JMIR Medical Education最新文献

{"title":"Collaborative Simulation-Based Learning to Develop Competencies in Physical Therapy Students: Randomized Controlled Trial.","authors":"Iván Cuyul-Vásquez, Hilda Aravena-Páez, Felipe Araya-Quintanilla, Sergio Guinez-Molinos, Rodrigo Núñez-Cortés, Héctor Gutiérrez Espinoza","doi":"10.2196/85442","DOIUrl":"10.2196/85442","url":null,"abstract":"<p><strong>Background: </strong>Although the implementation of simulation-based learning has progressed substantially in medicine, physical therapy education shows a lower level of development regarding standardized collaborative models.</p><p><strong>Objective: </strong>This study aims to determine the effectiveness of collaborative versus individual simulation-based learning in developing clinical competence in undergraduate physical therapy students in the care of standardized patients with low back pain flare-ups.</p><p><strong>Methods: </strong>A parallel-group randomized controlled trial was conducted. Thirty-one physical therapy undergraduate students were randomized (1:1) to a collaborative or individual simulation-based learning group. Students underwent 3 weeks of clinical simulation training. Standardized patient assessments were performed before and after the simulation-based learning intervention. Clinical competence was evaluated as the primary outcome, while effective communication and therapeutic alliance were evaluated as secondary outcomes.</p><p><strong>Results: </strong>For clinical competence, a significant main effect of time was observed (<i>η</i>_p²=0.739; <i>P</i><.001), with equivalent pre-to-post improvements in both the collaborative (<i>Δ</i>=+6.17; <i>P</i>_Bonf<.001) and individual simulation-based learning groups (<i>Δ</i>=+5.39; <i>P</i>_Bonf<.001); the interaction effect was nonsignificant (<i>P</i>=.61). For communication, no significant time (<i>P</i>=.91) or group × time interaction (<i>P</i>=.83) effects were observed. Therapeutic alliance scores demonstrated a significant main effect of time (<i>η</i>_p²=0.294; <i>P</i>=.008), indicating an overall decrease, though group-specific pre-to-post declines did not survive Bonferroni correction (both <i>P</i>_Bonf≥.11).</p><p><strong>Conclusions: </strong>Collaborative simulation-based learning effectively develops physical therapy clinical competencies, and its effects are comparable to those of individual simulation-based learning with a standardized patient. However, there were no differences between the groups for effective communication or therapeutic alliance. Further research is warranted to assess the efficacy of these modalities in diverse clinical settings.</p>","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e85442"},"PeriodicalIF":13.9,"publicationDate":"2026-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13592459/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149018043","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative AI for Personalized Learning Platforms in Otolaryngology-head and neck surgery (ENT) Education: Scoping Review.","authors":"Jiang-Tao Zhong, Zai-Zai Cao, Shui-Hong Zhou","doi":"10.2196/91785","DOIUrl":"10.2196/91785","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Generative artificial intelligence (AI), including large language models (LLMs), may enable personalized learning in medical education. In otolaryngology-head and neck surgery (ENT) education, AI may support knowledge retrieval, adaptive assessment, feedback, and scalable learning, but evidence remains fragmented.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This scoping review mapped generative AI research in otolaryngology education, summarized applications, performance, benefits, and risks, and identified directions for AI-enabled personalized learning.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;Eligibility Criteria: We included English-language empirical studies, preprints, and technical or developmental evaluations published from 2017 through 2026 with direct relevance to professional ENT education; nonempirical publications and studies without a professional educational focus were excluded. Sources of Evidence: PubMed, Embase, Web of Science, Scopus, ERIC, IEEE Xplore, and CINAHL were searched, supplemented by citation searching. The final search was conducted on July 6, 2026. Charting Methods: Data were charted using a standardized form by two reviewers and synthesized through a three-stage inductive thematic analysis, with third-reviewer adjudication. No formal critical appraisal or meta-analysis was performed because of study heterogeneity.