bims-librar 2026-09-20 papers

bims-librar

Biomed News

on Biomedical librarianship

Issue of 2026–09–20
29 papers selected by
Thomas Krichel, Open Library Society

What I'm Reading Now: Stories that Bind Us Together: Why Children's Literature Matters for Addiction Work.
Development, Implementation, and Evaluation of a Proof of Concept, Food- and Nutrition-Focused Science Education Program for Youth in Public Libraries.
FatPlants 2.0: an AI-powered platform integrating plant lipid genes, pathways, and literature.
Systematic assessment of text summarization methods for biomedical literature from frequency methods to language models.
Semantic Retrieval of Projects Funded by the Innovation Committee.
Retrieve-Then-Verify for Evaluating Evidence Support and Hallucination in Large Language Model-Generated Medical Information: Empirical Study.
Information Quality and Semantic Stability of AI Chatbot Responses to Complete Denture Questions: A Turkish-English Benchmarking Study.
How appropriate are ChatGPT-5's answers to acne questions?: A category-based evaluation by dermatologists.
Performance evaluation of large language models in bladder cancer patient education Q&A: a cross-sectional study.
Artificial intelligence responses to bronchiectasis patients' questions: the AIR-BE study.
Safety, accuracy, empathy, information quality, and readability of publicly accessible LLM-based chatbots for traumatic brain injury and concussion questions: a cross-sectional comparative study.
Assessing the accuracy and readability of large language models in answering Turkish-language patient-generated queries on radiological contrast agents.
Assessing the accuracy and usability of artificial intelligence-based language models in responding to common periodontal patient questions.
Large language models for tympanostomy patient education: readability and guideline adherence.
Evaluating ChatGPT's Effectiveness for Arabic Dry Mouth Patient Education.
Evaluation and Comparison of Large Language Model Responses to Frequently Asked Questions Regarding Patellofemoral Pain Syndrome: A Quality and Readability Assessment Study.
Comparing Chinese-language large language models for caregiver questions about developmental dysplasia of the hip.
A comparative evaluation of generative AI chatbots for patient-oriented advice on painful diabetic peripheral neuropathy: safety, accuracy, guideline concordance, actionability, and readability.
Quality of Online Information for Management of Leg Pain in Children and Adolescents With Hypermobility-Associated Conditions.
The readability of total elbow arthroplasty educational materials.
From hematuria warning to precision immunotherapy: an AI-assisted cross-platform analysis of bladder cancer information on social media.
Comparison of YouTube videos on single-port and multi-port robot-assisted radical prostatectomy using the PACE score.
Educational value assessment of YouTube surgical videos of bladder hydrodistention for surgeons.
Assessing the quality and reliability of health information on GLP-1 receptor agonists on TikTok and Bilibili: a cross-sectional analysis of the public health infodemic.
Correction: Quality and reliability assessment of systemic lupus erythematosus related health information on short video platforms.
Determinants of sexual health information-seeking on social media among Chinese college students: an application of the Comprehensive Model of Information Seeking.
Factors Associated with Digital Health Information-Seeking Behavior Among Healthcare Professionals in Izmir: A Cross-Sectional Study.
Supplement Safety Knowledge, Information Sources, and Self-Reported Safety-Related Behaviors Among Adults Aged 40 Years and Older: A Pilot Study.
Neural correlates of information gain during search revealed by fMRI.

J Am Psychoanal Assoc. 2026 Sep 16. 30651261477324

What I'm Reading Now: Stories that Bind Us Together: Why Children's Literature Matters for Addiction Work.

Linda C Mayes.

DOI: https://doi.org/10.1177/00030651261477324
Public Libr Q (N Y). 2026 Jul 15.

Development, Implementation, and Evaluation of a Proof of Concept, Food- and Nutrition-Focused Science Education Program for Youth in Public Libraries.

David H Holben, Melani W Duffrin.

  A four-lesson, hands-on, food- and nutrition-focused science education program was developed for a rural public library as proof of concept for science education in public libraries. Backpacks with materials, supplies, books, and evaluation postcards were provided. This manuscript summarizes program development, implementation, and evaluation, including perceived science learning gains and experiences of youth participants in kindergarten through eighth grade. This program improves perceived science learning gains of and provides positively-perceived science experiences for participants. Public libraries have the potential to improve accessibility of U.S. children to informal science education to potentially elevate both science literacy and good health.

Keywords:  STEM education; food; nutrition; public library; science; science education; science learning

DOI:  https://doi.org/10.1080/01616846.2026.2700934
Plant J. 2026 Sep;127(5): e71117

FatPlants 2.0: an AI-powered platform integrating plant lipid genes, pathways, and literature.

Yongfang Qin, Congyu Guo, Minseong Kim, Yichuan Zhang, Muhammad Azam, Tingyuan Xiao, Chunhui Xu, Basil Shorrosh, Timothy P Durrett, Doug K Allen, Trupti Joshi, Edgar B Cahoon, Jay J Thelen, Dong Xu.

  Plant lipid research depends on accessible databases that connect genes, pathways, and prior literature. However, currently available resources are fragmented, species-limited, and lack AI-powered interfaces for integrated querying. FatPlants 2.0 addresses these issues by integrating ARALIP, PlantFADB, and new Cuphea/Pennycress experimental data into a unified platform with 14 000 genes/proteins, 110 pathways across 5 species, and 57 000+ curated publications. The platform features LipidBot, an AI agent enabling natural language queries via graph-based pathway search and retrieval-augmented generation-powered literature retrieval. Users query complex relationships conversationally and receive answers with traceable citations. The graph database models 100+ biological pathways as queryable networks with large language model-guided Cypher generation. By evaluating more than 1000 curated questions, LipidBot achieved 95% accuracy on pathway queries and 92% recall in literature retrieval using optimized embeddings. The tool demonstrated robust performance across factual, numerical, and multi-hop queries on curated benchmarks. FatPlants 2.0 accelerates research by reducing the time spent on literature reviews. Database and AI agent freely available at https://fatplants.net with bulk downloads and quarterly updates.

Keywords:  AI chatbot; FatPlants; database; large language model; pathway analysis; plant lipids; retrieval‐augmented generation (RAG)

DOI:  https://doi.org/10.1111/tpj.71117
iScience. 2026 Sep 18. 29(9): 117438

Systematic assessment of text summarization methods for biomedical literature from frequency methods to language models.

Fabio Baumgärtel, Enrico Bono, Lucas Fillinger, Louiza Galou, Kinga Kęska-Izworska, Samuel M Walter, Peter Andorfer, Klaus Kratochwill, Paul Perco, Matthias Ley.

  The rapid expansion of biomedical literature demands automated summarization tools that reliably condense research articles into concise, accurate summaries. We benchmarked 62 summarization methods, ranging from frequency-based and TextRank extractors to encoder-decoder models (EDMs) and large language models (LLMs), on 1,000 biomedical abstracts from 20 journals across ScienceDirect and Cell Press, using author-written highlights as reference summaries. Models were evaluated with a composite suite of lexical, semantic, and factual metrics, including ROUGE, BLEU, METEOR, embedding-based similarity, and factuality scores. General-purpose models (e.g., Mistral, GPT, and Llama) achieved the highest overall performance across lexical and semantic dimensions, outperforming reasoning-oriented (e.g., DeepSeek and Magistral) and domain-specific (e.g., BioGPT and BioMistral) models. Notably, medium-sized models outperformed large-scale models, suggesting an optimal balance between model capacity and efficiency, while classical extractive methods lagged behind neural approaches. These findings provide a systematic reference for selecting biomedical summarization tools and highlight that broad pretraining outperforms narrow domain adaptation.

Keywords:  benchmarking; biomedical text summarization; large language models; natural language processing

DOI:  https://doi.org/10.1016/j.isci.2026.117438
Stud Health Technol Inform. 2026 Sep 17. 340 251-256

Semantic Retrieval of Projects Funded by the Innovation Committee.

