Abstract Background During the COVID-19 pandemic, social media platforms became major channels for public health communication and public sensemaking, but they also enabled the rapid circulation of rumors, false claims, conspiracy narratives, and misleading health information. Research on online COVID-19 misinformation expanded rapidly during the peak pandemic period, yet the evidence base was fragmented across public health, computing, communication, and information science. Objective This scoping review synthesizes empirical research on online COVID-19 misinformation on social media during the peak pandemic period from February 2020 to October 2022. It examines how misinformation was conceptualized, which platforms and public health contexts were emphasized, and how computational methods shaped the evidence base for infodemic management. Its distinct contribution is an integrated comparison of conceptual definitions, thematic and platform concentration, the computational research pipeline, and the gap between detection and actionable infodemic governance. Methods We conducted a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)-informed scoping review of empirical studies indexed in PubMed and IEEE Xplore in November 2022 using COVID-19, social media platforms, and misinformation-related terms. Eligible reports were English-language, peer-reviewed empirical studies published between February 2020 and October 2022 that analyzed social media data using computational, quantitative, or social media analytic methods; the language restriction applied to the publication, not to the underlying dataset. The final synthesis included 100 studies selected from 1688 records. Results Within the included corpus, the literature was concentrated around vaccine-related discourse, public health and social measures-related misinformation, and text-based data from X (X Corp), formerly known as Twitter. Studies commonly used misinformation detection, classification, sentiment analysis, topic modeling, and network analysis. Fewer studies examined diffusion pathways, source tracing, cross-platform movement, intervention effectiveness, or platform governance. The evidence base remains more developed for detecting and describing misinformation than for evaluating actionable responses to it. Conclusions Peak-pandemic COVID-19 misinformation research produced important computational and empirical knowledge, but it remains detection-centered and unevenly connected to public health action. Future infodemiology research should link computational outputs to intervention design, risk communication, fact-checking workflows, platform moderation, multilingual and multimodal analysis, and the evaluation of infodemic governance strategies.
BackgroundVitreous floaters are a prevalent phenomenon, with some patients experiencing persistent symptoms that substantially degrade their vision. Public interest and information-seeking behavior regarding vitreous surgery are challenging to quantify using claims alone. ObjectiveThe aim of this study is to evaluate temporal and geographic patterns of public interest in vitreous floaters and related interventions in the United States and to contextualize this interest relative to retina specialist availability. MethodsIn this retrospective observational study, we queried Google Trends for search terms originating in the United States from 2008 to 2025. We extracted the relative search volume (RSV) using a composite of terms combined with the Google Trends “+” operator. State-level counts of vitreoretinal surgeons from the American Society of Retina Specialists (ASRS) database were used to calculate surgeon density per 100,000 population. We calculated a search-to-surgeon index (SSI) as the ratio of state-level RSV to retina surgeon density to serve as an exploratory metric of public interest relative to specialist availability. We evaluated associations using the Mann-Kendall trend test, Sen slope, and Spearman correlation. ResultsFrom 2008 to 2025, nonsurgical RSV increased by 138%, demonstrating a significant temporal increasing trend (τ=0.600; P<.001). States with a higher proportion of residents aged 40 years or older demonstrated higher RSV (ρ=0.51; P<.01). The median SSI was 123 (IQR 100-127; range 48-311). Surgical RSV increased by 133% (τ=0.443; P<.001). The median surgical SSI was 133 (IQR 107-164; range 34-341). ConclusionsThis analysis demonstrates substantial geographic variability in public interest in vitreous floaters across the United States. Integrating Google Trends data with surgeon distribution provides an exploratory framework for evaluating relative interest in vitreoretinal care, though it does not directly measure clinical demand.
