Rheumatic diseases clinics of North AmericaJeffrey R Curtis, Emily Holladay, Tapan Mehta, John Osborne
Rheumatic diseases are chronic, heterogeneous, and longitudinal, and assembling real-world evidence for effectiveness and safety for their study is best served by integrating diverse data types. This article describes the infrastructure required to support scalable, trustworthy artificial intelligence (AI) in rheumatology, emphasizing data acquisition, harmonization, linkage, privacy protection, and computational environments. We outline computing infrastructure considerations relevant to rheumatology, including hybrid on-premises and cloud architectures. Sustained progress for AI applied to rheumatology will depend on deliberate investment in shared infrastructure, longitudinal data ecosystems, and governance models that balance innovation, privacy, reproducibility, and equitable clinical value.
The clinical teacherJohn M Cunningham, Luyang Jin, Iqra Kazi, Amith Skandhan, Ronald Rodriguez
BACKGROUND: The rapid introduction of AI into clinical practice shifts how we must teach resident trainees so they may become ethical patient-facing clinicians in an AI-integrated healthcare system. Currently, few published innovations assess outcomes beyond learner attitudes. We developed a pilot curricular innovation to equip postgraduate Internal Medicine resident trainees with the attitudes and knowledge needed to responsibly integrate AI tools into patient care decisions. APPROACH: In the 2025-2026 academic year, we piloted a curricular innovation to teach resident physicians the basics of prompting strategies for AI-assisted clinical reasoning, ethical AI use and legal considerations. The innovation consisted of an initial didactic followed by a hands-on, interactive session integrating AI prompts and outputs into clinical vignettes, thereby leveraging near-peer teaching and situated learning to achieve session objectives. EVALUATION: We assessed perceived knowledge and knowledge using a pre-post intervention strategy using the Wilcoxon Rank-Sum test. Fifty-nine/96 (61.5%) and 52/96 (54.2%) of residents participated in the presession and post-session survey, respectively. Perceived knowledge increased significantly across all five learning objectives with a moderate to large effect size. Fifty-one residents participated in the pre- and post-session knowledge test. The median pre-session score was 6/8 (interquartile range [IQR] 4-8), and the median post-session score was 7/8 (IQR: 5-8); p < 0.001, with a moderate effect size = 0.33. IMPLICATIONS: A combined didactic and small-group interaction session improved residents' perceived understanding and knowledge of ethical and legal considerations related to clinical AI use. Future work developing clinical assessments of trainee skills using AI tools is needed.
Renal failureKim Solez, Habba Mahal, Wisit Cheungpasitporn, Giuseppe Orlando
The purpose of this review is to summarize the most influential and conceptually significant publications from the past 2 years, including substantial 2026 publications, and to identify emerging directions likely to shape xenotransplantation and regenerative medicine in the near future. Advances in artificial intelligence (AI) now support more structured anticipation of future developments by integrating patterns across experimental, computational, and translational research. The field is approaching a potential inflection point in which increasingly capable AI systems, potentially approaching artificial general intelligence, may accelerate the design of stem-cell-derived tissues and progressively more complex organ constructs. In addition, scientific communication is evolving toward formats that support machine-assisted analysis and AI-driven knowledge synthesis. Multiple developments signal significant expansion across xenotransplantation and regenerative medicine, driven by innovations in gene editing, multimodal data integration, and AI-enabled prediction and decision-support systems. These advances will help to broaden access to transplantable organs and increase the scale and impact of the field across clinical practice, research, and workforce domains. Together, these trends suggest that AI-enabled regenerative and xenogeneic strategies may meaningfully reduce the organ shortage and support future progress toward precision-engineered organ replacement.
