Journal of the American Medical Informatics Association最新文献

{"title":"Towards an estimate of the impact of censorship on biomedical literature.","authors":"Clair Kronk, Os Keyes, Megh Marathe","doi":"10.1093/jamia/ocaf089","DOIUrl":"10.1093/jamia/ocaf089","url":null,"abstract":"<p><strong>Objective: </strong>To determine how much of the current biomedical literature would be flagged or require modification in relation to the presence of terms from leaked lists prepared by the Centers for Disease Control (CDC), the National Science Foundation (NSF), and the National Security Administration (NSA) in early 2025.</p><p><strong>Materials and methods: </strong>We searched PubMed (from 1996 to 2024) for all records that match at least one of the given terms, combined the terms and analyzed yearly and total frequency.</p><p><strong>Results: </strong>At least 36.3% of all biomedical literature analyzed, representing more than 10 million records, would be flagged for review or modification with the given term lists. It is conservatively estimated that such term lists could impact more than 2.7 million biomedical publications over the next four years.</p><p><strong>Discussion: </strong>Censorship of scientific findings and the use of term lists to judge the content of scientific materials could significantly impede scientific progress.</p><p><strong>Conclusion: </strong>Future research should investigate the long-term implications of, and interim strategies used to navigate, the imposition of censorship on the production and dissemination of scientific knowledge.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":"1199-1205"},"PeriodicalIF":4.7,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12204320/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144227401","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Discover important donor-recipient risk factors and interactions in heart transplant primary graft dysfunction with machine learning.","authors":"Sirui Ding, Yafen Liang, Chia-Yuan Chang, Cheryl Brown, Xiaoqian Jiang, Xia Hu, Na Zou","doi":"10.1093/jamia/ocaf066","DOIUrl":"10.1093/jamia/ocaf066","url":null,"abstract":"<p><strong>Objectives: </strong>Primary graft dysfunction (PGD) is an essential outcome after the heart transplant, which causes severe complications and symptoms for recipients. The in advance prediction of PGD can help the transplant physician better manage the risks of PGD occurrence for patients. Domain experts have identified some important risk factors leading to PGD. However, a widely accepted PGD prediction method is lacking from a computational perspective. In this work, we focus on the prediction of PGD after heart transplant with machine learning (ML).</p><p><strong>Materials and methods: </strong>With the strong power of artificial intelligence, we propose to design a ML algorithm to precisely predict the PGD with the donor and recipient features. Moreover, we apply the computational method to automatically identify important features and interactions between them.</p><p><strong>Results: </strong>To evaluate the effectiveness of the ML algorithm in PGD prediction, we curated a PGD patients' cohort from the United Network for Organ Sharing database, which contains 8008 recipients. 5 commonly used ML models are used for performance comparison. The multi-layer perceptron model achieves superior performance, as measured by area under the receiver operating characteristic curve (AUROC), at 0.868. We identify the top 20 important features and interactions between donors and recipients. Clinical analyses are conducted on the identified features and interactions.</p><p><strong>Discussion: </strong>We summarize the contributions of this work from three aspects including methodology, clinical analysis, and insights. We discuss the limitations of this work on data, model, and real-world implementation perspectives. Additionally, we further discuss the future directions to extend this work to more organ types and diseases.</p><p><strong>Conclusion: </strong>In summary, ML has promising applications in PGD prediction as a computational tool for clinical study. We can also use the ML model to help us identify and discover new risk factors and interactions between donor and recipient.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":"1101-1109"},"PeriodicalIF":4.7,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12199347/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144057100","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Effects of information framing cues and age on the comprehension of personal health records for self-care behaviors: an eye-tracking study.","authors":"Kaifeng Liu, Zhiyan Sun, Xinyuan Ren, Da Tao","doi":"10.1093/jamia/ocaf085","DOIUrl":"10.1093/jamia/ocaf085","url":null,"abstract":"<p><strong>Objective: </strong>To examine the effects of message framing and visualization framing on the comprehension of personal health records and subsequent self-care behavioral intention among young and middle-older aged adults.