Evaluating the reliability and quality of viral myocarditis educational content on TikTok and Bilibili

Abstract

Viral myocarditis (VM) affects over 1 million people yearly worldwide and may trigger arrhythmia, heart failure, or sudden death, bringing heavy clinical and economic burdens. TikTok and Bilibili are vital health information hubs, yet unregulated VM video content lacks quality evaluation on Chinese platforms. This cross-sectional study assessed VM video reliability and completeness on both platforms to support evidence-based health guidance. Data collection finished on February 15, 2026. We collected the top 100 default-ranked videos per platform via incognito new accounts searching “viral myocarditis.” After screening out ads, irrelevant, recent and duplicate videos, 180 clips (TikTok = 91, Bilibili = 89) were analyzed. Two trained cardiologists scored videos with the 0 to 5 modified DISCERN (instrument for evaluating health information quality) scale and 1 to 5 Global Quality Score and graded 6 disease domains’ completeness using a 3-point scale. Nonparametric tests and Spearman correlation were adopted, with P < .05 as the statistical significance threshold. TikTok videos had significantly higher median likes, comments, saves, and shares (all P < .001), while Bilibili videos were longer (139 vs 95 seconds, P = .002). Professional creators accounted for 71.43% of TikTok uploads versus 44.94% on Bilibili, which hosted more nonprofessional contributors. TikTok fully covered symptoms best (40.66%) but ignored epidemiology (83.52% uncovered); Bilibili offered more complete etiology, prevention, and symptom explanations yet lacked epidemiological content. Bilibili contained far more high-rating DISCERN (4–5) and 5-star Global Quality Score videos and no 1-point low-quality clips, unlike TikTok. Engagement metrics were strongly positively correlated on both platforms; video duration weakly negatively correlated with interaction on TikTok but moderately positively correlated on Bilibili. Inter-rater reliability was excellent (r = 0.81–0.96). Bilibili’s VM videos outperform TikTok in information completeness and overall quality. TikTok’s higher traffic often favors low-quality content, and preventive knowledge coverage is insufficient on both platforms. Such gaps arise from distinct user groups, content formats, and algorithms. Optimized algorithms, medical creator certification, and differentiated science communication can standardize VM online education and help patients access reliable disease management knowledge.

1. Introduction

Viral myocarditis (VM) is an inflammatory myocardial disease mediated by viral infection,^[1] with over 1 million new cases annually worldwide and distinct demographic characteristics. Its peak incidence occurs in adolescents and young adults, with males exhibiting a higher risk than females.^[2] As a clinically prevalent infectious cardiomyopathy, it can lead to serious adverse outcomes such as arrhythmias, heart failure, and even sudden death,^[3,4] imposing significant health and economic burdens on patients’ families and healthcare systems. With evolving viral infection patterns, changing lifestyles, and advances in myocardial injury diagnostics, the clinical detection rate and associated disease burden are projected to increase further.^[5,6] This underscores the urgent need for effective patient education and self-management strategies. Traditionally, healthcare providers served as the primary source of medical information, but the digital revolution has transformed this landscape, with patients increasingly turning to online platforms for health-related guidance.^[7]

In recent years, short-video platforms like TikTok and Bilibili have emerged as highly influential channels for health information dissemination due to their high accessibility and engaging nature.^[8] These platforms not only attract millions of users, including VM patients and health information seekers, but their visual and accessible content formats also lower the barrier to understanding medical knowledge. This makes them an ideal medium for people across ages and educational backgrounds to widely access health-related content.^[9] Notably, significant differences exist between the user bases and content on these 2 platforms. TikTok users are predominantly younger, with content creation emphasizing brevity and entertainment value.^[10,11] In contrast, Bilibili users span a broader age range and demonstrate higher receptivity to in-depth educational content in health domains. However, the lack of rigorous content review mechanisms on these platforms poses serious challenges to the accuracy and reliability of health information. A large number of previous studies focusing on other common diseases have confirmed that short-video platforms generally have problems with low-quality health content. False or misleading information about diseases such as diabetes and cancer is widely disseminated on short-video platforms, with some content lacking scientific support.^[12–14] Despite the continuous increase in the penetration rate of short-video platforms in health communication,^[15] research on the quality assessment of VM-related short videos remains scarce. Existing studies mostly focus on text or long-form video formats, and the presentation characteristics of information on VM etiology, clinical manifestations, diagnostic criteria, treatment plans, and rehabilitation guidance in short videos have not yet been clarified. Given the public’s increasing reliance on short-video health information.^[16,17] Particularly for VM patients who rely on digital resources for self-management, the clinical risks posed by misinformation cannot be overlooked. Therefore, systematically evaluating the quality of VM-related content on TikTok and Bilibili holds significant practical importance. It provides a basis for developing regulatory strategies to ensure patients access evidence-based, accurate health advice.

Thus, this study conducts a cross-sectional content analysis of VM-related videos on TikTok and Bilibili. The primary objective is to address the research gap in evaluating the quality of VM information on short-video platforms. Specifically, we will employ validated tools – such as the DISCERN (instrument for evaluating health information quality) scale and Global Quality Score (GQS) – to assess the quality and reliability of these videos. Through in-depth analysis of platform content, we aim to identify shortcomings in current health communication practices.

2. Materials and methods

This study performed content analysis on publicly accessible short video resources from online platforms. It did not involve human participants, human biological specimens, or any identifiable private personal information. For this reason, approval from the institutional review board and written informed consent were not required for this research.

