Quality and dissemination of thyroid cancer information on short-video platforms a cross-platform evaluation Authors: Xiaoran Zheng, Qiankuan Li, Lu Jin, Kaihang Shi Published: 2025 ABSTRACT Short-video platforms are becoming major sources of health information, but the quality and dissemination patterns of thyroid cancer–related content remain unclear. We conducted a large-scale cross-sectional analysis of 1,248 thyroid carcinoma videos retrieved from TikTok, Kwai, and Rednote between February 5–10, 2025. Video characteristics, including source, duration, and engagement metrics, were extracted, and quality was evaluated using JAMA, GQS, mDISCERN and PEMAT. TikTok videos had longer duration, higher engagement, and overall better quality compared with those on Kwai and Rednote. Videos from healthcare professionals and institutions achieved significantly higher information quality, whereas independent creators generated greater audience interaction, particularly comments and shares. Knowledge-focused content showed the higher reliable, while patient vlogs attracted more engagement. Engagement indicators were strongly intercorrelated but exhibited weak or negative associations with quality, suggesting that popularity does not fully reflect informational rigor. These findings highlight a divergence between video quality and audience reach, underscoring the need for professional participation, credential verification, and algorithmic refinement to enhance the visibility of evidence-based content and support public health literacy.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-27833-w. Introduction Thyroid cancer is among the fastest-growing malignancies worldwide in terms of incidence. According to the 2020 Global Cancer Statistics (GLOBOCAN), approximately 586,000 new cases were reported, representing 3% of all cancers. Thyroid cancer ranked fifth in incidence among women and fifteenth among men1. In China, thyroid cancer has become one of the most common malignant tumors among women2. Although the prognosis for most thyroid cancers is favorable, aggressive subtypes such as anaplastic and medullary carcinomas are prone to metastasis and confer poorer survival outcomes. These subtypes significantly impair patients’ quality of life3. Treatment-related complications, such as recurrent laryngeal nerve damage and hypoparathyroidism, further increase the disease burden4. Given these clinical challenges and the rising incidence, public demand for accessible and accurate health information on thyroid cancer has increased substantially. In China, the short-video market has expanded rapidly, with platforms such as TikTok and Kwai becoming dominant sources of health information5. These platforms attract massive user bases through free and easily accessible content, thereby serving as powerful tools for public health communication6. However, previous studies have reported significant variability in the quality of medical information across platforms, particularly regarding source credibility, comprehensibility, and reliability7,8. Most existing studies on thyroid cancer information have focused on traditional websites or single platforms, with limited cross-platform evaluations9. These platforms differ substantially in user demographics, recommendation algorithms, and content creators, all of which may influence both information quality and dissemination. Therefore, analyses limited to a single platform may fail to capture the full scope of public exposure, underscoring the need for systematic cross-platform comparisons. To address this gap, we conducted a cross-sectional study of thyroid cancer–related short videos on TikTok, Kwai, and Rednote. Using established evaluation tools, we assessed and compared content quality and dissemination characteristics, and identified influencing factors. The results were intended to guide digital health communication strategies, support professionals in delivering reliable online education, and facilitate public access to trustworthy information. Methods Ethical considerations Ethical approval was not required because this study analyzed only publicly accessible content from short-video platforms, without involving human participants or experimentation. Additionally, no identifiable user information or personal data were collected. Data retrieval methods and collection process We conducted searches from February 5 to 10, 2025, on TikTok, Kwai, and Rednote. Videos were ranked according to each platform’s default algorithm, which may incorporate relevance, popularity, and proprietary factors. TikTok was the only platform that publicly documented its recommendation system, which is based on artificial intelligence, machine learning, and deep learning. Fundamentally, the system operates through mathematical modeling, establishing statistical correlations between user interactions and content attributes, rather than interpreting semantic content. The model calculates a composite recommendation score based on weighted user behavior and video parameters, subsequently distributing the highest-ranked content to users10. In order to reduce algorithmic bias, we created a new account and searched keywords including “thyroid carcinoma” or “thyroid neoplasm” across these platforms. Precautions included clearing browsing histories each time, maintaining fixed IP addresses, and avoiding mobile number/personal information linkage. This protocol simulated non-personalized, standardized initial recommendation states to enhance impartiality and comparability of collected results. Previous research has shown that most audiences primarily focus on the initial search results11,12. Consequently, accounting for platform algorithms and user viewing patterns, we collected and recorded the top 250 videos for each search term across multiple platforms. Thus, the study size is 1500 in total. To ensure methodological rigor and