This cross-sectional study aimed to evaluate the quality, understandability, actionability, and readability of patient-oriented information generated by ChatGPT regarding overactive bladder (OAB). A total of 32 frequently asked patient questions related to OAB were categorized into 6 domains, including general information, diagnosis, lifestyle and behavioral management, medical treatment, minimally invasive treatments, and surgical treatment, and submitted to ChatGPT-4o. The responses were evaluated independently by 2 researchers using the DISCERN instrument to assess information quality and the Patient Education Materials Assessment Tool - Printable version to evaluate understandability and actionability. Readability was assessed using the Flesch-Kincaid Grade Level and Simple Measure of Gobbledygook formulas. The overall mean DISCERN score was 51, indicating moderate information quality across categories. Patient Education Materials Assessment Tool analysis demonstrated relatively high understandability (81.8%) but limited actionability (45.5%). Readability analysis revealed that the responses exceeded recommended patient education standards, with overall median Flesch-Kincaid Grade Level and Simple Measure of Gobbledygook scores of 14.00 (range: 7.61-17.11) and 12.34 (range: 5.01-14.93), respectively. Category-based analyses demonstrated variability across content domains, with treatment-related responses showing relatively higher information quality and lifestyle-related responses demonstrating greater actionability. Although ChatGPT-generated responses demonstrated moderate information quality and were generally understandable, limitations related to actionability and readability may restrict their practical use in patient education. These findings suggest that ChatGPT may serve as a supportive tool for patient education in OAB; however, it should be considered a complementary resource used under physician supervision rather than a substitute for clinical guidance.
Triple-negative breast cancer (TNBC), being one of the most aggressive subtypes of breast malignancies, is characterized by poor prognosis and limited treatment options. As a leading cause of cancer-related mortality among women, TNBC poses unique clinical challenges due to the absence of effective targeted therapies. Conventional treatment strategies, including chemotherapy, often suffer from significant drawbacks such as drug resistance and intolerable side effects, underscoring an urgent need for innovative approaches to improve therapeutic outcomes. A defining hallmark of TNBC is its markedly altered cellular metabolism, which not only supports tumor growth and survival but also contributes to therapy resistance. Elucidating these metabolic alterations could provide critical insights into potential vulnerabilities that may be exploited for therapeutic intervention. This review provides a comprehensive overview of the major metabolic pathways that are dysregulated in TNBC and their relevance to disease progression and therapeutic intervention. Finally, we explore recent advances in metabolomics-driven precision medicine, highlighting the integration of artificial intelligence and machine learning approaches, covering aspects such as data processing, feature selection, and model construction that can accelerate the advancement of personalized treatment strategies for TNBC.
JMIR cancerLaiba Husain, Megan Mullins, Bella Etingen, Raaed Mohammed Zafar, Mediha Siddiqui, George L Jackson
BACKGROUND: Tele-oncology addresses geographic barriers to cancer care, but implementation challenges persist in rural settings. AI-enhanced predictive analytics offer opportunities for optimizing deployment through personalized, data-driven strategies; however, evidence in rural tele-oncology contexts remains limited, and critical equity considerations remain underexamined. OBJECTIVE: This scoping review aimed to map evidence on AI-enhanced predictive analytics in tele-oncology implementation, with particular attention to rural and underserved populations, to identify research gaps and inform implementation science priorities. METHODS: We searched 5 databases (PubMed, Embase, CINAHL, Web of Science, and IEEE Xplore) using 4 concept domains (tele-oncology, rural implementation barriers, AI or predictive analytics, implementation science) from January 2015 through November 2025. Two independent reviewers screened 330 unique records (title or abstract; Cohen κ=0.78), with the principal investigator resolving conflicts. Of 138 full-text reviews (κ=0.82), 4 studies met inclusion criteria. Data extraction captured study characteristics, AI applications, implementation factors, and outcomes. We used narrative thematic analysis to map findings into three themes: (1) the current tele-oncology implementation landscape in rural and underserved settings, (2) potential AI applications addressing implementation challenges, and (3) implementation considerations for AI systems themselves. RESULTS: Four included studies (1 pilot feasibility study, 1 proof-of-concept validation study, 1 cross-sectional predictive study, and 1 platform development study; published 2019-2025) demonstrated limited evidence at the intersection of AI, tele-oncology, and rural health equity. Patient characteristics predicted telehealth modality preferences with 86.2% accuracy, revealing that male patients exhibited 66% increased odds of video selection versus female patients (P=.004), and urban residents showed 101% increased odds compared to rural counterparts (P<.001). Liu et al demonstrated that disadvantaged populations engaged with AI-generated health literacy content 2.52-fold more frequently than nondisadvantaged counterparts. However, all 4 studies documented substantial implementation barriers (patient, provider, organizational, and system levels) persisting despite technological sophistication. Organizational threshold effects, where remote monitoring interventions succeeded with adequate provider capacity but failed under resource constraints-suggest that algorithmic innovations cannot overcome structural limitations in rural facilities. No studies explicitly examined algorithmic bias, cross-population validation, or potential harms in rural contexts. Geographic concentration in high-resource countries (United States n=2, Greece n=1, and Singapore n=1) and limited oncology-specific focus underscore structural gaps in knowledge generation for underserved populations. CONCLUSIONS: Current evidence remains insufficient to support definitive practice recommendations. The observed evidence gap may reflect broader structural inequities in knowledge generation: populations with the greatest implementation challenges appear to remain substantially underrepresented in AI and digital health literature. Future research should prioritize comparative effectiveness studies in authentic rural contexts with implementation science outcomes, equity-centered cross-population validation, specification of translation mechanisms linking AI predictions to implementation strategies, health economic analyses, and mechanistic research on sociotechnical integration factors, ensuring technological innovation reduces rather than perpetuates disparities in cancer care.
