Assessing the Quality of YouTube Videos About Cannabinoid Hyperemesis Syndrome

Article type: Research Article

Keywords: cannabinoid hyperemesis syndrome, cannabis use, effects of social media, mixed methods research, qualitative content analysis

License: Copyright © 2026, Dean et al. CC BY 4.0 This is an open access article distributed under the terms of the Creative Commons Attribution License CC-BY 4.0., which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Article links: DOI: 10.7759/cureus.109113  |  PMC: PMC13274578

Relevance: Moderate: mentioned 3+ times in text

Full text: PDF (192 KB)

Introduction

Cannabis use continues to increase as medical and recreational legalization expands both in the United States and globally [ref. 1–ref. 3]. Although cannabis is used for several therapeutic indications, adverse effects have also become more apparent, including cannabinoid hyperemesis syndrome (CHS). CHS is characterized by cyclic vomiting in people with chronic, heavy cannabis use and is often accompanied by compulsive hot bathing for symptom relief [ref. 4–ref. 6]. The condition can lead to repeated emergency department visits, extensive diagnostic testing, and substantial healthcare costs, yet it remains underrecognized and inconsistently diagnosed [ref. 6,ref. 7]. Delayed recognition may delay counseling on cannabis cessation, the most reliable treatment for CHS.

YouTube (Alphabet Inc., Mountain View, CA) has become a major source of health information, and health-related videos increasingly influence patient understanding and decision-making [ref. 8–ref. 13]. Recent survey data indicate that approximately 87.6% of YouTube users report watching health-related content, and 84.7% state that these videos affect their health-related decisions, including whether to consult a clinician or change health behaviors [ref. 9,ref. 10]. At the same time, prior studies have shown that medical content on YouTube is often incomplete, biased, or inaccurate, particularly when produced by nonprofessional or commercial sources [ref. 14–ref. 17]. Cannabis-related videos are especially prominent, often attracting high engagement while providing limited discussion of risks or containing misinformation [ref. 8,ref. 18]. Prior work has also identified CHS-related misinformation online, including unsupported explanations for disease causation, minimization of symptom severity, and promotion of ineffective or potentially harmful self-management strategies [ref. 18–ref. 19]. For someone desperately searching for an explanation for their symptoms, these videos can be persuasive, even when incomplete or inaccurate.

This issue is clinically important because patients with suspected CHS may turn to online videos to interpret symptoms, consider treatment options, and decide whether to seek medical care [ref. 18]. Given that CHS may be unfamiliar to some clinicians and challenging for some patients to accept, online content may meaningfully shape risk perception and the willingness to cease cannabis use [ref. 13,ref. 18]. However, despite the growing volume of CHS-related videos, their accuracy, completeness, and clinical usefulness remain poorly characterized [ref. 19,ref. 20].

The objective of this study was to evaluate the accuracy, completeness, and clinical usefulness of YouTube videos about CHS. Specifically, the analysis assessed whether videos provided accurate information about diagnosis, treatment, prevention, complications, and indications for seeking medical care, and whether they contained misleading or unsubstantiated medical claims.

Materials and methods

Study design

This cross-sectional content analysis evaluated CHS-related videos on YouTube. The Michigan State University Institutional Review Board reviewed the study protocol and determined that it constituted non-human subjects research.

Search strategy

YouTube searches were conducted using 15 predefined CHS-related search prompts, including both clinical and lay terms (Table 1). The platform’s search function ranks videos based on a combination of keyword relevance, engagement metrics, and user-specific factors, such as watch history and location. This algorithm makes it difficult to obtain a completely unbiased or exhaustive list of videos on a given topic. To minimize personalization, all searches were conducted in a logged-out browser with a cleared cache and default settings. For each term, the first 50 relevance-ranked results were screened using protocol-specified inclusion and exclusion criteria. Screening was stopped when additional scrolling yielded only duplicates or clearly irrelevant content. This approach aligns with published methods for YouTube health content analyses and systematic gray literature searching [ref. 18,ref. 21–ref. 23].