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Forty-two studies published between 2023 and 2026 were included. Most were performance studies (33/42, 78.6%), followed by system development and educational intervention studies (3/42 each, 7.1%), surveys (2/42, 4.8%), and guideline development (1/42, 2.4%). Most studies evaluated question sets or model outputs rather than participants. Applications included knowledge retrieval, revision, board-style question practice, adaptive assessment, AI avatar-delivered instruction, and operative-planning support. Seven major thematic domains were identified: (1) Educational Support and Learning Enhancement, (2) Learning Outcomes, Educational Effectiveness, and Pedagogical Value, (3) Model Performance, Accuracy, and Variability, (4) Assessment, Examination Readiness, and Question-Type Sensitivity, (5) Reliability, Safety, and Risks, (6) Adoption, Usability, and Implementation Barriers, and (7) Advanced AI Architectures and Future Educational Design. Newer models performed strongly on selected structured knowledge tasks, but performance varied by task, difficulty, modality, and reasoning demands. Evidence remained heterogeneous and benchmarking-focused, with only 3 educational interventions and limited evidence for sustained learning, clinical reasoning, or real-world outcomes. Hallucinations, unreliable sourcing, confidently incorrect outputs, and contextual and multimodal reasoning limitations supported supervised rather than autonomous use.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;By moving beyond previous reviews centered primarily on AI applications or model performance,","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":" ","pages":""},"PeriodicalIF":13.9,"publicationDate":"2026-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149002784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The AWARE (AI Use, Why, Attachment, Reality and Risk, and Effect on Functioning) Framework for Assessing Patients' Use of AI in Mental Health Care.","authors":"Alexandre Hudon, Venkat Bhat, John-Jose Nunez, Vincent Paquin, Amylie Malouin-Lachance, François Lespérance","doi":"10.2196/108766","DOIUrl":"10.2196/108766","url":null,"abstract":"<p><p>Conversational AI has rapidly become integrated into everyday life, with many individuals using large language model-based systems, such as ChatGPT, Gemini, and Claude, for education, productivity, health information, decision-making, emotional support, and companionship. As patients increasingly incorporate conversational AI into their cognitive, emotional, and social lives, these interactions may influence coping strategies, treatment engagement, clinical decision-making, and mental health outcomes. However, although medical education has increasingly emphasized AI literacy and the responsible use of generative AI by clinicians, comparatively little attention has been given to preparing health care professionals to assess patients' use of conversational AI during routine clinical encounters. This Viewpoint introduces the AWARE (AI use, why, attachment, reality and risk, and effect on functioning) framework, a practical educational framework designed to help mental health professionals systematically assess patients' use of conversational AI. Rather than functioning as a diagnostic instrument or psychometric scale, AWARE provides a structured approach to psychiatric interviewing across 5 clinically relevant domains: AI use, why, attachment, reality and risk, and effect on functioning. Together, these domains guide clinicians in exploring patterns of AI use, the motivations underlying engagement, the emotional significance of patient-AI interactions, potential influences on reality testing and clinical risk, and the overall impact of AI on psychological well-being, daily functioning, relationships, and recovery. The framework is intended to complement existing psychiatric interviewing practices by supporting comprehensive history taking, clinical reasoning, risk assessment, documentation, and learner education. We discuss the rationale for routinely asking patients about AI use, review emerging evidence regarding both the potential benefits and risks of conversational AI, and describe how AWARE may be incorporated into undergraduate, postgraduate, and continuing professional education through simulation, objective structured clinical examinations, workplace-based assessment, and clinical supervision. We also outline priorities for future research, including validation, implementation, educational evaluation, and cross-cultural adaptation. As conversational AI becomes increasingly embedded within patients' daily lives, clinicians require practical approaches to understanding its role in mental health and health care. The AWARE framework offers a structured educational starting point for integrating assessment of patient-AI interactions into routine psychiatric interviewing while supporting patient-centered, evidence-informed clinical practice and health professions education.