Characterizing authoritative oncology-related key opinion leaders on Weibo: A social media profiling study

Abstract

Objective

This study aimed to explore the heterogeneity of authoritative oncology-related key opinion leaders (AOKOLs) on Weibo and to develop their user profiles that support precise health information dissemination and personalized online patient support.

Methods

Using Python-based web scraping, data were collected from Weibo (Nov 2023–Nov 2024) based on six oncology-related keywords. AOKOLs were profiled using K-means clustering across 12 indicators in four dimensions: content output, interaction, professional background, and social network features. Sentiment analysis of representative user comments was conducted to assess audience emotional responses.

Results

A total of 143 AOKOLs were included through clustering analysis, with 89,118 posts and 837,602 comments collected from their Weibo accounts. These AOKOLs were categorized into four distinct digital influence profiles: 31 expert knowledge communicators, 47 science information sharers, 47 emotional support providers, and 18 social engagement facilitators. These groups differed in content focus, engagement patterns, and audience interaction. Sentiment analysis of 62,154 comments from 37 representative AOKOLs revealed varying emotional responses across profiles, highlighting their differential impacts on audience well-being and digital health communication.

Conclusions

This study reveals distinct user profiles of AOKOLs on Weibo, highlighting the diverse communication strategies, engagement styles, and emotional influences within digital health ecosystems. Findings offer insights into how digital influencers can support patient-centered care and enhance quality of life through intelligent health communication.

Introduction

Cancer remains one of the leading causes of death globally. According to the World Health Organization, approximately 10 million people died from cancer in 2020, accounting for nearly one-sixth of all global deaths.¹ Although survival rates have improved due to advancements in early detection and treatment, side effects and long-term complications from surgery, chemotherapy, radiotherapy, and hormone therapy continue to significantly impair patients' quality of life.^2,3 Cancer patients and their families face complex medical decisions, including selecting appropriate treatment options and maintaining adherence, while also contending with the psychological and social disruptions caused by the disease.⁴ Studies have shown that cancer patients frequently experience multiple unmet needs—physical, psychological, informational, and accessibility-related.⁵ The greater the number of unmet needs, the higher the likelihood of psychological distress.⁶ However, the shortage of health care resources limits comprehensive support, and the restricted availability of in-person services makes it difficult to meet patients’ multifaceted needs, thereby exacerbating both physical and psychological burdens.^7,8

With the rapid development of information technology and the growing public demand for health information, an increasing number of cancer patients are turning to social media to seek or share disease-related content, making online communities an important channel for health communication.⁹ A Swedish study involving 282 cancer patients reported that 76.2% used the internet to search for cancer-related information following diagnosis.¹⁰ Online health communities now serve as vital digital platforms that connect health care providers, institutions, and patients, facilitating more accessible doctor-patient communication and peer-to-peer information exchange.¹¹ Sina Weibo, one of China's most influential social media platforms, reported 511 million monthly active users in 2020.¹² The platform allows users to instantly and interactively share information via text, images, and videos.¹³ Key Opinion Leaders (KOLs) on platforms like Weibo serve as trusted sources of advice and guidance throughout the cancer journey.¹⁴ KOLs are influential intermediaries who possess high levels of expertise or social status and play a pivotal role in disseminating knowledge and shaping public discourse within specific domains.¹⁵

Within online communities, KOLs function as both receivers and central disseminators of health information, using various multimedia formats to influence patients' perceptions, attitudes, and behaviors.¹⁶ Evidence suggests that social media interactions can bridge the psychological gap between authoritative KOLs and patients, fostering a sense of peer-like connection and increasing patient engagement.¹⁷ Distinct from patient opinion leaders, authoritative KOLs typically possess substantial clinical expertise and recognized reputations in health domains, enabling them to provide evidence-based guidance on managing treatment side effects and psychological challenges.¹⁸ Despite their credibility and expertise, the overwhelming volume of online health content makes it difficult for patients to obtain coherent and personalized guidance during complex decision-making processes.^19,20 Moreover, as patients’ expectations toward authoritative KOLs vary across treatment phases, a major challenge lies in translating KOL influence into meaningful behavioral outcomes.²¹

Currently, there is a lack of systematic research on the role of authoritative KOLs in supporting cancer patients through health information dissemination. Given that social media has become a key source of disease-related information and emotional support for these patients, a deeper understanding of authoritative oncology-related key opinion leaders (AOKOLs) in the digital health ecosystem is highly warranted. This study focuses on the Weibo platform and employs clustering and user profiling methods to explore the heterogeneity of AOKOLs in terms of content dissemination, interaction patterns, professional identities, and network structures. The findings aim to clarify the communicative roles of AOKOLs and offer empirical insights to inform the design of targeted, phase-specific support strategies for cancer patients in social media environments.