Journal of medical Internet researchAndy J King, Anthony Banks, Leandra H Hernández, Sabrina S Thompson, Leticia Stevens, Lindsey N Potter, Kimberly A Kaphingst, Paul A Estabrooks, Guilherme Del …
Large language models (LLMs) are increasingly embedded in clinical and population health workflows, including conversational agents such as health chatbots. As chatbots evolve from rule-based approaches to hybrid and LLM-enabled designs, risks and concerns about deployment readiness shift. Unlike rule-based chatbots, LLM outputs can be unpredictable, error-prone, and difficult to validate with traditional evaluation methods. Public health teams integrating customized LLMs into interventions face practical and ethical challenges related to performance variability, uncertainties about model behaviors, and inequitable performance across languages. Although existing frameworks address domains such as safety, ethics, effectiveness, engagement, and implementation, they often assume or imply-rather than operationalize-an explicit benchmark for deployment and implementation decisions. We propose an acceptance criteria framework (ACF) to determine implementation fit, defined as meeting prespecified minimum performance standards and demonstrating nonproblematic behavior under anticipated use. The ACF uses project-relevant and off-topic prompts, structured expert review, and prespecified thresholds to produce a documented decision record that can be iteratively rerun after model revisions. We demonstrate the framework through a case application in a tobacco cessation text messaging intervention, illustrating how the ACF can guide deployment decisions.
Prehospital and disaster medicineYaning Lai, Xiaoqin Lai, Hai Hu
OBJECTIVE: Traditional lecture-based learning (LBL) is often insufficient for cultivating the practical decision-making skills required in high-stakes environments like Emergency Medical Response (EMR). While game-based learning (GBL) offers an immersive alternative, it can lack immediate expert guidance. This study addresses this gap by exploring the integration of generative Artificial Intelligence (AI) as an "intelligent tutor" within GBL. The objective was to evaluate and compare the effectiveness of LBL, GBL, and generative AI-powered game-based learning (AI-GBL) on medical students' knowledge acquisition, retention, learning motivation, and cognitive load in an EMR course. METHODS: A retrospective, comparative study was conducted with 86 medical students from three consecutive cohorts (2022-2024), each exposed to one of the three teaching modalities (n = 29 LBL, n = 28 GBL, n = 29 AI-GBL). Knowledge was assessed via pre-test, post-test, and final-test scores with a maximum score of 10 points. Student feedback was collected for learning motivation, cognitive load, and technology acceptance. RESULTS: For immediate knowledge acquisition, both GBL (mean difference = 1.124/10 points; 95% CI [0.297, 1.952]; P = 0.008) and AI-GBL (mean difference = 0.897/10 points; 95% CI [0.076, 1.717]; P = 0.033) significantly outperformed LBL. For delayed knowledge retention, the AI-GBL group demonstrated significantly superior retention compared to both the GBL group (mean difference = 0.689 points; unadjusted 95% CI [0.080, 1.299]) and the LBL group (mean difference = 1.310 points; unadjusted 95% CI [0.706, 1.915]). The AI-GBL group also reported significantly lower cognitive load than the GBL group (mean difference = -0.273 points; unadjusted 95% CI [-0.456, -0.090]). Finally, students perceived the AI-powered approach as significantly more useful than the standard game-based approach (mean difference = 0.513 points; unadjusted 95% CI [0.137, 0.889]). CONCLUSION: The AI-enhanced GBL model for EMR training improves knowledge acquisition and retention while reducing cognitive load, representing a promising approach for developing proficiency in complex, high-stakes medical competencies.