Christopher Gundler, Mareile Beernink, Michael Blaß, Alexander Johannes Wiederhold.

   INTRODUCTION: Evidence-based development of healthcare innovations requires efficient access to existing knowledge. The Innovation Committee of the Federal Joint Committee (Gemeinsamer Bundesausschuss, G-BA) has funded more than 800 projects since 2016, yet the heterogeneity of topics and terminology limits the effectiveness of the current keyword-based search interface.
METHODS: To address vocabulary-mismatch problems and support more nuanced information needs, we developed and evaluated a semantic search tool using vector-based natural-language retrieval. The application embeds project descriptions and user queries and computes similarity directly in the browser, enabling privacy-preserving, offline operation. In a controlled comparative study with three experienced users, we assessed efficiency and effectiveness across four standardized search tasks.
RESULTS: Although based on a small sample size, the results indicate that semantic retrieval can meaningfully improve the discovery of projects funded by the Innovation Committee and support evidence-based research workflows.
CONCLUSION: Future work should validate these findings in larger cohorts and explore enhanced ranking strategies.

Keywords:  Federal Joint Committee; G-BA; Innovation Committee; Semantic search; evidence-based decision-making; health services research; information retrieval

DOI:  https://doi.org/10.3233/SHTI261010
JMIR AI. 2026 Sep 16. 5 e93761

Retrieve-Then-Verify for Evaluating Evidence Support and Hallucination in Large Language Model-Generated Medical Information: Empirical Study.

Zhaohui Liang, Cynthia Sheffield, Gisela Butera, Sameer Antani.

   BACKGROUND: Despite high reported accuracy on clinical and evidence appraisal tasks, AI-generated medical information may lack explicit support from source documents. This creates challenges for digital health practitioners regarding transparency, auditability, and trust when AI systems are used for evidence synthesis, guideline development, and clinical knowledge management. Large language models (LLMs) can generate fluent and seemingly correct outputs, but existing evaluations often rely on agreement with human judgments and do not directly assess whether AI-generated content is grounded in underlying evidence.
OBJECTIVE: This study measures evidence support and hallucination in AI-generated medical information by assessing the extent to which LLM-generated risk-of-bias assessments are supported by source clinical trial reports.
METHODS: We evaluated 3 LLMs (GPT-5, OpenAI o3-mini, and GPT-3.5) on risk-of-bias (RoB 2) assessment using all 97 randomized controlled trials for which the source Cochrane systematic review provided complete human RoB 2 annotations and full-text reports were accessible, constituting the complete reference set. No train-validation split was applied; all 97 studies were used for evaluation. Model outputs were constrained to structured RoB 2 signaling questions and domain-level judgments. For each generated claim, relevant text passages were retrieved from trial reports using the Okapi BM25 (Best Matching 25) algorithm. A verification step assigned evidence verdicts (supported, contradicted, not found, or out of scope) with verbatim quotations. We quantified evidence support rates and conservative and strict hallucination rates. Task performance was evaluated using exact and binary accuracy, sensitivity, specificity, F1-score, Youden J, and agreement with human reviewers using Cohen κ and Fleiss κ.
RESULTS: Binary accuracy of AI-generated risk-of-bias judgments was high across domains (90%-98%), whereas exact accuracy was substantially lower (42%-71%), reflecting frequent disagreements in severity classification despite correct directional classification. GPT-5 achieved the strongest overall performance, including perfect binary accuracy for overall risk-of-bias conclusions and the highest agreement with human reviewers (quadratic κ up to 0.81). However, evidence support rates across models ranged from only 60% to 65%, with conservative hallucination rates of 34%-37%. GPT-5 showed the highest mean evidence support (64.3%) and the lowest strict hallucination rate (35.7%). Mean top-1 BM25 retrieval scores were similar across models (approximately 30-31), suggesting that differences in hallucination were not primarily attributable to differences in retrieval strength.
CONCLUSIONS: AI-generated medical information can achieve high decision-level accuracy while still lacking documentary support in a substantial proportion of outputs. Measuring evidence support and hallucination reveals important limitations that are not captured by agreement metrics alone. Retrieval-based evidence verification provides a reproducible and transparent approach for evaluating the reliability of AI-generated medical information, with direct relevance to digital health practice, evidence-based medicine, and medical informatics.

Keywords:  AI; digital health; evidence-based medicine; hallucinations; information quality; information retrieval; large language models; medical informatics; risk of bias; systematic review

DOI:  https://doi.org/10.2196/93761
Healthcare (Basel). 2026 Sep 01. pii: 2797. [Epub ahead of print]14(17):

Information Quality and Semantic Stability of AI Chatbot Responses to Complete Denture Questions: A Turkish-English Benchmarking Study.

Sinan Coşkun, Fatma Nur Karaman, Hüseyin Ardıl Uytun, Gülcan Coşkun Akar.

  Background/Objectives: Artificial intelligence (AI)-based chatbots are increasingly used as sources of health information, yet the quality and semantic stability of their responses may vary across systems, languages, repeated queries, and prompt formulations. This study compared expert-rated general information quality and embedding-based semantic stability of responses generated by seven AI chatbot systems to expert-derived, patient-oriented complete denture questions in Turkish and English. Methods: Twenty-two complete denture-related questions were submitted to seven chatbot systems on three study days across scheduled morning, afternoon, and evening sessions. Two additional semantically equivalent Turkish variants were generated for each original question. Five prosthodontists assessed the original Turkish and English responses using the 5-point Global Quality Score (GQS). Semantic stability was quantified using the multilingual SentenceTransformer checkpoint paraphrase-multilingual-MiniLM-L12-v2 and cosine similarity. To incorporate complete responses, responses were divided into non-overlapping token-based chunks, chunk embeddings were combined into one normalized full-response representation, and the 36 pairwise similarities from the nine repeated responses were averaged to one stability estimate per question-model-language/variant condition. Linear mixed-effects models with question-level clustering were used, with Bonferroni adjustment and partial eta squared effect sizes with 95% confidence intervals. Results: AI model significantly affected embedding-based semantic stability (p < 0.001; ηp2 = 0.820, 95% CI 0.795-0.837) and GQS (p < 0.001; ηp2 = 0.249, 95% CI 0.152-0.316). Grok had the numerically highest overall semantic stability mean (0.952 ± 0.015), whereas GPT-4o had the numerically highest overall GQS (4.48 ± 0.85). English responses showed higher overall GQS than the original Turkish responses (3.96 ± 0.94 vs. 3.66 ± 1.23; p = 0.004) and higher embedding-based similarity than the Turkish conditions overall (p < 0.001). No significant overall effect of scheduled study day was observed (p = 0.504; ηp2 = 0.001), whereas semantic stability differed across scheduled query sessions (p < 0.001; ηp2 = 0.055), with means of 0.863 ± 0.106, 0.874 ± 0.092, and 0.841 ± 0.140 for morning, afternoon, and evening sessions, respectively. Inter-rater reliability was moderate for English GQS ratings (ICC = 0.675, 95% CI 0.581-0.752) and good for Turkish ratings (ICC = 0.771, 95% CI 0.687-0.833). Conclusions: The evaluated chatbot systems differed in expert-rated information quality and full-response embedding-based semantic stability. English responses showed higher overall values than Turkish responses, although cross-language measurement effects of the embedding model cannot be excluded. Differences across scheduled query sessions should be interpreted as run-to-run output variability rather than intrinsic temporal behavior. These findings characterize comparative chatbot performance under the tested conditions but do not establish clinical accuracy, safety, or patient education effectiveness.

Keywords:  artificial intelligence; chatbot; complete denture; digital health; health information; large language model; patient education; prosthodontics; semantic similarity

DOI:  https://doi.org/10.3390/healthcare14172797
Medicine (Baltimore). 2026 Sep 18. 105(38): e50664

How appropriate are ChatGPT-5's answers to acne questions?: A category-based evaluation by dermatologists.

Nazli Caf, Defne Özkoca, Tuğba Kevser Üstünbaş Uzunçakmak, Zekayi Kutlubay.