2.1. Search strategy and data extraction

The primary objective of this study was to collect short video data on the topic of “viral myocarditis” from the TikTok (https://www.douyin.com) and Bilibili (https://www.bilibili.com) platforms. All search and data collection work was finished on February 15, 2026. To ensure data comprehensiveness and representativeness, the Chinese term “viral myocarditis” was used as the search keyword. Data collection was conducted separately on both platforms. To minimize bias from personalized recommendations, all searches were performed using newly created accounts in incognito mode, ensuring historical data and personalized algorithms did not influence results. In the initial phase, the top 100 videos on each platform were collected based on the platform’s default ranking rules. Previous studies indicate that videos ranked below 100 have negligible impact on analysis.^[18,19] Videos were included if they directly addressed the etiology, symptoms, diagnosis, treatment, or prevention of VM and were publicly accessible (excluding private or restricted content). Videos were excluded if identified as advertisements or promotional materials, unrelated to VM, only superficially mentioning VM without substantive information, uploaded within the last week prior to data collection (to ensure data stability and timeliness), or duplicates/reuploads of previously analyzed content (specifically, 9 videos were excluded from the TikTok dataset and 11 from the Bilibili dataset; see Fig. 1 for details). We set this 1-week time cutoff mainly for data stability. The engagement metrics, including likes, comments, saves, and shares of newly released videos, usually fail to reach a stable state within 7 days after publication, which may interfere with the correlation analysis between video quality and user interaction data. Meanwhile, this study focuses on the overall quality and long-term dissemination characteristics of conventional health education content about viral myocarditis rather than real-time emergency information during sudden disease outbreaks. Even during disease outbreaks, most timely emergency guidance videos will continue to gain exposure and stable interaction data after 1 week, so this exclusion rule will not cause the loss of core relevant content. The following variables were recorded during data extraction: video source (TikTok or Bilibili), user nickname, user link, video link, video title, number of likes, number of comments, number of saves, number of shares, and video duration (seconds). All data was extracted via platform public APIs to ensure accuracy and consistency. These variables provide robust support for subsequent content analysis.

Figure 1.

Video inclusion flowchart across different platforms.

2.2. Video quality assessment

To evaluate the reliability and quality of collected short videos, this study employs a modified DISCERN assessment method and the GQS. Using the modified DISCERN, evaluators determine whether each video meets the following criteria: clarity, relevance, traceability, robustness, and fairness. Each criterion is rated as “Yes” (1 point) or “No” (0 points), with a cumulative score calculated (0–5 points). GQS employs a 5-point Likert scale to evaluate overall video quality, ranging from 1 (poor) to 5 (excellent). This scale considers dimensions such as video professionalism, information completeness, presentation clarity, and audience comprehension. Original scores of the modified DISCERN tool are presented in Table S1, Supplemental Digital Content 1, GQS scores are listed in Table S2, Supplemental Digital Content 2, and the full study Original Dataset is provided as supplementary data. Additionally, video completeness is assessed based on whether it includes the following information: epidemiology, etiology, symptoms, diagnosis, treatment, and prevention. Videos are also scored according to the uploader’s level of detail in explaining the relevant topic, categorized into 3 tiers: no explanation (0 point), partial explanation (1 point), and complete explanation (2 points). Distinctions were made between specialist physicians (including those specializing in the cardiovascular system) and nonspecialist physicians (primarily from other medical fields or traditional Chinese medicine [TCM]). All evaluations were conducted jointly by 2 assessors with relevant medical backgrounds. Both assessors are full-time cardiovascular specialists with extensive clinical experience in the diagnosis and management of viral myocarditis and other cardiovascular disorders. Both received standardized training prior to scoring to ensure consistency in assessment criteria and minimize bias.

2.3. Statistical analysis

Descriptive statistics and nonparametric tests were used to analyze video characteristics and quality indicators. Categorical variables are presented as frequencies and percentages, while continuous variables are described using medians and interquartile ranges. The Mann–Whitney U test was applied for pairwise group comparisons. The Kruskal–Wallis H test was used for evaluating comparisons among 3 or more groups. Spearman correlation analysis examined the association between GQS scores and modified DISCERN scores. Statistical calculations were performed using IBM SPSS Statistics 26.0 software (IBM Corporation), with visualizations generated by GraphPad Prism 10.1 software (GraphPad Software, part of Dotmatics). The significance threshold was set at P < .05.

3. Results

3.1. Video characteristics and uploader types

This study analyzed 180 videos from TikTok and Bilibili, comprising 91 videos (50.6%) in the TikTok group and 89 videos (49.4%) in the Bilibili group. The median number of likes in the TikTok group was higher than that in the Bilibili group (2189.00 [493.00, 5699.00] vs 146.00 [36.50, 861.00]), and the median number of comments was higher in the TikTok group (78.00 [21.00, 315.00] vs 21.00 [3.00, 139.00]). The median number of bookmarks was higher than that of the Bilibili group (479.00 [122.00, 1894.00] vs 82.00 [24.50, 308.50]), and median shares were also higher than in the Bilibili group (523.00 [89.00, 2788.00] vs 59.00 [15.50, 352.00]); while the median video duration in the Bilibili group was significantly longer than that in the TikTok group (139.00 [78.50, 285.00] seconds vs 95.00 [60.00, 177.00] seconds). All interaction metrics showed statistically significant intergroup differences (P < .001), and the difference in video duration was also statistically significant (P = .002). General video characteristics are summarized in Table 1.