</p>","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e108766"},"PeriodicalIF":13.9,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13628033/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148983961","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Students' Learning Outcomes in a Virtual Microscopic Online and Face-to-Face Anatomy Course: Cross-Over Comparative Study.","authors":"Katharina Sternecker, Christoph Schmitz, Markus Berndt","doi":"10.2196/82012","DOIUrl":"10.2196/82012","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Since 2021, the histology of organs in the microscopic anatomy course at the Department of Anatomy II of the Ludwig-Maximilians-Universität Munich has been held as a hybrid course using a 3D virtual microscopy system. Evidence comparing synchronous online versus face-to-face instruction using an advanced 3D virtual slide system remains limited. In particular, it is unclear whether learning outcomes are equivalent across instructional modes, and whether individual student characteristics influence learning outcomes under these conditions.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study systematically (1) compares the learning outcomes of preclinic students attending a microscopic anatomy course either face-to-face or synchronously online using a 3D virtual microscopy slide box generated from fully scanned histological glass slides previously used in a traditional course, and (2) investigates whether selected demographic characteristics and personal skills predict learning outcomes in either instructional mode. We hypothesized that learning outcomes would not differ between online and face-to-face participation when instructional conditions are identical.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;The students attended 2 consecutive course days on 2 topics (kidney and liver) in a crossover design. Half of the students attended the first day face-to-face, and the second day online, while the other half attended in the reverse order. Demographic characteristics (eg, self-organization and affinity for technology interaction) and learning outcomes were collected via survey and simulated examination on the day after the second course day. To minimize potential confounding factors (eg, learning by means of external resources) and to ensure that the results reflected the learning outcomes within the scope of this study, 2 particularly challenging course days were intentionally selected, and the time interval before the simulated examination was kept short. Learning outcomes were analyzed using a linear mixed-effects model, and potential predictors were explored using stepwise linear regression.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Overall, 740 student questionnaires and examinations (740/864, 86% of the total student cohort) were used to assess the items' quality of the simulated examination. Inferential analyses included data from 415 students who completed the crossover design as intended. No significant main effect of participation mode on learning outcomes was found (&lt;i&gt;F&lt;/i&gt;&lt;sub&gt;1, 413&lt;/sub&gt;=0.16; &lt;i&gt;P&lt;/i&gt;=.69). Although a significant effect of examination topic was observed (&lt;i&gt;F&lt;/i&gt;&lt;sub&gt;1, 413&lt;/sub&gt;=25.56; &lt;i&gt;P&lt;/i&gt;&lt;.001), there was no significant interaction between participation mode and examination topic (&lt;i&gt;F&lt;/i&gt;&lt;sub&gt;1, 413&lt;/sub&gt;=0.13; &lt;i&gt;P&lt;/i&gt;=.72). Regression analyses showed low predictive power overall (&lt;i&gt;R&lt;/i&gt;&lt;sup&gt;2&lt;/sup&gt;≥0.12), with self-organization emerging as a weak but consistent predictor across both instructional modes.&lt;","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e82012"},"PeriodicalIF":13.9,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13580892/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148983897","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Is This Still Medical Education? The Educational Use of Nude Pediatric Clinical Photographs: Bioethical Reassessment.","authors":"Iwona Bujek","doi":"10.2196/104023","DOIUrl":"10.2196/104023","url":null,"abstract":"&lt;p&gt;&lt;p&gt;Clinical photography can convey phenotypic, anatomical, and diagnostic information that may be difficult to replace with text. This also applies to highly sensitive pediatric clinical photographs when the scope of patient exposure is necessary to achieve a specific educational objective. This viewpoint addresses a more specific problem: \"How should the educational function of a highly sensitive pediatric clinical photograph be evaluated when the condition identified with the photograph is substantively described elsewhere in the educational material?