Laura Rosa Baratta, Linlin Xia, Daphne Lew, Elise Eiden, Y Jasmine Wu, Noshir Contractor, Bruce L Lambert, Sunny S Lou, Thomas Kannampallil
{"title":"Networked Behaviors Associated With a Large-Scale Secure Messaging Network: Cross-Sectional Secondary Data Analysis.","authors":"Laura Rosa Baratta, Linlin Xia, Daphne Lew, Elise Eiden, Y Jasmine Wu, Noshir Contractor, Bruce L Lambert, Sunny S Lou, Thomas Kannampallil","doi":"10.2196/66544","DOIUrl":"https://doi.org/10.2196/66544","url":null,"abstract":"<p><strong>Background: </strong>Communication among health care professionals is essential for effective clinical care. Asynchronous text-based clinician communication-secure messaging-is rapidly becoming the preferred mode of communication. The use of secure messaging platforms across health care institutions creates large-scale communication networks that can be used to characterize how interaction structures affect the behaviors and outcomes of network members. However, the understanding of the structure and interactions within these networks is relatively limited.</p><p><strong>Objective: </strong>This study investigates the characteristics of a large-scale secure messaging network and its association with health care professional messaging behaviors.</p><p><strong>Methods: </strong>Data on electronic health record-integrated secure messaging use from 14 inpatient and 282 outpatient practice locations within a large Midwestern health system over a 6-month period (June 1, 2023, through November 30, 2023) were collected. Social network analysis techniques were used to quantify the global (network)- and node (health care professional)-level properties of the network. Hierarchical clustering techniques were used to identify clusters of health care professionals based on network characteristics; associations between the clusters and the following messaging behaviors were assessed: message read time, message response time, total volume of messages, character length of messages sent, and character length of messages received.</p><p><strong>Results: </strong>The dataset included 31,800 health care professionals and 7,672,832 messages; the resultant messaging network consisted of 31,800 nodes and 1,228,041 edges. Network characteristics differed based on practice location and professional roles (P<.001). Specifically, pharmacists and advanced practice providers, as well as those working in inpatient settings, had the highest values for all network metrics considered. Four clusters were identified, representing differences in connectivity within the network. Statistically significant differences across clusters were identified between all considered secure messaging behaviors (P<.001). One of the clusters with 1109 nodes, consisting mostly of physicians and other inpatient health care professionals, had the highest values for all node-level metrics compared to the other clusters found. This cluster also had the quickest message read and response times and handled the largest volume of messages per day.</p><p><strong>Conclusions: </strong>Secure messaging use within a large health care system manifested as an expansive communication network where connectivity varied based on a health care professional's role and their practice setting. Furthermore, our findings highlighted a relationship between health care professionals' connectivity in the network and their daily secure messaging behaviors. These findings provide insights into the complexities of comm","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e66544"},"PeriodicalIF":3.1,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144610429","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jeremy A Balch, Sasank S Desaraju, Victoria J Nolan, Divya Vellanki, Timothy R Buchanan, Lindsey M Brinkley, Yordan Penev, Ahmet Bilgili, Aashay Patel, Corinne E Chatham, David M Vanderbilt, Rayon Uddin, Azra Bihorac, Philip Efron, Tyler J Loftus, Protiva Rahman, Benjamin Shickel