Table 1.

Video characteristics on TikTok and Bilibili platforms.

Variables	TikTok (n = 91)	Bilibili (n = 89)	Z	P
Likes, M (Q1, Q3)	2189.00 (493.00, 5699.00)	146.00 (36.50, 861.00)	−6.436	<.001
Comments, M (Q1, Q3)	78.00 (21.00, 315.00)	21.00 (3.00, 139.00)	−3.659	<.001
Collections, M (Q1, Q3)	479.00 (122.00, 1894.00)	82.00 (24.50, 308.50)	−4.966	<.001
Shared, M (Q1, Q3)	523.00 (89.00, 2788.00)	59.00 (15.50, 352.00)	−5.196	<.001
Duration (s), M (Q1, Q3)	95.00 (60.00, 177.00)	139.00 (78.50, 285.00)	−3.099	.002

Among 180 VM-related videos, 105 (58.33%) were uploaded by professional individuals, 32 (17.78%) by nonprofessional individuals, 7 (3.89%) by professional organizations, and 36 (20.00%) by nonprofessional organizations. This indicates that professional individuals dominate the creation and dissemination of VM-related content (Fig. 2). Additionally, on the TikTok platform, 71.43% of VM-related videos were uploaded by professional individuals, while nonprofessional individuals accounted for 8.79%. Professional and nonprofessional organizations contributed only 4.40% and 15.38%, respectively, indicating a relatively concentrated source of uploaders (Fig. 2A). In contrast, Bilibili exhibits greater diversity in uploader distribution: professional individuals accounted for 44.94% (remaining the primary group), nonprofessional individuals for 26.97%, professional organizations for 3.37%, and nonprofessional organizations for 24.72%. The proportion of nonprofessional uploaders was significantly higher than on TikTok (Fig. 2B).

Figure 2.

Distribution of videos by source type on (A) TikTok and (B) Bilibili platforms scores across different platforms.

3.2. Completeness of video content

Among 180 VM-related videos, we assessed information completeness across 6 dimensions (Table 2) – epidemiology, etiology, symptoms, diagnosis, treatment, and prevention – using a scoring system of “Not Covered (0 points), Partially Explained (1 point), Fully Explained (2 points).”

Table 2.

Content completeness of videos on TikTok and Bilibili platforms.

Platform	Video contents	Not involve (0 point)	Partial explanation (1 point)	Full explanation (2 points)
TikTok	Epidemiology, n (%)	76 (83.52)	11 (12.09)	4 (4.39)
	Etiology, n (%)	29 (31.87)	57 (62.64)	5 (5.49)
	Symptoms, n (%)	6 (6.59)	48 (52.75)	37 (40.66)
	Diagnosis, n (%)	37 (40.66)	32 (35.16)	22 (24.18)
	Treatment, n (%)	36 (39.56)	40 (43.96)	15 (16.48)
	Prevention, n (%)	38 (41.76)	41 (45.05)	12 (13.19)
Bilibili	Epidemiology, n (%)	61 (68.54)	24 (26.97)	4 (4.49)
	Etiology, n (%)	15 (16.85)	49 (55.06)	25 (28.09)
	Symptoms, n (%)	12 (13.48)	49 (55.06)	28 (31.46)
	Diagnosis, n (%)	29 (32.58)	43 (48.31)	17 (19.11)
	Treatment, n (%)	32 (35.96)	46 (51.69)	11 (12.35)
	Prevention, n (%)	25 (28.09)	38 (42.70)	26 (29.21)

On the TikTok platform, the epidemiology dimension showed the poorest coverage: 83.52% of videos did not address related content, only 12.09% provided partial explanations, and a mere 4.39% offered complete explanations. For etiology, 31.87% of videos lacked coverage, 62.64% provided partial explanations, and 5.49% offered complete explanations. In contrast, the symptom dimension showed the strongest information completeness: only 6.59% of videos omitted coverage, 52.75% provided partial explanations, and 40.66% delivered complete explanations. For the diagnosis dimension, 40.66% of videos did not cover it, 35.16% provided partial explanations, and 24.18% offered complete explanations. Regarding the treatment dimension, 39.56% of videos did not cover it, 43.96% provided partial explanations, and 16.48% offered complete explanations. Regarding the prevention dimension, 41.76% of videos did not cover it, 45.05% provided partial explanations, and 13.19% offered complete explanations.

On the Bilibili platform, information coverage for the epidemiology dimension was relatively poor: 68.54% of videos did not cover relevant content, 26.97% provided partial explanations, and only 4.49% offered complete explanations. In contrast, the etiology, symptoms, and prevention dimensions demonstrated superior information completeness. For etiology, 16.85% of videos did not cover the topic, 55.06% provided partial explanations, and 28.09% offered complete explanations. For symptoms, 13.48% of videos omitted information, 55.06% provided partial explanations, and 31.46% offered complete explanations. Regarding prevention, 28.09% of videos omitted information, 42.70% provided partial explanations, and 29.21% offered complete explanations. Regarding the diagnosis dimension, 32.58% of videos did not cover it, 48.31% provided partial explanations, and 19.11% offered complete explanations. The treatment dimension showed moderate information completeness: 35.96% of videos did not cover it, 51.69% provided partial explanations, and 12.35% offered complete explanations.