\" The analysis is grounded in a documented longitudinal observation of 28 editions and reprints of a single academic textbook lineage published between 2005 and 2019, across which the same image-condition configuration persisted. At the point of image presentation, the depicted clinical conditions were explicitly identified, while the accompanying text stated that they were described elsewhere in the textbook, where their substantive clinical descriptions were located in accordance with the textbook's declared curricular structure. On this basis, I propose educational disconnection to describe a situation in which the image-specific educational contribution at the point of presentation is difficult to reconstruct from the observable relationship among image location, the clinical condition identified with the image, the condition's declared curricular location, the location of its substantive clinical description, and the destination indicated in the text accompanying the photograph. The central distinction is that identification of a clinical condition, identification of where that condition is substantively described, and identification of the educational contribution of its photograph are not the same. A separate question therefore remains: \"What additional educational information does the image contribute here?\" The longitudinal dimension extends this analysis through the concept of educational value decay, referring to the possibility that the relationship between an image and the context-specific justification for its presence may weaken over time, without assuming that educational value automatically declines with the age or repeated reproduction of the image. In response, I propose an Educational Reassessment Framework-a practical, event-triggered process that proceeds from identifying the image's educational function and its relationship to the curricular structure, through assessment of its image-specific educational contribution and exposure-function proportionality, to the consideration of equivalent alternatives, the continuing validity of its justification, and the conditions of responsible continued use. Reassessment is not intended to automatically remove historical, repeatedly used, or highly sensitive photographs. Responsible reuse of an educationally valuable image may remain justified and may reduce the need to photograph additional patients, but educational value does not su","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e104023"},"PeriodicalIF":13.9,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13628035/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148983925","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Teaching Acute Coronary Syndrome High-Risk ECG Interpretation and Clinical Decision-Making Through FOAMed Videos and Podcast Versus Print-Based Materials Among Emergency Care Providers: Randomized Controlled Mixed Methods Trial.","authors":"Wiebke Turner, Klaus Fessele, Martin Fandler, Jan P Ehlers, Julia Nitsche","doi":"10.2196/87587","DOIUrl":"10.2196/87587","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Accurate interpretation of high-risk acute coronary syndrome (ACS) electrocardiograms (ECGs) is essential for early diagnosis and timely reperfusion, yet substantial deficits persist across health care professions. Digital self-learning formats such as FOAMed (Free Open Access Medical Education) are widely used, but their effectiveness has rarely been evaluated for complex, high-risk ACS ECG patterns. Existing ECG education studies often focus on students or single professional groups and established ST-segment elevation myocardial infarction (STEMI) criteria, leaving newer guideline-recognized STEMI equivalents, selected emerging occlusion myocardial infarction (OMI)-related patterns, and interprofessional emergency care underrepresented.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to compare the effectiveness of FOAMed podcast and videos versus traditional print-based materials for teaching high-risk ACS ECG patterns and related clinical decision-making in emergency providers.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We conducted a prospective, interprofessional, controlled mixed methods trial across 5 training sites in Germany. Paramedics, prehospital emergency physicians, and emergency department clinicians received either a FOAMed multimedia module or print-based materials through concealed allocation; deviations from the intended 1:1 ratio resulted from participant no-shows. The intervention consisted of a 30-minute supervised self-learning session. In total, 103 participants were allocated to FOAMed (n=45) or print-based materials (n=58). Two coprimary outcomes were assessed: ECG interpretation accuracy and text-based ACS clinical decision-making. Secondary outcomes included subjective confidence, learning experience, and exploratory qualitative free-text responses. Outcome assessment was automated and blinded; mixed ANOVA was the primary analysis. The study was not prospectively registered because it assessed educational outcomes in health care professionals rather than patient health outcomes.