  Online conversational tools have become increasingly common sources of health information, with many patients seeking acne advice through such platforms. Although they may enhance awareness and encourage early consultation, their medical reliability remains uncertain. Because acne is one of the most frequently searched dermatologic conditions, verifying the accuracy of automated responses is crucial to prevent misinformation. This study aimed to evaluate ChatGPT-5 responses to frequently asked acne questions in dermatology practice and to determine category-specific strengths and limitations based on blinded expert assessment. The analysis focused on identifying error patterns with potential implications for patient education and clinical safety. Twenty-five commonly asked patient questions were classified into 4 categories: skin care routine, nutrition and lifestyle, diagnosis and general information, and treatment and medications. Each question was entered into the publicly available ChatGPT-5 interface in 3 independent sessions, yielding 75 unique responses. Three dermatologists independently rated each response as "appropriate" or "inappropriate," and reasons for rejection were categorized using a predefined 9-item list. Of 75 responses, 49 (65.3%) were appropriate. Appropriateness was highest for skin care (83.3%) and nutrition (77.8%) and lowest for treatment (50.0%) and diagnosis (33.3%). Inter-rater agreement was low. The most frequent reasons for inappropriateness were misleading information and overly technical language, followed by lack of risk disclosure. ChatGPT-5 may provide helpful introductory information for patients; however, diagnostic and treatment guidance often lacked clarity and safety detail, underscoring the need for dermatologist oversight.

Keywords:  acne vulgaris; artificial intelligence; dermatology; patient education

DOI:  https://doi.org/10.1097/MD.0000000000050664
Front Oncol. 2026 ;16 1909227

Performance evaluation of large language models in bladder cancer patient education Q&A: a cross-sectional study.

Dian Wan, Youwen Li, Zheng Dong, Chen Dai, Jinghe Ye, Song Li, Sunlu Jiang.

   Background: Bladder cancer ranks among the most prevalent urological tumors worldwide, with its global incidence continuing to rise steadily. Although patient education materials (PEMs) play a crucial role in enhancing disease comprehension and supporting joint clinical decision-making, current online resources frequently surpass the readability thresholds recommended for the general public. Large language models (LLMs) hold promise for health communication, yet no systematic assessment has been conducted regarding their feasibility and trustworthiness specifically for bladder cancer patient education.
Objective: This study aimed to systematically benchmark five leading LLMs in producing question-and-answer content for bladder cancer science popularization, with a particular focus on readability, informational quality, and appropriateness for patient education.
Methods: In this cross-sectional simulation study, 20 common patient questions covering five disease domains were compiled. On January 15, 2026, each question was submitted identically to five publicly available LLMs (Doubao, DeepSeek, Kimi, Gemini, and ChatGPT). Readability was evaluated using seven conventional metrics. Two independent pharmacists, blinded to model identity, rated the responses using the Chinese version of the Patient Education Materials Assessment Tool for print materials (C-PEMAT-P) and the Global Quality Score (GQS). Additionally, two independent clinical specialists assessed factual accuracy and alignment with the Chinese Bladder Cancer Diagnosis and Treatment Guidelines (2024 edition) employing a 4-point scale. Cohen's kappa was used to determine inter-rater reliability.
Results: ChatGPT, DeepSeek, and Doubao outperformed Kimi and Gemini on both C-PEMAT and GQS (all P < 0.001), indicating superior understandability, actionability, and overall quality. Across all models, median C-PEMAT scores ranged from 8 to 10, suggesting broadly acceptable suitability for patient education. Readability varied significantly by content domain, with treatment-oriented texts showing the highest complexity. ChatGPT achieved the best alignment with clinical guidelines. No model produced harmful advice or directly contradicted guideline recommendations. Traditional readability measures correlated weakly with GQS, whereas C-PEMAT showed a moderate positive correlation (r = 0.34).
Conclusion: Current mainstream LLMs demonstrate initial potential for generating educational content on bladder cancer, albeit with considerable heterogeneity across models. Disease-specific evaluation instruments for patient education materials are more effective than general readability formulas in reflecting perceived quality. Our results advocate for a prudent, assistive role of LLMs in health communication under a human-AI collaborative model.

Keywords:  artificial intelligence; bladder cancer; large language models; patient education; readability

DOI:  https://doi.org/10.3389/fonc.2026.1909227
Ther Adv Respir Dis. 2026 Jan-Dec;20:20 17534666261464435

Artificial intelligence responses to bronchiectasis patients' questions: the AIR-BE study.

AIR-BE Task Force

   BACKGROUND: The popularity of Generative Artificial Intelligence (AI)-powered chatbots is growing, with more patients using AI to understand respiratory conditions, including bronchiectasis. The quality of AI's responses to bronchiectasis-related questions has not previously been evaluated.
OBJECTIVES: To evaluate the reliability, accuracy, comprehensiveness, and understandability of responses generated by different AI-based chatbots to bronchiectasis-related questions formulated by patients.
DESIGN: Cross-sectional international study.
METHODS: People living with bronchiectasis from the EMBARC/European Lung Foundation (ELF) patient advisory group formulated 15 bronchiectasis-related questions, categorised into 3 difficulty tiers. These questions were submitted to three AI-based chatbots: ChatGPT, Bard, and Copilot. An international group of 28 experts from the European Respiratory Societies (recruited through Assemblies and the CONNECT clinical research collaboration) and 33 patients evaluated the answers' reliability, accuracy, comprehensiveness, and understandability.
RESULTS: Of 45 outcomes, 37 were deemed reliable. Median accuracy scores ranged from 6.0 to 9.0, with ChatGPT scoring higher than Copilot and Bard (8.0 [7.0-9.0] vs 7.0 [6.0-8.0] vs 7.0 [6.0-8.0], p-value < 0.001). Median comprehensiveness scores ranged from 6.0 to 9.0, with ChatGPT providing more comprehensive answers compared to Bard and Copilot (8.0 [7.0-9.0] vs 8.0 [6.0-9.0] vs 7.0 [6.0-8.0], p-value < 0.001). Median understandability scores ranged from 7.0 to 10.0, with ChatGPT and Bard delivering more understandable answers than Copilot (9.0 [8.0-10.0] vs 9.0 [8.0-10.0] vs 9.0 [7.0-10.0], p-value < 0.001). Differences in understandability were noted in Bard's responses across difficulty tiers (easy 9.0 [8.0-10.0] vs medium 9.0 [7.0-10.0] vs difficult 8.0 [7.0-10.0], p-value = 0.012).
CONCLUSION: AI software, particularly ChatGPT, provided accurate, comprehensive, and understandable answers to our 15 bronchiectasis-related questions.

Keywords:  artificial intelligence; bronchiectasis; disease awareness; patient education

DOI:  https://doi.org/10.1177/17534666261464435
Front Public Health. 2026 ;14 1926980

Safety, accuracy, empathy, information quality, and readability of publicly accessible LLM-based chatbots for traumatic brain injury and concussion questions: a cross-sectional comparative study.

Xin Zuo, Huan Zuo, Min Zhang, Weihong Zheng.

   Background: Large language model (LLM)-based chatbots are increasingly used by the public to obtain health information, but their performance in answering questions related to traumatic brain injury (TBI) and concussion remains unclear. This study evaluated five publicly accessible LLM-based chatbots across safety, accuracy, empathy, information quality, transparency, and readability.
Methods: Sixty-five English-language, public-facing questions about TBI and concussion were submitted to ChatGPT, Gemini, Copilot, DeepSeek, and Doubao, generating 325 responses. Five blinded independent raters assessed the generated responses using guideline-informed criteria and established tools, including DISCERN, EQIP, JAMA benchmark criteria, the Global Quality Score, and readability indices.
Results: Inter-rater agreement was high. The proportions of responses rated as safe ranged from 89.2 to 95.4%, and all models achieved a median accuracy score of 4.00. However, 27 responses were classified as potentially harmful, mainly because of under-triage, overly reassuring advice regarding imaging findings, premature return-to-activity or driving guidance, and insufficient pediatric caution. Accuracy differences were statistically significant but small, whereas empathy, information quality, transparency, and readability showed clearer model-level variation. DeepSeek produced the easiest-to-read responses.
Conclusion: LLM-based chatbots generated responses that were generally rated as safe and informative by expert evaluators, but potentially harmful advice and readability problems remained. These findings characterize expert-rated response performance and do not establish patient comprehension, educational effectiveness, or clinical benefit. LLM-based chatbots may have potential as supplementary sources of patient-facing health information, but they should not replace professional medical evaluation.