Video content

Videos were collected from April to September 2023. Extracted variables included views, content author or publisher, presenter format, likes, dislikes, comments, and abstractor impressions. For each video, the total number of viewer comments was recorded and used as a quantitative indicator of engagement. Medical claims were classified as either substantiated or unsubstantiated by two board-certified toxicologists (BSJ, BDR) with extensive experience in cannabis-related toxicology.

Training of content reviewers

All investigators were trained in their protocol roles and jointly reviewed the abstraction form, coding scheme, and eligibility criteria at study initiation. Student reviewers completed protocol-specific training covering data extraction, use of the accuracy checklist, and content analysis principles. To standardize baseline knowledge and support consistent identification of CHS-related content, reviewers received a brief CHS fact sheet summarizing diagnostic features, clinical course, common misconceptions, and key management points. Final judgments about accuracy and substantiation were made by the toxicology experts.

Training included a piloting phase in which student reviewers independently coded a small set of videos, with discrepancies resolved through group discussions with senior investigators. Based on this iterative feedback, wording and operational definitions were refined before full data collection began. Throughout the abstraction period, seven student reviewers were mentored by a senior author, and a faculty member periodically reviewed forms to answer questions, monitor quality, and minimize interobserver variation.

Data abstraction

A standardized .docx-based abstraction form was organized into four sections. The first captured video characteristics, including URL, date accessed, date published, source, length, views, comments, likes, dislikes (when available), presentation format, and presenter demographics. The second section assessed content accuracy using a 16-item, open-ended response CHS checklist covering definition, epidemiology, symptoms, diagnosis, treatment, prevention, complications, and indications for seeking care (Table 2). Videos were classified based on the number of criteria met: ≥13 items as exemplary, 10-12 as useful, 7-9 as not useful, and ≤6 or any video containing unsubstantiated claims as misleading. These thresholds were developed a priori, informed by frameworks such as DISCERN and the Global Quality Scale [ref. 24–ref. 27], and were reviewed and approved by two medical toxicologists to ensure expert content validity.

The third section captured free-text responses documenting inaccurate or misleading information as well as helpful or unhelpful aspects. Each video received an overall usefulness rating (useful, not useful, misleading) as outlined in Table 3. All assessment tools were developed by the authors specifically for this study, informed by prior work evaluating online medical content, and have not been formally validated as independent instruments [ref. 28–ref. 30].

Reviewers often described a video as biased or misleading when it strongly promoted a particular viewpoint, minimized risks, or appeared to serve primarily as an advertisement, regardless of production quality or narrative style. Student impressions of bias were informed by their CHS training and fact sheet, but final classification of specific statements as unsubstantiated or misleading was determined by the toxicologist.

Following abstraction piloting, videos were assigned to student reviewers for parallel abstraction. Disagreements were discussed in review meetings and resolved by consensus. No conflicts required mediation by an additional author. Responses were then entered into a spreadsheet for thematic coding and quantitative analysis. To assess interrater reliability, 10% of videos were independently reviewed by senior physician investigators, and Cohen’s kappa was calculated to quantify agreement beyond chance [ref. 31].

Medical claim substantiation

To assess the veracity of medical statements in the video content, two board-certified toxicologists (BSJ, BDR) independently assessed medical statements. A statement was considered substantiated if clearly supported by rigorous, up-to-date scientific evidence and aligned with expert consensus [ref. 30,ref. 32,ref. 33]. Review criteria for unsubstantiated claims are shown in Table 4.

Study endpoint and analysis

The primary endpoint was overall clinical utility, defined as how clearly and accurately videos described CHS, including associated risks and complications. The main outcome was the reviewer-rated quality and educational value of each video for clinical information about CHS, as assessed by our trained reviewers and expert toxicologists. Descriptive statistics summarized key variables with 95% confidence intervals (CIs). Associations between continuous variables were examined using Pearson product-moment correlation coefficients.