clinical relevance, we selected content specifically about thyroid carcinoma. We limited our analysis to videos in Mandarin and English to match our target audience’s linguistic characteristics while also considering the international nature of peer review. We systematically excluded all commercial advertisements and promotional content to avoid bias and maintain focus on health education materials. Furthermore, to ensure content integrity and enable accurate thematic analysis, we applied strict exclusion criteria to remove videos containing irrelevant information, especially those discussing unrelated neck conditions. These standardized selection criteria were consistently followed throughout our study. The inclusion criteria consisted of1: content related to thyroid cancer and2videos in both Mandarin or English. The exclusion criteria comprised1: duplicate videos2, promotional materials, and3 videos with irrelevant information. A total of 1,248 eligible videos were included, and the following attributes were analyzed: title, number of likes, comments, saves, shares, time since upload, video duration, source, subscriber count, presentation format, and subject matter. Assessment of video material and its quality The Journal of the American Medical Association (JAMA) benchmarks are widely recognized as a robust framework for assessing the reliability of online health information12–16. Scores range from 0 to 4, with 4 indicating high accuracy and trustworthiness, and 0 representing low accuracy and reliability. The Global Quality Scale (GQS) uses a 5-point Likert scale to subjectively evaluate overall video quality, such as user-friendliness and functionality17–19. The modified DISCERN (mDISCERN) tool is a well-established instrument for assessing the reliability and quality of online health materials20,21, consisting of 5 items each worth one point. These metrics - JAMA, GQS, and mDISCERN are commonly employed in similar studies22–24. The Patient Education Materials Assessment Tool (PEMAT), developed by the Agency for Healthcare Research and Quality, evaluates health educational materials by assessing both understandability (PEMAT-U) and actionability (PEMAT-A)25. Detailed scoring criteria for all assessment tools are provided in Supplementary Tables 1–4. Two researchers (Z-XR and L-QK) independently rated all videos, with a third author (J-L) participating to resolve any discrepancies. Prior to formal scoring, Z-XR and L-QK first reviewed the application of assessment tools and jointly analyzed 20 videos and resolved discrepancies with other group members. Each video underwent three evaluations—twice by Z-XR and once by L-QK. This approach allowed us to calculate intraclass correlation coefficient (ICC)1 to measure both inter-rater and intra-rater reliability. We reported 95% confidence intervals (CI) for reliability. Statistical significance was set at p < 0.05, with reliability levels were defined as insufficient (ICC < 0.50), fair (ICC = 0.50–0.75), commendable (ICC = 0.75–0.90), or superior (ICC > 0.90)25. The intraclass correlation coefficient (ICC) was calculated as follows: 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\rm ICC} =\:\frac{M{S}_{R}-M{S}_{E}}{M{S}_{R}+(k-1)M{S}_{E}}$$\end{document} The two researchers demonstrated strong consistency in their ratings for JAMA scores (ICC = 0.931, 95%CI 0.921–0.940), GQS scores (ICC = 0.946, 95%CI 0.938–0.953), mDISCERN scores (ICC = 0.881, 95%CI 0.863–0.896), PEMAT-U (ICC = 0.886, 95%CI 0.869–0.901), and PEMAT-A (ICC = 0.924, 95%CI 0.912–0.934). The intra-rater reliability was excellent, with ICC values ranging from 0.976 (JAMA) to 0.998 (PEMAT-U). Statistical modeling Analyses were performed using IBM SPSS Statistics, version 22.0 (IBM Corp.), GraphPad Prism, version 10.1.2 (Dotmatics), and OriginPro, version 2021 (OriginLab). Categorical variables were summarized as frequencies and percentages, whereas normally distributed continuous variables were expressed as means ± standard deviations (SD). Before conducting group comparisons, the distribution of continuous variables was assessed for normality using the Shapiro-Wilk test. For variables with homogeneity of variance, independent-sample t tests or one-way analysis of variance (ANOVA) were applied; for those not meeting normality assumptions, Welch’s correction or non-parametric tests were employed. Post hoc multiple comparisons following ANOVA were adjusted using the Bonferroni method to control the type I error rate. Associations between variables were assessed using Spearman’s rank correlation analysis. A P value < 0.05 was considered statistically significant. Results General characteristics of the videos After exclusions, 1,248 thyroid carcinoma videos were included in the analysis (Fig. 1). The distributions of video types, sources, and upload dates were broadly similar across the three platforms (Table 1). TikTok videos were, on average, longer, more recently uploaded, and exhibited significantly higher engagement than videos on the other platforms (P < 0.001; Table 1). Fig. 1Flowchart of the study. Flowchart of the study. Table 1The general characteristics and scores of the thyroid carcinoma–related videos.ParametersTotal(N = 1248)TikTok(n = 409)Kwai(n = 435)Rednote(n = 404)P valueDuration (seconds), mean (SD)80.32(99.57)95.09(87.93)61.35(73.92)85.80(127.77)< 0.001Likes, mean (SD)2793.26(22859.70)6931.79(39034.69)1252.44(6459.03)262.57(1751.24)< 0.001Comments, mean (SD)254.60(1165.17)594.44(1948.08)139.90(393.80)34.04(107.83)< 0.001Saves, mean (SD)665.99(3025.82)1578.79(4776.65)306.51(1799.66)128.96(684.47)< 0.001Shares, mean (SD)1250.59(10749.05)2871.03(16455.10)837.10(8569.57)55.32(151.33)< 0.001Days since upload, mean (SD)283.90(314.01)203.62(243.69)432.60(392.99)205.08(205.00)< 0.001Fans of video uploaders, mean (SD)598554.16(5302049.21)1449683.24(9073415.83)336958.58(1520322.97)18559.82(18559.82)< 0.001JAMAa