Quality and reliability of cervical spondylotic myelopathy–related content on TikTok and Bilibili

Abstract

Short-video platforms such as TikTok and Bilibili have become widely used sources of health information. This study assessed the informational content, quality, and reliability of videos related to cervical spondylotic myelopathy (CSM) on these platforms. A cross-sectional search was conducted on October 1, 2025, using the Chinese-language equivalent of “cervical spondylotic myelopathy” as the sole search term. For each eligible video, data on duration, user interaction metrics, uploader identity, and content themes were collected. Video quality and reliability were evaluated using the Global Quality Scale (GQS) and the modified DISCERN (mDISCERN) instrument. Group differences were analyzed using the Mann-Whitney U and Kruskal-Wallis tests, and correlations were examined using Spearman analysis. In total, 274 videos met the inclusion criteria. Most videos addressed diagnosis (76.28%) and treatment (74.82%), while prognostic information was scarce (16.06%). The median GQS score was 3.00 (interquartile range [IQR]: 2.00–3.00), and the median mDISCERN score was 4.00 (IQR: 3.00–4.00). TikTok videos achieved significantly higher GQS and mDISCERN scores than those on Bilibili (P < .001). Videos produced by specialized healthcare professionals (SHPs) obtained the best evaluation scores (P < .05). Engagement indicators showed no significant correlation with GQS or mDISCERN scores (P > .05). The overall educational value of CSM-related videos was suboptimal. Content created by SHPs demonstrated superior quality and reliability. Video engagement metrics were not associated with video quality or reliability. Greater professional involvement and preferential presentation of evidence-based content warrant further investigation as potential approaches to improving the dissemination of high-quality CSM information on these platforms.

1. Introduction

Cervical spondylotic myelopathy (CSM) is the leading cause of spinal cord dysfunction in adults and typically results from age-related degenerative changes that produce progressive spinal cord compression.^[1] It represents the most severe form of cervical spondylosis and manifests as sensory loss, limb weakness, and gait disturbance, which may progress to paralysis if left untreated.^[2] The estimated prevalence ranges from 41 to 60 cases per 100,000 population, with substantially higher rates among adults older than 50 years.^[3] CSM is highly prevalent in Asian populations, particularly in China, affecting approximately 20% to 35% of patients with cervical spondylosis, and 10% to 25% of mild cases deteriorate over time.^[4,5] Given its progressive course and potential for irreversible neurological injury, improving public awareness and promoting early diagnosis are essential to reducing disease burden and improving long-term outcomes.^[6,7] High-quality and easily accessible health information, therefore, plays a vital role in prevention, early detection, and rehabilitation management.

Rapid advances in information technology have made short video platforms an influential medium for health communication.^[8] TikTok and Bilibili, in particular, have become prominent because their algorithm-driven recommendations, visual features, and interactive formats facilitate widespread dissemination of health information.^[9] These platforms can draw large audiences and satisfy the increasing demand for digital health education.^[10] However, their open and user-generated nature also raises concerns regarding the reliability, completeness, and credibility of the content they provide.^[11,12] Studies examining videos on diabetes,^[13] cataracts,^[14] and knee osteoarthritis^[15] have shown that although some content attracts considerable engagement, its educational quality and reliability are often low.^[13] Within the broader field of digital musculoskeletal health information, Venda Nova et al proposed a multidimensional framework for evaluating online information across readability, reliability, content, and quality, underscoring the importance of assessing informational attributes separately from engagement or popularity metrics.^[16] Considering the substantial impact of CSM on quality of life and occupational functioning, as well as its potential to cause permanent disability and psychological distress, it is important to assess the quality of CSM-related videos on short-video platforms.^[17]

This study systematically examined CSM-related videos on TikTok and Bilibili by assessing their informational content, educational quality, and overall reliability, and by analyzing variations across platforms and uploader groups. The aim was to characterize the present landscape of CSM information on short-video media, identify major deficiencies, and generate preliminary evidence to support more effective spinal health communication and strengthen the trustworthiness of online medical education.