{"title":"Language Models for Multilabel Document Classification of Surgical Concepts in Exploratory Laparotomy Operative Notes: Algorithm Development Study.","authors":"Jeremy A Balch, Sasank S Desaraju, Victoria J Nolan, Divya Vellanki, Timothy R Buchanan, Lindsey M Brinkley, Yordan Penev, Ahmet Bilgili, Aashay Patel, Corinne E Chatham, David M Vanderbilt, Rayon Uddin, Azra Bihorac, Philip Efron, Tyler J Loftus, Protiva Rahman, Benjamin Shickel","doi":"10.2196/71176","DOIUrl":"10.2196/71176","url":null,"abstract":"<p><strong>Background: </strong>Operative notes are frequently mined for surgical concepts in clinical care, research, quality improvement, and billing, often requiring hours of manual extraction. These notes are typically analyzed at the document level to determine the presence or absence of specific procedures or findings (eg, whether a hand-sewn anastomosis was performed or contamination occurred). Extracting several binary classification labels simultaneously is a multilabel classification problem. Traditional natural language processing approaches-bag-of-words (BoW) and term frequency-inverse document frequency (tf-idf) with linear classifiers-have been used previously for this task but are now being augmented or replaced by large language models (LLMs). However, few studies have examined their utility in surgery.</p><p><strong>Objective: </strong>We developed and evaluated LLMs for the purpose of expediting data extraction from surgical notes.</p><p><strong>Methods: </strong>A total of 388 exploratory laparotomy notes from a single institution were annotated for 21 concepts related to intraoperative findings, intraoperative techniques, and closure techniques. Annotation consistency was measured using the Cohen κ statistic. Data were preprocessed to include only the description of the procedure. We compared the evolution of document classification technologies from BoW and tf-idf to encoder-only (Clinical-Longformer) and decoder-only (Llama 3) transformer models. Multilabel classification performance was evaluated with 5-fold cross-validation with F1-score and hamming loss (HL). We experimented with and without context. Errors were assessed by manual review. Code and implementation instructions may be found on GitHub.</p><p><strong>Results: </strong>The prevalence of labels ranged from 0.05 (colostomy, ileostomy, active bleed from named vessel) to 0.50 (running fascial closure). Llama 3.3 was the overall best-performing model (micro F1-score 0.88, 5-fold range: 0.88-0.89; HL 0.11, 5-fold range: 0.11-0.12). The BoW model (micro F1-score 0.68, 5-fold range: 0.64-0.71; HL 0.14, 5-fold range: 0.13-0.16) and Clinical-Longformer (micro F1-score 0.73, 5-fold range: 0.70-0.74; HL 0.11, 5-fold range: 0.10-0.12) had overall similar performance, with tf-idf models trailing (micro F1-score 0.57, 5-fold range: 0.55-0.59; HL 0.27, 5-fold range: 0.25-0.29). F1-scores varied across concepts in the Llama model, ranging from 0.30 (5-fold range: 0.23-0.39) for class III contamination to 0.92 (5-fold range: 0.98-0.84) for bowel resection. Context enhanced Llama's performance, adding an average of 0.16 improvement to the F1-scores. Error analysis demonstrated semantic nuances and edge cases within operative notes, particularly when patients had references to prior operations in their operative notes or simultaneous operations with other surgical services.</p><p><strong>Conclusions: </strong>Off-the-shelf autoregressive LLMs outperformed fined-tuned, encoder-only","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e71176"},"PeriodicalIF":3.1,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12266303/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144602382","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Erica Y Lau, Linda Bird, Anthony Lau, Yau-Lam Alex Chau, Katherine Butcher, Susan Buchkowsky, Kira Gossack-Keenan, Cheryl Sadowski, Corinne M Hohl
{"title":"Developing a SNOMED CT-Based Value Set to Document Symptoms and Diagnoses for Adverse Drug Events: Mixed Methods Study.","authors":"Erica Y Lau, Linda Bird, Anthony Lau, Yau-Lam Alex Chau, Katherine Butcher, Susan Buchkowsky, Kira Gossack-Keenan, Cheryl Sadowski, Corinne M Hohl","doi":"10.2196/70167","DOIUrl":"10.2196/70167","url":null,"abstract":"<p><strong>Background: </strong>Adverse drug events (ADEs) lead to more than 2 million emergency department visits in Canada annually, resulting in significant patient harm and more than CAD $1 billion in health care costs (in 2018, the average exchange rate for 1 CAD was 0.7711 USD; 1 billion CAD would have been approximately 771.1 million USD). Effective documentation and sharing of ADE information through electronic medical records (EMRs) are essential to inform subsequent care and improve safety when culprit medications can be replaced and reexposures avoided. Yet, current systems often lack standardized comprehensive ADE value sets.</p><p><strong>Objective: </strong>This study aimed to develop a SNOMED CT value set for symptoms and diagnoses to standardize ADE documentation and improve ADE data integration into EMRs.