</p><p><strong>Materials and methods: </strong>A mixed design was employed with visualization framing (ie, black line graph, colored line graph, and colored area graph) and age (ie, young and middle-older aged adults) as between-group factors, and message framing (ie, gain and loss framing) as the within-group factor. Forty-eight participants were asked to comprehend a series of personal health records illustrated by different visualization framing and message framing formats. Data on comprehension performance, eye movement, and perception measures were collected.</p><p><strong>Results: </strong>Visualization framing exerted a significant effect on task accuracy, with colored area graph yielding higher task accuracy than black line graph. Participants perceived worse health status and had stronger behavioral intention to seek professional healthcare advice with loss framing messages compared with gain framing messages. The age effects on task performance turned non-significant when education level was considered, which significantly influenced task accuracy. Age significantly interacted with both visualization framing and message framing. Middle-older adults were more accurate with colored graphs and were more quickly attracted by loss framing messages.</p><p><strong>Discussion: </strong>Visualization framing appeared to play a more important role in user comprehension of personal health records compared with message framing. Color-based framing appears effective in facilitating comprehension, especially for middle-older aged adults. Education background may mediate how individuals in different age groups interpret health information with varied framing formats.</p><p><strong>Conclusions: </strong>This study investigates how the framing of health information influences the comprehension and decision-making processes across different age groups. The findings provide valuable insights for guiding the interface design of health information systems, ensuring that critical health information can be communicated clearly and effectively to patients.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":"1174-1185"},"PeriodicalIF":4.7,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12203526/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144217408","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Challenges using electronic health records to support unhealthy alcohol use screening and intervention in primary care practices in the Pacific Northwest.","authors":"James L McCormack, Tracey L Thomas, Chrystal Barnes, Victoria Sanchez, Erin S Kenzie, Jennifer Coury, Brigit A Hatch, Tiffany Weekley, Maya A Singh, Melinda M Davis","doi":"10.1093/jamia/ocaf083","DOIUrl":"10.1093/jamia/ocaf083","url":null,"abstract":"<p><strong>Objective: </strong>Screening Brief Intervention and Referral to Treatment (SBIRT) can reduce the health and social costs associated with unhealthy alcohol use (UAU). Electronic health records (EHRs) can support evidence-based screening practices for UAU and provide performance data needed for quality improvement. The objective of this study was to describe barriers faced by primary care clinics when using EHR systems to support UAU screening and delivery of recommended interventions.</p><p><strong>Materials and methods: </strong>The Partnerships to Enhance Alcohol Screening, Treatment, and Intervention program (ANTECEDENT) was designed to promote the adoption of SBIRT in primary care clinics through 15 months of tailored practice facilitation. Qualitative data about the participants' experiences were collected through clinic contact logs, periodic reflections, and interviews with practice facilitators and clinic project leads. Data were analyzed through qualitative content analysis to identify and describe the challenges encountered by clinics and facilitators using an 8-domain framework developed to describe socio-technical factors in EHR use.</p><p><strong>Results: </strong>Forty-eight clinics using 9 different EHRs participated in a tailored practice facilitation. Common EHR-related barriers to SBIRT implementation included an inability to report SBIRT performance data, a lack of reminders for screening, few built-in assessments, cumbersome documentation tasks, workflow variation, limited informatics support, and competing organizational priorities.</p><p><strong>Discussion: </strong>Sittig and Singh's framework provided a unique perspective on the challenges primary care clinics participating in ANTECEDENT experienced using a range of EHR systems to support and deliver quality improvements using the SBIRT framework. Our findings were consistent with previous studies evaluating EHR barriers in quality improvement work, and those identified by fellow grantees of the EvidenceNOW: Managing Unhealthy Alcohol Use initiative funded by the Agency for Healthcare Research and Quality.</p><p><strong>Conclusion: </strong>Clinics experience multiple challenges using EHRs to ensure that patients receive needed screening and follow-up for UAU. While vendors may provide relevant capabilities, research is needed to examine what factors affect clinics' awareness, adoption, and use of available EHR features and which are lacking.