Methods

Indicator selection for the conceptual profiling framework

This study developed a conceptual profiling framework comprising 12 evaluation indicators across four dimensions: content output, interaction, professional background, and social network characteristics, based on the posting behaviors and social influence of AOKOLs in online communities,²² as shown in Table 1. The content output dimension included posting frequency, educational content ratio, and content category, which together reflect the AOKOLs' activity and thematic focus, allowing for assessment of their contributions to health education. Interaction indicators comprised average number of comments, likes, reposts, and response rate, which capture both the visibility of AOKOLs' content and their willingness to engage with patients and other users.²³ Professional background indicators included education level, professional title, and years of experience, to ensure that AOKOLs possess adequate qualifications and credibility to influence medical decision-making. Social network indicators—follower count and mutual followings—reflected the KOLs’ online popularity and the degree of professional connectivity, highlighting their centrality in information dissemination networks.²⁴ Taken together, these four dimensions offer a comprehensive framework for evaluating the influence and utility of AOKOLs in cancer-related patient education and digital health communication.

Table 1.

Indicator framework for profiling authoritative oncology-related key opinion leaders (AOKOLs) on Weibo.

Dimension	Indicator	Definition & Description	Data type
Content output	Posting frequency	The average number of posts published per month by AOKOLs, reflecting their activeness in the community.	Numerical
	Educational content ratio	The proportion of posts containing educational content (e.g., disease explanations, treatment plans, recovery advice).	Percentage (%)
	Content category	Categorization of post content types, such as health education, treatment advice, psychological support, or symptom management.	Categorical (text)
Interaction	Average comments per post	The average number of comments received per post, indicating the level of attention the content attracts.	Numerical
	Average likes per post	The average number of likes per post, indicating audience approval or appreciation of the content.	Numerical
	Average reposts per post	The average number of shares per post, reflecting the breadth of content dissemination.	Numerical
	Response rate	The proportion of patient comments or questions replied to by AOKOLs, indicating their engagement level.	Percentage (%)
Professional background	Education level	The highest academic qualification of the AOKOL (e.g., Bachelor's, Master's, Doctorate).	Categorical (text)
	Professional title	The professional rank or title of the AOKOL (e.g., Attending physician, associate chief physician, chief physician).	Categorical (text)
	Years of experience	The total number of years the AOKOL has been professionally active, representing clinical experience.	Numerical
Social network	Number of followers	The number of followers the AOKOL has on the platform, reflecting their visibility and influence.	Numerical
	Number of mutual followers	The number of mutual followings with other AOKOLs or health care professionals, indicating network connectivity.	Numerical

Data collection

This study employed Python-based web scraping techniques to collect publicly available data from the Sina Weibo platform. Using six keywords — “oncology,” “cancer,” “oncology expert,” “cancer expert,” “oncologist,” and “oncology nurse” — relevant users were identified, and their posts and user interaction data (including post content, likes, comments, and reposts) published between November 2023 and November 2024 were retrieved. Data collection was implemented using Python and related libraries: GET requests were sent to target webpages via the Requests library, and structured data were extracted using the Etree library in combination with XPath syntax to parse HTML content. For pages with complex structures, multithreading and asynchronous programming were employed to enhance crawling efficiency. Custom headers and request delay settings were configured to prevent server blocking and filter out advertisements or irrelevant content. All structured data—such as user ID, post content, posting time, and interaction metrics—were stored in structured data tables for subsequent filtering, feature extraction, and sentiment analysis.

During initial screening, two trained researchers independently reviewed each account to document the username, registration date, content themes, and activity level. Screening criteria included consistent oncology-related content and demonstrated influence in health communication. Accounts that were irrelevant or inactive were excluded. Final AOKOLs were identified based on the following criteria: Inclusion criteria: medical or health-related background (e.g., tertiary hospital oncologists, oncology nurses, registered dietitians); content focused on oncology treatment, care, rehabilitation, or patient support. Exclusion criteria: content unrelated to oncology, presence of misinformation, or commercial intent that undermines objectivity or professionalism.