Journal of medical Internet researchHaoran Chen, Shenglan Xiao, Tong Wan, Gui Li, Yanhong Peng, Zhimin Wang
BACKGROUND: Hypertension is a major global health challenge, and effective health education is crucial for improving patients' self-management. Traditional health education approaches are often limited by insufficient personalization, accessibility, and scalability. Artificial intelligence (AI), including natural language processing, machine learning, and large language models (LLMs), offers promising solutions to address these limitations. However, evidence regarding AI applications in hypertension health education has not been comprehensively synthesized. OBJECTIVE: This scoping review aimed to summarize the current evidence on AI applications in hypertension health education, and identify research gaps to inform future research and practice. METHODS: This review followed the Joanna Briggs Institute methodology and PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines. Six databases (PubMed, Embase, Web of Science, Cochrane Library, CINAHL, and Scopus) were searched from January 2015 to June 2026. Eligibility criteria were developed using the participant-concept-context framework. Two reviewers independently conducted study screening and data extraction. Study designs were classified using the Mixed Methods Appraisal Tool framework. Consistent with scoping review methodology, no formal quality assessment was performed. Findings were synthesized narratively and presented using evidence gap maps, tables, and figures. RESULTS: A total of 24 studies from 11 countries were included, comprising 6 randomized controlled trials, 4 nonrandomized trials, 11 quantitative descriptive studies, and 3 mixed methods studies. Most studies were published between 2024 and 2026. In total, 3 AI application scenarios were identified: rule-based health education, data-driven adaptive health education, and generative AI-driven health education. Natural language processing was the most widely applied technology, and LLM-based applications increased rapidly after 2023. However, generative AI studies were predominantly proof-of-concept evaluations and lacked randomized clinical validation. Health education was rarely implemented as a standalone intervention and was typically embedded within multifunctional AI platforms. Outcomes were categorized using the Digital Health Scorecard Framework across 4 domains: technology, clinical, usability, and cost. Technical accuracy and blood pressure outcomes were the most frequently reported measures, whereas no study evaluated economic outcomes. CONCLUSIONS: This first scoping review of AI applications in hypertension health education identified a mismatch between rapid advances in generative AI and the limited availability of rigorous clinical evidence. Three major research gaps were identified: (1) the lack of standardized core outcome sets covering technical, behavioral, clinical, and implementation domains; (2) limited development of hybrid architectures integrating LLM with structured medical knowledge bases; and (3) the absence of evaluation frameworks that satisfy both regulatory and implementation requirements. AI appears most suitable as a complement to, rather than a replacement for, clinician-delivered education. Future research should prioritize rigorous clinical validation, economic evaluation, multicultural adaptation, and health literacy equity to ensure that AI-driven health education reduces rather than exacerbates disparities in hypertension control.
Journal of biomedical informaticsSumin Lee, Kyulee Jeon, Joon Seok Lim, Yiju Park, Min Seong Kim, Ju Hyun Jin, Chang Hoon Han, Soon Ho Yoon, Seng Chan You
OBJECTIVE: To develop an ontology-driven framework that standardizes heterogeneous local radiology procedure names by decomposing them into semantic components and aligning them to LOINC/RSNA Radiology Playbook codes using constrained large language model (LLM)-based parsing and selection. METHODS: Radiology procedure names from two tertiary hospitals in Korea were parsed into semantic components using LLM prompting with retrieval-augmented generation, aligned with the LOINC/RSNA Radiology Playbook (version 2.80). Ontology-based similarity scoring quantify correspondence between parsed components and Playbook candidates' attributes, and retrieve the Top 10 candidates, followed by LLM-based selection within this candidate set. Performance was evaluated against direct Playbook code name-based mapping and conventional similarity metrics using a radiologist-curated gold reference. RESULTS: A total of 3,326 local procedure names were analyzed. Ontology-based mapping approach substantially outperformed direct Playbook code name mapping across all evaluation metrics. At the candidate retrieval stage, ontology-based attribute matching achieved recall@5 of up to 0.78 (internal) and 0.89 (external), compared with 0.51 and 0.47 for direct mapping. After LLM-based selection, the ontology-based approach achieved a final selection recall@1 of up to 0.70 (internal) and 0.81 (external), exceeding direct mapping (0.48 and 0.50) more than 20 percentage points in both settings (p < 0.001). CONCLUSION: Decomposing procedure names into ontology-grounded semantic components enables robust handling of heterogeneous local terminology, while constraining LLM reasoning to structured selection tasks mitigates hallucination and preserves semantic fidelity. Ontology-driven knowledge encoding provides a scalable and reliable approach to standardizing radiology procedure names, supporting cross-institutional interoperability and secondary data use of imaging research.
