Khairat S, Morelli J, Yang S
… +8 more, Li Y, Herasevich V, Mohan D, Handzel R, Ratwani RM, Apathy NC, Boynton MH, Carson SS
J Am Med Inform Assoc
· 2026 Apr · PMID 41719162
·
Full text
OBJECTIVE: This study aimed to evaluate how differences in case complexity affect information-processing burden, as measured by eye-tracking, as well as the efficiency of electronic health record (EHR) use among healthca...OBJECTIVE: This study aimed to evaluate how differences in case complexity affect information-processing burden, as measured by eye-tracking, as well as the efficiency of electronic health record (EHR) use among healthcare providers in the intensive care unit. MATERIALS AND METHODS: This cross-sectional study recruited providers from 4 U.S. medical centers that use 2 prominent EHR systems (Epic and Oracle). After reporting demographic information, participants reviewed 2 complex cases and 2 standard cases in their institution's EHR system and then responded to 5 questions about each case, yielding a performance score. Information-processing burden was assessed by measuring the number of eye fixations via eye-tracking software. The efficiency of EHR use was assessed by measuring the task completion time, number of mouse clicks per minute, number of EHR screens viewed, and performance score. RESULTS: Eighty-one providers were included for analysis. Providers exhibited significantly more eye fixations (P < .001) and longer task completion times (P < .001) for complex cases than for standard cases. There were also significantly fewer mouse clicks per minute during complex cases (P < .001). Reviewing a complex case first led to significantly more eye fixations (P = .015) and longer task completion times (P < .01) than when a standard case was presented first. Case complexity did not significantly affect performance scores or the number of EHR screens viewed. DISCUSSION: Higher case complexity was shown to be associated with greater information-processing burden and less efficient EHR use. These findings have implications for enhancing the efficiency of EHR use, thereby leading to improved clinical decision-making and patient safety. Furthermore, reviewing complex cases first led to a greater information-processing burden, suggesting that providers could benefit from "warming up" with standard cases before reviewing complex cases.
Veinot TC, Wickens M, Hennessey E
… +14 more, Argentina M, Eggebrecht K, Zerkle A, Le V, Collins Damron K, Velez L, Bragg-Gresham J, Krein SL, Chatoth D, Heung M, Gillespie B, Murphy B, Zheng K, Saran R
J Am Med Inform Assoc
· 2026 May · PMID 41707178
·
Full text
OBJECTIVES: To detail patient challenges, and how technology support addressed them, in a remote patient activation intervention for hemodialysis patients (n = 93) from trained patient mentors (n = 26). MATERIALS AND MET...OBJECTIVES: To detail patient challenges, and how technology support addressed them, in a remote patient activation intervention for hemodialysis patients (n = 93) from trained patient mentors (n = 26). MATERIALS AND METHODS: Using digital divide theory-derived codes, content analysis of: technology support program delivery data, hemodialysis clinic staff interviews, and support staff reflection papers. Descriptive statistics from postintervention mentee/mentor surveys. RESULTS: All mentees and 46.2% of mentors received support. Motivational access was targeted with explanations, rapport, and support availability. Study-provided, data-capable tablets enhanced material access, but internet access barriers persisted. Skills access was addressed by training; password-related challenges initially dominated. For usage access, on-demand technology support was balanced by engagement support: proactive prementoring session calls and login monitoring. DISCUSSION: Interventionists should examine internet coverage in targeted areas, potentially using multiple carriers. A balance between password usability and security is required. Engagement support may be needed. CONCLUSION: Technology support can close patient digital divides.
Hill JB, Bernstam EV, Cui L
… +1 more, Killoran PV
J Am Med Inform Assoc
· 2026 May · PMID 41700937
·
Full text
OBJECTIVE: We present a set of definitions and a conceptual model to support primary and consulting physician schedule integration into the electronic health record (EHR) in an inpatient setting and show how utilization...OBJECTIVE: We present a set of definitions and a conceptual model to support primary and consulting physician schedule integration into the electronic health record (EHR) in an inpatient setting and show how utilization of this functionality supports patient-centered communication. MATERIALS AND METHODS: Our institution transitioned to the Epic EHR and implemented modules to connect primary and consulting provider schedules from external scheduling systems to secure messaging within an inpatient EHR context. We evaluated legacy functionality, met with provider groups to map their shifts to hospital teams, built a crosswalk tool to extract, transform, and load data from the scheduling systems to the EHR, evaluated the utilization, and assessed issues related to the implementation. We used our experience from the project to develop a set of definitions and a conceptual model for provider scheduling. RESULTS: We met with over 100 groups to map over 2000 shifts to nearly 700 teams across 15 facilities in our health system. Utilization was high with an average of 6500 on-call provider searches per day in the 30 days following implementation. The conceptual model for inpatient provider scheduling defines 11 terms. DISCUSSION: Our definitions and conceptual model sufficiently represent the inpatient provider scheduling domain as evidenced by high initial utilization and few reported defects. The standardized terminology for provider scheduling aids integration of scheduling data into the EHR. CONCLUSION: Our successful integration of real-time scheduling data within the EHR guided development of a provider scheduling conceptual model. Standardized provider scheduling terminology promotes interoperability of scheduling systems.
