JMIR Medical Informatics最新文献

{"title":"Building a National Interoperable Rare Eye Disease Data Warehouse: Methodological Framework and Implementation Report From the French Rare Eye Disease Database (FREDD) Initiative.","authors":"Camille Beluffi Marin, Marilyne Oswald, Laura Ratenet, Matthieu Stoll, Kirsley Chennen, Nathalie Chabin, Julie Hubert, Grégoire Rey, Guylène Le Meur, Vasily Smirnov, Isabelle Meunier, Isabelle Audo, Hélène Dollfus","doi":"10.2196/99206","DOIUrl":"10.2196/99206","url":null,"abstract":"<p><strong>Background: </strong>Rare eye diseases are characterized by low prevalence, clinical heterogeneity, and fragmented data collection, which limit the reliability of analysis results and multicenter research. In France, the development of health data warehouses is strictly regulated by national data protection authorities. While international initiatives aim to harmonize rare disease registries, capturing hyperspecialized, multimodal clinical records and ophthalmology-specific imaging within a fully compliant and sustainable infrastructure remains a major operational challenge.</p><p><strong>Objective: </strong>This study aims to describe the methodological framework, regulatory implementation, and early operational outcomes of FREDD (French Rare Eye Disease Database), a national interoperable health data warehouse dedicated to rare eye diseases, and to analyze its key success factors, technical bottlenecks, and long-term financial sustainability models.</p><p><strong>Methods: </strong>Developed under the oversight of the French National Institute of Health and Medical Research (Institut national de la santé et de la recherche médicale; INSERM), the warehouse implements a centralized 3-layer architecture encompassing data collection, processing, and research reuse environments. The dataset extends the national minimum rare disease dataset with detailed ophthalmology variables structured through a dynamic electronic case report form, as well as a dedicated collection of ophthalmic images. Interoperability is achieved using international domain ontologies and a custom parsing tool (FREDDEX) designed to automatically prefill clinical data from the existing French national rare disease registry. Centralized monitoring, data curation, and cross-center quality tracking are driven by a dedicated in-house dashboard (FREDDIn [FREDD Insights]), while unstructured retinal images are processed through a standardized pseudonymization and human-verified validation pipeline.</p><p><strong>Unlabelled: </strong>Official regulatory authorization was obtained in April 2024, and active data collection began in March 2025 across 5 pilot expert centers, focusing its initial phase on inherited retinal dystrophies, primarily retinitis pigmentosa. Over a 1-year period, the warehouse successfully integrated clinical data from 1649 patients-representing 20.3% of all retinitis pigmentosa cases registered nationally-and accumulated more than 10,000 ophthalmic images. The centralized dashboard maintained an overall data inconsistency rate of approximately 7%, chiefly reflecting logical dependencies. Among the included patients, 36.7% (605/1649) patients had associated imaging data, with fundus photography being the most widely available modality (768/1649, 46.6%).</p><p><strong>Conclusions: </strong>The implementation of this database demonstrates that a highly specialized, multimodal health data warehouse can be successfully deployed in a highly regulated environment","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e99206"},"PeriodicalIF":5.0,"publicationDate":"2026-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13507042/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148820373","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"GPT-4o-Powered Preanesthetic AI: Development and Validation Study.","authors":"Bo-Han Lin, Cheng-Wei Lu, I-Ching Hou, Tzu-Yu Lin, Chin-E Liu, Chong-Sun Khoi","doi":"10.2196/83893","DOIUrl":"10.2196/83893","url":null,"abstract":"<p><strong>Background: </strong>Accurate preanesthetic assessment is essential for perioperative risk stratification, but conventional tools such as the American Society of Anesthesiologists (ASA) physical status classification and postoperative nausea and vomiting (PONV) risk scores may be affected by subjective judgment, incomplete documentation, and fragmented clinical data. Large language models may support preanesthetic assessment by integrating structured and unstructured clinical information.</p><p><strong>Objective: </strong>This study aimed to develop and retrospectively validate a GPT-4o (OpenAI)-powered AI system for preanesthetic assessment. The primary validation focus was the agreement between AI-generated and clinician-assigned ASA physical status classifications. PONV risk stratification was evaluated as an additional clinically relevant performance outcome. A secondary objective was to assess the incremental contribution of National Health Insurance (NHI) cloud data to model performance.