Data cleaning and coding

To ensure data quality and methodological rigor, a systematic cleaning and screening process was undertaken after data collection. All crawled data were initially reviewed to remove blank entries, duplicate records, advertisements, and irrelevant content. Posting time, content fields, and interaction metrics were then standardized to ensure data consistency and comparability. Accounts were further assessed using inclusion and exclusion criteria, with low-activity or unprofessional profiles excluded. Text coding and AOKOL classification were independently conducted by the first author and a trained assistant, focusing on account attributes, content orientation, and initial labeling. Discrepancies were resolved through consensus, supplemented by review of historical posts or institutional affiliations when necessary, ensuring classification accuracy and data credibility.

Data preprocessing

After confirming eligible AOKOLs, data were preprocessed for quantitative analysis. Posts were cleaned by removing promotional phrases, hyperlinks, special characters, and formatting inconsistencies, and duplicate entries were also excluded. All data were structured into tables containing 12 indicators across content output, interaction, professional background, and social network features. Categorical variables were numerically encoded. Text segmentation was performed using Jieba, and stop words (e.g., “的”, “是”, “在”) were removed to reduce noise.

User profiling and characterization

K-means clustering was applied to classify AOKOLs using 12 indicators across the four predefined dimensions. The optimal number of clusters (K) was determined through the elbow method, scree plot, and silhouette coefficient to ensure interpretability and robustness. Subsequently, word clouds were generated using the wordcloud Python package to visualize keyword frequencies across AOKOL categories. Font size reflected relative term frequency (c-values), highlighting thematic emphases within each cluster.

Sentiment analysis of comments

Sentiment analysis of user comments on AOKOL posts was performed using the SnowNLP library. The process included text preprocessing and polarity classification. Jieba was used for tokenization, and regular expressions removed emojis, invalid characters, and stop words. Each comment received a sentiment score (0–1), with ≥ 0.6 defined as positive, ≤ 0.4 as negative, and scores in between considered neutral.²⁵ To validate model performance, a subset of comments was manually annotated and compared with automated results. Sentiment distributions were then calculated and visualized across the four AOKOL groups.

Results

Clustering analysis results

This study applied the K-means clustering algorithm to analyze the profile indicators of AOKOLs, with the results of the elbow method and silhouette coefficient shown in Fig. 1, Fig. 2.

Fig. 1.

Elbow method for optimal k.

Fig. 2.

Silhouette coefficient method for optimal k.

As shown in Fig. 1, the curve declines rapidly when the K value is less than 4, indicating greater clustering error; beyond K ​= ​4, the curve flattens, representing smaller incremental gains, forming a typical “elbow” shape, thus suggesting K ​= ​4 as the inflection point. In Fig. 2, the silhouette coefficient approaches 0.80 at K ​= ​4, indicating well-separated and compact clusters. Although slightly higher values are observed at K ​= ​2 and K ​= ​3, the resulting clusters are too coarse to capture structural heterogeneity. Therefore, K ​= ​4 was selected as the optimal number of clusters. Based on this optimal value, K-means clustering was performed on the sample data using Python's sklearn package, with results shown in Fig. 3.

Fig. 3.

K-means clustering into 4 clusters (PCA visualization with polygon boxes). PCA, principle component analysis.

Posting characteristics of AOKOLs

A total of 106,444 original posts and 847,404 comments from 265 Weibo users were initially collected. After data cleaning and eligibility screening, 143 AOKOLs were included, comprising 89,118 posts and 837,602 comments. Based on K-means clustering analysis, AOKOLs were categorized into four groups: expert knowledge communicators (n ​= ​31), science information sharers (n ​= ​47), emotional support providers (n ​= ​47), and social engagement facilitators (n ​= ​18). The characteristics of each group are presented in Table 2, which summarizes key indicators such as content output, interaction patterns, educational background, clinical title, years of experience, and social network reach, reflecting the full spectrum of influential oncology-related accounts on Weibo included in this study.

Table 2.

Persona profiles of authoritative oncology-related key opinion leaders (n ​= ​143).