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;All 103 participants completed the study. Both groups improved, with greater gains in the FOAMed group: ECG interpretation increased from 55% to 65.5% and text-based ACS clinical decision-making from 45% to 68%, versus 57% to 60% and from 47% to 63%, respectively, in the print-based group. Effect sizes were η²=0.055 for ECG interpretation and η²=0.044 for clinical decision-making. Exploratory subgroup analyses provided no evidence of differential effects across age, gender, or professional background and were likely underpowered. Qualitative responses (46 and 37 entries) provided contextual insights into perceived clarity, engagement, and practical relevance supporting the quantitative findings.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;This study is innovative in directly comparing a curated FOAMed multimedia module with selected print-based materials in an interprofessional emergency care po","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e87587"},"PeriodicalIF":13.9,"publicationDate":"2026-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13577616/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148977964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Teaching Model Context Protocol, Retrieval-Augmented Generation, and AI Agents to a Multidisciplinary Hospital Workforce: Single-Group Pre-Post Survey Study.","authors":"Gakyoung Baek, Hyunna Lee, Dong Hyun Yang, Minseo Kang, Kun Hee Lee, Minji Choi, Yura Lee, Kye Hwa Lee","doi":"10.2196/97822","DOIUrl":"10.2196/97822","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Hospitals worldwide need to upskill their workforce in advanced AI technologies; yet, published guidance on how to design and deliver such training, particularly in agent-level tools like retrieval-augmented generation (RAG) and the model context protocol (MCP), remains virtually absent.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;To describe the design, implementation, and lessons learned from an 8-week, 56-hour intensive generative AI training program for a multidisciplinary hospital workforce, drawing on both quantitative outcome data and participants' own reflections on their learning experience.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;The program was delivered on-site at Asan Medical Center with simultaneous online broadcast to 2 regional affiliate hospitals. The curriculum was built around the premise that MCP and AI agents would become the foundation of health care AI use, allocating 37% (11.5/31 hours) of on-site instructional time to MCP, and 71% (22/31 hours) to hands-on practice. Participants progressed from foundational concepts through RAG and MCP to team-based capstone projects, supported by funded AI tool subscriptions, a dedicated internal cloud platform, and 3-6 hours of weekly mentoring per team. A pre-post survey (pre: n=83; post: n=64) evaluated outcomes across Kirkpatrick levels 1-3, complemented by thematic analysis of open-ended reflections on self-perceived growth.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The technologies that received the greatest curricular investment were associated with the largest self-efficacy differences (MCP: &lt;i&gt;d&lt;/i&gt;=1.57; overall effect: &lt;i&gt;r&lt;/i&gt;=.574), and participants most frequently cited MCP and RAG when describing how abstract concepts \"became concrete and actionable.\" Non-IT professionals, clinicians, health information managers, researchers, and administrative staff showed consistently larger gains than IT specialists; several reported coding for the first time through vibe coding, challenging the assumption that advanced AI training requires technical backgrounds. Despite significant overall gains, a knowledge-practice gap persisted: job-specific competency remained below the scale midpoint, though participants spontaneously reported generating workplace application ideas. Curriculum pacing was rated lowest despite high overall satisfaction (4.03/5), signaling that even 56 hours may progress too quickly for mixed-expertise cohorts. Capstone projects with dedicated mentoring received the highest satisfaction ratings; 11 of 12 teams presented functional prototypes, and one has since entered active pilot use in clinical departments ahead of planned hospital-wide deployment.