2. Methods

2.1. Ethics approval

This study conducted a content analysis of publicly available videos on TikTok and Bilibili and did not recruit study participants. During retrieval and screening, publicly available account information was viewed only for video deduplication and classification. Usernames, account identifiers, profile images, facial images, and other directly identifiable personal information were not retained in the final analytical dataset or reported in the manuscript, and all results are presented in aggregate form. According to institutional policies for research using publicly available data, ethics committee approval and informed consent were not required.

2.2. Data sources, retrieval, and collection

A systematic search of TikTok and Bilibili was performed on October 1, 2025, in Yueyang, Hunan Province, China, using a personal computer with the interface and language settings configured in Chinese. The Chinese-language equivalent of “cervical spondylotic myelopathy” was entered as the sole search term. To minimize the influence of personalized algorithms, browsing history was deleted, and a new account was created specifically for data retrieval. Search results were reviewed in the platforms’ default order, and the first 150 videos returned by each platform were screened sequentially, yielding 300 videos for initial screening. The threshold was prespecified with reference to sample sizes used in comparable short-video studies to broaden content coverage while maintaining the same feasible screening procedure across both platforms; it was not intended as an exhaustive assessment of all available videos. For each eligible video, we recorded duration, number of likes, collections, comments, and shares, the uploader category, and the thematic content.

2.3. Inclusion and exclusion criteria

Videos were eligible for inclusion if they: were directly related to CSM, and were presented in Chinese. Exclusion criteria were: content unrelated to CSM; advertisements or commercially oriented promotional materials whose primary purpose was to promote drugs, medical devices, health products, medical services, or healthcare institutions and that contained explicit marketing information such as purchase, appointment, price, discount, or contact details; and duplicate or re-uploaded content, defined as videos with identical or highly similar visual, audio, or core content, including re-uploaded videos. After screening, 274 videos remained for analysis, including 145 from TikTok and 129 from Bilibili. Figure 1 summarizes the selection process.

Figure 1.

Flowchart of the video selection process. Of the initial 300 videos identified on TikTok and Bilibili, 274 videos (145 from TikTok and 129 from Bilibili) remained after excluding irrelevant, duplicate, and advertising content.

2.4. Uploader and content classification

Uploader identity was classified according to platform verification information, occupational descriptions on profile pages, affiliated healthcare institutions, and specialty information: specialized healthcare professionals (SHPs), defined as uploaders with a verifiable healthcare professional identity who worked in specialties directly related to CSM diagnosis and treatment, such as orthopedics, spine surgery, neurosurgery, or rehabilitation medicine; nonspecialized healthcare professionals (NSHPs), defined as uploaders with a verifiable healthcare professional identity whose specialty was not directly related to CSM diagnosis and treatment; and individual users (IUs), defined as uploaders who did not display or whose healthcare professional identity could not be verified from publicly available information. Uploaders with multiple occupational or specialty identities were classified according to the verifiable specialty most closely related to CSM diagnosis and treatment.

Video content was classified on the basis of spoken explanations, subtitles, and on-screen text. Content categories included etiology, clinical manifestations, diagnosis, treatment, and prognosis. These categories were not mutually exclusive, and a single video could be assigned to multiple categories. Detailed coding criteria are provided in Table S1, Supplemental Digital Content 1.

2.5. Quality and reliability assessment

Video quality and reliability were evaluated using the Global Quality Score (GQS)^[18] and the modified DISCERN (mDISCERN) tool.^[19] The GQS assesses the overall quality, organization, and usefulness of health-related content for patients, with higher scores indicating better overall quality. mDISCERN, adapted from the original DISCERN instrument, focuses on the reliability and trustworthiness of online health information. The scoring frameworks for both tools are presented in Tables 1 and 2. Two trained reviewers independently rated all videos, and disagreements were resolved through consultation with a third reviewer. Quadratic-weighted Cohen’s kappa coefficients and their 95% confidence intervals (CIs) were calculated to assess agreement between the 2 reviewers for GQS and mDISCERN scores before consensus was reached.

Table 1.

Criteria for assessing quality using the Global Quality Scale (GQS).