Keywords:  chatbots; concussion; large language models; patient education; traumatic brain injury

DOI:  https://doi.org/10.3389/fpubh.2026.1926980
Digit Health. 2026 Jan-Dec;12:12 20552076261491151

Assessing the accuracy and readability of large language models in answering Turkish-language patient-generated queries on radiological contrast agents.

Abdullah Enes Ataş.

   Objective: To assess the clinical accuracy, structural readability, and Ensuring Quality Information for Patients (EQIP-36) quality of Large Language Model (LLM) responses to real patient queries regarding radiological contrast agents and to define their safe boundaries in practice.
Methods: The 30 most popular Turkish patient questions reflecting fears and risk perceptions about contrast agents were identified via Google Trends and AlsoAsked.com. These queries were presented to ChatGPT-5.4, Gemini 3, and Claude 4.6 Sonnet using a zero-shot learning approach. Responses were evaluated by experts for clinical accuracy against the 2024/2025 American College of Radiology (ACR) Manual on Contrast Media. Information quality and readability were measured using the expanded EQIP-36 scale and the Ateşman and Bezirci-Yılmaz indices.
Results: Claude 4.6 Sonnet demonstrated the highest clinical accuracy and high compliance with guidelines (>90%). Gemini 3 exhibited a more conservative stance, occasionally resulting in over-triage by exaggerating risks. While all models scored exceptionally well in the Content and Structure dimensions of the EQIP-36, generating readable texts, they universally failed in the Identification dimension by omitting author names, update dates, or bibliographic sources. Readability indices indicated that comprehending the texts required an average of 10.5 to 11.6 years of formal education.
Conclusion: LLMs generate highly readable and subjectively reassuring Turkish-language texts in response to patient queries regarding radiological contrast agents. However, due to domain-specific structural gaps regarding quantitative thresholds and lack of citations, they should be positioned as hybrid communication tools subject to mandatory review by specialist physicians rather than standalone medical advisors.

Keywords:  artificial intelligence; community health education; contrast media; large language models; patient education; radiology

DOI:  https://doi.org/10.1177/20552076261491151
Clin Adv Periodontics. 2026 Sep 16.

Assessing the accuracy and usability of artificial intelligence-based language models in responding to common periodontal patient questions.

Sajad Jahantigh, Reza Amid, Anahita Moscowchi, Fazele Atarbashi-Moghadam, Alireza Akbarzadeh Baghban.

   BACKGROUND: Artificial intelligence-powered large language models (LLMs) are increasingly used by patients seeking quick information regarding dental and medical problems. Despite their growing popularity, concerns remain regarding the accuracy, clarity, and clinical usefulness of LLM-generated responses. This study aimed to comparatively evaluate the accuracy and usability of widely used LLMs in responding to frequently asked periodontal questions.
METHODS: In this analytical-comparative study, 15 commonly asked periodontal questions were selected from real patient encounters and administered to ten LLMs in both English and Persian. Responses were evaluated independently and blindly by two board-certified periodontists; disagreements were resolved by discussion or third-reviewer. Each response was rated on a 5-point Likert scale across six criteria: correlation with the question, adequacy, comprehensiveness, clarity/readability, usability for individuals with limited scientific literacy, and scientific accuracy. Statistical analyses included descriptive statistics, Mann-Whitney U test, Kruskal-Wallis test with Bonferroni correction, and univariate ANOVA. Significance level was set at p < 0.05.
RESULTS: Significant differences were observed among LLMS (χ2 = 294.78, p < 0.001). ChatGPT-4.5 with deep search enabled achieved the highest scores across most criteria, while ChatGPT-4o generally underperformed compared with other LLMs. English responses scored significantly higher than Persian (mean 4.25 vs. 3.96; U = 333,058, p < 0.001). Among evaluated criteria, correlation had the highest mean score (4.81), whereas usability for individuals with limited scientific literacy had the lowest score (3.80). LLMs performed significantly better on treatment/prevention questions than on diagnosis/pathogenesis questions. Only ChatGPT-4.5 (deep search) and Grok-3 (deep search) provided citations, with ChatGPT-4.5 referencing more specialized dental sources.
CONCLUSIONS: LLMs show variable performance in answering frequent periodontal patients' questions, influenced by model architecture, language, question type, and evaluative domain. Although advanced configurations such as deep-search ChatGPT-4.5 offer highly accurate, comprehensive, and well-referenced responses, significant limitations remain particularly in Persian language output and in usability for patients with limited scientific literacy. LLMs may serve as complementary tools for patient communication, but should not replace professional advice. Further development and language-specific optimization are needed to improve global accessibility and reliability.
KEY POINTS: While AI tools provide highly readable and generally accurate periodontal information, they often lack the clinical nuance required for personalized risk assessment and complex treatment planning. Practitioners should proactively discuss AI-generated health information with patients to address potential oversimplifications and ensure online advice is integrated into a professional, evidence-based care plan. As patients increasingly rely on AI for initial health guidance, practitioners must act as the authoritative "filter" to verify that online information remains clinically safe, accurate, and aligned with the patient's specific treatment goals.
PLAIN LANGUAGE SUMMARY: Many people now turn to artificial intelligence for help with their dental questions. This study found that these tools vary in how well they answer questions about gum disease. While the most advanced models (such as ChatGPT-4.5) provide accurate, high-quality information, there are two main limitations: first, performance is significantly better in English than in Persian; second, many explanations are too technical and hard for the average person to understand. Overall, while AI can be a helpful guide, it should never replace a professional consultation with a dentist.

Keywords:  artificial intelligence; large language models; natural language processing; periodontal diseases

DOI:  https://doi.org/10.1002/cap.70105
Eur Arch Otorhinolaryngol. 2026 Sep 16.

Large language models for tympanostomy patient education: readability and guideline adherence.

Shreeya Bahethi, Hetal Lad, Shrey Shah, Sudeepti Vedula, Brian Manzi.

   PURPOSE: To evaluate the accuracy and readability of large language model (LLM)-generated patient education materials regarding tympanostomy tube placement.
METHODS: Over a two-month period, ChatGPT 4o, Gemini 2.5 Flash, and Google Search AI were prompted daily using long-form and layered prompt formats covering typical concerns regarding tympanostomy. Responses were scored on a 12-point rubric adapted from the AAO-HNS Clinical Practice Guidelines (CPG), assessing diagnostic accuracy, procedural clarity, and postoperative care. Readability was evaluated using Flesch Reading Ease and Flesch-Kincaid Grade Level. For each model and prompt type, average scores and variability were analyzed with 95% confidence intervals. Between-model differences were tested with Welch's t-tests and Cohen's d; temporal trends were analyzed via linear regression.
RESULTS: For long-form outputs, Google Search AI and Gemini 2.5 Flash demonstrated the highest mean CPG adherence (96.4% and 96.6%, respectively), both significantly exceeding ChatGPT 4o (84.3%; both P < .001; Cohen's d ≈ 2.05). For layered prompt sessions, Google Search AI again demonstrated the highest adherence (91.7%), followed by Gemini 2.5 Flash (88.1%) and ChatGPT 4o (76.2%). Guideline adherence was temporally stable across all models (P > .05 for most). All outputs exceeded the recommended sixth-grade reading threshold (mean FKGL, 9.4 for long-form; 10.6 for layered), with no statistically significant readability differences between prompt strategies.
CONCLUSIONS: Google Search AI demonstrated the highest guideline concordance, though all models produced material too complex for typical patient comprehension. Structured prompting enhances clinical accuracy, but readability remains a key barrier to accessibility.