Name		Expert Knowledge Communicators	Science Information Sharers	Emotional Support Providers	Social Engagement Facilitators
Typedescription		• Large follower bases (typically > 500,000) • High average engagement (likes, comments, shares) • Mainly focuses on science popularization	• High proportion of educational content (> 30%) • Lower engagement levels • Focus primarily on disseminating professional knowledge rather than interacting with followers.	• High proportion of emotional support content (> 30%) • High average comments	• High responsiveness rate (typically > 5%) • High volume of comments and interactions • Prefer active interaction with followers
Content output indicators	Posting frequency	125 posts/month	16 posts/month	25 posts/month	93 posts/month
	Educational content ratio	22.46% (37.13% treatment plans; 40.00% disease explanations; 16.40% rehabilitation advice)	49.88% (47.00% treatment plans; 30.50% disease explanations; 22.50% rehabilitation advice)	6.83% (25.00% treatment plans; 36.40% disease explanations; 23.70% rehabilitation advice)	7.13% (38.70% treatment plans; 35.40% disease explanations; 25.90% rehabilitation advice)
	Content type (psychosocial support ratio)	25.09%	12.60%	43.89%	18.70%
Interaction indicators	Average comments per Post	16.75	0.95	2.78	6.77
	Average likes per Post	150.36	5.51	4.04	62.87
	Average reposts per Post	16.29	1.86	22.53	23.65
	Response rate	3.18%	3.65%	6.4%	4.7%
Professional background characteristics	Education level	PhD: 38.70%; Master: 6.45% Undisclosed: 41.90% Others: 12.95%	PhD: 42.60%; Master: 10.60% Undisclosed: 38.30% Others: 8.50%	PhD: 44.70%; Master: 12.80% Undisclosed: 42.50%	PhD: 50.00%; Master: 11.10% Undisclosed: 33.30% Others: 5.60%
	Professional title	Chief physician: 35.50% Associate chief physician: 6.45%	Chief physician: 40.40% Associate chief physician: 25.50%	Chief physician: 31.90% Associate chief physician: 27.70%	Chief physician: 44.40% Associate chief physician: 38.90%
	Years of experience	> 20 years: 41.90% 10–20 years: 29.00%	> 20 years: 48.40% 10–20 years: 29.00%	> 20 years: 21.40% 10–20 years: 29.00%	> 20 years: 50.00% 10–20 years: 27.80%
Social network characteristics	Number of followers	1.488 million	47,000	91,000	153,000
	Number of mutual followers	562	35	179	941

In terms of content output, expert knowledge communicators had the highest posting frequency, averaging 125 posts per month, with a relatively balanced share of educational content (22.46%) and psychological support content (25.09%). Educational content mainly focused on treatment plans (37.13%), disease interpretation (40.00%), and rehabilitation advice (16.40%). In contrast, the science information sharers had the lowest average monthly posting frequency at only 16 posts, yet educational content accounted for 49.88% of their output, primarily focused on professional knowledge dissemination, including treatment plans (47.00%), disease explanation (30.50%), and rehabilitation advice (22.50%). For example, nearly 70% of the posts by “Thoracic Surgeon Wang Jiyong” consisted of science popularization content, highlighting the group's educational emphasis. Emotional support providers posted less frequently (25 posts/month) but had the highest proportion of psychological support content (43.89%) among all categories, indicating a unique emotional companionship role. For instance, 39.68% of 2089 posts by “Dr. Li Jing” were emotionally supportive, with consistently high response rates. Social engagement facilitators had a relatively high posting frequency (93 posts/month), with moderate proportions of educational (7.13%) and psychological support (18.70%) content, and a more evenly distributed educational focus.

Regarding interaction indicators, expert knowledge communicators outperformed others in average comments (16.75), likes (150.36), and reposts (16.29) per post. For example, “Dr. Blueberry” posted over 400 times monthly, with an average of more than 400 likes per post, demonstrating substantial authoritative influence. Social engagement facilitators also showed high interaction levels, with 6.77 comments, 62.87 likes and 23.65 shares per post, and a response rate of 4.7%; for example, “Feng Ge for Health” averaged nearly 12 comments per post. In comparison, science information sharers and emotional support providers showed lower engagement, with the former averaging only 0.95 comments and 1.86 shares per post, and the latter slightly higher but still limited.

In terms of professional background, all AOKOL groups had a high proportion of doctoral degrees, particularly emotional support providers (44.70%) and social engagement facilitators (50.00%) types. For clinical titles, the proportion of chief physicians ranged from 31.90% to 44.40%, with the lowest in emotional support providers (31.90%) and the highest in social engagement facilitators (44.40%). Regarding years of experience, approximately 40% of AOKOLs in three groups had over 20 years of professional practice, except for emotional support providers, where only 21.40% had more than 20 years.