</p><p><strong>Methods: </strong>We used ADE data from ActionADE, a prospective reporting system implemented in 9 hospitals in British Columbia. We extract 5792 reports that yielded 827 unique ADE symptom and diagnosis terms based on Medical Dictionary for Regulatory Activities preferred terms. Two independent mappers used both automated and manual mapping approaches to match these terms to SNOMED CT concepts. Two clinical experts conducted validation, followed by a quality assurance review by a separate clinical team. Discrepancies were resolved through consensus discussions. Interrater reliability was assessed using Cohen κ.</p><p><strong>Results: </strong>The automated mapping process identified 63.1% (522/827) semantically equivalent matches from SNOMED CT's Clinical Finding hierarchy. Two mappers manually reviewed the automatically mapped terms and identified appropriate target concepts for the unmapped terms. After the manual mapping process, 95.3% (788/827) of the source terms were successfully mapped to SNOMED CT concepts, with 4.7% (39/827) remaining unmapped. Interrater reliability between the mappers was strong (κ=0.87, 95% CI 0.85-0.89). The validation phase identified and removed 1 irrelevant term, resulting in 98.4% (813/826) terms mapped, with 1.6% (13/826) unmapped, and a high interrater reliability (κ=0.88, 95% CI 0.80-0.95). During quality assurance, 6 terms were flagged for concerns regarding clinical relevance or safety risks and were resolved through discussions. The final value set comprised 813 SNOMED CT concepts, with 95.7% (778/813) of terms classified as semantically equivalent and 4.3% (35/813) as semantically similar. Thirteen additional terms remained unmapped and will be reviewed as new SNOMED CT codes are added.</p><p><strong>Conclusions: </strong>This study developed a SNOMED CT-based value set to document symptoms and diagnoses for ADEs observed in adults in EMRs. Adopting this value set can improve the consistency, accuracy, and interoperability of ADE documentation in EMRs, helping to reduce repeat ADEs and enhance patient safety. Ongoing refinement and improved clinical usability are essential ","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e70167"},"PeriodicalIF":3.1,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12262102/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144593013","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Toward Cross-Hospital Deployment of Natural Language Processing Systems: Model Development and Validation of Fine-Tuned Large Language Models for Disease Name Recognition in Japanese.","authors":"Seiji Shimizu, Tomohiro Nishiyama, Hiroyuki Nagai, Shoko Wakamiya, Eiji Aramaki","doi":"10.2196/76773","DOIUrl":"10.2196/76773","url":null,"abstract":"<p><strong>Background: </strong>Disease name recognition is a fundamental task in clinical natural language processing, enabling the extraction of critical patient information from electronic health records. While recent advances in large language models (LLMs) have shown promise, most evaluations have focused on English, and little is known about their robustness in low-resource languages such as Japanese. In particular, whether these models can perform reliably on previously unseen in-hospital data, which differs from training data in writing styles and clinical contexts, has not been thoroughly investigated.</p><p><strong>Objective: </strong>This study evaluated the robustness of fine-tuned LLMs for disease name recognition in Japanese clinical notes, with a particular focus on their performance on in-hospital data that was not included during training.</p><p><strong>Methods: </strong>We used two corpora for this study: (1) a publicly available set of Japanese case reports denoted as CR, and (2) a newly constructed corpus of progress notes, denoted as PN, written by ten physicians to capture stylistic variations of in-hospital clinical notes. To reflect real-world deployment scenarios, we first fine-tuned models on CR. Specifically, we compared a LLM and a baseline-masked language model (MLM). These models were then evaluated under two conditions: (1) on CR, representing the in-domain (ID) setting with the same document type, similar to training, and (2) on PN, representing the out-of-domain (OOD) setting with a different document type. Robustness was assessed by calculating the performance gap (ie, the performance drop from in-domain to out-of-domain settings).</p><p><strong>Results: </strong>The LLM demonstrated greater robustness, with a smaller performance gap in F1-scores (ID-OOD = -8.6) compared to the MLM baseline performance (ID-OOD = -13.9). This indicated more stable performance across ID and OOD settings, highlighting the effectiveness of fine-tuned LLMs for reliable use in diverse clinical settings.