Keywords:  Artificial intelligence; Clinical practice guidelines; Large language models; Patient education; Readability; Tympanostomy

DOI:  https://doi.org/10.1007/s00405-026-10612-2
Healthcare (Basel). 2026 Aug 24. pii: 2681. [Epub ahead of print]14(17):

Evaluating ChatGPT's Effectiveness for Arabic Dry Mouth Patient Education.

Abdullah Mohamed Alsoghier.

  Background/Objectives: The present study aimed to assess the understandability and actionability of Arabic text generated by a large language model for commonly searched Arabic queries on dry mouth. Methods: Using Google Trends, the top 10 searches worldwide related to 'oral dryness' were entered in OpenAI's Generative Pretrained Transformer 5.1. Generated texts were achieved independently. Assessments were performed using the Patient Education Materials Assessment Tool (PEMAT) to evaluate the content, word choice, and style. Results: Causes, symptoms, and treatment of dry mouth were the most common dry-mouth-related queries. The highly temporal distribution of search interests among Arabic-speaking countries peaked between 2020 and 2021, then remained high throughout 2023, before declining in November 2025. The mean PEMAT understandability and actionability scores were 89% and 80%, respectively. It was notable that all generated responses lacked visual aids, which could have made the content difficult to understand and insufficient for acting on the information. Moreover, the formal Arabic form of 'causes of dry mouth' with a glottal stop yielded lower actionability scores (60%) than the informal Arabic form (80%). Conclusions: Clinicians could actively supplement clinic-based discussions with advice on using large language models to help patients recognise dry mouth symptoms and improve self-care. Also, they could improve their effective adoption by clearly explaining expectations, limitations, and language/cultural differences when adopting these models.

Keywords:  generative artificial intelligence; health literacy; oral medicine; patient education; salivary gland diseases

DOI:  https://doi.org/10.3390/healthcare14172681
Healthcare (Basel). 2026 Aug 24. pii: 2694. [Epub ahead of print]14(17):

Evaluation and Comparison of Large Language Model Responses to Frequently Asked Questions Regarding Patellofemoral Pain Syndrome: A Quality and Readability Assessment Study.

Oktay Polat, Berk Koncalıoğlu, Mert Gündoğdu, Emrecan Akgün.

  Background: Patellofemoral pain syndrome (PFPS) is a common cause of anterior knee pain, and patients increasingly use large language models (LLMs) to obtain general medical information. However, the quality, reliability, and readability of LLM-generated responses to patient-oriented questions regarding PFPS remain uncertain. This study aimed to compare responses generated by four widely used LLMs. Methods: Seventeen frequently asked questions regarding PFPS were identified through Google searches and adapted into lay language. The questions were submitted to OpenAI GPT-5, Google Gemini 2.5 Pro, xAI Grok 4, and DeepSeek-V3.2-Exp using a standardized patient scenario. A total of 68 question-specific responses were independently evaluated by four orthopedic surgeons using the DISCERN instrument. Inter-rater reliability was assessed using the intraclass correlation coefficient. Readability was evaluated using the Gunning Fog Index, Coleman-Liau Index, and Flesch Reading Ease Score. Between-model comparisons were performed using the Friedman test, followed by Bonferroni-adjusted pairwise analyses. Results: The omnibus Friedman test showed a significant between-model difference in DISCERN scores (p = 0.002). In Bonferroni-adjusted pairwise comparisons, GPT-5 had lower DISCERN scores than Gemini 2.5 Pro (adjusted p = 0.006), Grok 4 (adjusted p = 0.021), and DeepSeek-V3.2-Exp (adjusted p = 0.036), whereas no significant differences were observed among the other three models. However, the absolute differences were small, and the between-model difference was not significant in the sensitivity analysis using the median evaluator score (p = 0.381). Inter-rater agreement was moderate for GPT-5 and DeepSeek-V3.2-Exp but poor for Gemini 2.5 Pro and Grok 4. Readability differed significantly among the models across all three indices. DeepSeek-V3.2-Exp generally showed more favorable numerical readability values, whereas Grok 4 tended to produce more difficult text; however, no model was consistently superior across all readability measures. The median Gunning Fog and Coleman-Liau scores for all four models exceeded the commonly recommended sixth- to eighth-grade reading level for patient education. Conclusions: The evaluated LLMs showed small and method-dependent differences in DISCERN-based information quality and variable differences in readability. Their responses may supplement general patient education, but the findings should not be interpreted as evidence of factual accuracy, clinical safety, or suitability for individualized decision-making. LLM-generated information should be critically reviewed and should not replace assessment by a qualified healthcare professional.

Keywords:  ChatGPT; DISCERN; Gemini; artificial intelligence; large language models; patellofemoral pain syndrome; patient education; readability

DOI:  https://doi.org/10.3390/healthcare14172694
Front Pediatr. 2026 ;14 1919701

Comparing Chinese-language large language models for caregiver questions about developmental dysplasia of the hip.

Fangyuan Wang, Xianhong Li, Yinghui Guan, Jiaqi Tian, Le Xu, Bing Huang, Jianhui Xie.

   Background: Large language models (LLMs) are increasingly used for caregiver-facing health information, but their reliability in Chinese-language pediatric orthopaedics remains uncertain. This study evaluated whether responses to developmental dysplasia of the hip (DDH) questions were clinically accurate, aligned with Chinese guidance, and educationally usable.
Methods: We compared ChatGPT-4o and DeepSeek-R1 using 53 Chinese-language DDH questions, including 31 caregiver-oriented frequently asked questions and 22 guideline-derived items. Each model generated one response per question using a standardized single-turn, five-sentence prompt. Six blinded pediatric orthopaedic surgeons rated clinical accuracy and guideline concordance. Paired model comparisons, inter-rater reliability, and exploratory formula-based readability were assessed.
Results: DeepSeek-R1 had higher clinical accuracy than ChatGPT-4o across 31 caregiver-oriented questions (mean question-level score 4.70 [SD 0.15] vs. 3.69 [0.28]; P < 0.001) and higher guideline concordance across 22 items (4.44 [0.22] vs. 3.82 [0.33]; P < 0.001). Average-rating reliability was good for clinical accuracy (ICC(2,6) = 0.826, 95% bootstrap CI 0.770-0.863) and moderate for guideline concordance (ICC(2,6) = 0.519, 95% CI 0.377-0.621). Formula-based readability findings were mixed: unadjusted tests indicated lower FKGL, FOG, and SMOG values for ChatGPT-4o, but only FOG and SMOG remained statistically significant after Holm adjustment. Surgery-related observations were exploratory and descriptive.
Conclusions: Under the tested single-query and five-sentence conditions, DeepSeek-R1 achieved higher expert-rated clinical accuracy and Chinese-guideline concordance. The findings describe specific model-platform configurations rather than a permanent model ranking. Professionally reviewed LLM responses may help clinicians reinforce routine DDH education, but individualized diagnostic and treatment guidance, especially for surgery-related questions, should remain clinician-led. Caregivers should use chatbot information only as a supplementary resource.

Keywords:  ChatGPT-4o; Chinese-language health information; DeepSeek-R1; caregiver education; developmental dysplasia of the hip; guideline concordance; large language models; patient education

DOI:  https://doi.org/10.3389/fped.2026.1919701
Front Public Health. 2026 ;14 1942799

A comparative evaluation of generative AI chatbots for patient-oriented advice on painful diabetic peripheral neuropathy: safety, accuracy, guideline concordance, actionability, and readability.

Lina Gu, Zhaole Gong, Xiaoli Qian, Ling Miao, Zhengfeng Gu.