As for social network characteristics, expert knowledge communicators had the largest follower base, averaging 1.488 million followers, indicating their strong influence on Weibo. Social engagement facilitators followed with 153,000 followers and the highest number of mutual follows (n ​= ​941), highlighting their advantages in social connectivity and interaction networks. Science information sharers had the smallest follower base (47,000) and the lowest number of mutual follows.

Description of user characterization

Expert knowledge communicators

The label cloud of expert knowledge communicators primarily features highly focused medical terminology and treatment-related themes. High-frequency terms include surgery, chemotherapy, treatment, radiotherapy, guideline, diagnosis, screening, and immunotherapy, indicating a strong emphasis on mainstream cancer therapies and standardized clinical pathways. Additionally, words like symptom, health, patient, and female reflect a focus on patient characteristics and tailored health needs in professional knowledge dissemination. Detailed word frequency distributions are shown in Fig. 4.

Fig. 4.

Expert knowledge communicators.

Science information sharers

The label cloud of science information sharers indicates a focus on disease understanding, screening practices, health management, and the promotion of health knowledge that integrates Western and Chinese medical systems. High-frequency keywords such as health, science popularization, traditional Chinese medicine (TCM), tumor, examination, patient, medication, risk, and hospital highlight the users’ emphasis on enhancing public health literacy as a core communication goal. Besides, the label cloud features terms like sharing, health news, and improvement reflecting their active role in science communication and public engagement. Detailed word frequency distributions are shown in Fig. 5.

Fig. 5.

Science information sharers.

Emotional support providers

The label cloud of emotional support providers demonstrates that their content primarily focuses on patient care, psychological support, and positive communication. High-frequency terms such as listening, emotion, strength, patient, mastery and sincerity, health, treatment, doctor, and reply highlight their emphasis on empathetic interaction and interpersonal engagement. Moreover, the presence of emotion-related terms such as hope, relief, anxiety, and depression suggests this group's strong concern with cancer patients' psychological conditions and emotional regulation. Detailed word frequency distributions are shown in Fig. 6.

Fig. 6.

Emotional support providers.

Social engagement facilitators

The label cloud of social engagement facilitators indicates that their content mainly revolves around TCM theory, health preservation knowledge, and symptom management. Frequently appearing keywords such as traditional Chinese medicine, health preservation, qi and blood, spleen and stomach, symptom, patient, medication, effect, efficacy, and treatment reflect their emphasis on the role of traditional medicine in disease management. Words such as relief, strength, clearing heat, and energetic also appear in the cloud, further demonstrating their tendency to promote TCM interventions, wellness concepts, and symptom relief strategies. Detailed word frequency distributions are shown in Fig. 7.

Fig. 7.

Social engagement facilitators.

Sentiment analysis results

Sentiment analysis was conducted on 62,154 comments from 37 representative AOKOLs, revealing notable differences in user emotional responses across the four categories (Fig. 8). Among the 37 accounts, the expert knowledge communicators (n ​= ​8) received the highest number of comments (n ​= ​39,955), with 39.08% classified as positive, 47.98% as neutral, and 12.94% as negative. For instance, comments on “Breast Surgeon Yang Qingfeng” included affirmative expressions such as “Very professional explanation,” “Well said,” and “Such a good doctor.” Negative responses often reflected concerns like “Fear of side effects,” “Sensationalism,” and “Inappropriate content,” while neutral comments typically consisted of factual or procedural inquiries, such as “Is iodine-131 therapy needed after benign tumor resection?” or “How soon can I jog after unilateral mastectomy?”

Fig. 8.

Comment sentiment analysis results.

Expert Knowledge Communicators (n ​= ​8, comments ​= ​39,955). Science Information Sharers (n ​= ​10, comments ​= ​2462). Emotional Support Providers (n ​= ​8, comments ​= ​7499). Social Engagement Facilitators (n ​= ​11, comments ​= ​12,238).

Among emotional support providers (n ​= ​8), the proportion of positive sentiment was the highest (43.47%), with neutral and negative sentiments accounting for 43.50% and 13.03%, respectively. For example, comments on “Dr. Yang, PhD in Oncology” commonly reflected appreciation for his empathetic content, with phrases such as “Well written, thank you for sharing,” “Hope is rekindled,” and “Love your post.” Negative comments frequently conveyed emotional distress, exemplified by expressions such as “This is terrifying” and “I feel hopeless.” In contrast, neutral responses were generally informational or cognitively driven, including statements like “This is my first time learning about this disease” or “This issue warrants attention—it appears to be increasingly prevalent among younger individuals.”