Item features	Points
Poor quality; poor organization; most key information missing; not useful for patients	1
Overall poor quality; includes limited information with many important topics missing; minimal value to patients	2
Moderate quality; suboptimal flow; some important topics are adequately discussed, while others are insufficiently addressed; somewhat useful for patients	3
Good quality with generally coherent flow; most of the relevant information presented, but some topics not covered; useful for patients	4
Excellent quality and flow; very useful for patients	5

Table 2.

Criteria used to assess information reliability using the Modified DISCERN (mDISCERN).

Reliability Score
1. Is the video clear, concise, and easy to understand?
2. Are the sources of information reliable?
3. Is the information balanced and unbiased?
4. Are additional sources of information listed?
5. Are areas of uncertainty mentioned?

2.6. Statistical analysis

Continuous variables were summarized as medians and interquartile ranges (IQRs), and categorical variables as frequencies and percentages. Between-group differences were assessed using the Kruskal–Wallis test or Mann–Whitney U test, as appropriate. Following a significant Kruskal–Wallis test, pairwise comparisons were performed using Dunn’s test with Holm adjustment for multiple comparisons. Tied observations were assigned average ranks, and appropriate tie corrections were applied in all rank-based analyses. Effect sizes were reported as rank-biserial correlation coefficients for Mann–Whitney U tests and epsilon-squared (ε²) values for Kruskal–Wallis tests. Associations between variables were assessed using Spearman’s rank correlation. To further evaluate whether video duration was independently associated with the scores, separate multivariable ordinal logistic regression models were constructed with GQS and mDISCERN scores as the dependent variables, adjusting for platform and uploader category. Video duration was entered as a continuous variable and expressed per 60-second increase. The proportional-odds assumption was assessed before model interpretation. A proportional-odds ordinal logistic regression model was used for GQS, whereas a partial proportional-odds model was used for mDISCERN because the proportional-odds assumption was violated for the NSHP effect, which was therefore allowed to vary across score thresholds. The partial proportional-odds model was fitted using the clm() function in the ordinal package in R. Adjusted odds ratios (aORs), 95% CIs, and P values were reported. Data completeness was assessed before analysis. Analyses involving variables with missing values were conducted using available cases, and no missing values were imputed. A two-tailed P < .05 was considered statistically significant. All statistical analyses and visualizations were performed using R, version 4.3.2.

3. Results

3.1. General characteristics of the videos

A total of 274 videos met the inclusion criteria. TikTok and Bilibili accounted for 52.92% and 47.08% of videos, respectively. Overall, the median video duration was 96.50 seconds (IQR: 56.25–184.00). Among all engagement metrics, likes were most common, with a median count of 157.50 (IQR: 18.25–527.75). Additional descriptive metrics are presented in Table 3. On TikTok, videos had a median duration of 67.00 seconds (IQR: 45.00–105.00), which was significantly shorter than that on Bilibili (P < .001). TikTok videos also demonstrated higher audience interaction, including a greater number of likes, collections, comments, and shares compared to Bilibili (P < .001). A comparison of platform-level parameters is shown in Table 4. For videos uploaded by SHPs, the median duration was 79.00 seconds (IQR: 51.75–118.50), and the median number of likes was 182.00 (IQR: 20.75–475.50). Characteristics stratified by uploader category are summarized in Table 5.

Table 3.

Summary of video characteristics, quality assessments, and reliability scores.

Variables	Total (n = 274)
General information
Video length (s)	96.50 (56.25–184.00)
Likes	157.50 (18.25–527.75)
Collections	60.00 (11.00–276.25)
Comments	16.00 (1.00–91.00)
Shares	28.50 (4.00–120.75)
Video content
Etiology	13 (4.74%)
Clinical manifestation	180 (65.69%)
Diagnosis	209 (76.28%)
Treatment	205 (74.82%)
Prognosis	44 (16.06%)
Video quality
GQS score	3.00 (2.00–3.00)
mDISCERN score	4.00 (3.00–4.00)

Data presented as median (interquartile range) or frequency (percentage).

Table 4.