Overall, significant disparities exist in information completeness across VM-related videos on the 2 platforms. TikTok demonstrated relatively better coverage in the symptom dimension but exhibited severe gaps in epidemiological information. Bilibili showed superior completeness in etiology, symptoms, and prevention dimensions, though its epidemiological coverage remains underdeveloped.

3.3. Video content analysis

Content analysis of 180 VM-related videos across TikTok (n = 91) and Bilibili (n = 89) platforms revealed the following thematic classifications: disease knowledge, prevention, treatment, TCM, daily care, and advertising/other. Among these, TikTok featured the highest coverage of disease knowledge content (n = 70), followed by prevention (n = 59), while also containing significant proportions of daily care (n = 54), treatment (n = 36), TCM (n = 10), and advertising/other (n = 13). The Bilibili platform also featured disease knowledge content as the primary category (n = 75), with prevention (n = 68) and daily care (n = 68) also holding significant proportions. Treatment-related content (n = 55) followed, alongside TCM (n = 12) and advertising/other content (n = 13), exhibiting an overall content distribution similar to TikTok.

Table 3 lists the median engagement metrics (interquartile range) for each content type across both platforms. On TikTok, the median number of likes ranged from 1317.00 (280.25, 3207.50) to 3756.00 (1256.00, 52,500.00), while the median number of comments ranged from 36.50 (4.75, 161.25) to 182.00 (74.00, 7230.50), the median number of saves ranged from 416.50 (95.00, 1398.75) to 702.00 (180.50, 4057.50), the median number of shares ranged from 334.50 (74.00, 1341.25) to 2454.00 (246.50, 61,531.00), and the median video duration ranged from 44.00 (25.00, 249.00) to 171.50 (92.25, 212.50) seconds. Further statistical analysis revealed no statistically significant differences (P > .05) in likes, comments, saves, shares, or video duration across content types on the TikTok platform. On the Bilibili platform, the median number of likes across content types ranged from 56.00 (11.00, 428.25) to 215.00 (50.00, 347.50), while the median number of comments ranged from 8.00 (2.25, 80.25) to 39.00 (6.50, 115.00), the median number of bookmarks ranged from 58.00 (31.00, 227.00) to 102.00 (28.50, 315.75), the median number of shares ranged from 49.00 (3.00, 111.00) to 62.00 (18.00, 211.00), and the median video duration ranged from 129.00 (78.25, 286.50) to 215.00 (86.25, 582.00) seconds. Statistical analysis indicates that differences in likes, comments, saves, shares, and video duration across content types on Bilibili are also not statistically significant (P > .05).

Table 3.

Characteristics of VM-related video content on TikTok and Bilibili platforms.

	Variables	Disease knowledge (n = 70)	Prevention (n = 59)	Treatment (n = 36)	Traditional Chinese medicine (n = 10)	Daily care (n = 54)	Advertisement and others (n = 13)	H	P
TikTok	Likes, M (Q1, Q3)	1888.00 (363.00, 4458.25)	2712.00 (663.00, 5699.00)	1888.00 (448.50, 4198.00)	1317.00 (280.25, 3207.50)	2428.00 (605.50, 4886.25)	3756.00 (1256.00, 52,500.00)	6.096	.297
	Comments, M (Q1, Q3)	64.50 (20.25, 222.25)	91.00 (27.00, 295.00)	46.00 (21.50, 156.50)	36.50 (4.75, 161.25)	72.00 (22.50, 270.25)	182.00 (74.00,7230.50)	9.496	.091
	Collections, M (Q1, Q3)	416.50 (95.00, 1398.75)	565.00 (154.00, 2014.00)	450.00 (115.75, 996.25)	671.00 (163.75, 1682.25)	455.50 (133.50, 1398.75)	702.00 (180.50, 4057.50)	2.349	.799
	Shared, M (Q1, Q3)	440.50 (77.75, 1947.00)	723.00 (84.00, 2405.00)	334.50 (74.00, 1341.25)	533.00 (29.00, 1356.00)	582.00 (82.50, 1888.25)	2454.00 (246.50, 61,531.00)	6.326	.276
	Duration (s), M (Q1, Q3)	97.00 (65.75, 176.25)	128.00 (84.00, 191.00)	171.50 (92.25, 212.50)	99.00 (52.50, 163.50)	141.00 (84.00, 193.50)	44.00 (25.00, 249.00)	10.094	.073
H		103.084	87.848	65.609	14.959	87.538	21.663
P		<.001	<.001	<.001	.005	<.001	<.001
Bilibili	Likes, M (Q1, Q3)	178.00 (29.00, 930.00)	196.50 (64.00, 911.25)	119.00 (26.00, 528.00)	56.00 (11.00, 428.25)	134.00 (31.00, 687.50)	215.00 (50.00, 347.50)	2.739	.740
	Comments, M (Q1, Q3)	21.00 (2.00, 140.00)	24.00 (4.25, 165.75)	16.00 (2.00, 91.00)	8.00 (2.25, 80.25)	20.00 (3.00, 121.25)	39.00 (6.50, 115.00)	3.161	.675
	Collections, M (Q1, Q3)	87.00 (22.00, 387.00)	102.00 (28.50, 315.75)	82.00 (25.00, 228.00)	71.00 (19.00, 259.75)	75.00 (21.25, 227.50)	58.00 (31.00, 227.00)	0.787	.978
	Shared, M (Q1, Q3)	55.00 (13.00, 411.00)	60.00 (26.50, 354.50)	52.00 (10.00, 219.00)	49.00 (3.00, 111.00)	54.50 (14.25, 265.50)	62.00 (18.00, 211.00)	1.922	.860
	Duration (s), M (Q1, Q3)	131.00 (78.00, 282.00)	149.50 (87.50, 321.00)	157.00 (87.00, 410.00)	215.00 (86.25, 582.00)	129.00 (78.25, 286.50)	197.00 (106.50, 816.50)	3.116	.682
H		32.559	30.208	38.993	13.188	31.335	12.608
P		<.001	<.001	<.001	.010	<.001	.013

VM = viral myocarditis.