</p><p><strong>Methods: </strong>This single-center retrospective validation study included 600 adult surgical patients randomly sampled from 4404 eligible inpatient surgical patients between January and May 2025. The system processed structured and unstructured electronic health record data, with and without NHI cloud data, to generate preanesthetic assessment outputs. ASA agreement was assessed using Cohen κ with 95% CIs. PONV prediction was evaluated using sensitivity, specificity, overall accuracy, positive predictive value, negative predictive value, confusion matrix analysis, the 2-sided Fisher-Freeman-Halton exact test, and Cramér V.</p><p><strong>Results: </strong>Among the 600 patients, clinician-assigned ASA classifications were ASA I in 39 patients, ASA II in 498 patients, ASA III in 57 patients, and ASA IV in 6 patients. Agreement between AI-generated and clinician-assigned ASA classifications was high when NHI data were incorporated (κ=0.883, 95% CI 0.841-0.946), whereas agreement was lower without NHI data (κ=0.518, 95% CI 0.412-0.691). A total of 49 (8.2%) patients experienced documented PONV within 24 hours after surgery. Using the high-risk category as the primary test-positive threshold, the model achieved a sensitivity of 34.7% (95% CI 22.9-48.7), a specificity of 99.1% (95% CI 97.9-99.6), a positive predictive value of 77.3%, and a negative predictive value of 94.5%. The association between the model-assigned PONV risk categories and observed PONV outcomes was statistically significant (<i>P</i><.001, Cramér V=0.531).</p><p><strong>Conclusions: </strong>The GPT-4o-powered preanesthetic assessment system demonstrated feasibility and high agreement with clinician-assigned ASA classifications when NHI cloud data were incorporated. For PONV risk stratification, the system showed high specificity and a high negative predictive value but modest sensitivity, indicating that it identified a small high-risk subgroup with a high observed","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e83893"},"PeriodicalIF":5.0,"publicationDate":"2026-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13506160/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148820406","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Public Reporting Systems in Health Care and the Underconceptualized Technical Substrate, a Core Information Systems Dimension: Scoping Review.","authors":"Elorm Damalie, Reima Suomi, Moustafa Mahmoud","doi":"10.2196/84820","DOIUrl":"10.2196/84820","url":null,"abstract":"<p><strong>Background: </strong>Public reporting systems (PRSs) in health care are defined as digital platforms that make comparative health care performance data available to the public (eg, Hospital Compare in the United States, NHS Choices in the United Kingdom, national quality registries in Europe, and MyHospital in Australia). These systems aim to improve transparency, accountability, and patient choice. While these systems have been widely studied from policy and clinical perspectives, the information system (IS) foundations that enable their operation, including architecture, interoperability, data governance, APIs, usability, and technical performance, remain underexplored.</p><p><strong>Objective: </strong>This study aims to map the extent and nature of the published literature addressing the technological foundations of PRSs in health care and to identify persistent gaps in IS scholarship.</p><p><strong>Methods: </strong>This scoping review followed the Joanna Briggs Institute guidelines and was reported in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews). Seven databases were searched: PubMed, Web of Science, Scopus, IEEE (Institute of Electrical and Electronics Engineers), ACM (Association for Computing Machinery), AIS (Association for Information Systems) eLibrary, and Cochrane. The search covered articles published between 2000 and 2026. Eligible articles included peer-reviewed journal articles and conference papers on PRSs in health care. Data were charted according to study characteristics, including 6 IS-specific coding dimensions: architecture, interoperability, data governance, APIs, usability, and technical performance. Descriptive statistics were used to summarize the findings, and a narrative synthesis was conducted to identify thematic trends and research gaps.</p><p><strong>Results: </strong>A total of 1882 records were identified across 7 databases. After deduplication, 1127 records were screened, and 233 studies were included in the review. Most of the included studies originated from the United States (n=157, 67.4%), followed by Europe (n=43, 18.4%). Of the 233 included studies analyzed, 60.5% (n=141) used quantitative methods, whereas 24.0% (n=56) used qualitative methods and 15.5% (n=36) used mixed methods. Applying a structured IS coding framework to all 233 studies revealed that the technical substrate of PRSs remains underexplored.