The social engagement facilitators (n ​= ​11) received a total of 12,238 comments, comprising 27.35% positive, 64.95% neutral, and 7.70% negative sentiments. For example, comments on “TCM Oncology Expert Li Zhong” reflected trust and satisfaction, such as “The video is easy to understand,” “Excellent medical skills and ethics,” “I've followed Professor Li for a long time,” and “TCM treatment is effective.” Negative responses included concerns over treatment efficacy or skepticism toward TCM principles, such as “Is TCM really scientific?” or “Multiple metastases occurred post-treatment, feeling helpless.” Neutral comments were mostly specific inquiries about TCM regimens, such as “Is this dampness-dispelling prescription suitable for general use?” and “What is the recommended dose of Panax notoginseng for health purposes?“.

Although science information sharers (n ​= ​10) received fewer comments (n ​= ​2462), the sentiment distribution remained largely neutral, with 72.48% of comments falling into the neutral category, 21.14% positive, and 6.38% negative. Taking “Thoracic Surgeon Wang Jiyong” as an example, positive comments reflected recognition of his professional tone and educational clarity, such as “Easy to understand and highly beneficial,” “Accurate language without exaggeration,” and “Patient and reassuring doctor.” Negative sentiments were mostly related to personal health anxieties or skepticism regarding information accuracy, for instance, “Is frequent testing risky?” or “Some statements are inaccurate.” Neutral comments largely involved knowledge-seeking questions, such as “Can CT scans detect gastric cancer?” and “How to manage side effects of targeted therapy?“.

Discussion

This study identified four core types of AOKOLs on the Weibo platform—expert knowledge communicators, science information sharers, emotional support providers, and social engagement facilitators—through clustering analysis and persona construction. Unlike previous classification approaches that primarily relied on follower counts, platform distribution, or topical focus,²⁶ our study incorporated professional background, content output, interaction patterns, and audience feedback, offering a more context-sensitive reflection of the heterogeneity in information and support provision. Notably, this classification not only highlights the diversity of KOLs in terms of professional credentials, communication styles, and social behaviors, but also maps onto the complex and dynamic support needs of cancer patients across their disease trajectory. Previous studies have demonstrated that cancer patients experience long and unpredictable care journeys, with shifting physical symptoms and psychosocial challenges across diagnosis, treatment, recovery, and recurrence stages—each associated with specific and layered support demands.²⁷ For instance, in addition to treatment-related information, breast cancer patients may experience persistent fear of recurrence and metastasis, while conventional health management models often fail to capture these changes in real time or offer personalized interventions.²⁸ In this context, the differentiation of AOKOL types is not merely relevant to dissemination strategies but is critical for providing stage-appropriate and personalized support throughout the cancer journey.

Our findings further illustrate how each AOKOL type corresponds to key patient needs. Expert knowledge communicators, often clinicians or researchers, build trust and enhance comprehension by sharing specialized medical knowledge, particularly during diagnosis and treatment decision-making. Science information sharers play a vital role in simplifying complex medical concepts into accessible language, delivering educational content suited to patients undergoing treatment and rehabilitation. Emotional support providers respond to patients' psychological distress through empathetic resonance, experience sharing, and positive reinforcement, consistently offering emotional support that accompanies patients through anxiety and fatigue. Social engagement facilitators foster online community cohesion by encouraging discussion and responding frequently, thereby enhancing peer interaction and social vitality. This classification framework for AOKOLs closely aligns with the real-world experiences of cancer patients throughout their care journeys. A grounded theory study of American breast cancer patients identified six core challenges throughout the care continuum—informational, psychosocial, financial, insurance-related, timeliness, and emotional support—emphasizing the importance of both access and continuity of care.²⁹ On social media, the identified AOKOL types correspond to these domains by acting as sources of information, emotional support, resource guidance, and community connection, thus helping to bridge persistent support gaps in traditional health care systems.³⁰ Moreover, prior researches further support the multidimensional roles of AOKOLs. For example, Chen et al.³¹ found that cancer-related Weibo discussions were concentrated on treatment information, consistent with the high-frequency terms such as “treatment” and “chemotherapy” in our expert knowledge communicators' word clouds. Research on online breast cancer support groups revealed that KOLs improve psychological well-being, underscoring their role in fostering not only information exchange but also deep emotional and social connections.³² Similarly, Han et al. found that “social support” was a recurring theme in HIV-related online communication, confirming the indispensable nature of emotional responses in patient support systems.³³ These studies collectively illustrate that in health communication, knowledge dissemination and emotional resonance are not mutually exclusive but intertwined, reflecting patients’ composite needs for cognitive and emotional support throughout different health stages.