   Background: Patients with painful diabetic peripheral neuropathy (PDPN) increasingly use generative artificial intelligence chatbots for information on symptoms, treatment, foot care, and when to seek professional help. Their usefulness depends on safety, accuracy, guideline concordance, actionability, and readability.
Objective: To compare five publicly accessible generative AI chatbots in answering standardized patient-oriented questions about PDPN. Methods: Sixty standardized English-language questions covering eight clinical domains were submitted once to ChatGPT, Gemini, Microsoft Copilot, DeepSeek, and Doubao in separate single-turn conversations, yielding 300 responses. Five reviewers independently assessed safety, accuracy, guideline concordance, and actionability using predefined criteria. Guideline concordance was scored against six mapped elements per question and converted to a percentage. Actionability was assessed using seven binary criteria with prespecified question-level applicability. Readability was evaluated using six established indices. Paired comparisons used Cochran's Q test for safety and Friedman tests for non-binary outcomes, followed by multiplicity-adjusted pairwise analyses.
Results: All 300 responses were analyzed. Inter-rater agreement was high for safety (Fleiss' κ = 0.874), accuracy [ICC (2,1) = 0.881], guideline concordance [ICC (2,1) = 0.874], and actionability [ICC (2,1) = 0.877]. Twenty-eight responses (9.3%) were classified as unsafe or potentially unsafe. Unsafe-response rates ranged from 5.0 to 15.0%, with no detected overall between-model difference (Cochran's Q = 4.462, p = 0.347). Accuracy, guideline concordance, and actionability differed across models (all p < 0.001; Kendall's W = 0.683, 0.730, and 0.556, respectively). ChatGPT generally achieved higher content-related scores, whereas Doubao scored lower. All readability indices also differed across models (all p < 0.001), with ChatGPT and Doubao producing less complex text and DeepSeek showing greater reading difficulty.
Conclusion: The five chatbots showed distinct performance patterns across content quality and readability. Although no overall safety difference was detected, every system generated at least one response with a plausible pathway to inappropriate self-management, delayed assessment, medication or product misuse, or preventable injury. Chatbots may support general patient education, but medication decisions, foot-risk assessment, and urgent-care triage require professional verification.

Keywords:  actionability; chatbot; generative artificial intelligence; guideline concordance; painful diabetic peripheral neuropathy; readability; safety

DOI:  https://doi.org/10.3389/fpubh.2026.1942799
J Paediatr Child Health. 2026 Sep 14.

Quality of Online Information for Management of Leg Pain in Children and Adolescents With Hypermobility-Associated Conditions.

R Quinlan, V Pacey, R Thomas, K Gray, L M Davies.

   AIM: Online health information seeking continues to gain popularity. Quality assessment of online information is particularly important for patient populations with low satisfaction and disengagement from health professionals and services, such as hypermobility-associated conditions. We aimed to assess the quality of online health information, for management of lower limb pain in children with hypermobility-associated conditions, available through search engines and corresponding AI-generated overviews.
METHODS: Webpages were collated from Google searches and the corresponding overviews presented by Gemini, using the search terms 'management of lower limb pain in children with hypermobility-associated conditions'. Webpage quality was assessed using the Health Information Website Evaluation Tool, while the Quality Assessment of Medical Artificial Intelligence was used to evaluate the quality of AI-generated overviews.
RESULTS: 20 webpages and four AI-generated overviews were evaluated. Three webpages were rated as good quality, with 14 webpages rated moderate quality. Three webpages were assigned a different rating (moderate or good) by the two reviewers. Quality of information in the AI-generated overviews, was considered good. The lowest scoring domain for webpages was accuracy, and for AI-generated overviews the completeness and provision of resources and references domains.
CONCLUSIONS: Availability of webpages on the management of lower limb pain in children with hypermobility-associated conditions is limited, with quality rated as moderate to good. All AI overviews were good quality. Quality improvements are possible for webpages and AI overviews. Online health information seekers should review references informing content, particularly when AI-generated, to ensure existence and relevance.

Keywords:  adolescent; child; consumer health information; internet; joint hypermobility; lower extremity; musculoskeletal pain

DOI:  https://doi.org/10.1111/jpc.70578
J Shoulder Elb Arthroplast. 2026 Dec;10(4): 100071

The readability of total elbow arthroplasty educational materials.

Ryan Terrany, William Cusma, Suleiman Sudah, Eric Kholodovsky.

   Background: The purpose of this study was to evaluate the readability of patient education materials related to total elbow arthroplasty (TEA) and to compare readability across clinical practice and general health information websites.
Methods: A Google search using the term "Total Elbow Arthroplasty patient information" was performed in September 2025. The first 25 websites containing written patient educational materials on TEA were included, excluding websites with only graphics, videos, or tables. Websites were categorized as clinical practice or general health information sources. Non-educational content was removed, and the remaining text was then analyzed using 9 validated readability formulas to calculate grade level scores for each source. Comparisons between clinical practice and general health information websites were performed using independent sample t-tests.
Results: Of the 25 patient education materials analyzed, 11 (44%) were from clinical practice websites and 14 (56%) from general health information websites. The mean reading level across all materials was 12.1 ± 2.5, ranging from 7.87 to 17.25. None of the educational materials met the recommended sixth-grade reading level set by the National Institutes of Health, American Medical Association, and the US Department of Health and Human Services. Clinical practice websites demonstrated significantly lower reading levels compared to general health information websites across most formulas (10.4 ± 2.2 vs. 13.4 ± 1.9; P = .002).
Conclusion: Online patient educational materials for TEA are written well above recommended reading levels. Although clinical practice websites are comparatively more readable than general health information sites, both remain above recommended standards. These findings underscore the need for plain language revisions and patient-centered design to improve comprehension and support informed decision-making among TEA candidates.

Keywords:  Clinical practice websites; Health literacy; Online health information; Patient education materials; Readability; Total elbow arthroplasty

DOI:  https://doi.org/10.1016/j.jsea.2026.100071
Front Med (Lausanne). 2026 ;13 1910134

From hematuria warning to precision immunotherapy: an AI-assisted cross-platform analysis of bladder cancer information on social media.

Peng Han, Xin Yang, Qiuxiang Li, Pengcheng Lu, Penghuan Wang, Zhonglei Deng, Zeping Gui, Yiping Zong, Mingyu Liu.

   Objective: Bladder cancer often presents with hematuria, while modern management increasingly depends on pathology-based risk stratification, immunotherapy, and biomarker-informed treatment. This study evaluated whether social media videos adequately connect early hematuria warning with precision bladder cancer care.
Methods: We conducted a cross-sectional, AI-assisted structured extraction and urologist consensus-scored content analysis of publicly available videos from Bilibili, Douyin, and YouTube. Videos were searched on May 17, 2026, using "" for the hematuria module and "" for the bladder cancer module, with relevance ranking. For each module, one predefined keyword was used for video retrieval to maintain a consistent and reproducible search strategy across platforms. DISCERN, GQS, PEMAT, general content completeness, the Bladder Cancer Warning Score (BCWS), and the Bladder Cancer Precision Immunotherapy Score (BC-PIS) were assessed. After rubric calibration, the three reviewer-level BC-PIS scores were averaged per video for the primary quantitative analysis, and videos with a mean BC-PIS ≥11 were classified as adequate.
Results: In the hematuria module, 368 of 400 videos were included. BCWS differed significantly across platforms, with mean scores of 6.59 ± 1.57 for Bilibili, 7.34 ± 1.09 for Douyin, and 7.99 ± 1.65 for YouTube (P < 0.001). Adequate warning information was present in 48/120, 79/99, and 120/149 videos, respectively. In the bladder cancer module, 363 of 395 videos were included. The three-rater mean BC-PIS differed significantly across platforms: 3.28 ± 3.45 for Bilibili, 10.21 ± 2.03 for Douyin, and 7.71 ± 4.75 for YouTube (P < 0.001). Adequate precision immunotherapy information was identified in 7/124, 33/90, and 44/149 videos, respectively. Reliability analysis yielded overall ICC(2,1) values of 0.693 for BCWS and 0.876 for BC-PIS; corresponding Fleiss' kappa values were 0.691 and 0.725. The calibrated Douyin reassessment improved BC-PIS reproducibility to an ICC(2,1) of 0.783 (95% CI, 0.653-0.864) and a Fleiss' kappa of 0.598 (95% CI, 0.459-0.718). All 149 included YouTube hematuria videos were matched by platform-specific sequence number and retained in the BCWS reliability analysis.
Conclusion: Social media videos incompletely connect hematuria warning with pathology-based precision immunotherapy in bladder cancer. Greater professional involvement and platform-level quality control are needed to improve online bladder cancer education.

