Frontiers in digital health最新文献

{"title":"Synthetic gatekeeping in health chatbots: configured answer systems as the unit of AI evaluation and governance for medical information.","authors":"Simona Wójcik, Justyna Domienik-Karłowicz, Anna Rulkiewicz","doi":"10.3389/fdgth.2026.1976313","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1976313","url":null,"abstract":"<p><p>Two health chatbots can give the same clinically appropriate advice and still constitute different interventions: one stops at a clinician referral, the other places an affiliated booking one click away. In this Hypothesis and Theory article we argue that the appropriate unit of health-AI evaluation is therefore not the model output alone, but the configured answer system - the large language model (LLM) together with source selection, retrieval, interface, action routing and downstream data return. We call this organizationally configured mode of mediation synthetic gatekeeping. Its necessary core comprises source selection and generative synthesis, with personalization and action routing as extending dimensions; its observable manifestations are content and pathway traces, and its proposed mechanism is infrastructural steering. Content parity does not guarantee pathway parity: two systems can name the same options while making only one easily actionable. Organizational position, revenue architecture and integration depth are treated as exposures that may shape these functions, not as elements of the construct, and owner-linked advantage is an empirical outcome to be demonstrated rather than assumed. Direct health evidence of owner- or revenue-linked steering remains sparse, and one comparative study found LLM answers less commercially biased than conventional search; the mechanisms nonetheless appear in adjacent domains, and health outputs are configurable at deployment level. We advance three core propositions - organizational alignment (P1), pathway asymmetry (P2) and behavioural mediation (P4) - with one conditional moderator hypothesis on conversion-linked funding (P3) and a measurement corollary on surface-pathway divergence (M1), each stated with its weakening condition. The framework yields a practical audit model based on configuration anchoring, paired content-pathway comparison, organizational disclosure and proportionate governance triggers. The unit of evaluation should move from the model output to the configured answer system in which information becomes action.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1976313"},"PeriodicalIF":5.2,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13677928/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225500","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":"A real-time, gamified approach for optimizing intra-hospital transport of time-critical biospecimens.","authors":"Adrian Bundschuh, Annalena Artinger, Rebecca Schönmehl, Christoph Brosius, Jens Langejürgen, Lina Winter, Christoph Brochhausen","doi":"10.3389/fdgth.2026.1920674","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1920674","url":null,"abstract":"<p><p>Intra-hospital transport of biological specimens is a time-critical, preanalytical step. However, especially for diagnostic or scientific critical biospecimens it remains largely manual, poorly tracked, and decoupled from performance feedback. Existing solutions commonly address isolated parts of the logistics chain but fall short in engaging the personnel who perform it. Therefore, we propose a mobile-first, gamified concept for real-time specimen tracking that combines QR-based documentation with behavioral incentives. Each transport is recorded automatically through start and end scans, closing a gap in the pre-analytical record without adding documentation effort. Beyond external rewards, the concept emphasizes determinants of intrinsic motivation such as task significance to support sustainable engagement. By combining technological and behavioral optimization, this perspective outlines a strategy for time saving, transparent and efficient specimen logistics.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1920674"},"PeriodicalIF":5.2,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13677696/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225530","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":"Correction: Selecting medical research data platforms for translational biomedical research: a five-tier overview and requirement-weighted assessment framework.","authors":"Marc Jacobs, Samad Goudarzi, Jan Stücke, Robin Röhm, Timo Kanninen, Tero Oinonen, Petra Ritter, Michael Schirner, Fabian Prasser, Arman Bhagwagar, Bernd Matzenbach, Hartmut Schultze, Joachim Schultze, Mathias Göschl, Henrik Matthies, Kenneth R Evans, Moyez Dharsee, Christian Stephan, Marvin Belz, Philip Gribbon, Katja Herzog, Chester Chen, David Ruau, Ittai Dayan, Adrish Sannyasi, Dieter Kranzlmüller, Naweiluo Zhou, Jens Warfsmann, Martin Hoffmann, Rudi Schmidt, Martin Hofmann-Apitius","doi":"10.3389/fdgth.2026.2006683","DOIUrl":"https://doi.org/10.3389/fdgth.2026.2006683","url":null,"abstract":"<p><p>[This corrects the article DOI: 10.3389/fdgth.2026.1814015.].