Schneider KR, Swann-Thomsen HE, Ribbens TG
… +4 more, Bahnmaier LA, Satterfield T, Pullicar R, Soni N
J Am Med Inform Assoc
· 2026 May · PMID 41691654
·
Full text
BACKGROUND AND SIGNIFICANCE: Physician and advanced practice provider (APP) well-being is a critical focus in healthcare. Emerging technology such as generative artificial intelligence (GAI) scribes reduces physician and...BACKGROUND AND SIGNIFICANCE: Physician and advanced practice provider (APP) well-being is a critical focus in healthcare. Emerging technology such as generative artificial intelligence (GAI) scribes reduces physician and APP administrative burden created by electronic health records. Early adopters of this technology have demonstrated promising improvements in clinical documentation, well-being, and cognitive load. However, further exploration across professional roles is warranted. OBJECTIVE: The goal of this quality improvement initiative was to explore how GAI scribes impacted well-being, cognitive load, and practice efficiency among physicians and APPs across professional roles. METHODS: A cross-sectional anonymous survey was conducted prior to implementation of GAI scribe technology and 3 months after physicians and APPs were onboarded. RESULTS: Physicians and APPs showed a reduction in cognitive task load following scribe technology implementation. Physicians reported reduced burnout and intent to leave; however, APPs did not have a significant reduction in burnout or intent to leave. CONCLUSION: Artificial intelligence scribe technology shows potential for improving well-being among physicians and APPs by reducing cognitive load and clinical documentation time. Although some differences were found, overall, the technology appears to hold promise across professional roles.
Farrag A, Soliman A, Hatef E
… +2 more, Goodin A, Rouhizadeh M
J Am Med Inform Assoc
· 2026 Mar · PMID 41691647
·
Full text
OBJECTIVE: This scoping review aimed to (1) map current applications of transformers and large language models (LLMs) for extracting social drivers of health (SDOH) from clinical text, (2) benchmark model performance acr...OBJECTIVE: This scoping review aimed to (1) map current applications of transformers and large language models (LLMs) for extracting social drivers of health (SDOH) from clinical text, (2) benchmark model performance across SDOH domains, and (3) evaluate methodological rigor to identify research gaps and inform clinical deployment. MATERIALS AND METHODS: We searched PubMed, Web of Science, Embase, Scopus, and IEEE Xplore for studies applying transformers or LLMs to detect SDOH in clinical narratives. We developed a novel methodological framework integrating (1) hierarchical classification of SDOH domains and transformer/LLM architectures, (2) systematic synthesis of performance metrics, and (3) a 7-domain instrument assessing internal validity, external validity, and reporting transparency. RESULTS: Forty-two studies met inclusion criteria. Performance varied substantially across SDOH domains. Behavioral Factors achieved the highest median F1-score (0.87), while Health Care Access and Quality showed the lowest performance and greatest variability (median F1 = 0.59). Research concentrated in the United States (85.7%), relied predominantly on private institutional datasets (69%), and focused primarily on critical care populations (45.2%). Methodological assessment revealed critical gaps; only 29% of studies provided annotation guidelines, 24% assessed fairness across demographic groups, and 21% performed external validation. DISCUSSION: Smaller open-source transformer models show promise for democratizing SDOH detection by achieving competitive performance at lower costs while enabling secure local deployment in resource-limited settings. Advancing clinical readiness requires standardized reporting practices, diverse benchmark datasets across care settings, and systematic equity evaluation to prevent perpetuating health disparities. CONCLUSION: Transformer and LLM performance for SDOH detection varied substantially across domains, with encoder-based models excelling at structured tasks and decoder-only models at linguistically complex tasks. Critical gaps in fairness assessment, external validation, and dataset diversity restrict generalizability and readiness for widespread clinical deployment.