</p><p><strong>Conclusions: </strong>Fine-tuned LLMs demonstrate superior robustness for disease name recognition in Japanese clinical notes, with a smaller performance gap. These findings highlight the potential of LLMs as reliable tools for clinical natural language processing in low-resource language settings and support their deployment in real-world health care applications, where diversity in documentation is inevitable.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e76773"},"PeriodicalIF":3.1,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12262928/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144593014","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Godwin Denk Giebel, Pascal Raszke, Hartmuth Nowak, Lars Palmowski, Michael Adamzik, Philipp Heinz, Marianne Tokic, Nina Timmesfeld, Frank Martin Brunkhorst, Jürgen Wasem, Nikola Blase
{"title":"Improving AI-Based Clinical Decision Support Systems and Their Integration Into Care From the Perspective of Experts: Interview Study Among Different Stakeholders.","authors":"Godwin Denk Giebel, Pascal Raszke, Hartmuth Nowak, Lars Palmowski, Michael Adamzik, Philipp Heinz, Marianne Tokic, Nina Timmesfeld, Frank Martin Brunkhorst, Jürgen Wasem, Nikola Blase","doi":"10.2196/69688","DOIUrl":"https://doi.org/10.2196/69688","url":null,"abstract":"<p><strong>Background: </strong>Artificial intelligence (AI)-based systems are receiving increasing attention in the health care sector. While the use of AI is well advanced in some medical applications, such as image recognition, it is still in its infancy in others, such as clinical decision support systems (CDSS). Examples of AI-based CDSS can be found in the context of sepsis prediction or antibiotic prescription. Scientific literature indicates that such systems can support physicians in their daily work and lead to improved patient outcomes. Nevertheless, there are various problems and barriers in this context that should be considered.</p><p><strong>Objective: </strong>This study aimed to identify opportunities to optimize AI-based CDSS and their integration into care from the perspective of experts.</p><p><strong>Methods: </strong>Semistructured web-based expert interviews were conducted. Experts representing the perspectives of patients; physicians; caregivers; developers; health insurance representatives; researchers (especially in law and IT); and experts in regulation, market admission and quality management or assurance, and ethics were included. The conversations were recorded and transcribed. Subsequently, a qualitative content analysis was performed. The different approaches to improvement were categorized into groups (\"technology,\" \"data,\" \"users,\" \"studies,\" \"law,\" and \"general\"). These also served as deductive codes. Inductive codes were determined within an internal project workshop.</p><p><strong>Results: </strong>In total, 13 individual and 2 double interviews were conducted with 17 experts. A total of 227 expert statements were included in the analysis. Suggestions were heterogeneous and concerned improvements: (1) in the systems themselves (eg, implementing comprehensive system training involving [future] users; using a comprehensive and high-quality database; considering usability, transparency, and customizability; preventing automation bias through control mechanisms or intelligent design; conducting studies to demonstrate the benefit of the system), (2) on the user side (eg, training [future] physicians could contribute to a more positive attitude and to greater awareness and questioning decision supports suggested by the system and ensuring that the use of the system does not lead to additional work), and (3) in the environment in which the systems are used (eg, increasing the digitalization of the health care system, especially in hospitals; providing transparent public communication about the benefits and risks of AI; providing research funding; clarifying open legal issues, eg, those related to liability; and standardizing and consolidating various approval processes).</p><p><strong>Conclusions: </strong>This study offers several possible strategies for improving AI-based CDSS and their integration into health care. These were found in the areas of \"technology,\" \"data,\" \"users,\" \"studies,\" \"law,\" and \"general.\" Sys","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e69688"},"PeriodicalIF":3.1,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144585740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Banafsheh Arshi, Laure Wynants, Eline Rijnhart, Kelly Reeve, Laura Elizabeth Cowley, Luc J Smits