</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"2006683"},"PeriodicalIF":5.2,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13678303/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225523","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":"Voice and speech as digital biomarkers of mental health: a scoping review of clinical interpretability, data collection and translational readiness.","authors":"Ralph Justin Uy, Ibrahim Shaaban, Angel Arnaout","doi":"10.3389/fdgth.2026.1964389","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1964389","url":null,"abstract":"<p><strong>Background: </strong>Voice and speech encode acoustic, linguistic, and interactional information associated with mental health. Translating these signals into clinically meaningful digital biomarkers depends not only on predictive performance but on how speech is elicited, how the target state is defined, whether findings generalize beyond the original dataset, and whether features can be measured reliably in real-world settings.</p><p><strong>Objective: </strong>To map voice, speech, language, and communication biomarkers studied across mental health outcomes and to examine three dimensions of translational readiness: clinical interpretability, evidentiary validity, and deployability.</p><p><strong>Methods: </strong>This scoping review followed Joanna Briggs Institute guidance and PRISMA-ScR. PubMed, PsycINFO, and Scopus were searched through June 18, 2026. Two reviewers independently screened records and charted data. Biomarkers were classified as directly clinically interpretable, software-derived with a perceptual correlate, or technical-only.</p><p><strong>Results: </strong>Of 1,524 records identified, 68 studies were included. Depression was examined in 56 studies (82%), followed by anxiety (<i>n</i> = 9), bipolar disorder or mania (<i>n</i> = 8), fatigue (<i>n</i> = 4), post-traumatic stress disorder (<i>n</i> = 3), stress (<i>n</i> = 2), and other outcomes including suicidal risk and loneliness (<i>n</i> = 10); no study examined burnout or emotional exhaustion. Controlled elicitation predominated (54/68, 79%) over naturalistic (16/68, 24%) and passive or ambient capture (3/68, 4%). One study focused on healthcare workers and one analyzed clinician-patient dyadic speech. Across depression studies, clinically interpretable patterns commonly included slower speech, more or longer pauses, lower or less variable pitch, reduced vocal intensity, flatter prosody, and reduced verbal output, although findings were not uniform. Validation on an independent or cross-dataset sample was reported by only 2 of 68 studies (3%).</p><p><strong>Conclusions: </strong>Voice-based mental health research identifies numerous candidate biomarkers, but translational readiness remains limited. Most evidence derives from depression and controlled speech tasks, and interpretability does not imply validity or deployability. Existing studies do not establish that candidate biomarkers remain valid in spontaneous, multispeaker, passive, or ambient clinical settings. Future work should prioritize validated outcome labels, longitudinal naturalistic capture, external validation, diverse populations, and explicit evaluation of interpretable and technical features across recording contexts.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1964389"},"PeriodicalIF":5.2,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13677924/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225514","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":"An agentic AI- and n8n automation-powered maternal care framework with predictive pregnancy monitoring and continuous learning.","authors":"Subhash Thippa, Nagesparan Ainarappan, R Suganya","doi":"10.3389/fdgth.2026.1840650","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1840650","url":null,"abstract":"<p><p>Conventional maternal remote patient monitoring has relied upon using immutable ranges for detection of conditions on maternal telemetry. However, this has created extreme alert fatigue in clinicians. The newly proposed Agentic AI Maternal Triage Framework combines patient-reported outcome data with quantitative telemetry values to deliver real-time, context-aware decisions regarding risk stratification to both the clinician and patient. The triage workloads are processed in an orchestrated manner via n8n, utilizing light gradient boosting machine models to determine the likelihood that the maternal telemetry is showing that the patient is developing cardiovascular or metabolic risks. The models produce continuous risk scores that are consumed by a retrieval-augmented generation-powered Google Gemini large language model to yield structured triaged outputs. The triage outputs are delivered to a conversational agent for low-risk patients and a physician for high-risk patients using a triage model of escalation. To allow for continued learning, expert validation is captured via stateless webhooks. In the simulation-based evaluation, the predictive models achieved classification metrics above 99.8%; however, these results reflect the controlled synthetic data-generating environment and do not establish clinical validity. The framework therefore represents a technical proof of concept for integrating predictive modeling, retrieval-grounded generative reasoning, context-aware routing, and human-in-the-loop feedback, with external clinical validation required before real-world deployment.