Summary of general information, and GQS and mDISCERN assessments of CSM videos on TikTok and Bilibili.

Variables	Bilibili (n = 129)	TikTok (n = 145)	rrb (95% CI)	P
General information
Video length(s)	181.00 (96.00–610.00)	67.00 (45.00–105.00)	0.667 (0.566, 0.758)	<.001
Likes	16.00 (3.00–128.00)	320.00 (138.00–839.00)	-0.596 (-0.711, −0.476)	<.001
Collections	19.00 (3.00–200.00)	100.00 (35.00–297.00)	-0.329 (-0.466, −0.190)	<.001
Comments	0.00 (0.00–13.00)	47.00 (16.00–154.00)	-0.645 (-0.747, −0.532)	<.001
Shares	6.00 (0.00–57.00)	52.00 (19.00–165.00)	-0.409 (-0.537, −0.275)	<.001
Video content
Etiology	5 (3.88%)	8 (5.52%)		-
Clinical manifestations	77 (59.69%)	103 (71.03%)		-
Diagnosis	72 (55.81%)	137 (94.48%)		-
Treatment	88 (68.22%)	117 (80.69%)		-
Prognosis	26 (20.16%)	18 (12.41%)		-
Video quality
GQS score	3.00 (2.00–3.00)	3.00 (3.00–3.00)	-0.367 (-0.475, −0.256)	<.001
mDISCERN score	3.00 (2.00–4.00)	4.00 (4.00–5.00)	-0.618 (-0.708, −0.521)	<.001

Data presented as median (interquartile range) or frequency (percentage).

Table 5.

Summary of video characteristics, and GQS and mDISCERN assessments across uploader groups on TikTok and Bilibili.

Variables	NSHPs (n = 15)	SHPs (n = 196)	IUs (n = 63)	ε2 (95% CI)	P
Video length	158.00 (88.50–531.50)	79.00 (51.75–118.50)	511.00 (149.50–681.50)	0.291 (0.212, 0.385)	<.001
Likes	521.00 (82.00–1098.00)	182.00 (20.75–475.50)	91.00 (2.00–562.00)	0.010 (0.000, 0.061)	.097
Collections	235.00 (46.50–811.50)	53.00 (13.00–186.50)	67.00 (3.00–678.50)	0.006 (0.000, 0.055)	.154
Comments	39.00 (8.00–118.00)	19.50 (2.00–84.25)	4.00 (0.00–123.75)	0.006 (0.000, 0.053)	.162
Shares	92.00 (16.50–365.00)	28.00 (4.00–106.25)	22.00 (0.00–160.50)	0.003 (0.000, 0.048)	.254
GQS score	3.00 (2.00–3.00)	3.00 (3.00–3.00)	2.00 (1.00–2.00)	0.517 (0.441, 0.595)	<.001
mDISCERN score	3.00 (3.00–4.00)	4.00 (4.00–4.00)	2.00 (1.00–3.00)	0.389 (0.310, 0.472)	<.001

Data presented as median (interquartile range).

IUs = individual users, NSHPs = nonspecialized healthcare professionals, SHPs = specialized healthcare professionals.

3.2. Uploader characteristics

On Bilibili, SHPs, NSHPs, and IUs accounted for 53%, 7%, and 40% of the videos, respectively. On TikTok, the corresponding proportions were 88%, 4%, and 8%, respectively. The distribution of uploader categories across both platforms is presented in Figure 2.

Figure 2.

Uploader distribution for TikTok and Bilibili. Categories included SHPs, NSHPs, and IUs. TikTok featured a higher representation of SHP-uploaded videos. IUs = individual users, NSHPs = nonspecialized healthcare professionals, SHPs = specialized healthcare professionals.

3.3. Video content

Diagnostic information (76.28%) and treatment-related content (74.82%) were the most frequently addressed topics. Prognostic information appeared least often, with only 16.06% of videos covering this aspect. Additional content categories are summarized in Table 4. Figure 3 illustrates the content distribution between the two platforms.

Figure 3.