Shea CM, Thomas SR, Khairat S
… +1 more, McSwain D
J Am Med Inform Assoc
· 2026 Apr · PMID 41677193
·
Full text
OBJECTIVE: Hybrid in-person and telehealth work environments are now common among health care providers. When in-person and telehealth services are not well integrated, provider workload could increase, negatively affect...OBJECTIVE: Hybrid in-person and telehealth work environments are now common among health care providers. When in-person and telehealth services are not well integrated, provider workload could increase, negatively affecting provider satisfaction and burnout and hindering implementation of interventions aimed at improving quality. A lack of measures of telehealth integration has hindered studies of such impacts. This article presents the Integration of Telehealth and In-Person Services (ITIPS) survey, developed to assess telehealth integration and its facilitators from the provider's perspective. The ITIPS survey represents a meaningful step toward measuring integration in hybrid healthcare environments, including indicators of the extent of telehealth integration and factors promoting telehealth integration. MATERIALS AND METHODS: Using an exploratory sequential mixed-methods design, this study included semi-structured interviews and a participant-driven, modified-Delphi survey development approach consisting of 2 rounds of Qualtrics surveys to obtain participant feedback. Some participants engaged in all phases, whereas others participated either in interviews or in the modified Delphi surveys. RESULTS: Interviews identified multiple indicators of telehealth integration related to decisions about the visit modality, provider and staff workflows, as well as influencing factors such as leadership priorities related to quality and access, physical space, scheduling systems, and staff support. Our study yielded a survey with 22 items measuring the extent of telehealth integration in a practice environment and 31 items assessing factors influencing the level of telehealth integration. CONCLUSION: This study presents the ITIPS survey, which has undergone assessments for content validity and is ready for psychometric assessment for additional types of validity.
Rincón-Guevara O, Olorukooba AA, Eau G
… +3 more, Ritchey MD, Conn LA, Knicely K
J Am Med Inform Assoc
· 2026 May · PMID 41671029
·
Full text
OBJECTIVE: Our study aims to assess the time-cost burden reduction of transitioning from manual case reporting to electronic case reporting (eCR) for COVID-19 among healthcare organizations (HCOs) over a 1-year period. M...OBJECTIVE: Our study aims to assess the time-cost burden reduction of transitioning from manual case reporting to electronic case reporting (eCR) for COVID-19 among healthcare organizations (HCOs) over a 1-year period. MATERIALS AND METHODS: The study included 3 HCOs from different states with a total of 204 healthcare facilities that had a mixture of hospital and ambulatory providers. We conducted semi-structured interviews, data collection, and analyses of healthcare provider hours spent manually reporting COVID-19 cases along with hours spent implementing and maintaining eCR over 1-year periods. We calculated the time-cost burden of both manual case reporting, and eCR implementation and maintenance to assess the burden reduction of switching from manual to eCR. RESULTS: The number of provider hours reduced by switching from manual case reporting to eCR was 16 942 hours for HCO1, 14 145 hours for HCO2, and 2933 hours for HCO3. Burden reduction of provider hours that eCR offered translated into an estimated $827 120 for HCO1, $569 424 for HCO2, and $78 540 for HCO3, where HCO3 is a small, Federally Qualified Health Center. DISCUSSION: To our knowledge, this is the first time a nationwide information technology innovation was shown to directly reduce the administrative provider time burden and cost of manual case reporting on HCOs. CONCLUSION: By automating patient case reporting, eCR reduced provider burden hours by at least 16-fold and cost savings of 4-fold or higher over a 1-year period.