score, mean (SD)1.75(0.56)1.93(0.57)1.67(0.49)1.64(0.58)< 0.001GQSb score, mean (SD)2.98(0.48)3.05(0.66)2.92(0.45)2.97(0.19)< 0.001mDISCERNc score, mean (SD)2.72(0.73)2.77(0.78)2.81(0.65)2.57(0.72)< 0.001PEMAT-Ud, mean (SD)66.66%(13.88%)68.46%(14.78%)64.53%(17.01%)67.15%(7.39%)< 0.001PEMAT-Ae, mean (SD)58.69%(21.31%)60.68%(24.83%)59.75%(20.59%)55.53%(17.57%)0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. Table 1The general characteristics and scores of the thyroid carcinoma–related videos.ParametersTotal(N = 1248)TikTok(n = 409)Kwai(n = 435)Rednote(n = 404)P valueDuration (seconds), mean (SD)80.32(99.57)95.09(87.93)61.35(73.92)85.80(127.77)< 0.001Likes, mean (SD)2793.26(22859.70)6931.79(39034.69)1252.44(6459.03)262.57(1751.24)< 0.001Comments, mean (SD)254.60(1165.17)594.44(1948.08)139.90(393.80)34.04(107.83)< 0.001Saves, mean (SD)665.99(3025.82)1578.79(4776.65)306.51(1799.66)128.96(684.47)< 0.001Shares, mean (SD)1250.59(10749.05)2871.03(16455.10)837.10(8569.57)55.32(151.33)< 0.001Days since upload, mean (SD)283.90(314.01)203.62(243.69)432.60(392.99)205.08(205.00)< 0.001Fans of video uploaders, mean (SD)598554.16(5302049.21)1449683.24(9073415.83)336958.58(1520322.97)18559.82(18559.82)< 0.001JAMAa score, mean (SD)1.75(0.56)1.93(0.57)1.67(0.49)1.64(0.58)< 0.001GQSb score, mean (SD)2.98(0.48)3.05(0.66)2.92(0.45)2.97(0.19)< 0.001mDISCERNc score, mean (SD)2.72(0.73)2.77(0.78)2.81(0.65)2.57(0.72)< 0.001PEMAT-Ud, mean (SD)66.66%(13.88%)68.46%(14.78%)64.53%(17.01%)67.15%(7.39%)< 0.001PEMAT-Ae, mean (SD)58.69%(21.31%)60.68%(24.83%)59.75%(20.59%)55.53%(17.57%)0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. The general characteristics and scores of the thyroid carcinoma–related videos. 598554.16 (5302049.21) 1449683.24 (9073415.83) 336958.58 (1520322.97) 18559.82 (18559.82) aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. Quality of videos across platforms Overall video quality varied by platform, with significant differences across measured indices (P < 0.001). TikTok outperformed other platforms on most quality indices, whereas Kwai showed higher median mDISCERN scores. Detailed metrics are presented in Table 1. Fig. 2The sources and content of the thyroid carcinoma–related videos. (A) Video source; (B) Different medical specialties; (C) Video content; (D) Different disease knowledge; and (E) Video presentation form. The sources and content of the thyroid carcinoma–related videos. (A) Video source; (B) Different medical specialties; (C) Video content; (D) Different disease knowledge; and (E) Video presentation form. Influence of source and specialty The majority of included videos were produced by physicians (Fig. 2). Videos from professional sources scored higher on quality measures, while non-professional videos elicited greater audience interaction, particularly in comment counts (Fig. 3; Table 2; Supplementary Table 5). Among medical specialties, videos labeled as Traditional Chinese Medicine (TCM) scored higher on mDISCERN and PEMAT-A (Fig. S1). Fig. 3Comparative quality assessment of thyroid carcinoma short-videos with different sources. Bar chart showing quality evaluation of thyroid carcinoma short videos with different sources. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Comparative quality assessment of thyroid carcinoma short-videos with different sources. Bar chart showing quality evaluation of thyroid carcinoma short videos with different sources. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Fig. 4Comparative quality assessment of thyroid carcinoma short-videos with different contents. Bar chart showing quality evaluation of thyroid carcinoma short videos with different contents. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Comparative quality assessment of thyroid carcinoma short-videos with different contents. Bar chart showing quality evaluation of thyroid carcinoma short videos with different contents. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Fig. 5Comparative quality assessment of thyroid carcinoma short-videos with different disease knowledge. Bar chart showing quality evaluation of thyroid carcinoma short videos with different knowledge. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Comparative quality assessment of thyroid carcinoma short-videos with different disease knowledge. Bar chart showing quality evaluation of thyroid carcinoma short videos with different knowledge. (A) JAMA score; (B) GQS score; (C) mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Fig. 6Comparative quality assessment of thyroid carcinoma short-videos with different presentation forms. Bar chart showing quality evaluation of thyroid carcinoma short-videos with different presentation form. (A) the JAMA score; (B) the GQS score; (C) the mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Comparative quality assessment of thyroid carcinoma short-videos with different presentation forms. Bar chart showing quality evaluation of thyroid carcinoma short-videos with different presentation form. (A) the JAMA score; (B) the GQS score; (C) the mDISCERN score; (D) PEMAT-U; and (E) PEMAT-A. *P < 0.05, **P < 0.01, ***P < 0.001, ns: non-significant. Table 2The evaluation from different sources with different contents and different presentation forms.VariablesJAMAaGQSbmDISCERNcPEMAT-UdPEMAT-Ae Video source Physician, mean (SD)1.79(0.46)2.97(0.47)2.93(0.45)67.47%(12.69%)60.78%(19.02%)Hospital, mean (SD)3(0)3.07(0.26)2.93(0.38)72.24%(3.32%)67.86%(11.04%)News agency, mean (SD)2.67(0.62)3.44(0.55)3.03(0.54)79.74%(7.57%)70.94%(13.64%)Independent user, mean (SD)1.02(0.14)2.88(0.51)1.23(0.47)57.27%(18.08%)40.73%(27.90%)Others, mean (SD)1.83(0.41)3(0)2.83(0.41)69.05%(5.83%)61.11%(13.61%)P value< 0.001< 0.001< 0.001< 0.001< 0.001 Different medical specialties Western medicine, mean (SD)1.74(0.57)2.98(0.48)2.70(0.74)66.60%(13.89%)58.20%(21.59%)TCMf, mean (SD)1.80(0.44)2.95(0.51)2.91(0.54)67.27%(13.83%)63.15%(18.01%)P