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;To our knowledge, this is the first program to teach 4 agent-level generative AI technologies, MCP, RAG, LangGraph orchestration, and AI agent design, to both IT and non-IT hospital staff. This program suggests that transforming a multidisciplinary hospital workforce into AI-capabl","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e97822"},"PeriodicalIF":13.9,"publicationDate":"2026-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13568689/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148945873","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance of Cloud-Hosted Large Vision-Language Models on the Japanese National Examination for Clinical Laboratory Technicians: Comparative Benchmarking Study.","authors":"Kota Murakami, Ryosuke Matsuzawa, Yuya Tahara-Arai, Haruka Ozaki","doi":"10.2196/84266","DOIUrl":"10.2196/84266","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Cloud-hosted large vision-language models (LVLMs) often outperform open-weight models on multimodal benchmarks, but their applicability to health care examinations that require both text and image reasoning remains unclear. Existing studies on Japan's National Examination for Clinical Laboratory Technicians mainly used text-only settings and a limited set of models.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to evaluate the accuracy and applicability of cloud-hosted LVLMs on this examination, quantify the contribution of images, examine alignment with human performance across medical subfields, and test generalizability on questions administered after the models' knowledge cutoff dates. We aim to provide a standardized benchmark of baseline multimodal performance under deployment-realistic conditions without additional fine-tuning.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;For this comparative benchmarking study, we built the Kensagishi question-answer dataset (Kensagishi QA), a benchmark of 800 questions from the examination (2020-2023), including text-only and image-based questions, in a standardized structured JSON format, and strict scoring criteria (all correct options required). We tested GPT-5, GPT-4o, GPT-4o-mini, Gemini 1.5 Pro, Gemini 2.5 Pro, and Neva-22B using a uniform \"numbers-only\" response prompt; images were provided as Base64-encoded data when available. Accuracy was computed overall, by item type, and by medical subfield. Human reference was drawn from published examination results. Spearman correlation assessed model-human alignment. Generalization was evaluated on 200 questions from the 71st examination (2025), released after the models' knowledge cutoffs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Top-performing models exceeded the 60% passing threshold. GPT-5 achieved 93.9% (95% CI 92.0%-95.4%) accuracy with images and 88.2% (95% CI 85.7%-90.2%) without images, Gemini 2.5 Pro 92.5% (95% CI 90.5%-94.1%) and 86.8% (95% CI 84.2%-88.9%), and GPT-4o 81.7% (95% CI 78.8%-84.2%) and 79.8% (95% CI 76.8%-82.4%), respectively. Providing images markedly improved accuracy for GPT-5, Gemini 2.5 Pro, and GPT-4o, whereas Neva-22B showed no improvement. Model-human rank correlations by medical subfield were moderate for GPT-4o (Spearman ρ=0.61; 95% CI -0.03 to 0.90; &lt;i&gt;P&lt;/i&gt;=.06 with images) and statistically significant for Gemini 1.5 Pro without images (ρ=0.67; 95% CI 0.07-0.91; &lt;i&gt;P&lt;/i&gt;=.03). On 200 postknowledge cutoff questions from the 71st examination (2025), model accuracies were comparable to historical averages (eg, GPT-5 91.5%, 95% CI 86.8%-94.6%; Gemini 2.5 Pro 90.7%, 95% CI 85.8%-94.0%; GPT-4o 82.0%, 95% CI 76.1%-86.7%), indicating no material performance degradation on unseen questions.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Cloud-hosted LVLMs show strong performance on this examination, and visual input is a key driver of accuracy for leading models. Models that better mirror human difficulty patterns (eg, GPT","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e84266"},"PeriodicalIF":13.9,"publicationDate":"2026-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13601776/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149031425","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Promise, Mimicry, and Surveillance: Responsibly Integrating Artificial Intelligence With Socratic Inquiry in Medical Education.","authors":"José Daniel Sánchez","doi":"10.2196/102958","DOIUrl":"10.2196/102958","url":null,"abstract":"<p><strong>Unlabelled: </strong>Large language models are increasingly proposed as a way to deliver Socratic dialogue at scale in medical education, offering personalized, always-available, and lower-stakes inquiry that human educators cannot logistically provide. In this viewpoint, I argue that this promise is plausible but unproven and that 2 failure modes deserve more attention than they currently receive. The first is the mimicry trap: a system that fluently generates Socratic-sounding questions can appear to cultivate reasoning while functioning as interactive content delivery. I position this as an educational instance of the ELIZA effect and of the proxy-outcome problem; separating conversational mimicry from genuine metacognitive gain remains a central unresolved evaluation problem. The second is what I term the Panopticon Paradox: the data collection that makes AI tutoring effective may erode the psychological safety