Content distribution of cervical spondylotic myelopathy-related videos. Comparison of major topics on TikTok and Bilibili, including clinical manifestations, diagnosis, etiology, prognosis, and treatment.

3.4. Video quality and reliability

Agreement between the 2 reviewers for the GQS and mDISCERN scores was high, with quadratic-weighted Cohen’s kappa values of 0.807 (95% CI: 0.739–0.869) and 0.852 (95% CI: 0.794–0.899), respectively. The overall median GQS score for all videos was 3.00 (IQR: 2.00–3.00), and the corresponding median mDISCERN score was 4.00 (IQR: 3.00–4.00). TikTok demonstrated significantly higher GQS and mDISCERN scores than Bilibili (P < .001). On TikTok, the median GQS score was 3.00 (IQR: 3.00–3.00), and the median mDISCERN score was 4.00 (IQR: 4.00–5.00) (Table 4). Figure 4A and B present the distribution of quality and reliability scores by uploader type. The overall differences in GQS and mDISCERN scores among uploader categories were statistically significant (both Kruskal–Wallis P < .001; Table 5). In the Dunn–Holm post hoc comparisons, SHP-uploaded videos had higher GQS scores than videos uploaded by NSHPs (Holm-adjusted P = .003) and IUs (Holm-adjusted P < .001), while NSHP-uploaded videos also had higher GQS scores than IU-uploaded videos (Holm-adjusted P = .003; Fig. 5A). For mDISCERN, SHP- and NSHP-uploaded videos did not differ significantly (Holm-adjusted P = .052), whereas both groups had higher scores than IUs (SHPs vs IUs, Holm-adjusted P < .001; NSHPs vs IUs, Holm-adjusted P = .001; Fig. 5B). Detailed scoring characteristics by uploader category are presented in Table 5. Detailed results of the post hoc pairwise comparisons are provided in Table S2, Supplemental Digital Content 2.

Figure 4.

GQS and mDISCERN score distribution by uploader type. (A) GQS distribution. (B) mDISCERN distribution. SHPs uploaded videos with higher scores than those uploaded by NSHPs and IUs. IUs = individual users, NSHPs = nonspecialized healthcare professionals, SHPs = specialized healthcare professionals.

Figure 5.

Comparison of GQS and mDISCERN scores across uploader categories. (A) GQS scores; (B) mDISCERN scores. Pairwise comparisons were performed using Dunn’s test with Holm adjustment for multiple comparisons. ns, not significant; *P < .05, **P < .01, ***P < .001, ****P < .0001.

3.5. Correlation analysis between duration, engagement, and quality

Unadjusted Spearman rank correlation analysis showed that video duration was negatively correlated with GQS scores (ρ = −0.35, P < .001) and mDISCERN scores (ρ = −0.44, P < .001). However, after adjustment for platform and uploader category, video duration was not independently associated with GQS scores (per 60-s increase: adjusted odds ratio [aOR] = 1.057, 95% CI: 0.998–1.118, P = .056) or mDISCERN scores (aOR = 1.032, 95% CI: 0.983–1.083, P = .198; Table 6). Engagement indicators were strongly interrelated. Likes demonstrated strong positive correlations with comments (ρ = 0.90, P < .05), shares (ρ = 0.93, P < .05), and collections (ρ = 0.92, P < .05). Comments were positively correlated with shares (ρ = 0.84, P < .05) and collections (ρ = 0.81, P < .05), and shares were strongly correlated with collections (ρ = 0.96, P < .05). No significant correlations were observed between engagement metrics and either GQS or mDISCERN scores (P > .05) (Fig. 6).

Table 6.

Adjusted associations between video duration and GQS and mDISCERN scores.

Outcome	aOR per 60-second increase	95% CI	P value
GQS score	1.057	0.998–1.118	.056
mDISCERN score	1.032	0.983–1.083	.198

Note: Both models were adjusted for platform and uploader category. Video duration was expressed per 60-second increase. aOR, adjusted odds ratio; CI, confidence interval.

Figure 6.