Health communication in the social media era is no longer a one-way dissemination of information but a comprehensive process integrating expertise, emotional engagement, and community participation, with AOKOLs playing multifaceted roles through varied communication approaches.³⁴ Our analysis of engagement indicators and comment sentiment revealed distinct patterns across the four AOKOL types. Expert knowledge communicators tend to attract more attention and engagement due to their prominence, although their comments are mostly neutral, suggesting that users are primarily information-seeking. In contrast, emotional support providers elicit stronger affective responses, with a higher proportion of both positive and negative sentiments, indicating that emotionally oriented content provokes deeper emotional engagement. Followers of science information sharers prioritize knowledge acquisition, leading to mostly neutral and infrequently negative responses, whereas social engagement facilitators sustain vibrant, positive interactions and foster an engaging community space. These variations in emotional response suggest that users are not passive recipients of information but active participants with emotional and informational needs, and the depth of interaction and affective feedback vary significantly depending on content and communication style. Towne et al.³⁵ reported that emotional support posts on Facebook achieved substantially higher interaction rates compared to educational ones, which, though more abundant, saw limited audience involvement—supporting our findings of stronger sentiment responses within the emotional support providers. Additionally, Goel et al. also reported that Reddit users were more inclined to engage with emotional content such as help-seeking and advice than static medical information, underscoring the importance of emotional needs.³⁶

Beyond the content itself, the narrative strategies and communication styles employed by AOKOLs play a crucial role in shaping patient engagement within online communities. While expert knowledge communicators attract more comments due to their professional authority, this does not necessarily correlate with higher interaction. In contrast, social engagement facilitators, through active participation in discussions, also achieve strong user engagement, demonstrating the importance of relational interaction in cultivating influence. Wu et al.³⁷ demonstrated that digital opinion leaders' effectiveness depends not only on informational authority but also on their ability to balance professionalism with audience accessibility. Participatory, interactive, and shareable characteristics are seen as key mechanisms for maintaining influence in social networks.³⁸ Likewise, a study on TikTok videos about breast cancer found that health care professionals enhanced trust and engagement by employing rhetorical strategies such as appropriate tone, clear messaging, and credibility framing, thereby creating emotional bonds with diverse audiences.³⁹ These findings emphasize that AOKOLs’ effectiveness depends not only on the professionalism of information but also on how they craft trusted, engaging, and emotionally resonant narratives.

In addition to Western clinical practices, the four AOKOL categories also cover complementary areas including TCM, dietary guidance, nutritional strategies, and rehabilitative care. In this study, several AOKOLs incorporated TCM principles into their posts, such as recommending herbal remedies for managing chemotherapy side effects. These discussions were often accompanied by user comments sharing experiences or seeking additional information, broadening the scope of community dialogue. A study on short videos about acupuncture found that science-based content and expert accounts were most favored by users, with TCM acupuncture KOLs boosting health awareness and trust through personal branding, reinforcing professional authority, and enhancing user interaction, thereby promoting the dissemination and acceptance of TCM on social platforms.⁴⁰ The popularity of TCM topics may be attributed to high cultural receptivity and audience appeal on Chinese social platforms, indicating a growing public interest in integrative and individualized health care strategies.⁴¹

