Journal of the American Medical Informatics Association最新文献

{"title":"Designing evaluation rubrics for the development of virtual human twins in healthcare.","authors":"Vaishnavi Krishnamurthi, Ruwini Edirisinghe, Sam E John, Nilmini Wickramasinghe, Magdalena Plebanski","doi":"10.1093/jamia/ocag174","DOIUrl":"https://doi.org/10.1093/jamia/ocag174","url":null,"abstract":"<p><strong>Objectives: </strong>Smart healthcare innovations driven by Virtual Human Twin (VHT) technology enable continuous patient monitoring. However, standards to assess their stage of development or a unified framework for performance comparisons are still lacking. This study aims to quantitatively analyze VHT-related studies, evaluate their development levels, and align them to the VHT roadmap.</p><p><strong>Materials and methods: </strong>A systematic review guided by the PRISMA methodology was conducted, in which 26 studies representing different stages of VHT development were systematically analyzed. Articles were evaluated against an established 5-level roadmap and classified according to their closest alignment with black box, grey box, or surrogate VHTs.</p><p><strong>Results: </strong>A rubric system was developed to assess VHT models to date, and their maturity was analyzed relative to the established roadmap. Among the 26 studies, all 5 roadmap levels were represented. Level 2 (Deductive stage) was the most frequently represented. Levels 4 and 5 (Evolutionary and Explainable stages) were represented only in 3 studies each.</p><p><strong>Discussion: </strong>Among the included studies, VHT development appears fragmented, with significant hurdles in building VHTs with advanced functionality.</p><p><strong>Conclusion: </strong>The proposed rubric system may serve as an assessment tool to evaluate progress, encourage discussion, and improve experimental approaches for future VHTs, helping address challenges in advancing VHT development along the established roadmap.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149225198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Longitudinal effects of ambient artificial intelligence scribe use on documentation burden and financial productivity in primary care: a quasi-experimental study.","authors":"R J Waken, Sunny S Lou, Mackenzie Hofford, Elise Eiden, Carrie Burk, Seunghwan Kim, Jared Esker, Linying Zhang, Thomas M Maddox, Joanna Abraham, Albert M Lai, Sam Bhayani, Deborah O'Dell, Jennifer Schmidt, Kayla Paynter, Michele Thomas, Matthew Gerling, Philip R O Payne, Thomas Kannampallil","doi":"10.1093/jamia/ocag151","DOIUrl":"https://doi.org/10.1093/jamia/ocag151","url":null,"abstract":"<p><strong>Objectives: </strong>To investigate the longitudinal effects of an ambient artificial intelligence (AI) scribe use in primary care on outcomes related to documentation time, note writing patterns, and clinician financial productivity.</p><p><strong>Materials and methods: </strong>We conducted a quasi-experimental study with an interrupted time-series design to ascertain the effect of AI scribe use on note writing time, time outside of scheduled hours (TOSH), pajama time (5:30PM-7:00aM), note length, note closure <24 h, and work relative value units (wRVU). For each primary care clinician (physician, advanced practice provider), date of initial access was day 0; each measure was calculated for 150 days prior (pre-period) and after (post-period) day zero.</p><p><strong>Results: </strong>Two hundred and twenty clinicians (mean age = 43.7, 70.9% females; 56.4% physicians) from 36 clinics, conducting 314 845 patient encounters were included. Compared to the pre-period, there was 7% decrease in note writing time at day 0 (incidence rate ratio [IRR] 0.93, 95% credible interval [CI] [0.89, 0.98]), increasing to 16% decrease by day 150 (IRR 0.84, 95% CI [0.82, 0.87]). There was no change in pajama time, TOSH, or wRVU at day 0; however, at day 150, there was a 19% decrease in pajama time (0.81, 95% CI [0.73, 0.90]), 13% decrease in TOSH (0.87, 95% CI [0.77, 0.98]), and a 2% increase in wRVU (1.02, 95% CI [1.01, 1.04]). At day 0, there was a 5% increase in note closures (1.05, 95% CI [1.02, 1.08]), with no difference to pre-period levels by day 150. There were no changes in note length at day 0 or day 150.</p><p><strong>Discussion and conclusion: </strong>There were persistent changes in AI scribe use measures over time as clinicians situated these tools within their workflows.