Results

YouTube content characteristics

Over 900 potentially eligible videos were identified in a pre-study search. Of these, 100 videos initially met the inclusion criteria. Three were excluded (one duplicate, one set to private, one irrelevant), yielding a final sample of 97 videos. The videos were collectively viewed a total of 795,264 times, with an average of 8,284 and a 1,050 median (interquartile range (IQR) of 8,284 (254-8,304) views noted per video. The mean video length was 9.8 ± 4.1 minutes (0.60-47.7 minutes). These video metrics fell within the site averages for all content uploaded to the platform [ref. 34,ref. 35].

The average age of the content (time from upload to access) was 30 ± 2.1 months (range: <1-85 months). Over half of the videos (n = 50, 52%) were accessed within two years of publication. Content “likes” had a mean of 257 (range: 0-8,400) and a median (IQR) of 28 (7-218) per video, with three videos having unavailable counts and two videos where uploaders had disabled counts. Dislike counts were available for only 26 videos uploaded before November 2021, when YouTube disabled dislike counts (mean: 1, range: 0-11).

Most videos (n = 82, 85%) featured on-camera presenters (character format), with 70 (72%) narrated by Caucasian presenters. Other demographic groups made up approximately 27 (28%) of the character videos. Non-character presentations (n = 15, 15%) included animations (n = 19, 20%) and photographs or slideshows (n = 3, 3%).

Engagement metrics and video quality

To assess the presence of relationships between content metrics, Pearson product-moment correlation coefficient (PPMCC) testing was performed. This analysis (Table 5) revealed no meaningful linear relationships between content age, video length, views, comments, or reviewer-rated usefulness (all correlations near 0).

Accuracy and completeness

No video met all 16 checklist criteria. Only two videos (2.1%) were classified as exemplary (≥13 accurate items), 25 (25.8%) as useful (10-12 items), 51 (52.6%) as not useful (seven to nine items), and 19 (19.6%) as misleading (≤6 items or containing unsubstantiated claims). Table 6 shows the coverage of individual checklist items; the most frequently addressed topics were hospital treatment (78%), definition (66%), and cure information (65%), while the least covered were fatal warnings (6%), addiction referrals (9%), and indications for emergency care (11%).

Video sources and formats

The study characterized the sources and perspectives of the videos, providing clinicians with context useful for discussing CHS with patients. Testimonials comprised 79% of videos (n = 77), primarily from current or former patients (n = 58, 60%). These candid discussions were often unstructured and included opinions and descriptions of clinical encounters. Ten testimonials functioned as advertisements for a for-profit substance-use coaching program. Didactic videos (n = 15, 15%) included six local news briefs that provided concise, useful overviews. Eight videos featured healthcare professionals (five physicians, one nurse, one pharmacist, and one EMT), although professional credentials did not guarantee the accuracy of the information presented.

Clinical usefulness, medical substantiation, and misleading claims

Strict definitions were applied to determine clinical usefulness. Content was required to satisfy at least 75% of the checklist (see topic accuracy assessment Table above) and meet standards for medical substantiation. Overall, 26% (n = 25) of videos were rated useful, 54% (n = 52) not useful, and 20% (n = 20) were considered misleading. Inaccurate information appeared in 55% of videos (n = 53). Interrater agreement was substantial for usefulness classification (κ = 0.78, 95% CI: 0.66-0.90) and for the presence of inaccurate information (κ = 0.71, 95% CI: 0.61-0.82).

Of the 58 patient testimonials, 15 (26%) contained unsubstantiated claims about CHS etiology or management, including denial of CHS existence, attribution to contaminated cannabis, claims that hot showers or home remedies permanently cure CHS, assertions that only heavy users are at risk, or statements that any vomiting in cannabis users indicates CHS.

Thematic analysis of reviewer impressions

An impartial qualitative assessment of the free-text responses was conducted by a faculty non-reviewer. The following themes and subcategories were identified for key reviewer questions. Reviewers identified both positive and problematic content features (Tables 7–8). Favorable elements included personal anecdotes that conveyed relatability and honesty about struggles with addiction (n = 95, 98% had some favorable features), accurate medical information, and high production quality with concise format (that is, less than <5 minutes in duration). Unfavorable aspects included off-topic or overly informal presentation, coarse language, poor audiovisual quality, and biased or misleading claims, particularly in videos promoting products or services. For 12 reviewers, 12% "no negative" impressions were documented.