Journal of biomedical informaticsDaniel J Tan, Jiayang Chen, Dilruk Perera, Kay Choong See, Mengling Feng
OBJECTIVE: Enteral nutrition (EN) delivery in the ICU remains suboptimal due to limited personalization and uncertainty regarding appropriate calorie, protein, and fluid targets under dynamic metabolic demands. We introduce DeepEN, a reinforcement learning (RL) framework for personalized EN optimization using electronic health record data. METHODS: DeepEN was trained on over 11,000 ICU patients from MIMIC-IV to generate 4-hourly, patient-specific caloric, protein, and fluid targets. The state representation incorporated demographics, comorbidities, vital signs, laboratory values, and recent interventions. A physiologically aligned reward framework balanced biomarker stability with long-term survival. Policy learning employed a dueling double deep Q-network with Conservative Q-Learning regularization to enable safe offline training. RESULTS: DeepEN achieved the highest estimated policy value (Vπ=9.48) and the lowest calibrated mortality (18.8 ± 1.0%), representing a 4.0 percentage-point absolute reduction compared with clinician practice (22.8%). The policy also demonstrated superior metabolic stability, achieving the highest proportion of glucose, phosphate, and sodium values within target range. Furthermore, deviation from the DeepEN policy was independently associated with increased mortality and biomarker instability, whereas deviation from a random policy showed no such association. Interpretability analyses further indicated that recommendations were conditioned on physiologically relevant markers of organ function and metabolic status rather than static dosing heuristics. CONCLUSION: DeepEN demonstrates the feasibility of conservative offline RL for safe, individualized EN optimization, highlighting the potential of data-driven personalization to complement guideline-based approaches in critical care.
Rheumatic diseases clinics of North AmericaShannon Amerilda Scielzo, Salahuddin Dino Kazi
Artificial intelligence (AI) is revolutionizing our approach to medical care in Rheumatology. From significant forthcoming changes in clinical care approaches, to changes in patient perspectives and the way they approach care, to our training approaches, changes in educational needs will be extensive. Similarly, AI will greatly support many components of training physicians but also comes with many potential concerns. Ethical expertise around utilization of AI, along with guidelines to guide practice are needed. Moreover, extensive research will be needed to help achieve these goals and support the successful integration of AI into rheumatology practice.
Rheumatic diseases clinics of North AmericaAugusto Garcia-Agundez, Megan Creasman, Gabriela Schmajuk, Jinoos Yazdany
Generative artificial intelligence (GenAI) is rapidly entering clinical workflows, yet its role remains incompletely defined. This review evaluates current and emerging applications of GenAI across common rheumatology activities using a task-based framework. The authors summarize evidence on mature tools such as AI scribes, emerging applications such as chart summarization and information extraction tools, and future opportunities in clinical prediction. While GenAI consistently reduces cognitive burden and improves documentation-related experience, efficiency gains are modest and important risks, especially automation bias and hallucinations, remain. With careful oversight and evaluation, GenAI has significant potential to enhance rheumatology practice.
Rheumatic diseases clinics of North AmericaJamie E Collins, Rayan Harari, Jeffrey Duryea
Artificial intelligence (AI) refers to a broad class of computational methods to perform tasks that typically require human intelligence, such as learning patterns, reasoning, and problem solving. AI is increasingly applied across rheumatology research and clinical practice to analyze complex data and support clinical decision-making, yet its underlying concepts are often perceived as opaque or inaccessible. This aim of this article is to demystify foundational AI concepts by defining core terminology and describing major learning paradigms and analytical methods, highlighted through clinically relevant examples from rheumatology. We explain how modern AI models differ from traditional analytical approaches.