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149220981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From data to personas: a multi-source approach to tailored online health information recommendation for ovarian cancer patients and caregivers.","authors":"Youjia Wang, Khushboo Thaker, Vivian Hui, Peter Brusilovsky, Daqing He, Heidi Donovan, Young Ji Lee","doi":"10.1093/jamia/ocag169","DOIUrl":"https://doi.org/10.1093/jamia/ocag169","url":null,"abstract":"<p><strong>Objective: </strong>Ovarian cancer (OvCa) patients and caregivers face complex, evolving information needs, yet digital health platforms often lack tailoring. This study developed and validated computational personas to support a tailored recommender system (Health E-Librarian with Personalized Recommendations [HELPeR]) and address the cold-start problem, in which limited information about new users reduces the system's ability to generate relevant recommendations.</p><p><strong>Materials and methods: </strong>We used a mixed-methods design integrating 4 data sources: (1) online OvCa forum posts (n = 457; 60 users), (2) semi-structured interviews with patients and caregivers (n = 20), (3) clinical case notes from a caregiver support program, and (4) expert input from gynecologic oncology nurses. Forum data were manually annotated and analyzed using k-modes clustering to identify user subgroups. Qualitative data were analyzed using thematic analysis, and findings were integrated to construct and refine personas.</p><p><strong>Results: </strong>Five distinct user clusters were identified from forum data. Integration of multi-source data yielded 15 personas (9 patients, 6 caregivers) representing trajectories from diagnosis to end-of-life care. Patient personas captured needs related to symptom management, treatment decision-making, financial concerns, and psychosocial support, while caregiver personas emphasized communication, emotional burden, self-care, and care coordination. Personas reflected dynamic, context-dependent needs and were validated for clinical realism through expert review.</p><p><strong>Discussion: </strong>Persona-based modeling provides an interpretable framework for tailored recommendation and approximates digital twin approaches, helping mitigate cold-start limitations.</p><p><strong>Conclusion: </strong>Multi-source persona development is a feasible strategy for enabling tailored digital health interventions and improving tailored information delivery in oncology care.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149220885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The missing layer: decision integrity as a verifiable assurance layer in interpretive clinical artificial intelligence.","authors":"James Thornton, Joshua M Henderson","doi":"10.1093/jamia/ocag152","DOIUrl":"https://doi.org/10.1093/jamia/ocag152","url":null,"abstract":"<p><strong>Objective: </strong>Interpretive clinical artificial intelligence (AI), in which a clinician interprets an image or signal and an AI system offers a competing or confirming read, is increasingly routine in radiology, pathology, dermatology, and endoscopy. Oversight of these systems assumes that independent human judgment occurred before AI influence entered the decision. The informatics infrastructure, meaning the electronic health record, its audit logs, and the reporting systems built on them, does not verify this.</p><p><strong>Discussion: </strong>We name the missing assurance layer decision integrity, situate it between cybersecurity verification and clinical quality review, and argue it is a measurement construct distinct from diagnostic accuracy. Drawing on the logic of preregistration, we propose a candidate architecture, a minimal six-event record under sequential disclosure, preserving tamper-evident evidence of pre-AI judgment, post-AI revision, and rationale.</p><p><strong>Conclusion: </strong>The architecture is a candidate for evaluation, not a validated standard; its purpose is to make clinician independence auditable rather than assumed.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149207978","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transparency features alone are insufficient for trust calibration in artificial intelligence diagnostic support.","authors":"Cassiopeia Frank, Alexander Stevens, Yea-Jen Hsu, Gordon Gao, Michael B Streiff, Lilly Engineer","doi":"10.1093/jamia/ocag176","DOIUrl":"https://doi.org/10.1093/jamia/ocag176","url":null,"abstract":"<p><strong>Objective: </strong>Clinician trust in artificial intelligence (AI) is a frequently cited barrier to adoption, yet there is limited empirical evidence on how transparency features influence trust. This study aimed to address this evidence gap.