Furthermore, comparing overall engagement across both platforms reveals that TikTok’s median counts for likes, comments, saves, and shares were significantly higher than Bilibili’s across all content categories. This difference was statistically significant (P < .001) for all content types. Only Bilibili’s “Traditional Chinese Medicine” category and “Advertisement and others” category videos exhibited relatively higher values in the duration metric.

3.4. Comparative study of GQS and DISCERN score distributions on TikTok and Bilibili platforms

A comparative analysis of GQS and DISCERN scores was conducted for 180 VM-related videos on TikTok and Bilibili platforms. Results indicate significant differences in the DISCERN score distribution for Reviewer A: TikTok videos showed higher proportions of 1-point (8.79% vs 1.12%) and 2-point (26.37% vs 5.62%) ratings, while the 3-point rating was similar (25.28% vs 24.72%); Bilibili videos showed significantly higher proportions of 4-point (32.58% vs 28.57%) and 5-point (35.96% vs 10.99%) ratings. Reviewer B’s DISCERN scores showed similar trends: TikTok videos had higher proportions of 1-star ratings (8.79% vs 1.12%), 2-star ratings (30.77% vs 6.74%), and 3-star ratings (18.68% vs 15.73%); Bilibili videos showed significantly higher proportions of 4-point (28.09% vs 24.18%) and 5-point (48.32% vs 17.58%) ratings.

Reviewer A’s GQS score distribution showed significant differences: TikTok videos had higher proportions of 1-star ratings (12.09% vs 0%), 2-star ratings (24.17% vs 5.62%), and 3-star ratings (29.67% vs 23.59%); while Bilibili videos had significantly higher proportions of 4-star (39.33% vs 25.28%) and 5-star (31.46% vs 8.79%) ratings. Notably, TikTok contained videos rated 1 star, whereas Bilibili had none. Reviewer B’s GQS ratings also showed significant differences: TikTok videos had higher proportions of 1-star (12.09% vs 0%), 2-star (28.57% vs 4.50%), and 3-star (18.68% vs 16.85%) ratings; while Bilibili videos had significantly higher proportions of 4-star (38.20% vs 29.67%) and 5-star (40.45% vs 10.99%) ratings. TikTok still had 1-star videos, while Bilibili had none (Table 4). Furthermore, box plots and mountain plots provide clearer visualization of the concentration trends in ratings across the 2 evaluators (Fig. 3).

Table 4.

Rating distribution of videos on TikTok and Bilibili platforms by different evaluators.

Reviewers’ score	TikTok (n = 91)	Bilibili (n = 89)	Total (n = 180)
A (DISCERN, score), n (%)
1	8 (8.79%)	1 (1.12%)	9 (5.00%)
2	24 (26.37%)	5 (5.62%)	29 (16.11%)
3	23 (25.28%)	22 (24.72%)	45 (25.00%)
4	26 (28.57%)	29 (32.58%)	55 (30.56%)
5	10 (10.99%)	32 (35.96%)	42 (23.33%)
B (DISCERN, score), n (%)
1	8 (8.79%)	1 (1.12%)	9 (5.00%)
2	28 (30.77%)	6 (6.74%)	34 (18.89%)
3	17 (18.68%)	14 (15.73%)	31 (17.22%)
4	22 (24.18%)	25 (28.09%)	47 (26.11%)
5	16 (17.58%)	43 (48.32%)	59 (32.78%)
A (GQS, score), n (%)
1	11 (12.09%)	0	11 (6.11%)
2	22 (24.17%)	5 (5.62%)	27 (15.00%)
3	27 (29.67%)	21 (23.59%)	48 (26.67%)
4	23 (25.28%)	35 (39.33%)	58 (32.22%)
5	8 (8.79%)	28 (31.46%)	36 (20.00%)
B (GQS, score), n (%)
1	11 (12.09%)	0	11 (6.11%)
2	26 (28.57%)	4 (4.50%)	30 (16.67%)
3	17 (18.68%)	15 (16.85%)	32 (17.78%)
4	27 (29.67%)	34 (38.20%)	61 (33.89%)
5	10 (10.99%)	36 (40.45%)	46 (25.55%)

DISCERN = DISCERN instrument for evaluating health information quality, GQS = Global Quality Score.

Figure 3.

Visualization of evaluation scores. (A) Box plots of DISCERN scores on TikTok and Bilibili platform; (B) mountain plots of DISCERN scores from different reviewers on TikTok platform; (C) mountain plots of DISCERN scores from different reviewers on Bilibili platform; (D) box plots of GQS scores on TikTok and Bilibili platform; (E) mountain plots of GQS scores from different reviewers on TikTok platform; (F) mountain plots of GQS scores from different reviewers on Bilibili platform. DISCERN = DISCERN instrument for evaluating health information quality, GQS = Global Quality Score.