{"title":"Number of Publications on New Clinical Prediction Models: A Bibliometric Review.","authors":"Banafsheh Arshi, Laure Wynants, Eline Rijnhart, Kelly Reeve, Laura Elizabeth Cowley, Luc J Smits","doi":"10.2196/62710","DOIUrl":"10.2196/62710","url":null,"abstract":"<p><strong>Background: </strong>Concerns have been expressed about the abundance of new clinical prediction models (CPMs) proposed in the literature. However, the extent of this proliferation in prediction research remains unclear.</p><p><strong>Objective: </strong>This study aimed to estimate the total and annual number of CPM development-related publications available across all medical fields.</p><p><strong>Methods: </strong>Using a validated search strategy, we conducted a systematic search of literature for prediction model studies published in Pubmed and Embase between 1995 and the end of 2020. By taking random samples for each year, we identified eligible studies that developed a multivariable model (ie, diagnostic or prognostic) for individual-level prediction of a health outcome across all medical fields. Exclusion criteria included development of models with a single predictor, studies not involving humans, methodological studies, conference abstracts, articles with unavailable full text, and those not available in English. We estimated the total and annual number of published regression-based multivariable CPM development articles, based on the total number of publications, proportion of included articles, and the search sensitivity. Furthermore, we used an adjusted Poisson regression to extrapolate our results to the period 1950-2024. Additionally, we estimated the number of articles that developed CPMs using techniques other than regression (eg, machine learning).</p><p><strong>Results: </strong>From a random sample of 10,660 articles published between 1995 and 2020, 109 regression-based CPM development articles were included. We estimated that 82,772 (95% CI 65,313-100,231) CPM development articles using regression were published, with an acceleration in model development from 2010 onward. With the addition of articles that developed non-regression-based CPMs, the number increased to 147,714 (95% CI 125,201-170,226). After extrapolation to the years 1950-2024, the number of articles increased to 156,673 and 248,431 for regression-based models and total CPMs, respectively.</p><p><strong>Conclusions: </strong>Based on a representative sample of publications from the literature, we estimated that nearly 250,000 articles reporting the development of CPMs across all medical fields were published until 2024. CPM development-related publications continue to increase in number. To prevent research waste and close the gap between research and clinical practice, focus should shift away from developing new CPMs to facilitating model validation and impact assessment of the plethora of existing CPMs. Limitations of this study include restriction of search to articles available in English and development of the validated search strategy prior to the popularity of artificial intelligence and machine learning models.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e62710"},"PeriodicalIF":3.1,"publicationDate":"2025-07-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12252138/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144565475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Nurses' Experience Regarding Barriers to Providing Internet Plus Continuous Nursing: Mixed Methods Study.","authors":"Huanhuan Huang, Zhiyu Chen, Lijuan Chen, Xingyao Du, Qi Huang, Wenbi Jia, Qinghua Zhao","doi":"10.2196/65445","DOIUrl":"10.2196/65445","url":null,"abstract":"<p><strong>Background: </strong>The novel medical model of \"Internet Plus continuous nursing\" has received much attention under the dual background of aging and digitalization in China. However, there is a scarcity of studies that report on the potential barriers and challenges associated with the implementation of this practice.</p><p><strong>Objective: </strong>This study aimed to investigate and understand nurses' experience regarding barriers to providing Internet Plus continuous nursing.</p><p><strong>Methods: </strong>A sequential mixed methods design was adopted. In the first phase, a self-made questionnaire was used to quantify the barriers and challenges into 3 domains: management, relational, and information continuity. In the second phase, nurses who participated in the Internet Plus continuous nursing program were invited to attend semistructured interviews to further explore, explain, and understand the complexities behind these data, obtaining more detailed information on participants' experiences, perspectives, and meanings.