dataset comprised both objective and subjective information. Objective information consisted of the doctor’s name, title, medical specialty, affiliated hospital, and the number of online consultations. Subjective information primarily comprised the original narrative reviews. We collected data from 1362 doctors in these 4 specialties at 2-time points, July 5, 2019, and October 3, 2019. We matched these 2 datasets to analyze the changes in service amounts per doctor and manually removed some doctors’ data if it was incomplete. The final dataset encompassed records for 747 doctors and 105 032 reviews, which were then used for subsequent analyses.

Topic Extraction and Statistical Analysis

Topic modeling is a widely used text-mining method that aims to uncover latent semantic structures in document sets by identifying potential topics. ²⁵ It can statistically capture those topics with the use of different algorithms, ²² such as Principal Components Analysis (PCA) and Latent Dirichlet Allocation (LDA). LDA is a popular algorithm in natural language processing (NLP) and was employed in this study. Firstly, a Python Kit was used to parse the reviews, followed by the exclusion of meaningless words (eg, “I” and “we”) and high-frequency words (eg, “doctor” and “patient”) from the texts. Using LDA, we selected the optimal number of topics in the corpus based on its perplexity evaluation criteria, which measures the quality of the model. We experimented with different topic numbers, ranging from 2 to 10, running the LDA model 10 times in each iteration through Anaconda 3. After evaluating perplexity statistics, we found that the minimum perplexity occurred for 3 topics.

Data management was performed using Microsoft Excel 2017, while SPSS software version 22.0 was used for all statistical analyses. Given the exploratory nature of our study, stepwise regression was deemed suitable as it can screen for significant independent variables affecting the dependent variable and simplify the regression equation. Specifically, independent variables were only introduced if their partial regression sum-of-squares were significant. Any independent variables deemed to have little influence on the dependent variable were eliminated to identify the optimal regression subset. Our models were constructed hierarchically, with control variables included in model 1, followed by independent variables in model 2, and interaction terms in model 3. All reported P-values were 2-sided, and a p-value of less than .05 was considered statistically significant. The regression equation was expressed as follows in equation (1), where β₀ represents the constant term, β₁ through β₂₀ represent the regression coefficients, and ε represents the error term. The term Topic_i_t-1×Spe_dummyi_t-1 (i = 1, 2, 3) reflects the moderating effect of the specialty.

$$\begin{array}{l}\Delta LnSelection\\ =LnSelectio{n}_t-LnSelectio{n}_{t-1}\\ ={\beta }_0+{\beta }_{1}Topic\_1_{t-1}+{\beta }_{2}Topic\_2_{t-1}+{\beta }_{3}Topic\_3_{t-1}\\ +{\beta }_{4}Spe\_dummy{1}_{t-1}+{\beta }_{5}Spe\_dummy{2}_{t-1}+{\beta }_{6}Spe\_dummy{3}_{t-1}\\ +{\beta }_{7}Topic\_1_{t-1}\times Spe\_dummy{1}_{t-1}+{\beta }_{8}Topic\_1_{t-1}\times Spe\_dummy{2}_{t-1}\\ +{\beta }_{9}Topic\_1_{t-1}\times Spe\_dummy{3}_{t-1}+{\beta }_{10}Topic\_2_{t-1}\times Spe\_dummy{1}_{t-1}\\ +{\beta }_{11}Topic\_2_{t-1}\times Spe\_dummy{2}_{t-1}+{\beta }_{12}Topic\_2_{t-1}\times Spe\_dummy{3}_{t-1}\\ +{\beta }_{13}Topic\_3_{t-1}\times Spe\_dummy{1}_{t-1}+{\beta }_{14}Topic\_3_{t-1}\times Spe\_dummy{2}_{t-1}\\ +{\beta }_{15}Topic\_3_{t-1}\times Spe\_dummy{3}_{t-1}+{\beta }_{16}Title\_dummy{1}_{t-1}\\ +{\beta }_{17}Title\_dummy{2}_{t-1}+{\beta }_{18}Title\_dummy{3}_{t-1}+{\beta }_{19}H\_Leve{l}_{t-1}\\ +{\beta }_{20}LnN\_Review{s}_{t-1}+\epsilon \end{array}$$ (1)

Results

Descriptive Statistics

Table 2 presents the descriptive statistics of the variables included in our study. Two of the variables, patient choice and total reviews for doctors may not have followed normal distributions. Therefore, we applied the logarithmic transformation (ln) to these 2 variables to generate normal distributions, which were used for further statistical analyses.

Table 2.