</p><p><strong>Materials and methods: </strong>We conducted a randomized controlled experiment in which 220 clinicians were randomized to 1 of 3 conditions: no transparency (control), numeric confidence scores, or narrative explanations with citations, using simulated AI diagnostic assistance across clinical case vignettes.</p><p><strong>Results: </strong>Transparency did not influence tool-specific trust [F(2, 214) = 0.38, P = .68], global trust [F(2, 217) = 0.13, P = .88], or perceived information credibility [F(2, 210) = 0.42, P = .66]. Confidence increased following any AI assistance (+6.7%), but narrative explanations were associated with greater change in confidence compared with both control (P = .024) and numeric confidence (P = .0051). Participants changed or refined diagnoses after reviewing AI output 45.8% (SD 23.4%) of the time, with participants in the control group demonstrating higher levels of behavioral influence than transparency conditions (P = .0006).</p><p><strong>Discussion: </strong>Trust, confidence, and behavioral influence represent distinct constructs and are impacted differently by transparency methods. The difference in these contributions to clinician-AI partnerships should be considered when designing for safe trust calibration, with particular consideration for the dissociation between trust and confidence.</p><p><strong>Conclusion: </strong>These findings suggest common interface transparency features alone failed to produce a detectable, meaningful shift in clinician trust in AI assistance and that changes may instead alter behavioral influence and confidence without a corresponding change in trust.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149185685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A language model framework for sequence modeling of audit logs to characterize clinician-electronic health record interactions.","authors":"Seunghwan Kim, Sunny S Lou, Adam Cobb, Susmit Jha, Thomas Kannampallil","doi":"10.1093/jamia/ocag163","DOIUrl":"10.1093/jamia/ocag163","url":null,"abstract":"<p><strong>Objective: </strong>Electronic health record (EHR) audit logs capture clinician-EHR interaction patterns, but most audit log research relies on aggregated measures (eg, total time). We investigated how audit logs could be modeled using large language model (LLM) architectures to learn underlying workflow sequences during clinician-EHR interactions.</p><p><strong>Materials and methods: </strong>Using >295 million EHR-based audit log actions from inpatient settings spanning 2019 to 2024, we fine-tuned Llama-3-8B under 2 encodings: (1) symbolic (field-based) tokens and (2) semantic (word-based) natural language-based audit log action descriptions. A first-order Markov model, using only the immediate prior action, served as a baseline minimal context comparator. Model representation was assessed using next-action prediction accuracy in an early-period test set and 2 temporally distinct out-of-sample (OOS) periods.</p><p><strong>Results: </strong>In the early-period test set, the word-based LLM achieved the highest accuracy (0.7473; 95% CI, 0.7468-0.7478) compared to the field-based LLM (0.3464; 95% CI, 0.3456-0.3472) and Markov baseline (0.4230; 95% CI, 0.4223-0.4236). The word-based approach retained a substantial performance advantage over the Markov baseline across the OOS periods, although accuracy declined over time (word-based LLM vs Markov accuracy, OOS-1: 0.6555 vs 0.3139, OOS-2: 0.6286 vs 0.2720).</p><p><strong>Discussion: </strong>Word-based LLM relying on audit log action descriptions yielded the highest next-action prediction accuracy and demonstrated relative robustness to temporal drift, suggesting that longer sequential context and word-based action descriptions may improve audit log-based sequence modeling.</p><p><strong>Conclusions: </strong>These findings support the development of natural language-based LLM models for audit log research for task identification, workflow characterization, and safety-focused analyses of clinician-EHR interaction patterns.