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":"1157-1163"},"PeriodicalIF":4.7,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12198768/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144235765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Electronic patient-reported outcome measures for triaging and scheduling outpatient appointments: a systematic review and meta-analysis.","authors":"Chen He, Yuelin Xia, Ying Shan Cheung, Sze Tung Lam, Suephy C Chen, Jose M Valderas, Ellie Choi","doi":"10.1093/jamia/ocaf078","DOIUrl":"10.1093/jamia/ocaf078","url":null,"abstract":"<p><strong>Objectives: </strong>This study aims to review the effectiveness of electronic patient-reported outcome measures (ePROMs) to triage and schedule appointments for adult patients with chronic medical conditions.</p><p><strong>Materials and methods: </strong>A structured search was implemented in electronic databases for randomized controlled trials that compared use of ePROMs to facilitate flexible scheduling of appointments (intervention) with conventional scheduling practices (control) in adult outpatients with chronic medical condition. The primary outcome was the difference in healthcare utilization, measured by the number of outpatient physical appointments. Secondary outcomes include disease control and implementational outcomes. A meta-analysis using random effects modeling was performed.</p><p><strong>Results: </strong>The search strategy yielded 3769 citations and 1 additional article from hand search; 17 randomized controlled trials (6469 patients) were included. Most studies focused on cancer (n = 9) or rheumatoid arthritis (n = 3). Six out of 10 studies comparing the number of physical appointments showed that ePROMs significantly reduced the mean number of physical appointments, while 1 study reported increased appointments. A meta-analysis of 6 studies with sufficient data for pooling indicated that the ePROMs group had fewer appointments, with a mean difference of -1.12 (CI, -1.87 to -0.37). Among 10 studies evaluating disease control, 2 showed improved disease control with ePROMs, 2 reported improved survival in cancer patients, while 6 found no significant differences.</p><p><strong>Discussion: </strong>Current evidence supports the feasibility and acceptability of incorporating ePROMs in outpatient visit scheduling, with a reduction in physical appointments without compromising disease outcomes.</p><p><strong>Conclusion: </strong>ePROMs can be used to support and guide decisions regarding outpatient appointment scheduling.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":"1219-1226"},"PeriodicalIF":4.7,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12202323/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144129480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Identifying family structures from obituaries and matching them to patients in an electronic heath record.","authors":"John Mayer, Brooke Delgoffe, Scott Hebbring","doi":"10.1093/jamia/ocaf102","DOIUrl":"10.1093/jamia/ocaf102","url":null,"abstract":"<p><strong>Objectives: </strong>Family data are a valuable data source in bioinformatic research. This is because family members often share common genetic and environmental exposures. Collecting this family data is traditionally very labor intensive but advances in electronic health record (EHR) data mining has proven useful when identifying pedigrees linked to longitudinal health histories. These are called e-pedigrees. Unfortunately, e-pedigrees tend to miss the oldest patients who inherently have the longest and richest health histories. A good source of family data from older generations includes obituaries, as they have a formulaic nature making them a good candidate for natural language processing (NLP) that can extract relationships to the decedent. While there have been several studies on obtaining such data from obituaries, we demonstrate for the first time approaches that tie that information to an EHR.</p><p><strong>Methods: </strong>Natural language processing extraction resulted in 8 166 534 family members being abstracted from 567 279 obituaries published in the state of Wisconsin. After matching decedent and family members to patients in the EHR, we identified 200 033 unique patients that were put in 53 640 pedigrees.</p><p><strong>Results: </strong>The largest pedigree consisted of 21 individuals. Heritability of adult height was quantified (H2=0.51±0.04, P<1.00e-07) demonstrating these data's use in genetic research. The heritability data, coupled with overlapping data in a biobank, suggested 80%-90% of familial relationships were accurately defined.</p><p><strong>Conclusion: </strong>The totality of these findings demonstrate obituaries with the oldest people in society can be highly informative for bioinformatic research.