Keywords:  artificial intelligence; bladder cancer; hematuria; immunotherapy; precision medicine; social media

DOI:  https://doi.org/10.3389/fmed.2026.1910134
J Robot Surg. 2026 Sep 15. pii: 993. [Epub ahead of print]20(1):

Comparison of YouTube videos on single-port and multi-port robot-assisted radical prostatectomy using the PACE score.

Yunus Colakoglu, Arda Meric, Nazim Furkan Gunay.

  This study aimed to compare the PACE-based procedural demonstration quality of YouTube videos demonstrating single-port robot-assisted radical prostatectomy (SP-RARP) and multi-port robot-assisted radical prostatectomy (MP-RARP). A cross-sectional analysis of publicly available YouTube videos was performed between January 1 and January 15, 2026. Videos containing real operative footage of SP-RARP or MP-RARP were included only if all seven PACE domains could be assessed. Selected procedural-segment videos were defined as edited operative videos that still included all PACE steps; videos with missing or non-assessable PACE domains were excluded. Video characteristics, source, procedure type, narration, and on-screen annotations were recorded. Comparisons between groups and correlations with total PACE score were analyzed. A total of 141 videos were included: 43 SP-RARP and 98 MP-RARP videos. MP-RARP videos were significantly longer (60 vs. 28 min, p < 0.001), older (2096 vs. 1095 days, p < 0.001), and received more likes (p = 0.029). Academic or official sources were more common among SP-RARP videos (69.8% vs. 51.0%, p = 0.044). Total PACE scores were similar between groups (25.86 ± 5.48 vs. 26.50 ± 4.37, p = 0.501). SP-RARP videos had higher scores for apical dissection and urethro-vesical anastomosis. In MP-RARP videos, video length correlated positively with total PACE score. SP-RARP and MP-RARP videos showed comparable overall PACE-based demonstration quality, although SP-RARP better demonstrated apical dissection and urethro-vesical anastomosis. Step-based assessment appears more useful than engagement metrics when selecting RARP videos.

Keywords:  Multi-port surgery; PACE score; Robot-assisted radical prostatectomy; Single-port surgery; Surgical education; YouTube

DOI:  https://doi.org/10.1007/s11701-026-03961-x
Transl Androl Urol. 2026 Aug 31. 15(8): 283

Educational value assessment of YouTube surgical videos of bladder hydrodistention for surgeons.

Hai-Rui Li, Jin Li, Tao Liu, Liao Peng, De-Yi Luo.

   Background: The increasing reliance on online surgical learning underscores the need to evaluate the quality and reliability of available educational resources. This study aimed to assess the quality and educational value of YouTube videos related to bladder hydrodistention.
Methods: A cross-sectional analysis was performed on 33 eligible bladder hydrodistention videos, which were searched on YouTube in December 2025. Videos were categorized according to upload source, content type, and region. Video quality, reliability, and educational value were assessed using validated instruments: the Journal of the American Medical Association (JAMA) score (0-4), Global Quality Scale (GQS) score (1-5), Modified Discriminant (M.DISCERN) score (0-5), and Laparoscopic Surgery Video Educational Guidelines (LAP-VEGaS, 0-18). Correlations between video characteristics and assessment scores were statistically analyzed.
Results: Among all included videos, the majority were uploaded by urologists (48.5%). The content category "surgery + information" accounted for the lowest proportion (18.2%), but demonstrated the highest reliability, quality, and educational value (JAMA: 3 vs. 1, P=0.02; M.DISCERN: 4 vs. 1, P=0.001; GQS: 4 vs. 2, P<0.001; LAP-VEGaS: 10.5 vs. 5, P=0.005). Videos with high reliability were significantly longer in duration (P=0.02). Only JAMA scores showed a positive correlation with engagement metrics (P<0.05), whereas GQS and M.DISCERN scores were not significantly associated with these indicators (P>0.05).
Conclusions: Despite the large number of bladder hydrodistention videos uploaded by urologists, their overall quality and educational value are generally suboptimal. Videos combining surgical procedures with patient-oriented information show higher reliability and quality. Our findings highlight the importance of integrating concise summaries, subtitles, and surgical footage to improve the educational quality of such videos for surgeons.

Keywords:  Bladder hydrodistention; YouTube; educational value; social media; video recording

DOI:  https://doi.org/10.21037/tau-2026-0421
Ther Adv Endocrinol Metab. 2026 ;17 20420188261467887

Assessing the quality and reliability of health information on GLP-1 receptor agonists on TikTok and Bilibili: a cross-sectional analysis of the public health infodemic.

Gang Peng, Chen Wang, Hong-Wei Zhang, Ting Xu, Jian-Zhong Di.

   Background: Glucagon-like peptide-1 (GLP-1) receptor agonists have transitioned from clinical tools for diabetes to a global cultural phenomenon for weight loss, triggering an "infodemic." While previous studies have characterized this landscape on Western platforms, the digital health ecosystem in China remains distinct. China is currently navigating a unique market shift driven by the June 2025 approval of the domestic dual agonist mazdutide, an impending 2026 patent cliff for semaglutide, and a specific market for counterfeit products.
Objectives: To assess and compare the quality, reliability, and content completeness of health information regarding GLP-1 agonists on China's two dominant video platforms, Douyin (the Chinese version of TikTok) and Bilibili; to examine how platform architecture and regulatory context influence content reliability; and to contrast these findings with recent benchmarks from Western social media studies.
Design: Cross-sectional observational study.
Methods: A systematic search for "GLP1" was conducted on Douyin and Bilibili. Due to divergent algorithms, an asymmetric sampling strategy was used: all 106 videos returned by Douyin were screened, while the top 200 videos from Bilibili's default ranking were assessed. After applying exclusion criteria, 128 videos (56 Douyin, 72 Bilibili) were analyzed. Quality was evaluated using the modified DISCERN tool, Global Quality Score, and JAMA benchmarks. No human subjects were involved; therefore, ethical approval was not required.
Results: Bilibili videos were significantly longer and demonstrated superior quality, scoring significantly higher than Douyin videos on both the DISCERN (p < 0.001) and JAMA benchmark (p = 0.003). Unlike findings on TikTok and Instagram, where quality was uniformly low, Bilibili hosted high-quality, in-depth content comparable to patient education materials. However, a safety analysis revealed videos promoting fraudulent products that co-opt the drug's brand name, posing a severe risk unique to the unregulated domestic supply chain.
Conclusion: The Chinese digital landscape exhibits a quality tier on Bilibili that is absent in Western short-video counterparts. While Bilibili serves as a reliable "digital library," Douyin remains a high-risk environment for misinformation and counterfeit sales. As China approaches the 2026 patent cliff, regulatory bodies must address the specific threat of fraudulent "GLP-1" branding on illicit products.
Trial registration: Not applicable.

Keywords:  Bilibili; Douyin; GLP-1 receptor agonists; counterfeit drugs; digital health; infodemic; mazdutide; semaglutide

DOI:  https://doi.org/10.1177/20420188261467887
Front Digit Health. 2026 ;8 1970952

Correction: Quality and reliability assessment of systemic lupus erythematosus related health information on short video platforms.

Lingling Gao, Yujie Yang, Yuchen Wang, Kunling Wang, Wanqing Li, Hongrui Wen, Yunan Zhang, Xiangfeng Li, Wei Bai.