value0.1250.545< 0.0010.6100.005 Video content Disease knowledge, mean (SD)1.87(0.52)3.01(0.49)2.92(0.51)68.30%(12.83%)61.97%(18.76%)Outpatient scenarios, mean (SD)1.67(0.54)2.92(0.39)2.86(0.38)65.70%(12.80%)57.43%(20.02%)Personal experience, mean (SD)1.03(0.18)2.87(0.50)1.21(0.44)57.70%(17.74%)39.74%(27.40%)P value< 0.001< 0.001< 0.001< 0.001< 0.001 Different disease knowledge Treatment, mean (SD)1.69(0.57)2.99(0.35)2.71(0.65)67.89%(10.37%)59.00%(17.80%)Prevention, mean (SD)1.82(0.57)3.08(0.59)2.78(0.65)69.01%(15.37%)64.29%(20.85%)Complication, mean (SD)2.19(0.67)2.97(0.61)2.92(0.73)67.92%(14.21%)56.71%(25.11%)Symptom, mean (SD)1.69(0.52)2.87(0.48)2.73(0.65)63.56%(16.10%)56.02%(24.51%)Definition, mean (SD)2.05(0.36)2.98(0.61)3.09(0.54)66.19%(16.21%)59.69%(20.68%)Post treatment considerations, mean (SD)1.44(0.50)2.97(0.56)2.10(0.98)62.63%(18.19%)54.88%(29.86%)Re-examination, mean (SD)1.71(0.59)3.29(0.59)2.82(0.73)71.09%(13.09%)70.59%(26.04%)Others, mean (SD)1.15(0.37)2.77(0.65)1.69(0.79)59.27%(19.52%)43.59%(26.28%)P value< 0.0010.014< 0.0010.0020.003 Video presentation form Expert monologue, mean (SD)1.87(0.48)3.01(0.46)2.96(0.46)68.76%(11.62%)62.69%(18.25%)Dialogue, mean (SD)1.75(0.53)2.97(0.40)2.84(0.44)67.20%(11.25%)58.92%(19.12%)Visual pictures and literature, mean (SD)1.90(0.74)2.84(0.61)2.83(0.60)62.36%(20.93%)55.12%(21.18%)Surgery demonstration, mean (SD)1.62(0.65)3.15(0.38)2.85(0.55)72.54%(6.52%)66.67%(0%)Animation, mean (SD)1.67(1.21)2.83(0.75)2.50(0.84)57.09%(17.80%)44.44%(17.22%)Vlogs of patients, mean (SD)1.04(0.20)2.89(0.51)1.23(0.47)57.95%(17.89%)40.11%(27.89%)Others, mean (SD)1.57(0.53)2.57(0.98)2.43(0.79)53.40%(27.18%)47.62%(37.80%)P value< 0.0010.031< 0.001< 0.001< 0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. fTCM: Traditional Chinese Medicine. Table 2The evaluation from different sources with different contents and different presentation forms.VariablesJAMAaGQSbmDISCERNcPEMAT-UdPEMAT-Ae Video source Physician, mean (SD)1.79(0.46)2.97(0.47)2.93(0.45)67.47%(12.69%)60.78%(19.02%)Hospital, mean (SD)3(0)3.07(0.26)2.93(0.38)72.24%(3.32%)67.86%(11.04%)News agency, mean (SD)2.67(0.62)3.44(0.55)3.03(0.54)79.74%(7.57%)70.94%(13.64%)Independent user, mean (SD)1.02(0.14)2.88(0.51)1.23(0.47)57.27%(18.08%)40.73%(27.90%)Others, mean (SD)1.83(0.41)3(0)2.83(0.41)69.05%(5.83%)61.11%(13.61%)P value< 0.001< 0.001< 0.001< 0.001< 0.001 Different medical specialties Western medicine, mean (SD)1.74(0.57)2.98(0.48)2.70(0.74)66.60%(13.89%)58.20%(21.59%)TCMf, mean (SD)1.80(0.44)2.95(0.51)2.91(0.54)67.27%(13.83%)63.15%(18.01%)P value0.1250.545< 0.0010.6100.005 Video content Disease knowledge, mean (SD)1.87(0.52)3.01(0.49)2.92(0.51)68.30%(12.83%)61.97%(18.76%)Outpatient scenarios, mean (SD)1.67(0.54)2.92(0.39)2.86(0.38)65.70%(12.80%)57.43%(20.02%)Personal experience, mean (SD)1.03(0.18)2.87(0.50)1.21(0.44)57.70%(17.74%)39.74%(27.40%)P value< 0.001< 0.001< 0.001< 0.001< 0.001 Different disease knowledge Treatment, mean (SD)1.69(0.57)2.99(0.35)2.71(0.65)67.89%(10.37%)59.00%(17.80%)Prevention, mean (SD)1.82(0.57)3.08(0.59)2.78(0.65)69.01%(15.37%)64.29%(20.85%)Complication, mean (SD)2.19(0.67)2.97(0.61)2.92(0.73)67.92%(14.21%)56.71%(25.11%)Symptom, mean (SD)1.69(0.52)2.87(0.48)2.73(0.65)63.56%(16.10%)56.02%(24.51%)Definition, mean (SD)2.05(0.36)2.98(0.61)3.09(0.54)66.19%(16.21%)59.69%(20.68%)Post treatment considerations, mean (SD)1.44(0.50)2.97(0.56)2.10(0.98)62.63%(18.19%)54.88%(29.86%)Re-examination, mean (SD)1.71(0.59)3.29(0.59)2.82(0.73)71.09%(13.09%)70.59%(26.04%)Others, mean (SD)1.15(0.37)2.77(0.65)1.69(0.79)59.27%(19.52%)43.59%(26.28%)P value< 0.0010.014< 0.0010.0020.003 Video presentation form Expert monologue, mean (SD)1.87(0.48)3.01(0.46)2.96(0.46)68.76%(11.62%)62.69%(18.25%)Dialogue, mean (SD)1.75(0.53)2.97(0.40)2.84(0.44)67.20%(11.25%)58.92%(19.12%)Visual pictures and literature, mean (SD)1.90(0.74)2.84(0.61)2.83(0.60)62.36%(20.93%)55.12%(21.18%)Surgery demonstration, mean (SD)1.62(0.65)3.15(0.38)2.85(0.55)72.54%(6.52%)66.67%(0%)Animation, mean (SD)1.67(1.21)2.83(0.75)2.50(0.84)57.09%(17.80%)44.44%(17.22%)Vlogs of patients, mean (SD)1.04(0.20)2.89(0.51)1.23(0.47)57.95%(17.89%)40.11%(27.89%)Others, mean (SD)1.57(0.53)2.57(0.98)2.43(0.79)53.40%(27.18%)47.62%(37.80%)P value< 0.0010.031< 0.001< 0.001< 0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. fTCM: Traditional Chinese Medicine. The evaluation from different sources with different contents and different presentation forms. aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. fTCM: Traditional Chinese Medicine. Content and format differences Knowledge-oriented videos were the most common content type and scored higher in quality than outpatient scenario or personal-experience videos (Fig. 4). Dissemination differed significantly across categories (P < 0.001): disease knowledge videos showed relatively higher engagement, and personal experience content generated more comments (Table 3). Within knowledge videos, videos addressing complications and definition-focused topics achieved the highest JAMA and mDISCERN scores. Re-examination content received the highest GQS ratings (Fig. 5). Prevention-oriented videos performed best on PEMAT-A and, along with treatment content, were associated with higher engagement (Table 3; Supplementary Table 5). Expert monologues were the most common format and exhibited higher quality ratings than dialogues, animations, or patient vlogs (Fig. 6). Although patient vlogs scored lower on measured quality, they elicited more comments and shares, reflecting greater interactive appeal (Table 3; Supplementary Table 6). Table 3The popularity of videos from different sources with different contents and different presentation forms.VariablesLikesCommentsSavesShares Video source Physician, mean (SD)2153.32(13292.65)178.45(885.72)562.57(2505.29)1007.30(8659.93)Hospital, mean (SD)227.07(622.20)42.43(118.71)113.61(339.54)139.29(456.41)News