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1840650"},"PeriodicalIF":5.2,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13677860/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225496","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":"Algorithmic hyperindividualisation and the transformation of vulnerability in substance use prevention.","authors":"Christoph Dockweiler, Carlotta Louisa Westfeld, Tessa Schulenkorf","doi":"10.3389/fdgth.2026.1956219","DOIUrl":"10.3389/fdgth.2026.1956219","url":null,"abstract":"<p><p>Generative artificial intelligence is entering substance use prevention through applications that tailor health information to the individual in real time. These systems promise to reach vulnerable groups more precisely than mass communication, and they rest on individual-level models of behaviour change. This article asks what happens to the concept of vulnerability once it becomes something such a system can read and act upon. It draws on the public health literature on vulnerability as a multilevel construct, the theory of fundamental causes, Rose's prevention paradox, and critical work on algorithmic systems and precision public health. On this basis it develops the Vulnerability Transformation Hypothesis. The article first distinguishes hyperindividualisation from the tailoring it intensifies. What is new is the inference of undeclared attributes, continuous re-profiling, open-ended generation, and operation at near-zero marginal cost per message. The hypothesis is stated as three claims of decreasing certainty. A near-analytic addressability claim holds that a system addressing users as individuals has only individual-level levers. A substantive and conditional re-encoding claim states that, for any system presenting itself as a means of reaching the vulnerable, the operationalised concept of vulnerability is reconstructed into a constellation of individual attributes. Structural information that enters an individual model returns as a predictor of personal risk rather than as a target of action. This transformation is representational rather than ontological, since the structural causes persist even as they recede from what the system can represent and change. A contingent displacement claim adds that, at the level of the field rather than the single system, the success of individualised tools may displace structural effort, a tendency documented as lifestyle drift. The article separates its claim from the precision public health debate and offers four falsifiable predictions. Among them is a discursive test of whether systems that invoke vulnerability then present individual tailoring as the means of addressing it. Such a pattern would evidence substitution of the concept, not mere inconsistency. The article illustrates the coding scheme across contrasting systems, so that others can test and revise the hypothesis.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1956219"},"PeriodicalIF":5.2,"publicationDate":"2026-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13672495/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149216721","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":"Harmonizing Vietnamese traditional and western medicine: a digital health blueprint and the IMKE architectural framework.","authors":"Anh Nguyen Thi Dieu, Hieu Le Ngoc","doi":"10.3389/fdgth.2026.1928965","DOIUrl":"10.3389/fdgth.2026.1928965","url":null,"abstract":"<p><p>The concurrent use of traditional and Western medicine is a global healthcare reality. In countries like Vietnam, dual-paradigm clinical guidelines actively mandate integrated treatments for high-prevalence chronic diseases. However, safe integration at the point of care is severely hindered by data fragmentation, profound epistemological differences, and a critical \"shadow pharmacovigilance\" crisis regarding underreported herb-drug interactions (HDIs). This paper analyzes the state-of-the-art of Artificial Intelligence (AI) in integrative healthcare and proposes the Integrative Medicine Knowledge Engine (IMKE), a novel architectural blueprint designed to safely bridge Vietnamese traditional medicine (VTM) and Western biomedicine. Shifting from siloed technical models to a clinical-first perspective, the IMKE conceptualizes a three-stage digital health pipeline: (1) advanced medical text mining to digitize unstructured traditional practitioner notes; (2) an Integrative Knowledge Graph (IKG) that maps holistic VTM diagnostic patterns to Western pharmacological databases and ICD-10 standards; and (3) an AI-driven Clinical Decision Support System (CDSS) for proactive HDI safety monitoring. We illustrate this architectural data flow through three clinical scenarios-Essential Hypertension, Type 2 Diabetes, and Rheumatoid Arthritis. These scenarios demonstrate the system's theoretical capacity to retrieve culturally appropriate treatments, identify cross-paradigm therapeutic synergies, and generate explainable, mechanistic safety alerts that augment physician judgment while mitigating clinical alert fatigue. Ultimately, the IMKE framework provides a scalable, culturally inclusive roadmap for digital health. By translating centuries of experiential medical wisdom into an interoperable, computationally safeguarded infrastructure, this blueprint sets the foundation for evidence-based integrative care, future retrospective clinical validation, and enhanced patient safety.