Descriptive Statistical Analysis of Variables.

Variables	Number	Range	Minimum	Maximum	Mean	Std. Deviation
ΔLn Selection	747	3.115	0.000	3.115	0.180	0.259
Topic_1	747	0.259	0.197	0.456	0.263	0.051
Topic_2	747	0.302	0.317	0.619	0.455	0.053
Topic_3	747	0.253	0.158	0.411	0.283	0.046
Spe_ dummy1	747	1.000	0.000	1.000	0.365	0.482
Spe_ dummy2	747	1.000	0.000	1.000	0.207	0.406
Spe_ dummy3	747	1.000	0.000	1.000	0.133	0.339
Title_dummy1	747	1.000	0.000	1.000	0.360	0.480
Title_dummy2	747	1.000	0.000	1.000	0.367	0.482
Title_dummy3	747	1.000	0.000	1.000	0.229	0.420
H_ level	747	1.000	0.000	1.000	0.865	0.342
Ln N _reviews	747	6.684	1.792	8.476	4.642	1.442

Topic Extraction From Patient Narrative Reviews

The results of the topic extraction are presented in Table 3, where the top 20 keywords for each topic are provided.

Table 3.

Topics Extracted From Narrative Reviews in OHCs by LDA.

Topic_1	Topic_2	Topic_3
Clinical skills and effects	Service attitude and trust	Convenience
Professional	Careful	Timely
Helpful	Dedicated	Questions
Operation	Earnest	Easy
Detailed	Kind	Ask
Medical skills	Enthusiastic	Hospital
Excellent	Attentive	One-to-one
Solve	Relax	Need
Experienced	Warmhearted	Rapid
Illness	Mood	Find Spare Time
Meticulous	Depressed	For the first time
Work	Worried	Anytime
Clear	Answer all the questions	Anywhere
Inspect	Great	Medical record
Guidance	Trust	Photograph
Comprehensive	Consultation	WeDoctor
Consummate	Nice	App
Opinion	Responsible	Quickly
Effect	Patient	Worth
Revisit	Happy	Consult
Come Again	Standard	Appointment

Topic 1, named “clinical skills and effects,” is characterized by keywords such as professional, operation, solve, experienced, excellent, visit, and come again, reflecting patients’ evaluation of doctors’ technical competence. This includes assessments of the doctors’ clinical skills, knowledge of treatment options, follow-up care, decision-making abilities, and treatment effectiveness. Topic 2, labeled “service attitude and trust,” encompasses patient-doctor communication and interpersonal skills, as indicated by keywords such as careful, dedicated, kind, and warmhearted. Patients value doctors who are empathetic, friendly, trustworthy, and able to explain medical information clearly while also reassuring patients. Topic 3, which we have named “convenience,” is primarily related to the online healthcare platform used by patients to find suitable doctors. Keywords such as questions, ask, need, one-to-one, timely, rapid, easy, anytime, and anywhere reflect patients’ desire for convenience when using the online platform WeDoctor. Table 4 provides an example of LDA results for a doctor’s reviews, illustrating that this particular doctor’s reviews are predominantly relevant to Topic 2 (47%), reflecting the doctor’s interpersonal manner, as the keywords associated with this topic include trust, nice, worried, standard, among others. The probability of the doctor’s reviews pertained to Topics 1 and 3 were 33% and 20%, respectively.

Table 4.

An Example of LDA Results of a Doctor’s Reviews.

Review Document	Content	The Probability of Topic (%)	Topic Keywords
Document: xxx.txt (Note: part of the reviews of a doctor)	Wang**: After taking the medicine for two months, my condition has been greatly improved. I was able to sleep and my temper and mood are becoming better. Thank you, Dr. Liu. Chen **: Dr. Liu is warmhearted with patients. Dr. Liu has helped my mother control her blood glucose for a long time, and the effect is well. Thank you for your kindness. Chen **: Dr. Liu is very patient, nice, and dedicated to his work. I will come again if necessary. Chen **: Dr. Liu handles cases with good care, but it’s hard to make an appointment. Ding **: Dr. Liu is kind and polite. He respects patients and is attentive to the elderly. Considering for patients, excellent medical skills. After the doctor’s treatment, my condition continues to improve~ Chen * *: WeDoctor is really convenient. It allows me to ask doctors questions anytime and anywhere. It’s great. Liu **: Doctor Liu treats patients earnestly and carefully. He is professional, and the medication is accurate. I am confident in his medical skills.	33%	Professional, helpful, operation, detailed, medical skills, excellent, solve, experienced, illness, meticulous, work, clear, inspect, guidance, comprehensive, consummate, opinion, effect, revisit, come again
		47%	Careful, dedicated, earnest, kind, enthusiastic, attentive, relax, warmhearted, mood, depressed, worried, answer all the questions, great, trust, consultation, nice, responsible, patient, happy, standard
		20%	Timely, questions, easy, ask, hospital, one-to-one, need, rapid, find spare time, for the first time, anytime, anywhere, medical record, photograph, WeDoctor, app, quickly, worth, great, consult, appointment