agency, mean (SD)25372.62(108767.30)1371.67(4366.16)3852.64(10092.86)11477.44(40849.29)Independent user, mean (SD)1902.17(3763.64)511.99(816.09)666.32(2187.35)541.59(1502.17)Others, mean (SD)22.67(22.93)1.83(2.79)17.50(21.30)25.83(44.06)P value< 0.001< 0.001< 0.001< 0.001 Different medical specialties Western medicine, mean (SD)2894.40(24024.02)261.20(1214.09)624.80(2785.77)1262.92(11243.96)TCMa, mean (SD)1868.18(4792.92)194.23(540.67)1042.75(4682.45)1137.90(4024.17)P value0.220.2710.3330.8 Video content Disease knowledge, mean (SD)3443.88(26797.16)252.79(1329.17)797.58(3497.32)1615.31(12610.65)Outpatient scenarios, mean (SD)590.65(1816.97)101.77(305.25)172.31(567.52)196.65(599.72)Personal experience, mean (SD)1796.21(3516.98)481.87(694.95)532.42(1333.04)436.30(940.40)P value< 0.001< 0.001< 0.001< 0.001 Different disease knowledge Treatment, mean (SD)1722.04(11047.08)204.66(1030.88)507.23(2581.33)1077.65(10794.02)Prevention, mean (SD)9520.39(31907.39)596.04(2273.04)1990.18(5790.75)3239.88(10644.72)Complication, mean (SD)2405.94(7153.54)258.83(505.00)633.19(2109.30)921.42(3176.47)Symptom, mean (SD)1007.42(2972.18)142.83(435.91)322.46(987.25)422.97(1874.08)Definition, mean (SD)4854.77(1726.98)299.00(1383.00)885.90(3857.77)1971.56(17048.71)Post treatment considerations, mean (SD)1045.30(1726.98)232.07(407.70)340.02(725.50)334.08(781.13)Re-examination, mean (SD)696.29(1091.27)65.24(99.58)480.65(1036.33)290.18(461.23)Others, mean (SD)1346.81(2426.56)433.42(748.77)286.08(518.78)273.73(531.14)P value0.048< 0.0010.0360.044 Video presentation form Expert monologue, mean (SD)3535.40(28551.40)222.62(1203.57)751.84(3348.19)1585.45(13212.02)Dialogue, mean (SD)970.21(2891.65)145.74(426.72)385.32(1652.07)424.40(1980.54)Visual pictures and literature, mean (SD)2835.80(14078.96)344.05(2222.31)538.05(1646.00)1771.93(9509.07)Surgery demonstration, mean (SD)1021.54(1911.81)197.08(408.42)275.92(421.50)770.38(1363.51)Animation, mean (SD)570.67(1055.87)5.50(7.64)142.83(225.01)121.67(129.01)Vlogs of patients, mean (SD)1772.49(3388.06)548.51(857.89)497.14(1197.26)419.28(882.80)Others, mean (SD)3292.29(743172)95.57(90.41)6102.86(15760.64)3092.29(7760.31)P value0.577< 0.0010.0130.002aTCM: Traditional Chinese Medicine. Table 3The popularity of videos from different sources with different contents and different presentation forms.VariablesLikesCommentsSavesShares Video source Physician, mean (SD)2153.32(13292.65)178.45(885.72)562.57(2505.29)1007.30(8659.93)Hospital, mean (SD)227.07(622.20)42.43(118.71)113.61(339.54)139.29(456.41)News agency, mean (SD)25372.62(108767.30)1371.67(4366.16)3852.64(10092.86)11477.44(40849.29)Independent user, mean (SD)1902.17(3763.64)511.99(816.09)666.32(2187.35)541.59(1502.17)Others, mean (SD)22.67(22.93)1.83(2.79)17.50(21.30)25.83(44.06)P value< 0.001< 0.001< 0.001< 0.001 Different medical specialties Western medicine, mean (SD)2894.40(24024.02)261.20(1214.09)624.80(2785.77)1262.92(11243.96)TCMa, mean (SD)1868.18(4792.92)194.23(540.67)1042.75(4682.45)1137.90(4024.17)P value0.220.2710.3330.8 Video content Disease knowledge, mean (SD)3443.88(26797.16)252.79(1329.17)797.58(3497.32)1615.31(12610.65)Outpatient scenarios, mean (SD)590.65(1816.97)101.77(305.25)172.31(567.52)196.65(599.72)Personal experience, mean (SD)1796.21(3516.98)481.87(694.95)532.42(1333.04)436.30(940.40)P value< 0.001< 0.001< 0.001< 0.001 Different disease knowledge Treatment, mean (SD)1722.04(11047.08)204.66(1030.88)507.23(2581.33)1077.65(10794.02)Prevention, mean (SD)9520.39(31907.39)596.04(2273.04)1990.18(5790.75)3239.88(10644.72)Complication, mean (SD)2405.94(7153.54)258.83(505.00)633.19(2109.30)921.42(3176.47)Symptom, mean (SD)1007.42(2972.18)142.83(435.91)322.46(987.25)422.97(1874.08)Definition, mean (SD)4854.77(1726.98)299.00(1383.00)885.90(3857.77)1971.56(17048.71)Post treatment considerations, mean (SD)1045.30(1726.98)232.07(407.70)340.02(725.50)334.08(781.13)Re-examination, mean (SD)696.29(1091.27)65.24(99.58)480.65(1036.33)290.18(461.23)Others, mean (SD)1346.81(2426.56)433.42(748.77)286.08(518.78)273.73(531.14)P value0.048< 0.0010.0360.044 Video presentation form Expert monologue, mean (SD)3535.40(28551.40)222.62(1203.57)751.84(3348.19)1585.45(13212.02)Dialogue, mean (SD)970.21(2891.65)145.74(426.72)385.32(1652.07)424.40(1980.54)Visual pictures and literature, mean (SD)2835.80(14078.96)344.05(2222.31)538.05(1646.00)1771.93(9509.07)Surgery demonstration, mean (SD)1021.54(1911.81)197.08(408.42)275.92(421.50)770.38(1363.51)Animation, mean (SD)570.67(1055.87)5.50(7.64)142.83(225.01)121.67(129.01)Vlogs of patients, mean (SD)1772.49(3388.06)548.51(857.89)497.14(1197.26)419.28(882.80)Others, mean (SD)3292.29(743172)95.57(90.41)6102.86(15760.64)3092.29(7760.31)P value0.577< 0.0010.0130.002aTCM: Traditional Chinese Medicine. The popularity of videos from different sources with different contents and different presentation forms. 2153.32 (13292.65) 178.45 (885.72) 562.57 (2505.29) 1007.30 (8659.93) 25372.62 (108767.30) 1371.67 (4366.16) 3852.64 (10092.86) 11477.44 (40849.29) 1902.17 (3763.64) 511.99 (816.09) 666.32 (2187.35) 541.59 (1502.17) 2894.40 (24024.02) 261.20 (1214.09) 624.80 (2785.77) 1262.92 (11243.96) 1868.18 (4792.92) 194.23 (540.67) 1042.75 (4682.45) 1137.90 (4024.17) 3443.88 (26797.16) 252.79 (1329.17) 797.58 (3497.32) 1615.31 (12610.65) 1796.21 (3516.98) 481.87 (694.95) 532.42 (1333.04) 436.30 (940.40) 3292.29(7431 72) aTCM: Traditional Chinese Medicine. Correlation analysis Engagement metrics were strongly intercorrelated (P < 0.001; Fig. 7). Associations between engagement and video duration, upload time, and uploader follower count were comparatively weak. Uploader follower count showed modest positive associations with several quality measures, whereas comment volume tended to correlate negatively with multiple quality scores—indicating that popularity and rigor are not fully aligned. Video duration was positively correlated