BackgroundPublic health agencies increasingly use social media platforms such as Twitter (subsequently rebranded X, X Corp) to monitor public responses to official health communications and support timely decision-making during health crises. Public replies to communications from official agencies provide valuable insights into population-level discourse and engagement with health policies. However, analyzing large-scale short-text responses remains challenging because social media replies are often brief, informal, ambiguous, and linguistically fragmented. Topic modeling offers a scalable approach for identifying thematic patterns in such data, but its performance is often limited when applied to social media short texts. Recent advances in large language models (LLMs) provide new opportunities to improve text normalization before topic modeling, but their value and limitations for public health infoveillance remain underexplored. ObjectiveThis study develops and evaluates TM-Rephrase, a model-agnostic LLM-based rephrasing framework designed to improve the performance of topic models for short public health social media texts while ensuring rephrasing preserves the original semantic meaning, stance or intent, and tone. MethodsWe analyzed 25,027 public replies to official Centers for Disease Control and Prevention tweets on Twitter collected between May 2020 and November 2022. TM-Rephrase transformed informal and noisy short texts into more standardized expressions using 2 prompt-guided schemes: general rephrasing and colloquial-to-formal rephrasing. Original and rephrased texts were analyzed using multiple topic models, with rephrasing generated by Gemini 2.5 Flash, GPT‑4o mini, and Mistral-7B-Instruct. Topic quality was evaluated using coherence, uniqueness, redundancy, and diversity. We also conducted an expert semantic-fidelity validation study in which 4 expert raters evaluated whether rephrased texts preserved the original meaning, stance or intent, and tone. ResultsTM-Rephrase generally improved topic quality across models, rephrasing schemes, and LLM backbones. For latent Dirichlet allocation, Cv coherence increased from 0.3094 without rephrasing to 0.5004 with colloquial-to-formal rephrasing. For BERTopic (BERT-based topic modeling), Cv coherence improved from 0.4078 to 0.4734. Diversity-related metrics also improved in most rephrasing settings. Additional checks using CNPMI, CUCI, and UMASS generally supported the main improvement pattern. Sensitivity analysis across different topic numbers further suggested that the coherence gains were not limited to the main setting. Expert validation showed favorable semantic preservation overall, with a mean rating of 3.90 (SD 1.01) and median of 4.00 (IQR 3.00-5.00) of 5.00, and acceptable interrater reliability, intraclass correlation coefficient =0.736. General rephrasing received stronger semantic-fidelity ratings than colloquial-to-formal rephrasing, suggesting that stronger formalization may improve topic readability while introducing greater risk of changes to the original meaning, stance or intent, and tone. ConclusionsTM-Rephrase can enhance the coherence, distinctiveness, and interpretability of topic modeling outputs for short, noisy public health social media data. However, it is a text-normalization preprocessing strategy for topic enhancement rather than a fully neutral replacement for original public discourse.
Abstract Background Patients with seizure clusters and their families often use YouTube videos as resources for information about the illness and how to use concomitant rescue medications; however, many videos include misinformation or incorrect information that could potentially cause harm. Objective This study aims to examine YouTube videos intended as educational and instructional resources for seizure clusters and treatment options posted since the approval of intranasal benzodiazepine rescue medications in 2019 and to assess their accuracy and quality using a 3-item rating scale. Methods We searched YouTube for 10 specific terms related to seizure clusters and treatment options and analyzed videos from the first 20 results for each to mimic internet user behavior when choosing results with search engines. Videos were rated on 3 criteria: accurate seizure cluster definition, appropriate US Food and Drug Administration (FDA)–approved treatment recommendation, and correct visual demonstration of rescue medication use. A single author rated the videos based on the presence of specific subcriteria of all 3 primary criteria in each video. Score assignment was based on the number of present subcriteria. Results Seizure clusters and trade names generated the highest number of videos meeting criteria; other search terms, including nonproprietary names, garnered <8 videos each. Only 19.5% (n=39) of the 200 collected videos provided accurate descriptions of seizure clusters, while 61.0% (n=122) offered no definition, 37.0% (n=74) recommended at least 1 FDA-approved rescue medication, and 37.5% (n=75) offered no rescue therapy at all; 26.5% (n=53) provided a full demonstration of rescue medicine, while 62.5% (n=125) did not. Only 4.5% (n=9) of videos included all subcriteria of the 3 primary criteria. Search results for specific trade names yielded more results, while less specific terminology returned unrelated or unhelpful videos. Conclusions Accurate YouTube videos of seizure cluster management are limited and represent a knowledge gap that should be addressed to avoid patient harm. To facilitate better navigation of educational and instructional resources, clinicians should have a ready list of videos available to give to patients or provide a list of search terms that are known to produce accurate results to facilitate better searches. Reputable health care sources could also produce brief, helpful videos of each rescue therapy with search terms that direct patients to their content.