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1928965"},"PeriodicalIF":5.2,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13668813/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149217097","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":"Co-designing embodied extended reality telehealth for veteran mental health: from movement care to whole health.","authors":"Aviv Elor, Maxim Kuznetsov, Laura Kusumoto, Jared Duval, Kedron Burnett, Felicia Skelton, Hilary Touchett","doi":"10.3389/fdgth.2026.1952753","DOIUrl":"10.3389/fdgth.2026.1952753","url":null,"abstract":"<p><strong>Introduction: </strong>Veterans in rural and underserved communities face persistent care barriers from provider shortages, travel burdens, and the digital divide. Within the Department of Veterans Affairs (VA), the Clinical Resource Hub (CRH) National Telehealth Program delivers primary, mental health, specialty, surgical, and rehabilitative care into veterans' homes and rural clinics, while VA Immersive scales extended reality (XR) across the system. Embodied XR telehealth, where provider and patient meet as full-body avatars in a shared immersive environment, intersects these programs and introduces new design requirements. This study extends our embodied XR codesign series from physical rehabilitation to mental health and whole health.</p><p><strong>Methods: </strong>Stakeholders from VA, academia, and industry ( <math><mi>N</mi> <mo>=</mo> <mn>30</mn></math> ), drawn from Boise and other participating CRHs, joined a 3-day codesign at the Boise VA Medical Center in June 2026, including shadowing telehealth workflows, a full-day workshop with five teams, and a half-day debrief with program leads. Seven usability task ratings, 164 four-axis ideas (Raise, Reduce, Eliminate, Create), 95 debrief proposals, 188 prioritization votes, and 19 discussion transcripts were analyzed through descriptive statistics and reflexive thematic analysis.</p><p><strong>Results: </strong>Perceived task ease was high among a cohort with mixed VR experience across seven tasks (tutorial, joining a telehealth room, exercise completion, movement recording, environment and interactable customization, and spatial drawing), each above 4.1 of 5 on the Single Ease Question, led by customizing virtual environments ( <math><mi>M</mi> <mo>=</mo> <mn>4.84</mn> <mo>±</mo> <mn>0.37</mn></math> ). Stakeholders most endorsed minimalist defaults, onboarding resources, and out-of-the-box readiness, followed by embodiment with personalized environments, disability representation, and veteran identity markers. The debrief prioritized breathing and grounding, graded exposure, psychoeducation, and whole-health movement, delivered through headsets paired to an in-clinic tablet or monitored home deployment. Infrastructure compatibility, workforce fit, and governance emerged as pervasive constraints. Workshop satisfaction was high ( <math><mi>M</mi> <mo>=</mo> <mn>4.82</mn> <mo>±</mo> <mn>0.53</mn></math> of 5) with 88% requesting follow-up; platform-specific ratings were favorable but lower and more variable (e.g., user-friendliness <math><mi>M</mi> <mo>=</mo> <mn>3.76</mn> <mo>±</mo> <mn>0.97</mn></math> of 5).</p><p><strong>Discussion: </strong>We propose candidate design principles for extending an embodied XR telehealth platform, built for physical rehabilitation, to mental health and whole health care for veterans, and discuss implications for platform design, piloting, and deployment.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1952753"},"PeriodicalIF":5.2,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13670515/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149217099","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":"Beyond detection: developing probabilistic hemodynamic phenotyping after a positive point-of-care pulmonary hypertension test.","authors":"Farhad Fathieh, Navid Nemati, Timothy Burton, Horace R Gillins, Ian Shadforth, Shyam Ramchandani, Charles R Bridges","doi":"10.3389/fdgth.2026.1955810","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1955810","url":null,"abstract":"<p><strong>Background: </strong>Pulmonary hypertension (PH) is a heterogeneous condition in which hemodynamic subtype directs management and specialist referral. PH can be identified non-invasively at the point of care by a previously developed, regulatory-cleared machine-learned diagnostic (area under the receiver operating characteristic curve 0.95, sensitivity 82%, specificity 92%), but a positive test result raises a second clinical question: which patients have pre-capillary disease and therefore warrant referral to a specialist PH center. A probabilistic post-test phenotyping model was developed to estimate hemodynamic subtype probabilities and support rule-out of pre-capillary PH among test-positive patients.