Inaccurate information (n = 55, 57%) most commonly involved unsubstantiated etiologies (pesticides, genetic changes, tetrahydrocannabinol (THC) release theories), unproven claims about cannabidiol (CBD) products, downplaying symptom severity, recommendations for continued cannabis use (including one advocating increased intake), and suggestions that lifestyle or diet changes alone are sufficient treatment instead of complete cessation. These impressions are outlined in Table 9 below.

Discussion

CHS is increasingly recognized as a consequence of chronic, heavy cannabis use, yet timely diagnosis and acceptance of cannabis cessation as the primary treatment remain challenging. Patients with suspected CHS commonly search online for explanations of their symptoms before or between clinical encounters, often turning to YouTube for first-person accounts and advice. This study provides, to our knowledge, the first systematic evaluation of CHS-related YouTube videos. The findings reveal substantial gaps in reliable information, a high prevalence of misleading claims, and weak correlations between video popularity and medical quality.

Main findings

Of 97 CHS-related YouTube videos analyzed, only a small minority were clinically useful or exemplary; most were rated not useful or misleading. Videos frequently omitted essential information, including explicit diagnostic criteria, indications for urgent evaluation, evidence-based management strategies, and the central role of sustained cannabis cessation. Even when videos accurately described some CHS features, they often lacked critical content domains, limiting their value as patient education resources.

Personal testimonial videos were common and often valued by reviewers for their honesty and relatability, conveying the emotional toll of recurrent vomiting, emergency visits, and stigma associated with cannabis use. However, testimonials also accounted for a substantial proportion of unsubstantiated claims, including speculative etiologies, minimized symptom severity, and promotion of unproven remedies. This duality illustrates how patient narratives can both validate lived experiences and amplify misinformation about CHS [ref. 32].

Standard YouTube engagement metrics (views, likes, comments) showed no meaningful linear association with reviewer-rated usefulness. Several highly viewed videos contained substantial misinformation, suggesting that basic engagement metrics are unreliable proxies for content quality. This disconnect underscores the challenge patients face when using surface cues to judge credibility.

Clinical and educational implications

Given the high rates of online health information seeking, clinicians should assume that many patients have viewed CHS-related videos and directly ask about content exposure to clarify misconceptions, align expectations, and build trust. Common themes requiring discussion include attributing CHS to contaminated cannabis rather than chronic use, overemphasizing hot showers or home remedies as curative, and suggesting symptom control without cannabis cessation.

Because engagement metrics do not reliably indicate content quality, clinicians should recommend specific, vetted resources rather than expecting patients to identify high-quality content independently [ref. 28,ref. 30]. Professional societies and educators have an opportunity to address this gap by creating concise, evidence-based, patient-facing content that is easy to access. Such resources should detail diagnostic features, risks, management rationale, and indications for urgent care, and ideally be readily available on the same content platforms [ref. 27]. Collaboration between toxicologists, emergency physicians, and communications professionals could also engage active patients to discuss their experiences with seeking health information, which may enhance both the accuracy of CHS patients’ lived experiences and the content’s audience appeal.

Strengths and limitations

This study employed a transparent, systematically applied search strategy with a structured CHS content checklist informed by expert toxicologists and supported by standardized reviewer training. It also demonstrated substantial interrater agreement (κ = 0.78). Additionally, the mixed-methods approach to reviewer ratings and impressions provided a nuanced understanding of how CHS is portrayed on YouTube.

At the same time, several limitations should be considered when interpreting these findings. The author-developed checklist and usefulness scale, while supported by expert input and high kappa values, were study-specific and have not undergone formal psychometric validation. Classification thresholds were set a priori and deliberately conservative, and different cutoffs could alter the distribution of categories.

YouTube search results are dynamic and influenced by personalization factors that cannot be fully controlled. Despite using standardized search terms and logged-out browsing, the dataset represents a time-limited snapshot (April-September 2023) and may reflect selection and platform algorithm biases. The analysis was restricted to English-language content on a single platform, limiting generalizability to other languages, regions, or platforms such as TikTok or Instagram. The modest sample size (n = 97) is adequate for descriptive analysis but constrains statistical power for detecting small associations or subgroup analyses. Our analyses are therefore primarily descriptive, supplemented by simple correlation tests, and are not intended to support complex inferential modeling.