that Socratic inquiry requires, pushing learners toward performative rather than authentic engagement, in a manner analogous to gaming behaviors documented in intelligent tutoring systems. I present this second construct as a testable causal model with specified mediators, moderators, and falsifiable predictions rather than as an established finding. Because the evidence base is dominated by proof-of-concept tools, cross-sectional surveys, and short-term evaluations, I argue that AI is complementary rather than a replacement technology, and I propose a 3-pillar framework (governance, curriculum, and faculty development) tied explicitly to the 2 failure modes. I distinguish 4 modalities of delegation and argue that ethical deliberation, emotionally complex communication, and ambiguous clinical judgment cannot presently be recommended for autonomous or primary AI delivery.</p>","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e102958"},"PeriodicalIF":13.9,"publicationDate":"2026-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13561443/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148931272","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Attitudes Toward and Preconditions for Digital Game-Based Learning Among Public Health and Health Sciences Students: Cross-Sectional Online Survey Study.","authors":"Kamil J Wrona, Joanna Albrecht, Anne Krümmel, Leona Aschentrup, Christoph Dockweiler, Timothy Mc Call","doi":"10.2196/83347","DOIUrl":"10.2196/83347","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Digital game-based learning (DGBL) is gaining traction in medical and nursing education, but its integration into public health and health sciences curricula remains limited. This study addresses this research gap by exploring student perspectives, use patterns, and expectations regarding DGBL within German public health study programs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;The study investigated how public health and health sciences students perceive DGBL and identified conditions that promote or hinder its integration into academic training for public health professionals.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A cross-sectional online survey was conducted from January 2021 to March 2024. A total of 444 students participated, with 326 completing the questionnaire. The instrument combined closed- and open-ended questions on sociodemographic characteristics, digital media use, attitudes toward DGBL, and structural and pedagogical requirements. Quantitative data were analyzed descriptively using SPSS; qualitative data were evaluated using MAXQDA and Kuckartz structured content analysis.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The sample consisted predominantly of full-time students enrolled in bachelor's or master's programs in public health and health communication. Digital media are widely used for communication (n=320, 98.2%), information search (n=291, 89.3%), and entertainment (n=268, 82.2%) in daily student routines. In contrast, video gaming and online shopping play a minor role. Students expressed a generally positive attitude toward DGBL, particularly in practical learning contexts such as exam preparation, visualizing complex content, and knowledge checks. High agreement was found for using DGBL in modules such as Prevention and Health Promotion (n=277, 84.9%), Epidemiology and Statistics (n=271, 83.2%), and Population Medicine (n=251, 76.7%). Preferred devices include desktop PCs, tablets, and smartphones, while immersive technologies such as virtual reality or augmented reality glasses or consoles are less favored. Students highlighted specific prerequisites for DGBL integration: a functioning technical infrastructure (n=321, 98.5%), media-literate teaching staff (n=311, 95.4%), and compliance with data protection standards (n=303, 93%). Furthermore, students valued user-friendly design (n=255, 78.2%), technical reliability (n=299, 91.7%), clear feedback (n=236, 72.4%), and flexibility regarding the time and place of use (n=307, 94.2%). Game elements such as animations, multimedia, and reward systems were appreciated, whereas features such as 3D graphics, avatars, and multiplayer modes were considered less important.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;The findings show a high willingness among students to engage with DGBL, especially when its design emphasizes clarity, usability, and relevance to course content. Emotional safety, flexible access, and pedagogical integration are more decisive than immersive game features. ","PeriodicalId":36236,"journal":{"name":"JMIR Medical Education","volume":"12 ","pages":"e83347"},"PeriodicalIF":13.9,"publicationDate":"2026-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13561400/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148930740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}