</p><p><strong>Methods: </strong>Subjects with right heart catheterization (RHC)-confirmed PH [mean pulmonary artery pressure (mPAP) ≥25 mmHg] enrolled in a prospective multisite trial (NCT04031989) who tested positive on the previously validated point-of-care PH diagnostic formed the analysis cohort (<i>n</i> = 139). Hemodynamic subtypes were assigned per 2015 European Society of Cardiology/European Respiratory Society (ESC/ERS) criteria. A one-vs.-one (OvO) Gaussian mixture model (GMM) framework, combining unsupervised clustering with <i>post hoc</i> supervised assignment of subtype probabilities was applied to features derived from non-invasive orthogonal voltage gradient and photoplethysmographic signals to generate subject-level subtype probability estimates. Blinded validation in a large dataset has not yet been performed; development results are reported, supported by a small blinded independent dataset.</p><p><strong>Results: </strong>At an exemplar <i>post-hoc</i> rule-out threshold of 16% pre-capillary probability, the GMM OvO approach yielded sensitivity 86% (95% CI: 71%-95%), specificity 51% (95% CI: 41%-61%), negative predictive value 88% (95% CI: 73%-96%) at an assumed pre-capillary prevalence of 34%, and a negative likelihood ratio (LR-) of 0.27 (95% CI: 0.07-0.70) on the development cohort. Supportive out-of-sample discrimination was observed in the blinded independent cohort.</p><p><strong>Conclusions: </strong>A GMM OvO probabilistic model applied to non-invasive point-of-care signals can generate PH subtype probability estimates and may support rule-out of pre-capillary disease among true-positive patients. The model reuses signals already acquired for PH detection and can therefore be implemented as a software addition to an existing point-of-care workflow. Independent blinded validation is required to comprehensively characterize model performance.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1955810"},"PeriodicalIF":5.2,"publicationDate":"2026-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13662907/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149198325","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":"A multi-agent large language model (LLM) platform for structured medical anamnesis in outpatient care.","authors":"Elena Zaitseva, Simon Ochotnicky, Miriam Fandakova, Vitaly Levashenko, Jan Rabcan","doi":"10.3389/fdgth.2026.1882397","DOIUrl":"https://doi.org/10.3389/fdgth.2026.1882397","url":null,"abstract":"<p><strong>Introduction: </strong>Medical anamnesis is fundamental to clinical decision-making, yet outpatient consultations often provide limited time for systematic information collection. This study investigates whether structured anamnesis acquisition can be redistributed to a pre-consultation phase using a framework-driven large language model (LLM) platform.</p><p><strong>Methods: </strong>We developed Symptio, a proof-of-concept multi-agent LLM platform that operationalises the SOCRATES, Review of Systems, and SAMPLE frameworks as machine-interpretable constraints within a human-in-the-loop workflow. The system was evaluated in 100 simulated outpatient scenarios. Generated anamneses were compared with reference-standard representations using structural agreement and field-level performance metrics.</p><p><strong>Results: </strong>Full structural agreement across all three frameworks was achieved in 87% of scenarios. Framework-specific agreement reached 96% for SOCRATES, 88% for Review of Systems, and 93% for SAMPLE. Micro-averaged precision, recall, and F1-score were 99.04%, 98.54%, and 98.79%, respectively. Although no high-severity errors were identified, 58 field-level discrepancies, including 28 medium-severity errors, were observed, supporting the need for physician review of all outputs.</p><p><strong>Discussion: </strong>The findings demonstrate the technical feasibility of acquiring structured anamnesis in a framework-driven manner within a process-oriented multi-agent LLM architecture under controlled conditions. The proposed approach enables the relocation of routine information collection to a pre-consultation phase while preserving physician authority through human-in-the-loop validation. Further studies are required to evaluate clinical safety, workflow impact, and effectiveness in real-world settings.</p>","PeriodicalId":73078,"journal":{"name":"Frontiers in digital health","volume":"8 ","pages":"1882397"},"PeriodicalIF":5.2,"publicationDate":"2026-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13662122/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149198355","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}