Additionally, there were a few YouTube policy changes that affected the content included in this analysis. YouTube disabled the “dislike” reaction in November 2021, which affected 27 videos (26%) of our dataset. Further, a Community Guidelines update in August 2023, during the data extraction period, provided a framework for prevention, treatment, and disinformation categorization for reporting violations [ref. 36]. While our dataset was not actively affected by these changes, future content analyses will likely be influenced by these updates.

Judgments of usefulness and misleading content involve some degree of subjectivity despite structured tools, reviewer training, and consensus procedures. This may contribute to misclassification at the margins and should be considered when interpreting the exact proportions of videos in each category. Finally, the study characterizes content but does not assess how viewers interpret or act on videos, precluding causal inferences about clinical outcomes. Our recommendations to clinicians and educators should therefore be understood as reasoned implications of the observed content landscape rather than conclusions about measured behavioral effects.

Future research directions

Future research could analyze larger, more diverse samples, including non-English content and videos on other platforms; examine trends over time to determine whether content quality improves in response to platform policy changes or professional education initiatives; and formally validate assessment tools against established measures, such as DISCERN, to enhance comparability across studies.

Conclusions

Most CHS-related YouTube videos in this sample were not clinically useful and frequently contained incomplete or unsubstantiated information regarding diagnosis, risks, and management. Personal testimonials often blend accurate symptom descriptions with speculative etiologies and non-evidence-based advice. Because engagement metrics are poor indicators of content quality, patients relying on popular videos may encounter misleading messages that downplay the importance of cannabis cessation and medical evaluation. Clinicians should anticipate patient exposure to such content, address misconceptions directly, and guide patients toward vetted resources. High-quality, expert-developed CHS educational videos are urgently needed to improve the reliability of information available on social media platforms.