with PEMAT-A, whereas shares were positively correlated with PEMAT-U. Likes and saves showed a small inverse relationship with mDISCERN scores, suggesting that more popular items were, on average, slightly less rigorous (Fig. 8; Table 4). Fig. 7Spearman correlation analysis between the video variables. *P < 0.05, **P < 0.01, ***P < 0.001. Spearman correlation analysis between the video variables. *P < 0.05, **P < 0.01, ***P < 0.001. Fig. 8Spearman correlation analysis between video variables and the video quality. *P < 0.05, **P < 0.01, ***P < 0.001. Spearman correlation analysis between video variables and the video quality. *P < 0.05, **P < 0.01, ***P < 0.001. Table 4The correlation analysis between video variables and the video quality.VariablesJAMAaGQSbmDISCERNcPEMAT-UdPEMAT-Ae Likes ρρ0.0270.059−0.0650.045−0.001P value0.3430.0370.0210.1130.983 Comments ρρ−0.0890.044−0.190−0.150−0.075P value0.0020.121< 0.0010.6070.008 Saves ρ0.0240.069−0.0800.032−0.014P value0.4020.0150.0040.2570.628 Shares ρ0.0330.067−0.0390.0650.023P value0.2460.0190.1730.0220.407 Duration ρ−0.1320.249−0.140−0.0320.076P value< 0.001< 0.001< 0.0010.2580.007 Days since uploaded ρ−0.135−0.011−0.035−0.047−0.006P value< 0.0010.6990.2230.0940.842 Fans ρ0.2290.1310.2810.1850.216P value< 0.001< 0.001< 0.001< 0.001< 0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. Table 4The correlation analysis between video variables and the video quality.VariablesJAMAaGQSbmDISCERNcPEMAT-UdPEMAT-Ae Likes ρρ0.0270.059−0.0650.045−0.001P value0.3430.0370.0210.1130.983 Comments ρρ−0.0890.044−0.190−0.150−0.075P value0.0020.121< 0.0010.6070.008 Saves ρ0.0240.069−0.0800.032−0.014P value0.4020.0150.0040.2570.628 Shares ρ0.0330.067−0.0390.0650.023P value0.2460.0190.1730.0220.407 Duration ρ−0.1320.249−0.140−0.0320.076P value< 0.001< 0.001< 0.0010.2580.007 Days since uploaded ρ−0.135−0.011−0.035−0.047−0.006P value< 0.0010.6990.2230.0940.842 Fans ρ0.2290.1310.2810.1850.216P value< 0.001< 0.001< 0.001< 0.001< 0.001aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. The correlation analysis between video variables and the video quality. aJAMA: Journal of American Medical Association. bGQS: Global Quality Scale. cmDISCERN: Modified DISCERN. dPEMAT-U: Patient Education Materials Assessment Tool-Understandability. ePEMAT-A: Patient Education Materials Assessment Tool-Actionability. Discussion Main discoveries Short-video platforms are increasingly important sources of health information and can influence public health behaviors26–28. However, prior studies of diabetes, respiratory disease, and various cancers have documented substantial variability in content reliability and source credibility18,27,29. In this context, we analyzed thyroid carcinoma videos collected from three major Chinese short-video platforms. We found that videos produced by experts and institutions generally scored higher on quality measures, while independent creators elicited higher interaction—demonstrating a divergence between informational quality and audience appeal (see Results; Tables 2 and 3). High reach, measured by views and interactions, did not reliably indicate higher informational quality, underlining the risk that lower-quality content can spread widely30,31. Factors influencing video quality Platform differences in moderation policies, creator composition, and recommendation algorithms likely contribute to the observed quality variation. TikTok demonstrated advantages in reach and measured quality, with over 100 million health videos and 500 billion views in the first half of 202332. TikTok’s verification policy may enhance perceived credibility by prioritizing certified professionals for medical content while permitting general wellness content from a broader uploader base33,34. However, the moderate overall quality scores observed in our sample suggest that platform governance alone is insufficient; creator incentives and audience preferences also influence content quality. This pattern echoes findings from the study of heart-failure videos on TikTok showing that moderation can reduce, but not eliminate, misinformation33. Accordingly, short videos should be considered a complement to, not a substitute for, professional medical consultation. Similar coexistence of high engagement and moderate reliability has been reported for diabetes, COPD35, inflammatory bowel disease20, gallstones17 and atopic eczema36. Source type exerted a major influence on quality in our sample: videos from professional sources scored higher on quality measures, consistent with prior studies of diabetes-related content on TikTok37. News agencies often produce edited, accessible explanations of complex topics, which may improve perceived clarity and usability32. Independent creators often produced emotionally engaging, anecdote-driven content, which may carry a greater risk of incomplete or misleading information38. Prior studies have likewise reported higher reliability for professionally produced videos and more subjective content in non-professional uploads19,39–42. Content type affected quality: disease-knowledge videos were generally more reliable than outpatient-scenario or personal-experience videos, consistent with previous short-video studies on thyroid topics43. This likely reflects that knowledge videos are more often evidence-based and include explanatory visual aids that enhance clarity44,45. Complication-related videos often achieved higher transparency and verifiability because they address acute risks and warning signs, prompting clinicians and institutions to provide explicit authorship, timestamps, and source references. Definition-focused videos scored higher in reliability and balance by delivering concise, factual explanations of medical terms or conditions in a structured, evidence-based format with clear scope and limitations. Re-examination videos were rated higher in overall quality, as they