  [This corrects the article DOI: 10.3389/fdgth.2026.1875817.].

Keywords:  health information; quality assessment; short video; social media; systemic lupus erythematosus

DOI:  https://doi.org/10.3389/fdgth.2026.1970952
Front Public Health. 2026 ;14 1874571

Determinants of sexual health information-seeking on social media among Chinese college students: an application of the Comprehensive Model of Information Seeking.

Zhuowen Feng, Wei He, Jiaru Xie, Guancheng Wang, Wanfang Lin.

   Background: Chinese college students often use social media to access sexual health information in a context of uneven formal sex education and cultural sensitivity surrounding sexuality. Understanding the factors associated with this behavior can inform sexual health communication and education.
Methods: A cross-sectional online survey was conducted with 283 Chinese college students. Structural equation modeling was used to examine the associations among perceived direct or vicarious experience with sexual health issues, risk-oriented salience, beliefs, perceived information carrier characteristics, perceived utility of sexual health information (SHI), and sexual health information-seeking behavior (SHISB).
Results: Perceived direct or vicarious experience, risk-oriented salience, beliefs, and perceived information carrier characteristics were positively associated with the perceived utility of social media for SHI. Perceived information carrier characteristics and perceived utility of SHI were positively associated with SHISB. Perceived utility served as an indirect statistical pathway linking perceived information carrier characteristics and SHISB.
Conclusion: The findings suggest that the Comprehensive Model of Information Seeking (CMIS) provides a useful but bounded framework for examining individual-cognitive and information-evaluation factors related to SHISB among Chinese college students. Improving the credibility, clarity, accessibility, and relevance of online sexual health information may better support students' sexual health information needs. Given the cross-sectional design and risk-oriented measurement, the findings should be interpreted as statistical associations rather than causal effects or a comprehensive account of sexual health practices.

Keywords:  Chinese college students; Comprehensive Model of Information Seeking; health behavior; new media; sexual health information seeking; social media

DOI:  https://doi.org/10.3389/fpubh.2026.1874571
Healthcare (Basel). 2026 Sep 04. pii: 2854. [Epub ahead of print]14(17):

Factors Associated with Digital Health Information-Seeking Behavior Among Healthcare Professionals in Izmir: A Cross-Sectional Study.

Gökben Yaslı, Levent Uğurlu.

  Background: Recently, artificial intelligence-based tools such as large language models (LLMs) have further transformed health information-seeking practices. This study aimed to assess digital health information-seeking behaviors among healthcare professionals working in İzmir, Türkiye, and to examine individual and environmental factors associated with these behaviors. Methods: A cross-sectional study was conducted among 380 healthcare professionals in İzmir. Data were collected using an online questionnaire comprising sociodemographic and health information-seeking items and the eight-item e-health literacy scale (eHEALS). Descriptive statistics and non-parametric tests were used for unadjusted comparisons. Multivariable binary logistic regression was performed to identify factors independently associated with LLM use, while multivariable linear regression with HC3-robust standard errors was used to identify factors independently associated with e-health literacy. To reduce sparse-data instability, conceptually compatible small categories were collapsed before refitting the multivariable models. Statistical significance was set at p < 0.05. Results: Overall, 73.2% of participants reported using LLMs (e.g., ChatGPT) for health-related information seeking. After multivariable adjustment, daily internet use of 3-6 h was associated with higher odds of LLM use compared with ≤3 h/day (adjusted OR = 2.0437, 95% CI: 1.1467-3.6423, p = 0.0153). E-health literacy was not independently associated with LLM use (adjusted OR = 0.9875, 95% CI: 0.9599-1.0158, p = 0.3825). In the e-health literacy model, the combined divorced/widowed group had lower adjusted scores than married participants (B = -4.4087, 95% CI: -7.4003 to -1.4171, p = 0.0039). Uncertainty about institutional scientific database access was also associated with lower e-health literacy (B = -6.1805, 95% CI: -8.9255 to -3.4356, p < 0.0001), whereas often/always reading online health information was associated with higher scores compared with never/rarely reading it (B = 2.6264, 95% CI: 0.5112-4.7416, p = 0.0149). Conclusion: LLM use was common among healthcare professionals, but it was not independently associated with e-health literacy. Patterns of internet use, marital status, institutional database awareness, and frequency of reading online health information showed independent associations with the study outcomes. These findings support targeted digital health and AI-literacy initiatives.

Keywords:  ChatGPT; e-health literacy; health information; health information-seeking behavior; healthcare professionals

DOI:  https://doi.org/10.3390/healthcare14172854
Nutrients. 2026 Sep 07. pii: 2921. [Epub ahead of print]18(17):

Supplement Safety Knowledge, Information Sources, and Self-Reported Safety-Related Behaviors Among Adults Aged 40 Years and Older: A Pilot Study.

Beata Walczyk, Agnieszka Białek.

   BACKGROUND/OBJECTIVES: Dietary supplement use is common in middle and later adulthood, but factual safety knowledge may not necessarily translate into safety-related behavior. This pilot study examined supplement use, supplement safety knowledge, information-source pathways, and self-reported safety-related behaviors among adults aged 40 years and older.
METHODS: A cross-sectional questionnaire study was conducted in Poland between December 2025 and April 2026 in a convenience sample of 150 adults aged ≥40 years. Multivariable logistic regression models assessed associations of a five-item factual supplement safety knowledge score and professional or digital information-source use with proactive supplement safety information seeking and other supplementation behaviors.
RESULTS: Current supplement use was reported by 113/150 participants (75.3%); among current users with available supplement count data, 86/111 (77.5%) used at least two supplements simultaneously. The knowledge score was not independently associated with proactive supplement safety information seeking (aOR 0.93 per one-point increase, 95% CI 0.69-1.25; p = 0.641). Professional information-source use was associated with the original composite information-seeking outcome (aOR 4.03, 95% CI 1.51-10.77; BH q = 0.033) and with lower odds of self-initiated supplementation (aOR 0.32, 95% CI 0.13-0.79; BH q = 0.042). However, in an additional non-overlap sensitivity analysis restricted to internet/leaflet safety information seeking, professional-source use was not independently associated with the outcome (aOR 0.80, 95% CI 0.33-1.94; p = 0.627).
CONCLUSIONS: In this convenience sample, the study-specific factual knowledge score did not show a clear independent association with proactive supplement safety information seeking. Associations involving professional information sources depended partly on outcome definition, while the association with lower self-initiation remained. Given the pilot nature, modest sample size, non-probability recruitment, self-reported measures, and lack of product-level verification, the findings are hypothesis-generating and require confirmation in larger, prospectively characterized samples.

Keywords:  dietary supplements; health knowledge; healthcare professionals; information sources; pilot study; poly-supplementation; supplement safety; supplement–medicine interactions

DOI:  https://doi.org/10.3390/nu18172921
iScience. 2026 Sep 18. 29(9): 117445

Neural correlates of information gain during search revealed by fMRI.

Sakrapee Paisalnan, Yashar Moshfeghi, Frank Pollick.

  When confronted with uncertainty, humans engage in information-seeking behavior to gather sufficient information. Identifying the brain regions involved offers insight into the underlying cognitive mechanisms. Using Shannon information theory, we quantified the information contained in the image tiles presented in an fMRI experiment. Nineteen participants judged whether a caption described an image by viewing a sequence of tiles and signaling when it matched. Random-effects ANOVA and representational similarity analysis revealed that regions associated with information gain resulted in two complementary patterns. The first, involving right Ph2 and left hIP5, showed a graded relationship with information gain, consistent with tracking currently available perceptual evidence. The second, involving bilateral PGp with left hIP3 and left FG2, was more sensitive to intermediate information gain, possibly reflecting integrative processes relating accumulated visual information to task-relevant semantic expectations. These findings offer a neurocognitive framework for how information is acquired and evaluated during a sequential search.

Keywords:  cognition; fMRI; information gain; information retrieval

DOI:  https://doi.org/10.1016/j.isci.2026.117445