References

1. United Nations Office on Drugs and Crime – World Drug Report. 2025
2. Substance Abuse and Mental Health Services Administration – Results from the 2014 National Survey on Drug Use and Health: summary of national findings. 2025
3. Manzanette Manzanette, (2021 Z.. Marijuana legalization continues to grow: 2021 laws map. 2025
4. PF Whiting, RF Wolff, S Deshpande. Cannabinoids for medical use: a systematic review and meta-analysis. JAMA, 2015. [PubMed]
5. HS Kim, JD Anderson, O Saghafi, KJ Heard, AA Monte. Cyclic vomiting presentations following marijuana liberalization in Colorado. Acad Emerg Med, 2015. [PubMed]
6. JH Allen, GM de Moore, R Heddle, JC Twartz. Cannabinoid hyperemesis: cyclical hyperemesis in association with chronic cannabis abuse. Gut, 2004. [PubMed]
7. G Perrotta, J Miller, T Stevens. Cannabinoid hyperemesis: relevance to emergency medicine. Acad Emerg Med, 2012
8. Team GMI Research. YouTube Statistics 2026: users by country and demographics. YouTube Statistics, 2026
9. F Mohamed, A Shoufan. Users’ experience with health-related content on YouTube: an exploratory study. BMC Public Health, 2024. [PubMed]
10. J Lee, K Turner, Z Xie, B Kadhim, YR Hong. Association between health information‒seeking behavior on YouTube and physical activity among U.S. adults: results from Health Information Trends Survey 2020. AJPM Focus, 2022. [PubMed]
11. YC Zhao, M Zhao, S Song. Online health information seeking behaviors among older adults: systematic scoping review. J Med Internet Res, 2022
12. T Deb. Social media in healthcare statistics by data, insights, engagement. 2026
13. SW Oh. YouTube, health information, and health literacy. Korean J Fam Med, 2023. [PubMed]
14. B Etumuse, M Greer, J Onyemachi. Medical misinformation and quality of public video content on cannabis for chronic pain management: a cross-sectional analysis of the YouTube platform. J Pain Res, 2024. [PubMed]
15. B Lewis, E Leach, LB Fomum Mugri. Community-based study of cannabis hyperemesis syndrome. Am J Emerg Med, 2021. [PubMed]
16. X Liu, A Susarla, R Padman. Promoting health literacy with human-in-the-loop video understandability classification of YouTube videos: development and evaluation study. J Med Internet Res, 2025
17. L Ouellette, M Cearley, B Judge, B Riley, J Jones. Cooking with cannabis: the rapid spread of (mis)information on YouTube. Am J Emerg Med, 2018. [PubMed]
18. S Khare, S Erridge, S Chidambaram, MH Sodergren. Misinformation about medical cannabis in YouTube videos: systematic review. JMIR Form Res, 2025
19. L Ouellette, S Farley, J Rieth, B Riley, B Judge, J Jones. Monitoring emerging toxicology trends using social media: eyeballing, vaportinis, and funneling. Am J Emerg Med, 2018. [PubMed]
20. T Sun, C Lim, G Chan. Sun T, Lim C, Chan G, Leung J – High times for cannabis-related videos on YouTube during the COVID-19 lockdown: implications on the risk of cannabis use disorders – Proceedings of the 3rd International Electronic Conference on Environmental Research and Public Health—Public Health Issues in the Context of the COVID-19 Pandemic. Proceedings of the 3rd International Electronic Conference on Environmental Research and Public Health —Public Health Issues in the Context of the COVID-19 Pandemic, 2026
21. CC Lim, J Leung, JY Chung. Content analysis of cannabis vaping videos on YouTube. Addiction, 2021. [PubMed]
22. Google Help. YouTube. How YouTube search works. 2026
23. K Godin, J Stapleton, SI Kirkpatrick, RM Hanning, ST Leatherdale. Applying systematic review search methods to the grey literature: a case study examining guidelines for school-based breakfast programs in Canada. Syst Rev, 2015. [PubMed]
24. D Charnock, S Shepperd, G Needham, R Gann. DISCERN: an instrument for judging the quality of written consumer health information on treatment choices. J Epidemiol Community Health, 1999. [PubMed]
25. A Zainab, R Sakkour, K Handu, S Mughal, V Menon, D Shabbir, A Mehmood. Measuring the quality of YouTube videos on anxiety: a study using the Global Quality Scale and DISCERN Tool. Int J Community Med Public Health, 2023
26. S Shepperd, D Charnock, A Cook. A 5-star system for rating the quality of information based on DISCERN. Health Info Libr J, 2002. [PubMed]
27. H Hakyemez Toptan, A Kizildemir. Quality and reliability analysis of YouTube videos related to neonatal sepsis. Cureus, 2023
28. E Gabarron, L Fernandez-Luque, M Armayones, AY Lau. Identifying measures used for assessing quality of YouTube videos with patient health information: a review of current literature. Interact J Med Res, 2013
29. B Drozd, E Couvillon, A Suarez. Medical YouTube videos and methods of evaluation: literature review. JMIR Med Educ, 2018
30. M Khalil, F Mohamed, A Shoufan. Evaluating the quality of medical content on YouTube using large language models. Sci Rep, 2025. [PubMed]
31. ML McHugh. Interrater reliability: the Kappa statistic. Biochemia Medica, 2012
32. D Gurler, I Buyukceran. Assessment of the medical reliability of videos on social media: detailed analysis of the quality and usability of four social media platforms (Facebook, Instagram, Twitter, and YouTube). Healthcare (Basel), 2022
33. D Mandrioli, EK Silbergeld. Evidence from toxicology: the most essential science for prevention. Environ Health Perspect, 2016. [PubMed]
34. Bora Bora, S. S., Atasoy Atasoy, I. I., Uslu &, (2025 A.. StatsUp: latest YouTube statistics. 2025
35. (2025 Bynder.. Average video length predicted to decline to 65 seconds in 2024. 2025
36. Graham Graham, G. G., Halprin &, (2023 M.. A long-term vision for YouTube’s medical misinformation policies. 2025