Table 5 presents the results of all regression analyses. The adjusted R-squared value and the significance of the F-value (P < .001) indicated that the independent variables explained the dependent variable well. As shown in Table 5, two independent variables had a positive effect on patient choice in OHCs: Topic 1, focusing on clinical skills and effects related to technical competence (β₁ = .243***, P < .001), and Topic 2, emphasizing interpersonal aspects with an emphasis on service attitude and trust (β₂ = .130*, P < .05). However, this study did not support the impact of Topic 3, convenience. Furthermore, the results of model 3 indicated a significant moderating effect on the specialties of the doctors. Specifically, the Internal Medicine specialty positively moderated the relationship between the Topic 1 clinical skills and effects of narrative reviews and patient choice (β₉ = .087*, P < .05).

Table 5.

Stepwise Regression Model Results.

	Model 1	Model 2	Model 3
Ln N_reviews	−0.309***	−0.331***	−0.331***
	(–0.056)	(0.007)	(0.007)
Title_dummy3	0.079*	0.115**	0.115**
	(0.021)	(0.022)	(0.022)
Spe_ dummy3	0.095**	−0.076*	—
	(0.027)	(0.027)
H_ level	—	−0.075*	−0.076*
		(–0.057)	(0.028)
Topic_1	—	0.243***	0.237***
		(0.238)	(0.238)
Topic_2	—	0.130*	0.132*
		(0.257)	(0.257)
Topic1*Spe_ dummy3	—	—	0.087*
			(0.098)
Constant	0.417***	0.103***	−0.126***
R 2	0.120	0.153	0.155
Adjusted–R2	0.117	0.147	0.148
F–statistics	33.929***	22.342***	22.650***

Note. The number in parentheses is the standard error.

^***P < .001; **P < .01; *P < .05.

Discussion

This paper focuses on investigating the relationship between unstructured narrative reviews and patient e-doctor choice behavior to shed light on the role of patient reviews in shaping other patients’ decision-making when seeking online medical consultations.

First, the results of our study indicate that patients in the OHC field prioritize service quality when seeking online medical consultations, consistent with previous studies in offline physical hospitals and other online service domains. Specifically, better service quality offered by e-doctors leads to an increase in online consultation numbers. This finding aligns with the research of Lu and Wu, ¹⁸ who reported that both technical and functional quality significantly affect patient choice. In our study, the topic “clinical skills and effects” represents technical quality feedback, while “service attitude and trust” reflects functional quality statements. Online patients seek better doctors to address their health concerns, and their reviews reflect doctors’ knowledge, clinical skills, attention to detail, and treatment effectiveness. Patients are willing to share their opinions to assess doctors’ technical competence online. Additionally, patients tend to evaluate doctors’ empathy, friendliness, trustworthiness, helpfulness, and ability to make patients feel comfortable. A suitable interpersonal manner helps establish trust between doctors and patients in virtual spaces. ²⁶ Trust is a critical element in the online context and plays a vital role in reducing risks and uncertainties and increasing customers’ willingness to make purchases. ²⁷ Several studies on patient choice in OHCs have also found that good service attitude and trust influence patients’ decisions when selecting e-doctors.^28,29 In general, e-doctors who provide better treatment outcomes and establish trust with patients will likely attract more patients in the future.

Second, our study results revealed that the convenience topic conveyed in patient reviews did not significantly influence the choice of other patients. Patients often opt for online health services due to the convenience that OHCs offer by transcending the limitations imposed by time and geography. ³⁰ In other words, convenience is a fundamental prerequisite for patients to consider online health services. More research in this area can be warranted, considering the differences in contexts, patient populations, or healthcare systems.