</p><p><strong>Conclusions: </strong>Despite the centrality of IS infrastructure to public reporting, existing scholarship predominantly frames these systems as policy or social tools while neglecting their technical underpinnings. This oversight has practical implications, including poor system usability, weak interoperability, and noncompliance with data regulations, which can undermine patient trust and system effectiveness. Addressing this gap requires interdisciplinary approaches that integrate IS frame","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e84820"},"PeriodicalIF":5.0,"publicationDate":"2026-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13502770/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148814960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Patient Stratification for Improving Acute Chest Pain Management and Mitigating Emergency Department Crowding: Machine Learning Model Development and Validation.","authors":"Chih-Chia Hsieh, Shao-Chung Chu, Jung-Ting Lee, Chih-Hao Lin, Chung-Yao Kao","doi":"10.2196/83099","DOIUrl":"10.2196/83099","url":null,"abstract":"<p><strong>Background: </strong>Acute chest pain (ACP) is one of the most common chief complaints in the emergency department (ED), accounting for approximately 8% of all ED visits. However, among patients presenting with chest pain suggestive of cardiac origin, fewer than 10% are ultimately diagnosed with acute coronary syndrome (ACS).</p><p><strong>Objective: </strong>AI models were developed to support clinical decision-making for triage level-2 ED patients presenting with ACP. These models aim to accurately detect ACS and reliably identify low-risk patients based on a single high-sensitivity cardiac troponin T test result. By integrating this AI-assisted strategy into clinical workflows, we aim to reduce ED length of stay, alleviate crowding, and improve health care efficiency while maintaining high safety standards.</p><p><strong>Methods: </strong>We conducted a retrospective study using single-center data from a tertiary teaching hospital between January 2016 and December 2022. Models based on artificial neural networks (ANNs) were trained to classify patients with ACP at triage level 2 into 3 clinical classes: critical patients with ACS (subgroup GA), critical patients without ACS (subgroup GB1), and low-risk patients (subgroup GB2). The models were trained and internally validated via a 5-fold cross-validation protocol using data from 2016 to 2020. Model performance was then evaluated on the hold-out testing data (2021-2022) using area under the receiver operating characteristic curve, area under the precision-recall curve, and subgroup-specific metrics. Feature selection methods and the Shapley Additive Explanations value analysis were applied to identify features that drive the model's predictive power. The study was approved by the institutional review board of the National Cheng Kung University Hospital, Tainan, Taiwan (A-ER-111-199).</p><p><strong>Results: </strong>After excluding ED visits with missing triage data, incomplete medical histories, or unsuitable dispositions, 17,935 visits were included, with 1209 GA, 3873 GB1, and 12,853 GB2. Rendering prediction based on 24 feature variables (2 demographics, 6 vital signs, 4 blood test results, and 12 medical history), all ANN models demonstrated strong testing AUROC (95% CI) performance of 0.942 (0.920-0.965) for GA classification, 0.824 (0.808-0.841) for GB1, and 0.893 (0.884-0.902) for GB2. Regarding multiclassification, ANN-S3 achieved balanced performance, with ACS sensitivity of 0.941 (95% CI 0.908-0.973) and low-risk positive predictive value and sensitivity of 0.911 (95% CI 0.9-0.921) and 0.837 (95% CI 0.824-0.85), respectively. A sensitivity-prioritized variant, ANN-S3-L, increased ACS sensitivity to 0.966 (95% CI 0.94-0.991) and negative predictive value to 0.998 (95% CI 0.996-0.999), but at the cost of lower specificity and reduced low-risk sensitivity, indicating a safety-efficiency trade-off.</p><p><strong>Conclusions: </strong>These findings suggest that ANN-based classi","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e83099"},"PeriodicalIF":5.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13496927/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148801573","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrated Clinical-Molecular Risk Stratification in Diffuse Large B-Cell Lymphoma: Machine Learning Survival Analysis.","authors":"Jin Zhao, Xiaolian Wen, Li Ma, Liping Su","doi":"10.2196/84636","DOIUrl":"10.2196/84636","url":null,"abstract":"<p><strong>Background: </strong>The clinical outcomes of diffuse large B-cell lymphoma (DLBCL) are highly heterogeneous. While clinical indices like the international prognostic index (IPI) are widely used, their predictive accuracy remains limited. The integration of molecular features with clinical characteristics holds promise for developing more precise prognostic models to improve risk stratification and personalize treatment strategies.