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149170025","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A harmonized SNOMED Clinical Terms components quality assurance framework.","authors":"Dennis H Lee, Simon R Minshall, Francis Y Lau","doi":"10.1093/jamia/ocag158","DOIUrl":"https://doi.org/10.1093/jamia/ocag158","url":null,"abstract":"<p><strong>Objectives: </strong>The objective of this study was to construct a harmonized SNOMED Clinical Terms (SNOMED CT) components quality assurance framework for identifying, categorizing, and assessing data quality issues in SNOMED CT components.</p><p><strong>Materials and methods: </strong>We constructed the framework through an iterative process that analyzed SNOMED CT's core components, synthesized quality factors from prior literature, reviewed inactivation reasons, and SNOMED CT release validation framework, and refined error categories through a targeted review of SNOMED CT error studies. The framework was tested using a case study evaluation.</p><p><strong>Results: </strong>The resulting harmonized SNOMED CT components quality assurance framework consists of 2 dimensions: (1) SNOMED CT core components (concept, description, subtype relationships, attribute relationships) and (2) quality factors (clarity, conciseness, consistency, correctness, comprehensive domain coverage, currency, completeness, conformance). The quality factors were also defined by error categories (omission, commission).</p><p><strong>Discussion: </strong>The framework integrates previously reported quality factors, SNOMED CT inactivation reasons, release validation methods, and includes component-specific error definitions to standardize how quality issues are categorized and assessed. It extends prior frameworks, supports consistent terminology governance, enables measurable quality indicators, and provides a structured foundation for auditing methods and comparative evaluation across SNOMED CT versions and releases.</p><p><strong>Conclusion: </strong>The harmonized framework provides a common vocabulary and structure for describing and classifying quality issues in SNOMED CT. It can guide the development and comparison of auditing methods, support terminology governance, and inform quality-improvement activities undertaken by SNOMED International and national release centers.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149170378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Five safes task execution service: a standards-based architecture and reference implementation for secure federated analytics across trusted research environments.","authors":"Tim Beck, Hazel Lockhart-Jones, Vasiliki Panagi, Justin Biddle, Mervin Chan, Jonathan Couldridge, Jonaa Eva, Grazziela Figueredo, Anwar Gaungoo, Judy Hong, Neerja Karnani, Mingshi Koh, Russell Kua, Mukkesh Kumar, Alex Lee, James Mitchell-White, Alvin Ng, Win Pin Ng, Lee Ning Ng, Tri Thien Nguyen, Chris Orton, Benjamin Partridge, Pin Yan Low, Andy Rae, Thomas Rowlands, Karthikeyan Sivakumar, Stian Soiland-Reyes, Emma Squires, Erwin Tantoso, Haopeng Tong, Wing Cheong Wong, Xing Yi Woo, Carole Goble, Simon Thompson, Philip Quinlan","doi":"10.1093/jamia/ocag160","DOIUrl":"https://doi.org/10.1093/jamia/ocag160","url":null,"abstract":"<p><strong>Objective: </strong>Design, implement, and evaluate a standards-based architecture for secure federated analytics across heterogeneous Trusted Research Environments (TREs) that aligns with the Five Safes Framework, without requiring TREs to modify their internal governance, infrastructure, or operational practices.</p><p><strong>Materials and methods: </strong>The Five Safes task execution service (TES) architecture standardizes the format of analysis requests and handles analysis execution and release of outputs consistent with the Five Safes Framework. The Global Alliance for Genomics and Health TES standard is used for submitting analyses to TREs. Three logical layers support federated analytics: a Submission and Orchestration Layer outside the TRE accepts analyses and holds outputs for collection; a Controller Layer inside the TRE enforces local governance policies and manages disclosure control; and an Analysis Execution Layer governs the technical execution of approved analytics against sensitive data. An open-source reference implementation was developed and deployed in 2 real-world case studies.</p><p><strong>Results: </strong>Five Safes TES was deployed to enable federated analytics within the Dementias Platform UK national research infrastructure. In a transcontinental deployment, federated analytics across 2 TREs hosted in Singapore was enabled by Submission and Orchestration Layers hosted in the United Kingdom, demonstrating multi-jurisdictional federated analytics without centralizing control or moving sensitive data.