Last, our study findings indicate that Topic 1, “clinical skills and effects,” has a more significant impact on the choice of patients in the Internal Medicine specialty, as seen in the results of model 3. This may be due to related policies on OHCs. ³¹ The Administrative Measures for Internet Diagnosis and Treatment (for Trial Implementation) in China states that Internet diagnosis and treatment activities are not open for first-time patients due to medical risks. The Chinese government encourages OHCs to offer premedical consultations or provide medical consultations for revisiting patients with chronic diseases such as hypertension, coronary heart disease, diabetes, stroke, liver disease, and chronic respiratory diseases, which are mainly under the Internal Medicine specialty. The traditional patient-doctor relationship is based on the assumption that healthcare professionals know what is best for patients. ³² However, knowledge gained from OHCs informs and empowers patients. In addition, patients with chronic diseases often have a greater understanding of their conditions and the ability to evaluate the treatment effects of Internal Medicine doctors. Hence, these patients may be particularly sensitive to the clinical skills and treatment outcomes of e-doctors in this specialty area. This may help explain why Topic 1 significantly affects the choice of Internal Medicine patients in this study.

Theoretical Implications

Our research makes several theoretical contributions to the study of patient choice behavior in OHCs. Firstly, our study extends prior research by exploring the impact of unstructured and subjective narrative content on patient choice in OHCs using text-mining technology. While previous studies have examined the impact of subjective information on patient choice based on statistical characteristics such as the number of reviews, few have focused on the narrative content of reviews. Our study provides a more nuanced understanding of how patients make e-doctor choices by demonstrating the importance of unstructured feedback in patient reviews. Secondly, our study differs in the selection of the dependent variable from prior literature. For instance, James et al ²³ employed an overall numeric rating per doctor to measure patient choice behavior. In contrast, we used the 3-month change in consultation numbers per doctor to evaluate patient choice, which provides a more precise indicator of patient selection behavior. Our study contributes to developing more accurate and reliable measures of patient choice behavior in OHCs. Lastly, our study considers patients’ online search habits by incorporating the doctor’s specialty/department as a moderator variable, whereas previous studies have commonly used disease risk as a moderator. This change supplements and enriches patient behavior analysis in OHCs by providing insights into how the doctor’s specialty influences patient choice behavior.

Practical Implications

Our study has several valuable practical implications. First, the COVID-19 pandemic has led to the widespread adoption of telehealth globally, thus exposing gaps in the readiness of health services to deliver telehealth routinely. ³³ Our study findings provide evidence regarding the importance of developing a skilled and competent workforce to sustain telehealth beyond the pandemic. Doctors providing care through telehealth require additional skills and adequate support. ³⁴ Access to training programs is crucial to support current staff with online health technologies. By integrating telehealth use into course curricula, universities can ensure that future health professionals are proficient in telehealth skills. Second, our study’s findings can raise doctors’ awareness of the importance of communication quality in the doctor-patient interaction, which plays a crucial role in online health services. Patients value doctors’ informational and emotional support to enhance their health status. By offering more emotional input, e-doctors can improve patients’ satisfaction with their service attitude. Lastly, the absence of visual cues, eye contact, and body language poses a significant challenge to building online trust. Doctors and patients need to learn how to establish online trust in a virtual context. Policymakers should invest in technologies that can facilitate virtual contact during telehealth consultations to mitigate the challenge to trust-building.

Conclusions

This study examined the impact of narrative reviews on patient choice in OHCs, utilizing data from 747 doctors on China’s WeDoctor online health platform. Employing the LDA method, we extracted 3 topics from unstructured patient reviews, namely clinical skills and effects (Topic 1), service attitude and trust (Topic 2), and convenience (Topic 3). We then used stepwise regression analysis to investigate the relationship between the topics and patient choice behavior. The results indicate that a doctor delivering more trust or better treatment outcomes is more likely to attract patients in OHCs. At the same time, convenience was not a significant driver for patient e-doctor choice. Furthermore, clinical skills and effects were found to be more significant for patients in the Internal Medicine specialty. This study provides new insights into word of mouth and patient doctor choice in OHCs and offers valuable references for the sustainable development of telehealth services.

This study also has several limitations. First, the data samples were obtained solely from an online health platform in China. To enhance the generalizability of the findings, future studies can include narrative reviews from various online health platforms for comparison. Second, we measured patient choice using an increase in the consultation number of doctors over 3 months. Future studies may consider selecting different time intervals to capture patient doctor choices dynamically. Third, we used LDA as an unsupervised machine learning method to derive topics from reviews. Future research could adopt other machine learning methods to obtain topics and their probabilities to improve the accuracy of the results. Lastly, while some scholars have studied the effect of sentiments expressed in consumer reviews on sales in e-commerce, future research could investigate the impact of emotions expressed in reviews on patient e-doctor choice in OHCs.