3.5. Spearman correlation analysis

Spearman correlation analysis revealed that on the TikTok platform, user engagement metrics such as likes, comments, saves, and shares exhibited significant strong positive correlations, with correlation coefficients ranging from 0.85 to 1.00, indicating high consistency. Video duration showed extremely weak correlations with all interaction metrics, with coefficients ranging from −0.14 to 0.05. Some metrics even exhibited slight negative correlations. Notably, the DISCERN scores and GQS scores assigned by the 2 reviewers showed mostly weak negative correlations with user engagement metrics (correlation coefficients ranging from −0.20 to −0.06). However, a strong positive correlation was found between Reviewer A’s and Reviewer B’s DISCERN scores (r = 0.83) and GQS scores (r = 0.96), indicating good consistency in their ratings (specific correlation coefficients shown in Fig. 4A). GQS scores (r = 0.96) showed significant strong positive correlations, indicating good inter-reviewer agreement (see Fig. 4A for specific correlation coefficients). On the Bilibili platform, user interaction metrics also exhibited significant strong positive correlations, with correlation coefficients ranging from 0.67 to 0.89. Unlike TikTok, video duration showed a moderately strong positive correlation (r = 0.27) with the number of bookmarks, while exhibiting weak correlations with other interaction metrics. Additionally, both evaluators’ DISCERN scores and GQS scores showed predominantly weak positive correlations with user interaction metrics (correlation coefficients ranging from 0.01 to 0.51). Furthermore, strong positive correlations were observed between Evaluator A’s and Evaluator B’s DISCERN scores (r = 0.82) and GQS scores (r = 0.81), respectively, GQS scores (r = 0.81) between Reviewers A and B also showed significant strong positive correlations, indicating good inter-reviewer consistency (see Fig. 4B for specific correlation coefficients).

Figure 4.

Spearman correlation analysis. (A) Spearman correlation analysis on TikTok platform; (B) Spearman correlation analysis on Bilibili platform. DISCERN = DISCERN instrument for evaluating health information quality, GQS = Global Quality Score.

4. Discussion

This study systematically evaluated the characteristics, information integrity, and quality of VM-related short videos on TikTok and Bilibili through cross-sectional content analysis. Core findings are closely linked to each platform’s user demographics, content ecosystem, and algorithmic mechanisms. Bilibili’s VM videos demonstrated significantly superior information integrity (across etiology, prevention, and symptom dimensions) and quality scores (higher proportions of high GQS and DISCERN ratings) compared to TikTok, while TikTok’s user engagement metrics (likes, comments, saves, and shares) were significantly higher than Bilibili’s (e.g., median likes: 2189.00 vs 146.00), yet showed no significant correlation with content quality. Conversely, lower-quality content sometimes garnered higher engagement. These findings not only fill a research gap in assessing the quality of VM health information on Chinese short-video platforms but also resonate with global concerns about the reliability of medical information on social media, providing empirical evidence for optimizing future health communication strategies.

4.1. Differentiated impact of platform characteristics on VM content quality

The significant disparity in VM content quality between the 2 platforms fundamentally stems from the combined effects of their core positioning, content formats, and user demographics. TikTok’s core competitive advantage lies in its fast-paced, entertaining content.^[20] With a median video length of 95 seconds, this format objectively limits the space available for conveying information about acute infectious cardiomyopathies like VM. As a virus-mediated myocardial inflammation, VM’s etiology involves multiple viral subtypes (e.g., Coxsackievirus, adenovirus, etc).^[21,22] Diagnosis requires integrating clinical manifestations, cardiac enzyme profiles, electrocardiograms, and imaging evidence,^[23] while treatment encompasses individualized approaches, including symptomatic support and immunomodulation.^[24] Systematic explanations of such specialized content typically demand extended context and interpretation. Consistent with previous research on social media algorithmic bias in health communication, TikTok’s algorithmic recommendation system prioritizes high-engagement, short-duration content over comprehensive and professional medical content.^[25] To capture traffic, creators are compelled to oversimplify complex medical knowledge, sometimes sacrificing professionalism to align with the platform’s entertainment-oriented tone.^[26,27] This explains why content covering specialized dimensions like epidemiology (83.52% not addressed) accounts for <10% of TikTok’s epidemiological coverage. Even though 71.43% of videos are uploaded by professionals, striking a balance between traffic dissemination and professional expression remains challenging.

In contrast, Bilibili’s platform ecosystem is more conducive to disseminating in-depth health content.^[28] With a median video length of 139 seconds, creators have ample space to systematically explore professional topics such as the etiology and pathogenesis of VM, diagnostic criteria, treatment pathways, and prevention guidelines. For instance, some Bilibili videos analyze key differential diagnosis points for VM using clinical cases or demonstrate the pathological process of virus-mediated myocardial injury through animations. While such content requires higher production costs, it aligns with the platform users’ demand for in-depth educational material. Furthermore, Bilibili boasts a broader age distribution among its user base, including VM patients and their families, as well as healthcare professionals. This demographic demands higher standards of professionalism and accuracy in health information, creating a positive feedback loop where high-quality content garners significant recognition.^[29] Relevant studies have confirmed that platforms favoring long-form educational content tend to form a more rigorous environment for health information dissemination.^[30] In this study, the proportion of videos scoring 5 on the GQS scale on Bilibili (Reviewer A: 31.46%, Reviewer B: 40.45%) was significantly higher than on TikTok (Reviewer A: 8.79%, Reviewer B: 10.99%), directly reflecting this ecosystem.