Rheumatic diseases clinics of North AmericaMegan Creasman, Augusto Garcia-Agundez, Jinoos Yazdany, Gabriela Schmajuk
Rheumatology machine-learning models are limited by preexisting, technical, and emergent biases; the interaction of data constraints, design choices, and real-world clinical workflows, rather than from isolated technical errors. Across the model lifecycle, optimization objectives can encode patterns of care, access, and documentation, producing hidden subgroup failures that are obscured by aggregate performance metrics. Given the heterogeneity of rheumatic disease and current disparities in care delivery, addressing bias requires deliberate design choices before, during, and after a model is built, as well as a commitment to transparency, and sustained oversight.
Rheumatic diseases clinics of North AmericaElena Myasoedova, Cynthia S Crowson
This article highlights Mayo Clinic's pioneering efforts to integrate artificial intelligence (AI) and machine learning into rheumatology, focusing on genomics, imaging, pathology, and clinical data science to improve diagnosis, treatment and operational efficiency. Key innovations include transformer-based models for genomic analysis, autonomous ultrasound devices, multimodal imaging solutions, and generative AI tools for clinical documentation and patient education, all aimed at bridging the gap between research and routine clinical care. The article emphasizes the need for rigorous validation, explainable AI, electronic health records integration, clinician training, and global collaboration to ensure safe and effective adoption of AI-powered tools in clinical practice.
Rheumatic diseases clinics of North AmericaReece Blay, Amanda E Nelson
This article summarizes key advancements of artificial intelligence (AI) for rheumatic and musculoskeletal disease imaging in the diagnosis and classification, and predictive modeling of rheumatoid arthritis, psoriatic arthritis, spondyloarthritis, and osteoarthritis since 2020. AI applications are emerging in disease diagnosis, severity classification, and prediction of incidence and progression, with ongoing challenges related to external validation, mitigation of bias, data privacy, transparency, and clinical integration. In the near term, AI could assist clinicians with diagnostic interpretation and disease monitoring. Future applications include improved prognostic modeling and identifying candidates for targeted interventions and clinical trials.
Rheumatic diseases clinics of North AmericaIlana M Usiskin, Maria I Danila, Tianxi Cai, Katherine P Liao
Clinical decision support systems (CDSS) have the potential to enhance rheumatology practice by assisting with differential diagnosis, treatment decisions, and predicting patient outcomes. Rheumatic conditions are complex diseases largely diagnosed clinically rather than with a predefined set of clinical, laboratory, or imaging findings. Newer artificial intelligence (AI)-driven CDSS, known as machine-learning-based systems, may be better suited for applications to rheumatic conditions compared to knowledge-based systems, which rely on a set of predefined rules. This report reviews existing studies on CDSS in rheumatology and highlights the potential benefits and pitfalls of incorporating AI-driven CDSS into rheumatology practice.
Rheumatic diseases clinics of North AmericaVictor Mocanu, Farbod Moghaddam, Mina Aminghafari, May Y Choi
The growing implementation of machine learning (ML) has extended into autoantibody research for the study of systemic autoimmune rheumatic diseases (SARDs). ML methods offer a promising approach for efficiently handling and identifying important signals within the big data generated by modern autoantibody technologies. The novel biomarkers identified through advanced ML techniques show promise in outperforming current clinical tools, bringing us closer to the goal of precision medicine. In this article, we will provide an overview of ML approaches and how they have been applied in autoantibody research to improve the diagnosis and characterization of SARDs.
Rheumatic diseases clinics of North AmericaSuzanne Tamang
Artificial intelligence (AI) is increasingly embedded in clinical tools used in rheumatology, including imaging interpretation, longitudinal disease monitoring, and electronic health record-based decision support. AI has moved from the periphery of biomedical research to an operational component of clinical care, increasingly embedded in electronic health records, imaging platforms, and decision support systems. In rheumatology, where care is longitudinal, AI systems offer substantial promise-but also poses distinct risks. Together, these commitments-advancing humanity, ensuring equity, engaging impacted individuals, improving workforce well-being, monitoring performance, innovating and learning, and promoting sustainability-operationalize trustworthy AI systems across a patient's care trajectory.