J Am Med Inform Assoc
· 2026 May · PMID 41671017
·
Full text
OBJECTIVES: Efficient exchange of health information requires consistent representation of clinical concepts across laboratories, hospitals, and public health systems. LOINC supports this interoperability by standardizin...OBJECTIVES: Efficient exchange of health information requires consistent representation of clinical concepts across laboratories, hospitals, and public health systems. LOINC supports this interoperability by standardizing laboratory test codes, but mapping remains difficult when datasets are incomplete, inconsistently formatted, or structurally diverse. These challenges often create a mismatch between algorithmic performance in controlled settings and real-world deployment. This study aimed to develop a biomedical natural language processing (NLP) approach for mapping heterogeneous laboratory test strings to LOINC v2.81 and to compare its performance with established algorithms in the Open Concept Lab (OCL) Mapper. MATERIALS AND METHODS: We implemented a ScispaCy-based pipeline (ScispaCy-LOINC) that identifies clinical entities, links them to UMLS Concept Unique Identifiers, assembles LOINC codes from LOINC parts, and ranks candidates using a weighted scoring system. Overall and ranked performance was evaluated against 2 OCL algorithms, Elasticsearch Keyword Retrieval (OCL-Keyword) and MiniLM Semantic Search (OCL-Semantic), on 2 datasets: MIMIC-IV lab_d_items and a LOINC-mapped subset of the CIEL interface terminology v2025-07-15. RESULTS: In MIMIC-IV, the ScispaCy-LOINC achieved the highest coverage, correctly identifying the LOINC code in 42.3% of cases, outperforming OCL-Keyword (19.5%) and OCL-Semantic (21.4%). In the CIEL dataset, OCL-Semantic achieved the highest coverage (54.4%), followed by OCL-Keyword (46.9%) and ScispaCy-LOINC (28.4%). DISCUSSION: These results indicate that ScispaCy-LOINC is particularly effective for noisier or structurally sparse inputs, whereas OCL-based approaches perform better for more standardized terminologies, highlighting complementary algorithmic strengths. CONCLUSION: ScispaCy-LOINC offers a flexible approach to LOINC mapping and demonstrates complementary strengths relative to existing OCL algorithms. These findings support the development of an integrated framework that combines algorithmic strategies to improve robustness across diverse clinical datasets.
Wang Y, Song Y, Siu R
… +11 more, Nimma IR, Yan Y, Savage TR, Wang Y, Li Z, Ramai D, Wang J, Badurdeen D, Tao C, Kumbhari V, Huang Y
J Am Med Inform Assoc
· 2026 May · PMID 41670956
·
Full text
OBJECTIVES: To comprehensively evaluate the validity of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes for both prevalent diagnoses and less common diseases, and to asse...OBJECTIVES: To comprehensively evaluate the validity of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes for both prevalent diagnoses and less common diseases, and to assess the performance of a large language model (LLM)-based system in validating these codes. MATERIALS AND METHODS: This retrospective study analyzed hospital admissions from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. We developed a validated LLM-based system using GPT-4o, refined through iterative prompt engineering, to assess ICD-10-CM code validity. We measured the positive predictive value (PPV) of ICD-10-CM codes, PPV of principal and secondary diagnoses, and the performance of an LLM-based system in code validation. RESULTS: Among 865 079 assigned codes, the PPV was 84.6% (95% CI, 84.5%-84.6%). Principal diagnoses had a PPV of 93.9% (95% CI, 93.7%-94.1%), while secondary diagnoses had a PPV of 83.8% (95% CI, 83.7%-83.9%). The LLM system demonstrated high performance in validating ICD codes, achieving 93.6% accuracy, 95.4% sensitivity, and 85.2% specificity. Among correctly assigned secondary diagnoses, the majority (67.9%) represented historical or baseline conditions, while 32.1% reflected active conditions that deviated from baseline status; 22.3% of these emerged after hospital admission. PPV decreases with later diagnosis positions, with the largest decline occurring between principal and secondary diagnoses. DISCUSSION AND CONCLUSION: In this large-scale evaluation, ICD-10-CM codes exhibited generally high accuracy, though variability existed by position and condition type. A validated LLM system performed comparably to physician review and offers a scalable means to improve coding accuracy. These findings support the potential for integrating LLM-based auditing into routine workflows to strengthen the quality of administrative and research data.
Stults CD, Martinez MC, Szwerinski NK
… +2 more, Rabbani N, Jones VG
J Am Med Inform Assoc
· 2026 Apr · PMID 41666081
·
Full text
OBJECTIVE: Understand the qualitative impact of an ambient artificial intelligence (AI) documentation platform on clinicians' experiences and workflows. MATERIALS AND METHODS: A quality improvement (QI) qualitative study...OBJECTIVE: Understand the qualitative impact of an ambient artificial intelligence (AI) documentation platform on clinicians' experiences and workflows. MATERIALS AND METHODS: A quality improvement (QI) qualitative study using semi-structured interviews after pilot implementation of an ambient AI documentation platform at a large healthcare organization in Northern and Central California. Pragmatic thematic analysis was used to code and analyze the interviews. RESULTS: 100 clinicians were invited and 42 (42%) participated in an interview. 23 (54.8%) were males and 28 (66.7%) in primary care. Many respondents noted that ambient AI had decreased their cognitive burden by eliminating the need to remember as many specific visit details and saved time for other tasks. They also liked how ambient AI generated transcripts in multiple languages and then created an English-language progress note. However, clinicians also reported challenges, particularly missed or inaccurate information that required them to review the transcript/audio and edit the note. Additionally, many clinicians, particularly specialists, disliked the note formatting and the inability to customize the note template as this resulted in additional manual editing. DISCUSSION: Results from these QI qualitative interviews suggest that ambient AI improved clinicians' overall experience at work. CONCLUSION: While there are many similarities, there may be key differences in clinician experience between ambient AI documentation platforms depending on unique features of each. Future research is needed to understand the potential range of experiences based on type of ambient AI platform and if these findings continue with longer use and broader expansion of this new and evolving technology.