4.2. Correlation between creator types and content quality

Although professional individuals dominate VM content creation on both major platforms (accounting for 58.33% overall), the impact of creator types on content quality varies significantly across platforms. On TikTok, the proportion of content uploaded by professional individuals (71.43%) far exceeds that on Bilibili (44.94%), yet content quality does not correspondingly improve. This phenomenon likely stems from platform constraints on content formats.^[31,32] To adapt to TikTok’s short-form, high-tempo dissemination characteristics, professional creators often fragment and simplify specialized knowledge or even resort to attention-grabbing expressions to gain engagement. This approach leads to insufficient scientific rigor in some content. For instance, some TikTok videos oversimplify VM treatment as “cured by certain Chinese herbal medicine” or “healed through rest alone,” disregarding the critical need for hospitalized, standardized treatment for severe cases. While such content garners high engagement, it carries a significant risk of misinformation.

Conversely, Bilibili hosts a higher proportion of nonprofessional uploaders (26.97% vs TikTok’s 8.79%), yet this hasn’t significantly lowered overall content quality. This contrast stems from Bilibili’s longer content format, which allows nonprofessional creators to cite authoritative literature and integrate professional resources – enabling nonexperts to publish expert-level content. For instance, some nonprofessional Bilibili creators combine their personal viral myocarditis recovery experiences with core insights from current viral myocarditis treatment guidelines, sharing long-term management strategies following standardized treatment. Such content balances the emotional impact of personal narratives with the assurance of accurate information. Furthermore, Bilibili’s “bullet chat interaction” mechanism provides additional safeguards for content quality. Professional users supplement key technical points and correct misstatements within the chat stream, creating a dynamic optimization process where creators produce content while users collectively refine it. This stands in stark contrast to TikTok’s predominantly 1-way dissemination model.^[33]

4.3. Correlation between engagement levels and content quality and its clinical risks

The mismatch between engagement levels and content quality observed in this study – where low-quality VM content on TikTok garners higher engagement – is not unique to the VM field but represents a common issue in health information dissemination across social media.^[34,35] Previous studies on short videos concerning chronic diseases like diabetes and hypertension have reached similar conclusions.^[36,37] Content promoting exaggerated folk remedies or oversimplifying complex treatments often garners higher engagement metrics like likes and shares, while evidence-based professional content, due to its higher accessibility barriers and perceived lack of entertainment value, tends to receive relatively lower interaction rates. This phenomenon stems from the combined influence of ordinary users’ cognitive limitations regarding medical information and platform traffic algorithms. VM patients and the general public often lack sufficient medical knowledge to discern content accuracy, preferring easily digestible, clearly effective information. Platform algorithms, prioritizing user engagement metrics like dwell time and interaction frequency, further amplify the reach of such content.

The clinical risks arising from this trend cannot be overlooked. As an acute cardiomyopathy capable of triggering arrhythmias, heart failure, or even sudden death, the core of VM treatment lies in early recognition, standardized intervention, and long-term management.^[38] Any misleading information may influence patients’ treatment decisions. For instance, some short videos exaggerate the efficacy of TCM or alternative therapies while downplaying the critical need for ECG monitoring and supportive care in severe cases, potentially delaying medical intervention. Simplified descriptions of VM diagnostic criteria may cause early-stage patients with atypical symptoms (such as fatigue or palpitations alone) to miss optimal intervention windows. Furthermore, this study found that while preventive content coverage improved on both platforms (TikTok: 13.19% complete explanations; Bilibili: 29.21%), it remains insufficient. Modifiable risk factors for VM (e.g., preventing viral infections, boosting immunity) significantly influence disease onset.^[39] The absence of such content may cause patients to overlook the importance of primary prevention, thereby increasing their risk of developing the disease.

4.4. Relevance to existing research

The core findings of this study resonate with and complement conclusions from global research on the quality of health information on social media.^[40–42] For instance, a study on myocarditis-related videos on YouTube found that only a small proportion of content met evidence-based medicine standards, with video length positively correlated with quality.^[43] This aligns with our finding that longer videos on Bilibili exhibit superior quality. Research on TikTok indicates a high proportion of misleading health content, consistent with our observation of a high prevalence of low GQS scores on TikTok.^[44] Notably, VM content on Chinese platforms exhibits distinct characteristics. TCM-related content holds a relatively high proportion (TikTok: 10.99%, Bilibili: 13.48%), reflecting TCM’s widespread application in infectious disease management. However, some TCM content exaggerates efficacy and lacks scientific validation – a phenomenon rarely addressed in previous foreign-language studies, constituting a unique contribution of this research.

Furthermore, the study’s findings of “good inter-rater reliability” (DISCERN and GQS scoring correlations ranging from 0.8 to 0.96) validate the applicability of the modified DISCERN and GQS scales for evaluating health information in Chinese short videos, providing a methodological reference for future similar research. The divergent correlations observed across platforms (weak negative correlation between duration and engagement on TikTok versus moderate positive correlation on Bilibili) further illuminate the influence of platform algorithms on content dissemination. Specifically, TikTok’s algorithm favors shorter content, while Bilibili users demonstrate higher receptivity to longer, in-depth content. This distinction provides crucial insights for developing targeted health communication strategies.