Spearman correlation heatmap depicting correlations among video characteristics, engagement indicators, and evaluation scores. The figure displays correlations between video duration, engagement metrics (likes, comments, shares, and collections), and evaluation scores (GQS and mDISCERN). A mild negative association was identified between duration and both scoring indices.

4. Discussion

This study evaluated the content, quality, and reliability of CSM-related videos on TikTok and Bilibili. Most videos concentrated on diagnostic and therapeutic information, while prognostic content remained notably underrepresented. Overall, video quality and reliability were modest, although TikTok videos performed better in both domains. Content produced by SHPs demonstrated the highest quality and reliability. Engagement indicators, including likes, comments, and shares, showed no association with video quality or reliability, suggesting that popularity does not necessarily reflect educational value. These findings suggest that strengthening platform-level content quality management and providing greater guidance to content creators may be feasible approaches to improving the quality and reliability of online spinal health information.

The engagement performance observed in this study aligns with findings from other health-related short-video analyses. CSM-related videos received substantial audience interaction, particularly on TikTok, where engagement metrics were consistently higher than those on Bilibili. Similar patterns have been documented across various medical topics. For example, a Chinese study of pulmonary nodule-related videos similarly found that TikTok videos received more likes, comments, and shares.^[20] Zheng S et al likewise found that liver cancer educational videos generated higher interaction levels on TikTok and Bilibili, reflecting the ability of these platforms to stimulate user participation and facilitate wide content dissemination.^[11] Such strong engagement may be related to algorithm-based recommendations, visually appealing formats, and large active user bases that accelerate information spread.^[21] Future efforts should capitalize on these advantages to promote the dissemination of high-quality medical content, thereby enhancing public awareness and facilitating early intervention for CSM.

From a content perspective, this study showed that CSM-related videos most often address diagnostic features (76.28%) and treatment options (74.82%), whereas prognostic information was relatively limited, appearing in only 16.06% of videos. Delayed or inadequate diagnosis and treatment of CSM can result in progressive neurological impairment, including spinal cord compression, limb weakness, and gait instability, all of which markedly affect daily functioning and work capacity.^[22] In more advanced or improperly managed cases, patients may experience lasting adverse outcomes, such as irreversible muscle atrophy, loss of independent mobility, chronic pain, and significant psychological distress.^[7,23–26] Although existing short videos frequently highlight clinical presentation and therapeutic approaches, future content should devote greater attention to prognosis, particularly long-term recovery, rehabilitation strategies, and posttreatment care. Emphasizing these aspects may help patients avoid severe consequences and improve their quality of life.

Overall, the quality and trustworthiness of CSM-related videos on both platforms were moderate, and some videos remained deficient in information sourcing, balance, and reliability. Although most videos were found to contain some useful health information, a considerable proportion lacked thorough explanations, adequate referencing, or balanced presentation of information. Similar observations have been reported in previous evaluations of medical videos on social media platforms. For example, a study of brain tumor-related videos found that their educational value was moderate or suboptimal.^[27] Collectively, these findings indicate that despite the rapid growth of health communication through short videos, the scientific rigor and completeness of online medical content remain insufficient.

When comparing the two platforms, the median GQS and mDISCERN scores for TikTok videos were significantly higher than those for Bilibili. This difference may be related to differences between TikTok and Bilibili in content governance, medical creator verification, and content recommendation mechanisms.^[11,27] By contrast, Bilibili hosted a broader range of creators, including a higher proportion of IUs. Although its videos tended to be longer, this diversity may also have contributed to greater variability in content quality and reliability. These findings highlight the need to explore platform-specific improvements in content governance. TikTok could further enhance educational value by encouraging more comprehensive explanations and clearer citation of evidence, whereas Bilibili could strengthen verification of medical content produced by IU creators to improve consistency in content quality.

Videos uploaded by SHPs demonstrated higher quality and reliability than those produced by NSHPs or IUs. Similar patterns have been reported in studies of laryngeal carcinoma-related videos, in which SHP-generated content achieved the highest educational value,^[28] and in analyses of breast cancer videos, which likewise showed superior performance among professional medical uploaders.^[12] This trend may be related to the formal medical training and subject-matter expertise of SHPs, which may enable them to provide more complete, detailed, and reliable information.^[21] Encouraging greater participation of SHPs in creating CSM-related content may, therefore, help expand the availability of high-quality videos, enhance public awareness of the condition, and facilitate early intervention for the disease.