typically offered practical, stepwise guidance on monitoring intervals, required tests, and preparation steps, combining rationale with actionable instructions. Prevention-focused videos scored higher on actionability measures, as they provide actionable strategies that can be incorporated into daily routines and support long-term health management46,47. Expert monologues and dialogues were associated with higher quality, whereas patient vlogs, visual pictures and literature, and animation were tended to be fragmented and less coherent, potentially reduce their usefulness for delivering practical guidance. TCM-related content may have contributed to variation in video quality. Although comprising a small proportion of our sample (< 10%), TCM content scored relatively high on reliability and practicality, likely reflecting production by licensed practitioners presenting integrated approaches. TCM’s cultural resonance and holistic approaches (e.g., diet, acupuncture, herbal therapy) can foster trust among certain audiences based on perceived efficacy48,49. Prior work has documented patient acceptance of TCM across cardiovascular, oncologic, and neurological contexts50. However, findings are inconsistent across diseases and platforms. For example, Chen et al. observed that TCM uploaders accounted for a larger share of thyroid-nodule videos on Bilibili than on TikTok, yet their DISCERN and GQS scores were lower than those of Western medicine physicians43. Similarly, Zheng et al. reported generally low quality for many TCM-related liver-cancer videos. These discrepancies may reflect platform differences, disease specificity, and uploader heterogeneity, underscoring the need for larger, cross-platform studies to clarify contextual factors influencing TCM content quality19. Relevant factors in the effectiveness of video dissemination Beyond quality, we examined factors associated with dissemination and audience engagement. Engagement indicators (likes, comments, saves, shares) measure audience interaction but do not necessarily reflect information accuracy or subsequent downstream behavior change51,52. In this study, thyroid-carcinoma videos reached broad audiences, with popularity and interaction levels similar to those reported for related thyroid-nodule short-video content43. Source type shaped dissemination: news agencies achieved the highest engagement, likely reflecting editorial resources, institutional credibility, and larger follower network53. Patient vlogs generated relatively more comments and shares, driven largely by emotional resonance and community exchange. Interactive comment sections appear to facilitate experience sharing and peer support, which helps explain the popularity of patient vlogs41,54. However, these experience-sharing videos often scored lower on quality measures, illustrating a public health challenge: engaging narratives can spread widely despite limited evidentiary support38. These findings suggest that short-form health communication should seek to integrate scientific rigor with empathetic storytelling, and that public-health efforts should promote critical health literacy so viewers can better distinguish engaging narratives from evidence-based information. Dissemination varied across content categories. The predominance of disease-knowledge videos likely reflects strong public demand for health information and the internet’s accessibility as a convenient, low-cost channel. Treatment-related content was more frequent in our sample, possibly because of information needs for specific conditions that drive active searching and incentivize creators. The popularity of treatment and prevention videos indicates that online health content serves two distinct audiences: individuals seeking solutions for existing illnesses and those aiming to maintain health and prevent disease. These patterns carry significant public health implications. High demand for medical information creates opportunities for education but also increases the risk of misinformation when accuracy is not assured. Variations in health literacy may further influence how information is interpreted and applied, potentially reinforcing disparities. These considerations underscore the importance of healthcare institutions providing reliable, accessible resources and of platform mechanisms that align content visibility with scientific validity. Comparable dissemination patterns have been observed in studies of thyroid cancer and transarterial chemoembolization videos on TikTok and Bilibili, emphasizing the ongoing need for high-quality health information across digital platforms3,55–57. Overall, quality does not necessarily determine reach: emotionally engaging but less reliable content can capture disproportionate attention. Addressing this imbalance will require coordinated governance, algorithmic refinement, and active professional participation. To further explore these dynamics, we conducted correlation analyses linking video features, quality dimensions, and engagement patterns. Correlation analysis Correlation analyses provided additional insight into relationships among content features, quality, and engagement on short-video platforms. Strong intercorrelations among engagement metrics suggest that user behavior and platform design jointly effect content circulation on social media18,51,58. Likes, comments, saves and shares can amplify video visibility independently of content quality. Prior studies across TikTok and Bilibili—covering thyroid nodules, acute pancreatitis, and knee osteoarthritis—similarly report that videos with higher interactions reach larger audiences43,59,60. We observed a negative association between some quality scores and engagement, underscoring a tension between popularity and scientific rigor. Simplified or entertainment-oriented presentations may attract attention at the expense of accuracy, helping to explain