</p><p><strong>Objective: </strong>This study aimed to systematically identify key factors influencing overall survival (OS) and relapse in patients with DLBCL by leveraging publicly available transcriptomic data and clinical information. The goal was to construct and validate a high-precision risk-prediction model by using machine learning methods to aid in individualized clinical decision-making.</p><p><strong>Methods: </strong>We curated clinical and transcriptomic data from the GSE31312 cohort. A baseline clinical model was first constructed using multivariate Cox regression. Key genes associated with prognosis were identified through univariate Cox and survival analyses. Subsequently, 3 machine learning survival models, namely, fast survival support vector machine (FastSurvivalSVM), gradient boosting survival analysis (GBSurvival), and random survival forest (RSF), were trained and evaluated using 5-fold cross-validation. The interpretability of the optimal model was further elucidated using Shapley Additive Explanations (SHAP) methodology.</p><p><strong>Results: </strong>The baseline clinical model confirmed age, elevated lactate dehydrogenase, Eastern Cooperative Oncology Group score, Ann Arbor stage, and B symptoms as independent risk factors for OS and relapse-free survival, with a C-index of 0.65-0.67. At the molecular level, genes such as PSMG4 and CRY1 were significantly associated with poor OS, while TMEM182 and SPIRE1 were prominent in relapse prediction. Among the machine learning models, FastSurvivalSVM demonstrated the best overall performance, achieving an area under the curve of 0.791 for 1-year OS prediction and 0.774 for 1-year relapse prediction. SHAP analysis revealed that both clinical (eg, IPI and age) and molecular (eg, PSMG4 and SPIRE1) features were critical drivers of the model's predictions.</p><p><strong>Conclusions: </strong>This study successfully developed a multidimensional risk prediction model that integrates clinical and molecular characteristics for DLBCL. The FastSurvivalSVM model showed superior performance in predicting mortality and relapse risks. The interpretability analysis uncovered key prognostic factors, providing a valuable tool for personalized risk management and new theoretical insights for future mechanistic research.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e84636"},"PeriodicalIF":5.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13545529/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148802291","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of Prompt Design and Internal Reasoning in Chatbot-Based Medical History Taking: Simulation Study.","authors":"Nattawipa Thawinwisan, Chang Liu, Goshiro Yamamoto, Kazumasa Kishimoto, Yukiko Mori, Tomohiro Kuroda","doi":"10.2196/94614","DOIUrl":"10.2196/94614","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;A persistent discrepancy exists between patient-reported information and physician documentation. While conversational agents have been developed to collect medical histories prior to consultations, existing evaluations have largely focused on diagnostic accuracy or user satisfaction rather than on the completeness and clinical relevance of the information collected. There remains a need to assess the extent to which clinically relevant information is captured through chatbot-based interviews, and to understand how model configurations and instructional strategies influence this coverage.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aimed to evaluate the extent to which a chatbot can obtain clinically useful patient history information, and to examine how prompt detail and internal reasoning influence information coverage during chatbot-based medical interviews.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We developed a medical history-taking chatbot using the Qwen3-14B-Instruct model and evaluated 4 configurations in a 2×2 factorial design: detailed and thinking mode, detailed and nonthinking mode, minimal and thinking mode, and minimal and nonthinking mode. These configurations were compared against a rule-based system baseline (choice mode) using 66 standardized primary care clinical cases, with simulated patients interacting with the chatbot according to predefined case scripts. Information coverage (%) was assessed using a checklist inspired by Objective Structured Clinical Examination (OSCE) frameworks. Three physicians independently evaluated transcript coverage, with interrater agreement assessed using full agreement rates and Fleiss κ. For the 4 LLM configurations, coverage was analyzed using 2-way repeated-measures ANOVA to examine the effects of prompt detail, internal reasoning, and their interaction. All 5 configurations, including the rule-based baseline, were additionally compared using 1-way repeated-measures ANOVA with post hoc 2-tailed paired &lt;i&gt;t&lt;/i&gt; tests.