Unadjusted analyses showed a weak negative correlation between video duration and GQS and a moderate negative correlation between video duration and mDISCERN, consistent with findings from previous studies of similar topics.^[29,30] However, after controlling for platform and uploader category, these associations were no longer significant, suggesting that video duration itself may not be an independent indicator of quality or reliability. Therefore, evaluations of medical educational videos should place greater emphasis on information completeness, clarity of presentation, and content reliability rather than relying solely on video duration. In addition,engagement metrics were not associated with either quality or reliability. A study of gonorrhea-related videos similarly found no association between user engagement and educational value.^[31] Another study of videos on thyroid eye disease treatment further supported our findings.^[17] Collectively, these results suggest that engagement metrics are poor proxies for information quality and may inadvertently increase the visibility of misleading, exaggerated, or inaccurate content. More effective strategies for health information dissemination should therefore be explored. Previous research has proposed that platforms could improve the visibility of concise, reliable, and evidence-based medical content by optimizing recommendation mechanisms.^[32]Although this strategy may have practical value, whether it improves the educational utility of online health information or public health outcomes requires further evaluation.

Several limitations should be considered. First, this study analyzed only Chinese-language videos on TikTok and Bilibili and did not include other platforms such as YouTube. In addition, only the Chinese-language equivalent of “cervical spondylotic myelopathy” was used as the search term; therefore, videos posted using other related terms may have been missed, limiting the applicability of the findings to other platforms and language settings. Second, although browsing history was deleted and new accounts were created to minimize personalized recommendations, the search results may still have been influenced by platform ranking algorithms and their dynamic changes, resulting in selection bias. Third, only the first 150 search results from each platform were screened; therefore, the sampling scope was limited and may not fully represent all relevant videos available on the platforms. Fourth, although standardized and validated instruments were used, some degree of subjective judgment was unavoidable. Pre-consensus inter-rater agreement was quantified only for the GQS and mDISCERN scores; agreement for uploader and content classifications could not be assessed because the independent classification records were not retained. Finally, because this was a cross-sectional analysis conducted at a single time point, it reflects only the video content and engagement status under the specific search time and conditions and cannot evaluate changes over time. Therefore, the findings primarily apply to Chinese-language CSM-related videos on TikTok and Bilibili under the present search conditions and should not be directly generalized to other platforms, language settings, search strategies, or time points. Future studies should include additional short-video platforms, use multiple CSM-related search terms, conduct repeated searches at different time points, and incorporate disease-specific accuracy assessments and viewer surveys to further evaluate the stability and real-world impact of the findings.

5. Conclusion

This study assessed CSM-related videos on TikTok and Bilibili in terms of content composition, quality, and reliability. Prognostic information was notably underrepresented, and the overall educational value of the videos was modest. TikTok hosted videos with higher quality and reliability than those on Bilibili, and content produced by SHPs consistently ranked higher across assessment tools. Engagement metrics showed no correlation with video quality or reliability. Prospective studies are needed to determine whether strengthening the participation of healthcare professionals, enhancing platform-level regulation, and improving content review mechanisms can promote the dissemination of high-quality CSM-related videos.

Acknowledgments

The authors thank all colleagues who contributed to this study. We thank Medjaden Inc. for scientific editing of this manuscript.

Author contributions

Conceptualization: Dawen Ying, Qifang Zeng.

Data curation: Dawen Ying, Minzheng Ying, Min Yi.

Formal analysis: Dawen Ying.

Investigation: Dawen Ying, Minzheng Ying, Min Yi.

Methodology: Dawen Ying, Qifang Zeng.

Project administration: Dawen Ying, Qifang Zeng.

Writing – original draft: Dawen Ying.

Writing – review & editing: Dawen Ying, Qifang Zeng.

Supervision: Jiongbiao Zhong, Qifang Zeng.