5. Limitations

This study has several limitations that warrant clarification to provide directions for future research improvements. First, the sample selection exhibits certain selection biases. This study utilized the top 100 videos in the platform’s default ranking as research subjects. While this approach aligns with common strategies in similar cross-sectional studies,^[45,46] the default ranking is influenced by multiple factors such as algorithmic recommendations, publication time, and interaction volume. Consequently, it cannot fully represent all VM-related content on the platform. For instance, some high-quality videos published earlier or with lower interaction volumes may have been excluded. Furthermore, algorithmically recommended popular videos may be overly concentrated among specific creators or content types. Future research could employ multidimensional sampling strategies (e.g., combining keyword searches with stratified sampling by creator type) to enhance sample representativeness. Second, the study scope is limited to Chinese-language content, excluding VM short videos in other languages. Although TikTok and Bilibili are predominantly Chinese-user platforms, some foreign-language content circulates on them. Furthermore, the dissemination of VM health information may vary across languages and regions (e.g., Western platforms may emphasize antiviral treatments, while Southeast Asian platforms may focus more on traditional therapies). Therefore, the generalizability of this study’s conclusions is restricted to Chinese-speaking audiences. Future cross-national, multilingual comparative studies could reveal global commonalities and differences in VM short video content. Additionally, content quality assessment carries inherent risks of subjective bias. Although 2 evaluators with medical backgrounds conducted the assessments after standardized training and demonstrated good inter-rater reliability, their medical expertise may influence judgments of content professionalism. For instance, cardiovascular specialists may prioritize diagnostic criteria and treatment protocol accuracy, while physicians from other disciplines might emphasize content comprehensibility. Furthermore, this study excluded lay users from evaluation, failing to capture nonprofessional audiences’ actual perceptions of content quality. Future research could adopt a dual-evaluation system combining expert assessment with user feedback to enhance result comprehensiveness.

Furthermore, the cross-sectional study design fails to capture dynamic shifts in content quality or long-term impacts. Data collection was limited to a specific time point, omitting tracking of video dissemination trends, the evolution of user feedback, and the assessment of content quality’s actual influence on VM patients’ health behaviors (e.g., treatment adherence and healthcare decisions). For instance, some low-quality content may achieve high engagement in the short term but could be corrected by professional users or removed by platforms over time, while high-quality content may gradually gain influence through word-of-mouth dissemination. Future longitudinal studies could explore the dissemination patterns of VM short videos and their impact on patient health outcomes. Finally, this study did not consider the influence of supplementary information on video quality, as assessments were based solely on the video content itself, excluding elements like video descriptions, comment section interactions, or creator profiles. For instance, some videos might be incomplete in content but include physician contact details in their descriptions. Comments sections may feature expert users supplementing key knowledge points to enhance overall value. Additionally, creators’ professional certification status could influence user trust in content, yet this study excluded such variables from analysis. Future research should expand evaluation dimensions to comprehensively assess the overall quality of videos and their supplementary information. Notably, TCM-related content accounts for a noticeable proportion of videos on both platforms, but this study did not separately evaluate the accuracy and scientific validation of relevant claims. A small portion of such content may lack rigorous scientific support, which is another limitation of the present work. In view of this, conducting targeted evaluations on the rationality and credibility of TCM-related health information can serve as an important direction for subsequent research. Additionally, this study only divided uploaders into professional and nonprofessional categories and did not further verify their formal medical licenses or official institutional affiliations. Examining the authenticity of uploaders’ professional credentials will help improve the rigor of relevant research and is worthy of consideration in future investigations.

6. Conclusions and outlook

This study systematically reveals for the 1st time the quality characteristics of VM-related short videos on TikTok and Bilibili platforms. Bilibili significantly outperforms TikTok in information completeness (across etiology, prevention, and symptom dimensions) and quality scores (higher proportions of GQS and DISCERN high ratings). While TikTok achieves higher engagement (median likes: 2189.00 vs 146.00), it exhibits a pattern where lower-quality content gains greater dissemination. Both platforms still have room for improvement in covering preventive content. Future directions: from a platform perspective, TikTok should optimize algorithmic weighting, establish professional certification and content review mechanisms, and increase exposure for high-quality content. Bilibili could further develop a dedicated VM science popularization section, strengthen creator empowerment and support, and enhance the depth of prevention-related content. Healthcare professionals should tailor content creation to platform characteristics: prioritizing concise, actionable science communication (e.g., early symptom recognition and emergency protocols) on TikTok, while developing evidence-based in-depth content (e.g., guideline interpretations and long-term management strategies) on Bilibili. Additionally, the public must enhance critical evaluation skills when accessing such information, favoring content from creators with verified professional credentials.

This study provides empirical evidence for standardizing health information dissemination on short-video platforms. Future efforts should involve multi-stakeholder collaboration among platforms, healthcare professionals, the public, and relevant authorities. This approach will ensure both dissemination efficiency and scientific rigor, empowering patients to access accurate and reliable disease knowledge, improve disease management outcomes, and reduce the risk of adverse events.

Acknowledgments

The authors wish to express their gratitude to all participants who contributed to this study.

Author contributions

Conceptualization: Zipeng Zhu.

Data curation: Zipeng Zhu.

Formal analysis: Zipeng Zhu, Xuexu Su.

Project administration: Xuexu Su.

Supervision: Xuexu Su.

Writing – original draft: Zipeng Zhu.

Writing – review & editing: Zipeng Zhu, Xuexu Su.