</p><p><strong>Discussion: </strong>By aligning open technical standards with established governance frameworks, the architecture supports heterogeneous TREs participating in federated analyses while preserving data sovereignty, institutional autonomy, and compliance with local governance requirements.</p><p><strong>Conclusion: </strong>Five Safes TES demonstrates how TRE federated analytics can be deployed as a routine research capability, rather than requiring bespoke technical development.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149138882","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Health information technology deregulation jeopardizes patient safety and public health.","authors":"Jacob Reider, Raj Ratwani, Rollin J Fairbanks, Matthew B Weinger","doi":"10.1093/jamia/ocag168","DOIUrl":"https://doi.org/10.1093/jamia/ocag168","url":null,"abstract":"<p><p>In this Perspective, we provide the rationale for the retention of safety-critical certification criteria for health information technology by the Office of the National Coordinator for Health Information Technology (ONC). To not do so puts all future American patients at risk. The proposed deregulatory actions are based on faulty premises, will lead to serious adverse patient outcomes, and sets a dangerous precedent. Public safety regulations should be held to a higher standard than other regulations, as a central role of government is to protect the general population from the actions of individuals and entities who may have incentives to create potentially harmful products or services.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149129181","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Lost in transformation: diagnostic aggregation and semantic drift in ICD-derived kidney phenotypes.","authors":"Jordan G Nestor, Sarath Babu Krishna Murthy, Casey Ta, Teena Zachariah, Mariela Navarro-Torres, Natasha S Freeman, Heedeok Han, Sindhuri Prakash, Wooin Ahn, Yilu Fang, Atlas Khan, Jai Radhakrishnan, Krzysztof Kiryluk, Chunhua Weng","doi":"10.1093/jamia/ocag167","DOIUrl":"https://doi.org/10.1093/jamia/ocag167","url":null,"abstract":"<p><strong>Objectives: </strong>To evaluate how cross-terminology mapping of International Classification of Diseases (ICD)-coded diagnoses is associated with preservation of diagnostic distinctions and traceability in electronic health record (EHR)-derived phenotypes for genomic research.</p><p><strong>Materials and methods: </strong>We conducted a retrospective, cross-sectional study of real-world ICD code utilization and semantic fidelity among kidney-related diagnoses. Kidney-relevant ICD-9-CM and ICD-10-CM codes were identified using clinician-curated OHDSI ATLAS searches and evaluated in a large academic health system. ICD codes were mapped to Phecodes, SNOMED-CT, the Human Phenotype Ontology (HPO), and Online Mendelian Inheritance in Man (OMIM) using standardized crosswalks. Mapping coverage, richness, and multiplicity were quantified, with multiplicity operationalizing diagnostic aggregation. Three nephrologists independently rated label-level semantic fidelity (0-2 scale). Ordinal logistic regression evaluated associations between mapping multiplicity and semantic fidelity.</p><p><strong>Results: </strong>Among 585 nephrology-relevant ICD codes (571 435 patients), mapping coverage and richness varied across terminologies. Phecodes showed high coverage and mapping multiplicity, whereas SNOMED-CT, HPO, and OMIM mappings were predominantly one-to-one. Among ICD-Phecode mappings, 39% were rated poor, whereas no poor matches were observed among mapped SNOMED-CT, HPO, or OMIM concepts. Greater mapping multiplicity was associated with lower semantic fidelity (OR, 0.33; 95% CI, 0.29-0.37; P < .001). Most ICD-9-CM codes (96%) were represented in at least 1 external cohort.</p><p><strong>Discussion: </strong>Diagnostic aggregation was associated with reduced semantic fidelity and traceability, with potential implications for phenotype construction and interpretation in GWAS, PheWAS, and biobank-scale studies.</p><p><strong>Conclusion: </strong>This study provides an empirical framework for evaluating semantic drift in EHR-derived phenotypes and highlights a central tradeoff: abstraction enables scalability, whereas traceability enables translation.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":7.1,"publicationDate":"2026-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149117781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}