why engaging but less rigorous information can dominate public attention. Strategies such as expert endorsements, credential tags, and inclusion of authoritative references may help mitigate this imbalance and support both accessibility and rigor61. Comparable trends have also been noted for lung cancer and brain tumor content on YouTube and Bilibili, suggesting a potentially broader pattern across diseases and media environments61,62. Nevertheless, the limited sample size and cross-sectional design of our study warrant cautious interpretation. Thus longitudinal or experimental research is need to test the robustness and causal mechanisms of these associations. Content-specific dimensions also influenced dissemination: understandability facilitated sharing, while actionability often required more detailed content to be conveyed effectively. This implies that practical guidance may require longer or structured formats, whereas easily digestible content spreads more readily. These patterns echo findings from coronary artery disease and pediatric pneumonia research, underscoring that quality assessment must be multidimensional because different aspects are driven by distinct factors. Combining understandability with actionable guidance may therefore enhance both dissemination and real-world utility of health videos60,63. Finally, we observed a modest positive association between content quality and creator follower count, consistent with prior observations for coronary heart disease information on TikTok32. This finding underscores the potential for verified expert accounts to enhance reliability, guide audience interpretation, and improve health literacy and decision-making at scale. Implications and significance Practical Implications. Our findings underscore the need for audience-centered health communication strategies. Healthcare professionals should optimize videos by integrating case-based narratives, visual models, and interactive features. Translating complex medical concepts into clear, actionable information enhances video quality, audience engagement, and public health literacy. For culturally relevant content (e.g., TCM), ensuring information originates from credible sources and includes practical guidance can help strengthen audience trust. Individual creators should clearly distinguish personal experiences from professional advice and emphasize psychosocial support and practical navigation information where appropriate. News organizations and institutions should produce evidence-based content and consider credential verification, tiered labeling, or preferential recommendation to improve content engagement. Policy and Governance Implications. System-level, multi-stakeholder approaches—including professional engagement and platform-level measures—may help better align reach with reliability. Measures such as optimizing recommendation algorithms, piloting AI-assisted screening, and promoting cross-platform collaboration could reduce amplification of low-quality content. Structured cues, clear case analyses, and visualized information can integrate personal narratives with professional knowledge to preserve engagement while improving informational accuracy. Overall Significance. By mapping cross-platform short-video content about thyroid carcinoma, this study clarifies relationships among content quality, audience engagement, and platform dynamics, offering evidence to inform creators, platforms, and future research aimed at improving online health information. Limitations This study has several limitations. First, our analysis was limited to Chinese-language videos on TikTok, Kwai, and Rednote. Although selected for their dominance in the Chinese market and distinct short-video mechanisms, these platforms differ from cross-language platforms (e.g., YouTube, Instagram), so our findings may not generalize globally. Future research should include cross-language comparisons to capture broader patterns in digital health communication. Second, data were collected within a fixed time window, which may not capture the dynamic and rapidly evolving nature of thyroid-carcinoma—related content. This temporal constraint could introduce bias and limit interpretation of longitudinal trends. Third, we used proxy indicators—such as likes, comments, saves, and shares—to assess engagement and dissemination. Although commonly used in digital-media research, these measures do not directly capture behavioral change, clinical decision-making, or health outcomes. Finally, we retrieved the top 250 ranked videos per search query to maximize coverage of highly visible content; however, platform-specific ranking algorithms and temporal fluctuations in popularity may have influenced video selection and thus limit generalizability. Moreover, the cross-sectional observational design precludes causal inference about relationships between content features and engagement. Conclusions This cross-platform evaluation demonstrates that despite substantial reach, content quality for thyroid carcinoma videos is uneven. Healthcare professionals and institutions generally produced more accurate, reliable content but received fewer interactive comments than patient vlogs; independent creators attracted more comments yet typically scored lower on JAMA and mDISCERN indices. These patterns varied across dissemination metrics, highlighting the complex interplay between content quality and audience behavior. These findings support efforts to promote authoritative sources, foster institution–media collaborations, and refine platform mechanisms to better balance accessibility with rigor. Interventions should address both content quality and audience engagement to be effective. Our results can inform policies and practices aimed at aligning audience reach with evidence-based information on short-video platforms. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1