</p><p><strong>Availability and implementation: </strong>Code is available on GitHub at https://github.com/jgmayer672/ObituaryNLP.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144509198","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":"In vitro to in vivo translation of artificial intelligence for clinical use: screening for acute coronary syndrome to identify ST-elevation myocardial infarction.","authors":"Gabrielle Bunney, Kate Miller, Anna Graber-Naidich, Rana Kabeer, Sean M Bloos, Alexander J Wessels, Melissa A Pasao, Marium Rizvi, Ian P Brown, Maame Yaa A B Yiadom","doi":"10.1093/jamia/ocaf101","DOIUrl":"https://doi.org/10.1093/jamia/ocaf101","url":null,"abstract":"<p><strong>Objective: </strong>The integration of predictive models into live clinical care requires scientific testing before implementation to ensure patient safety. We built and technically implemented a model that predicts which patients require an electrocardiogram (ECG) to screen for heart attacks within 10 minutes of their arrival to the Emergency Department. We developed a structured framework for the in vitro to in vivo translation of the model through implementation as clinical decision support (CDS).</p><p><strong>Materials and methods: </strong>The CDS ran as a silent pilot for 2 months. We conducted (1) a Technical Component Analysis to ensure each part of the CDS coding functioned as planned, and (2) a Technical Fidelity Analysis to ensure agreement between the CDS's in vivo and the model's in vitro screening decisions.</p><p><strong>Results: </strong>The Technical Component Analysis indicated several small coding errors in CDS components that were addressed. During this period, the CDS processed 18 335 patient encounters. CDS fidelity to the model reflected raw agreement of 95.5% (CI, 95.2%-95.9%) and kappa of 87.6% (CI, 86.7%-88.6%). Additional coding errors were identified and were corrected.</p><p><strong>Discussion: </strong>Our structured framework for the in vitro to in vivo translation of our predictive model uncovered ways to improve performance in vivo and the validity of risk assessment decisions. Testing predictive models on live care data and accompanying analyses is necessary to safely implement a predictive model for clinical use.</p><p><strong>Conclusion: </strong>We developed a method for the translation of our model from in vitro to in vivo that can be utilized with other applications of predictive modeling in healthcare.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144509199","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":"Supporting clinical reasoning through visual summarization and presentation of patient data: a systematic review.","authors":"Hao Fan, Angela Hardi, Po-Yin Yen","doi":"10.1093/jamia/ocaf103","DOIUrl":"https://doi.org/10.1093/jamia/ocaf103","url":null,"abstract":"<p><strong>Objectives: </strong>Clinicians retrieve data from electronic health record (EHR) systems and summarize them into clinical information to accomplish clinical reasoning and decision-making tasks. Visualization, using meaningful summarization methods and intuitive presentation approaches, can enhance this process. This systematic review examines how EHR data are summarized, visualized, and aligned with the 7 clinical reasoning and decision-making tasks shared by clinicians.</p><p><strong>Materials and methods: </strong>We searched 7 databases for research articles on individual patient EHR related to visualization, clinical decision-support, and patient summaries. Evidence from included studies was extracted for EHR data types, information summarization methods, visualization strategies, clinician characteristics, and evaluations. The synthesized evidence generated data-information-visualization (data-info-vis) flows.</p><p><strong>Results: </strong>We included 112 studies of which 70 (62.5%) conducted detailed usability evaluations, while 42 (37.5%) did not report any evaluations. Gaps remain in deriving actionable insights from EHR data, particularly for tasks requiring data quality reports. Three representative data-info-vis flows emerge. The first uses structured data to generate patterns for temporal visualizations, supporting tasks such as diagnosis and patient management. The second abstracts data into miniature charts, aiding situation-aware understanding and knowledge synthesis. The third features high-level visual metaphors for complex and overarching tasks, such as achieving better care.</p><p><strong>Discussion and conclusion: </strong>This review identifies 2 primary visualization strategies: (1) timeline-based presentations emphasizing temporal trends and longitudinal tracking, and (2) snapshot-based approaches focusing on status overviews and rapid assessments. The identified critical design approaches and distinct data-info-vis flows are tailored to clinical reasoning and decision-making tasks, offering insights for developing visualization-based decision-support tools.