Foryciarz A, Srivathsa N, Sedan O
… +2 more, Goldman Rosas L, Rose S
J Am Med Inform Assoc
· 2026 Apr · PMID 41666048
·
Full text
OBJECTIVES: Incomplete or incorrect causal theories are a key source of bias in machine learning (ML) algorithms. Community-engaged methodologies provide an avenue for mitigating this bias through incorporating causal in...OBJECTIVES: Incomplete or incorrect causal theories are a key source of bias in machine learning (ML) algorithms. Community-engaged methodologies provide an avenue for mitigating this bias through incorporating causal insights from community stakeholders into ML development. In health applications, community-engaged approaches can enable the study of social drivers of health (SDOH), which are known to shape health inequities. However, it remains challenging for SDOH to inform ML algorithms, partially because SDOH variables are known to be interrelated, yet it is difficult to elucidate the causal relationships between them. Community-based system dynamics is a community-engaged methodology that can be used to cocreate formal causal graphs, called causal loop diagrams, with patients. MATERIALS AND METHODS: We used community-based system dynamics to create a causal graph representing the impacts of SDOH on the progression of chronic kidney disease, a chronic condition with SDOH-driven health disparities. We conducted focus groups with 42 participants and a day-long model building workshop with 11 participants. RESULTS: Our model building workshop resulted in a final graph comprising 16 variables, 42 causal links, and 5 subsystems of semantically related SDOH variables. CONCLUSION: This final graph, representing the causal relationships between social variables relevant to chronic kidney disease, can inform the development of clinical ML algorithms and other technological interventions.
Lee S, Jung S, Park JH
… +3 more, Cho H, Moon S, Ahn S
J Am Med Inform Assoc
· 2026 Apr · PMID 41649331
·
Full text
OBJECTIVES: Accurate triage in emergency departments (ED) is critical for appropriate resource allocation. While artificial intelligence (AI) has been explored for triage, prior models relied on summarized clinical scena...OBJECTIVES: Accurate triage in emergency departments (ED) is critical for appropriate resource allocation. While artificial intelligence (AI) has been explored for triage, prior models relied on summarized clinical scenarios. We aimed to develop and evaluate large language models (LLMs) trained on real-world clinical conversations to classify patient urgency. MATERIALS AND METHODS: We used a nationally curated dataset of anonymized triage-level conversations from 3 tertiary Korean hospitals. Two BERT-based models were developed to classify urgency per the Korean Triage and Acuity Scale (KTAS) into urgent (KTAS 3) or non-urgent (KTAS 4-5). One model tokenized the entire conversation, while the other applied a hierarchical structure with sentence-level tokenization and speaker-role embeddings. Performance metrics included accuracy, precision, recall, and F1-score. We compared our models against ChatGPT GPT-4o and ClinicalBERT, and assessed explainability using SHapley Additive exPlanations (SHAP). RESULTS: A total of 5244 clinical conversations, 1057 triage-level dialogues were used, with 950 for training and 107 for testing. Our model with hierarchical structure achieved accuracies of 75.94%, significantly outperforming ChatGPT (56.68%) or fine-tuned ClinicalBERT (69.42%). For urgent cases, the best model achieved a recall of 0.9610, outperforming ChatGPT (0.5352). SHapley Additive exPlanations analysis confirmed that our model focused on clinically relevant cues aligned with KTAS criteria. CONCLUSION: BERT-based LLMs trained on real-world ED conversations significantly outperform general-purpose models like ChatGPT in triage accuracy. This approach demonstrates the potential for enhancing clinical decision support with interpretable and efficient AI.