In summary, while the four AOKOL types identified in this study reflect different stages of cancer patients' support needs, discrepancies remain between KOL content and actual patient demands. Previous studies have shown that cancer patients typically express intense and more easily fulfilled demands for disease-related information, treatment plans, and adverse effects during the initial stages, whereas unmet needs concerning psychological care, emotional adjustment, and social bonding tend to increase as the illness advances.⁴² Goerling et al. highlighted that although most patients felt adequately informed, only half found the information helpful, and psychological support awareness was even lower.⁴³ A multi-center study in Germany found that only one-third of patients felt sufficiently informed about emotional support resources.⁴⁴ These findings align with the dominance of medical information among expert knowledge communicators and science information sharers in our study, yet highlights the structural limitations in the emotional content from the emotional support providers. Furthermore, cancer patients' engagement in online health communities reflects diverse support needs across emotional regulation, peer support, treatment decision-making, clinical process navigation, understanding medical outcomes, and drug management.⁴⁵ While social engagement facilitators and emotional support providers partly address emotional resonance and social needs, their content is fragmented and lacks structure. This situation echoes the findings of Krishnasamy et al.,⁴⁶ who highlighted that emotional support is often overlooked and lacks systematic intervention in real-world clinical contexts. Moreover, the rising reliance of young breast cancer patients on social media for information and emotional support, contrasted with the commercialization and emotional content scarcity on platforms, further highlights gaps in online support resources.¹⁶ Thus, the proposed AOKOL classification framework not only reveals digital health communication functions but also exposes the potential gap between KOL-generated content and actual patient needs. Future work should enhance KOL content structure and stage-specific accuracy, especially in emotional support, by developing more systematic and targeted communication strategies to better fulfill cancer patients’ multifaceted needs across their care continuum.

Limitations

Despite its contributions, this study has several limitations. First, the data were collected exclusively from the Weibo platform, and AOKOL behavioral characteristics may vary across different social media platforms. Second, the analysis focused primarily on textual data, without accounting for multimodal content such as videos and livestreams. Third, while sentiment analysis can reveal audience emotional tendencies, it does not directly uncover behavioral transformation pathways influenced by AOKOLs. Future research should investigate how different AOKOL types impact patient decision-making, health care behaviors, or digital health literacy through multi-platform and longitudinal studies to enhance the practical application of persona profiling in health communication. Lastly, we did not conduct a systematic evaluation of the accuracy of medical content disseminated by AOKOLs. While accounts prone to misinformation or commercial bias were excluded during screening, verifying clinical correctness at scale remains challenging due to data volume, content variability, and context dependency. This study prioritized profiling and audience responses. Therefore, future studies could enhance rigor by incorporating expert review or AI-based credibility assessments to better evaluate content quality across AOKOL types.

Conclusions

This study developed a multidimensional profiling framework for AOKOLs on the Weibo platform, encompassing content features, interaction behaviors, professional background, and social network structures. By identifying four representative influencer types, the study highlighted functional heterogeneity in health information dissemination, emotional support, and user engagement, offering new perspectives on the diverse roles of AOKOLs in digital health communication. The findings underscore the dynamic relationship between expertise, emotional resonance, and user participation, advancing theory on patient-centered digital communication. Future work should explore integrating user profiling with artificial intelligence and big data analytics to optimize content recommendation, refine emotional expression strategies, and enhance platform responsiveness to diverse user needs, ultimately promoting an empathetic and adaptive digital health ecosystem.

CRediT authorship contribution statement

Yiwen Duan: Conceptualization, Methodology, Formal analysis, Data curation, Writing – original draft, Visualization. Yi Chen: Conceptualization, Methodology, Formal analysis, Data curation, Writing – original draft. Qi Sun: Data curation, Writing – review & editing. Jialin Chen: Data curation, Writing – review & editing. Jiaojiao Zhang: Writing – review & editing. Bei Yun: Writing – review & editing. Tingting Cai: Conceptualization, Supervision, Funding acquisition, Writing – review & editing. Changrong Yuan: Conceptualization, Supervision, Project administration, Funding acquisition, Writing – review & editing. All authors have read and approved the final version of the manuscript.

Ethics statement

Not applicable.

Data availability

The data presented in this study are available on reasonable request from the corresponding author, TC, upon reasonable request.

Declaration of generative AI and AI-assisted technologies in the writing process

No AI tools/services were used during the preparation of this work.

Funding

This study was supported by the National Natural Science Foundation of China (Grant No. 72374048) and the Ministry of Education of the People's Republic of China Humanities and Social Sciences Youth Foundation, China (Grant No. 23YJC630002). The funders had no role in considering the study design or in the collection, analysis, interpretation of data, writing of the report, or decision to submit the article for publication.

Declaration of competing interest

The authors declare no conflict of interest. The corresponding author, Dr. Tingting Cai, is an editorial board member of Asia–Pacific Journal of Oncology Nursing. The article was subject to the journal's standard procedures, with peer review handled independently of Dr. Cai and their research groups.