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144530766","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":"Enhancing diagnostic precision for rare diseases using case-based reasoning.","authors":"Richard Noll, Alexandra Berger, Carlo Facchinello, Katharina Stratmann, Jannik Schaaf, Holger Storf","doi":"10.1093/jamia/ocaf092","DOIUrl":"https://doi.org/10.1093/jamia/ocaf092","url":null,"abstract":"<p><strong>Objective: </strong>This study aims to enhance the diagnostic process for rare diseases using case-based reasoning (CBR). CBR compares new cases with historical data, utilizing both structured and unstructured clinical data.</p><p><strong>Materials and methods: </strong>The study uses a dataset of 4295 patient cases from the University Hospital Frankfurt. Data were standardized using the OMOP Common Data Model. Three methods-TF, TF-IDF, and TF-IDF with semantic vector embeddings-were employed to represent patient records. Similarity search effectiveness was evaluated using cross-validation to assess diagnostic precision. High-weighted concepts were rated by medical experts for relevance. Additionally, the impact of different levels of ICD-10 code granularity on prediction outcomes was analyzed.</p><p><strong>Results: </strong>The TF-IDF method showed a high degree of precision, with an average positive predictive value of 91% in the 10 most similar cases. The differences between the methods were not statistically significant. The expert evaluation rated the medical relevance of high-weighted concepts as moderate. The granularity of ICD-10 coding significantly influences the precision of predictions, with more granular codes showing decreased precision.</p><p><strong>Discussion: </strong>The methods effectively handle data from multiple medical specialties, suggesting broad applicability. The use of broader ICD-10 codes with high precision in prediction could improve initial diagnostic guidance. The use of Explainable AI could enhance diagnostic transparency, leading to better patient outcomes. Limitations include standardization issues and the need for more comprehensive lab value integration.</p><p><strong>Conclusion: </strong>While CBR shows promise for rare disease diagnostics, its utility depends on the specific needs of the decision support system and its intended clinical application.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2025-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144509197","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":"SDoH-GPT: using large language models to extract social determinants of health.","authors":"Bernardo Consoli, Haoyang Wang, Xizhi Wu, Song Wang, Xinyu Zhao, Yanshan Wang, Justin Rousseau, Tom Hartvigsen, Li Shen, Huanmei Wu, Yifan Peng, Qi Long, Tianlong Chen, Ying Ding","doi":"10.1093/jamia/ocaf094","DOIUrl":"10.1093/jamia/ocaf094","url":null,"abstract":"<p><strong>Objective: </strong>Extracting social determinants of health (SDoHs) from medical notes depends heavily on labor-intensive annotations, which are typically task-specific, hampering reusability and limiting sharing. Here, we introduce SDoH-GPT, a novel framework leveraging few-shot learning large language models (LLMs) to automate the extraction of SDoH from unstructured text, aiming to improve both efficiency and generalizability.</p><p><strong>Materials and methods: </strong>SDoH-GPT is a framework including the few-shot learning LLM methods to extract the SDoH from medical notes and the XGBoost classifiers which continue to classify SDoH using the annotations generated by the few-shot learning LLM methods as training datasets. The unique combination of the few-shot learning LLM methods with XGBoost utilizes the strength of LLMs as great few shot learners and the efficiency of XGBoost when the training dataset is sufficient. Therefore, SDoH-GPT can extract SDoH without relying on extensive medical annotations or costly human intervention.</p><p><strong>Results: </strong>Our approach achieved tenfold and twentyfold reductions in time and cost, respectively, and superior consistency with human annotators measured by Cohen's kappa of up to 0.92. The innovative combination of LLM and XGBoost can ensure high accuracy and computational efficiency while consistently maintaining 0.90+ AUROC scores.</p><p><strong>Discussion: </strong>This study has verified SDoH-GPT on three datasets and highlights the potential of leveraging LLM and XGBoost to revolutionize medical note classification, demonstrating its capability to achieve highly accurate classifications with significantly reduced time and cost.</p><p><strong>Conclusion: </strong>The key contribution of this study is the integration of LLM with XGBoost, which enables cost-effective and high quality annotations of SDoH. This research sets the stage for SDoH can be more accessible, scalable, and impactful in driving future healthcare solutions.</p>","PeriodicalId":50016,"journal":{"name":"Journal of the American Medical Informatics Association","volume":" ","pages":""},"PeriodicalIF":4.7,"publicationDate":"2025-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144267837","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}