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Interrater agreement was substantial (Fleiss κ=0.75). Across all 66 simulated cases, information coverage differed significantly among configurations (&lt;i&gt;P&lt;/i&gt;&lt;.001), with the detailed prompt with thinking (detailed and thinking) mode achieving the highest mean coverage (72.3%, SD 14.3%), compared with moderate coverage in configurations using either thinking or detailed prompts alone (approximately 60%) and lower coverage in minimal nonthinking and rule-based configurations (approximately 51%-54%). In the factorial analysis of the 4 LLM configurations, both prompt detail and internal reasoning were significantly associated with improved information coverage, with a significant interaction between prompt detail and internal reasoning interaction. Mode-related differences were most pronounced for past medical and family history domains. Symptom-level analyses revealed substantial variability, with higher coverage for symptom","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e94614"},"PeriodicalIF":5.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13501399/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148802293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Correction: Enhanced Prediction of Atrial Fibrillation in Patients With Ischemic Stroke Through Electronic Medical Records and Text Mining: Algorithm Development and Validation.","authors":"","doi":"10.2196/107705","DOIUrl":"10.2196/107705","url":null,"abstract":"","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e107705"},"PeriodicalIF":5.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13496389/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148802240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Discordance Between Textual Reasoning and Visual Interpretation in Large Language Models for Low Back Pain: Cross-Sectional Quantitative Evaluation and Exploratory Multimodal Stress Test.","authors":"Ziyu Zhang, Longhao Chen, Zhizhen Lv, Hanzhe Lv, Wei Sheng, Zicheng Wei, Binghao Wang, Ying Shen, Yu Tian, Jingwen Hu, Zhifang Shen, Lijiang Lv","doi":"10.2196/93522","DOIUrl":"10.2196/93522","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Large language models (LLMs) are rapidly evolving from text-based agents to multimodal systems capable of interpreting medical images. While their textual reasoning has improved, the safety implications of this shift remain underexplored, specifically regarding the alignment between visual interpretation and textual advice in low back pain (LBP) management.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This study aims to evaluate the performance of 10 commonly used and highly representative LLMs, including paid models such as ChatGPT 5 Plus (OpenAI) and reasoning-oriented models like Qwen3-Thinking (Alibaba/Qwen team) in the context of LBP. Specifically, it assesses their clinical accuracy, readability, and the \"asymmetric evolution\" in cross-modal diagnostic consistency.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;We conducted a cross-sectional quantitative evaluation integrated with an exploratory multimodal stress test. First, 10 LLMs, including ChatGPT 5 and ChatGPT 5 Plus, Gemini 3 Pro (Google DeepMind), Claude 4.5 Sonnet (Anthropic), Grok 4 (xAI), DeepSeek, Kimi K2 (Moonshot AI), Doubao (ByteDance Seed [Doubao Team]), Qwen3 (Base), and Qwen3-Thinking, addressed 25 standardized LBP inquiries. A multidisciplinary expert panel extracted 1816 individual recommendations by segmenting the models' point-by-point responses. These were independently coded for clinical accuracy against international guidelines on a 5-point Likert scale. Outcomes also included readability (Flesch Reading Ease [FRE] and Simple Measure of Gobbledygook [SMOG]), understandability (Patient Education Materials Assessment Tool [PEMAT]), and the presence of disclaimers. Second, to probe multimodal capabilities, we conducted an exploratory stress test (N=5) using a curated set of challenging clinical cases, including representative but complex LBP pathologies to assess baseline image recognition, and adversarial cases with deliberate clinical-radiological mismatches to evaluate cross-modal alignment.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;In textual tasks, models achieved an overall accuracy of 88.4% (1605/1816). ChatGPT 5 Plus demonstrated near-perfect performance, with zero serious errors. Within the only directly matched reasoning comparison, Qwen3-Thinking showed a lower serious-error rate than Qwen3 (Base). However, textual advice remained difficult to read, although understandability scores were robust. In contrast, multimodal performance in the exploratory stress test exhibited a negative divergence. In pure imaging tasks featuring complex but clinically representative pathologies, the average diagnostic score was poor, with models failing to identify core pathologies. In comprehensive diagnostic analysis tasks (Q29-Q30), models exhibited \"textual masking,\" aligning visual findings with text cues rather than imaging evidence. Crucially, safety disclaimer coverage dropped precipitously from 82.8% (207/250) in textual tasks to 28% (14/50) in multimodal interactions.