J Am Med Inform Assoc
· 2026 Apr · PMID 41649326
·
Full text
OBJECTIVES: To explore the complexities of eliminating race correction in clinical artificial intelligence (AI), the pitfalls of naive solutions, and to propose systematic strategies for equitable model development. BACK...OBJECTIVES: To explore the complexities of eliminating race correction in clinical artificial intelligence (AI), the pitfalls of naive solutions, and to propose systematic strategies for equitable model development. BACKGROUND AND SIGNIFICANCE: Race correction in clinical AI, as in traditional medicine, introduces biases with potentially harmful consequences. Simple removal of race from models is insufficient due to the lasting influence of historically biased data. APPROACH: We analyze 4 standardized scenarios to demonstrate how race correction manifests in clinical AI: use of race-corrected variables, explicit inclusion of race, inference via proxy variables, and use of race-specific models. RESULTS: For each scenario, the intuitive solution to removing race correction fails to eliminate bias, often due to legacy effects embedded in the data. More thoughtful approaches are required. DISCUSSION: Ending race correction in clinical AI requires deliberate, context-sensitive interventions, inclusion of diverse stakeholders, and strategies to make model reasoning more transparent and auditable.
Syed M, Hamidi M, Bikkanuri M
… +4 more, Dierschke NA, Katragadda HV, Zozus M, Teixeira AL
J Am Med Inform Assoc
· 2026 Apr · PMID 41637159
·
Full text
OBJECTIVES: To evaluate the performance of a locally deployed adaptation of TrialGPT, a large language model (LLM) system for identifying trial-eligible patients from unstructured electronic health record (EHR) data. MAT...OBJECTIVES: To evaluate the performance of a locally deployed adaptation of TrialGPT, a large language model (LLM) system for identifying trial-eligible patients from unstructured electronic health record (EHR) data. MATERIALS AND METHODS: TrialGPT was re-engineered for secure, deployment at UT Health San Antonio using a locally hosted LLM. It was optimized for real-world data needs through a longitudinal patient-encounter-note hierarchy mirroring EHR documentation. Performance was evaluated in two stages: (1) benchmarking against an expert-adjudicated gold corpus (n = 149) and (2) comparative validation against manual screening (n = 55). RESULTS: Against the expert-adjudicated corpus, the system achieved 81.8% sensitivity, 97.8% specificity, and a positive predictive value of 75.0%. Compared with manual screening, it identified more than twice as many truly eligible patients (81.8% vs 36.4%) while preserving equivalent specificity. CONCLUSION: The adapted TrialGPT framework operationalizes trial matching, translating EHR data into actionable screening intelligence for efficient, scalable clinical trial recruitment.
Zhou H, Chow L, Harnack L
… +5 more, Panda S, Manoogian ENC, Li M, Xiao Y, Zhang R
J Am Med Inform Assoc
· 2026 Apr · PMID 41617202
·
Full text
OBJECTIVES: This study explores the use of advanced natural language processing (NLP) techniques to enhance food classification and dietary analysis using raw text input from a diet tracking app. MATERIALS AND METHODS: T...OBJECTIVES: This study explores the use of advanced natural language processing (NLP) techniques to enhance food classification and dietary analysis using raw text input from a diet tracking app. MATERIALS AND METHODS: The study was conducted in 3 stages: data collection, framework development, and application. Data were collected from a 12-week randomized controlled trial (RCT: NCT04259632), in which participants recorded their meals in free-text format using the myCircadianClock app. Only de-identified data were used. We developed nutrition-focused retrieval-augmented generation (NutriRAG), an NLP framework that uses a retrieval-augmented generation approach to enhance food classification from free-text inputs. The framework retrieves relevant examples from a curated database and then leverages large language models, such as GPT-4, to classify user-recorded food items into predefined categories without fine-tuning. NutriRAG was then applied to data from the RCT, which included 77 adults with obesity recruited from the Twin Cities metro area and randomized into 3 intervention groups: time-restricted eating (TRE, 8-hs eating window), caloric restriction (CR, 15% reduction), and unrestricted eating. RESULTS: NutriRAG significantly enhanced classification accuracy and helped to analyze dietary habits, as noted by the retrieval-augmented GPT-4 model achieving a micro-F1 score of 82.24. Both interventions showed dietary alterations: CR participants ate fewer snacks and sugary foods, while TRE participants reduced nighttime eating. CONCLUSION: By using artificial intelligence, NutriRAG marks a substantial advancement in food classification and dietary analysis of nutritional assessments. The findings highlight NLP's potential to personalize nutrition and manage diet-related health issues, suggesting further research to expand these models for wider use.