&lt;/p&gt;&lt;","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e93522"},"PeriodicalIF":5.0,"publicationDate":"2026-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13494679/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148802274","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prediction of Atrial Fibrillation Occurrence With Handheld Mobile Electrocardiogram: Deep Learning Model Development Using Real-World Data.","authors":"Minje Park, Hyun Jin Ahn, Yeongyeon Na, Sunghoon Joo, Young Ho Lee, Seongwoo Han, Myung Soo Park, Dae Young Cheon, Jeen Hwa Lee, Ki Hong Lee","doi":"10.2196/87142","DOIUrl":"10.2196/87142","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Atrial fibrillation (AF) is a common arrhythmia associated with an increased risk of stroke and heart failure. To improve prevention, recent studies have used deep learning models to identify at-risk individuals early from normal sinus rhythm (NSR). However, studies using mobile electrocardiogram (mECG) in outpatient, real-world settings remain underexplored.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;The study aimed to develop and evaluate deep learning models using a real-world limb-lead mECG database to predict the short-term occurrence of AF from NSR recordings.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;mECG data were collected from real-world users of commercially available handheld mECG devices capable of capturing 6 limb leads. AF occurrence was defined as an AF event within a predefined time window (7, 14, or 31 d) from the date of the NSR recording. Transformer-based prediction models were developed for limb-lead and lead I input configurations using a multistage training approach with self-supervised pretraining and domain adaptation, drawing on both open, large-scale clinical 12-lead ECG and proprietary real-world mECG databases. The models were evaluated in an internal real-world cohort and explored in an external cohort as a proof of concept via time-to-event analysis.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Between March 2023 and November 2024, 386,519 mECGs were acquired from 8206 users. There were 18,949, 25,206, and 33,524 AF incidences within the 7-, 14-, and 31-day time windows. The models were pretrained with 787,257 12-lead ECGs and 202,689 mECGs, then fine-tuned to predict AF occurrence using 97,447 labeled mECGs. The limb-lead models achieved areas under the receiver operating characteristic curves (AUROCs) of 0.793, 0.785, and 0.787 for 7-, 14-, and 31-day predictions on the internal cohort, respectively, with a user-level AUROC of 0.702 for the 31-day prediction. These models significantly outperformed the lead I models (&lt;i&gt;P&lt;/i&gt;&lt;.001), supporting the value of multilead configurations. The multistage pretraining was essential, as single-source pretraining yielded lower AUROCs of 0.555 with mECGs only and 0.761 with 12-lead ECGs only for the 31-day prediction. In the subgroup analysis, AUROC values were consistent across age, PR interval, and corrected QT interval, but showed disparities (&lt;i&gt;P&lt;/i&gt;&lt;.001) by sex (0.713 in females vs 0.794 in males) and by QRS duration (0.583 in ≥120 ms vs 0.796 in &lt;120 ms). In the external cohort (n=144), the 31-day model stratified all 5 new-onset AF events, showing significantly different survival functions between the positively and negatively predicted groups (&lt;i&gt;P&lt;/i&gt;=.03); Cox proportional hazards regression yielded a hazard ratio of 1.49 (95% CI 1.06-2.09) per 0.1 increase in model output.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Our findings elucidate the feasibility of deep learning-based AF risk prediction using single-NSR recordings from mobile devices, highlighting the potential ","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e87142"},"PeriodicalIF":5.0,"publicationDate":"2026-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13492302/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148801584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Retraction: A Deep Neural Network for Estimating Low-Density Lipoprotein Cholesterol From Electronic Health Records: Real-Time Routine Clinical Application.","authors":"","doi":"10.2196/108367","DOIUrl":"10.2196/108367","url":null,"abstract":"","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"14 ","pages":"e108367"},"PeriodicalIF":5.0,"publicationDate":"2026-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13539173/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148801596","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}