</p><p><strong>Results: </strong>A total of 4638 participants from 312 hospitals were selected for the final analysis; the adjusted mean score of the survey was 3.49 (SD 0.83). Among the 3 domains, management continuity had the lowest score (mean 3.32, SD 0.97), followed by relational continuity (mean 3.44, SD 0.9) and information continuity (mean 3.62, SD 0.92). The results of the multivariable analysis showed that age, education level, and a greater number of working years were predictors of continuity for Internet Plus continuous nursing (P<.001). Following the qualitive study, 8 subthemes emerged from 72 initial codes and were grouped into 3 themes: organizational changes, practice changes, and future directions.</p><p><strong>Conclusions: </strong>This mixed methods study revealed that Chinese nurses may have differential challenges when providing Internet Plus continuous nursing, particularly in management continuity. To better benefit patients and improve health care delivery, health care organizations and policymakers should implement strategies to improve interdisciplinary relationships, establish and perfect organizational management, and enhance communication.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e65445"},"PeriodicalIF":3.1,"publicationDate":"2025-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12239685/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144555967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jeong Heon Kim, A Reum Choe, Ju Ran Byeon, Yehyun Park, Eun Mi Song, Seong-Eun Kim, Eui Sun Jeong, Rena Lee, Jin Sung Kim, So Hyun Ahn, Sung Ae Jung
{"title":"Enhancing the Predictions of Cytomegalovirus Infection in Severe Ulcerative Colitis Using a Deep Learning Ensemble Model: Development and Validation Study.","authors":"Jeong Heon Kim, A Reum Choe, Ju Ran Byeon, Yehyun Park, Eun Mi Song, Seong-Eun Kim, Eui Sun Jeong, Rena Lee, Jin Sung Kim, So Hyun Ahn, Sung Ae Jung","doi":"10.2196/64987","DOIUrl":"10.2196/64987","url":null,"abstract":"<p><strong>Background: </strong>Cytomegalovirus (CMV) reactivation in patients with severe ulcerative colitis (UC) leads to worse outcomes; yet, early detection remains challenging due to the reliance on time-intensive biopsy procedures.</p><p><strong>Objective: </strong>This study explores the use of deep learning to differentiate CMV from severe UC through endoscopic imaging, offering a potential noninvasive diagnostic tool.</p><p><strong>Methods: </strong>We analyzed 86 endoscopic images using an ensemble of deep learning models, including DenseNet (Densely Connected Convolutional Network) 121 pretrained on ImageNet. Advanced preprocessing and test-time augmentation (TTA) were applied to optimize model performance. The models were evaluated using metrics such as accuracy, precision, recall, F1-score, and area under the curve.</p><p><strong>Results: </strong>The ensemble approach, enhanced by TTA, achieved high performance, with an accuracy of 0.836, precision of 0.850, recall of 0.904, and an F1-score of 0.875. Models without TTA showed a significant drop in these metrics, emphasizing TTA's importance in improving classification performance.</p><p><strong>Conclusions: </strong>This study demonstrates that deep learning models can effectively distinguish CMV from severe UC in endoscopic images, paving the way for early, noninvasive diagnosis and improved patient care.</p>","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e64987"},"PeriodicalIF":3.1,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12236115/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144546348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zhigang Tian, Kayla Wierts, Lirije Hyseni, Beth Gerritsen, Kim Lynch, Russell Buchanan, Mohamed Alarakhia
{"title":"A Unique Way to Axe the Fax Through Using Business Automation Workflow to Expedite eReferral Adoption, Bridging eReferral, and Fax: Proof-of-Concept Study.","authors":"Zhigang Tian, Kayla Wierts, Lirije Hyseni, Beth Gerritsen, Kim Lynch, Russell Buchanan, Mohamed Alarakhia","doi":"10.2196/62983","DOIUrl":"10.2196/62983","url":null,"abstract":"<p><strong>Background: </strong>It is estimated that 88% of Ontario physicians still use fax technology to share patient information. Transitioning to electronic referral (eReferral) has been shown to have numerous benefits, but the major barrier to adoption of eReferral is the need for both sending and receiving clinicians on the same platform to enable information sharing. The traditional onboarding process takes time and effort. An innovative method is required to improve eReferral adoption by bridging the gap between eReferral senders and fax referral receivers.</p><p><strong>Objective: </strong>This study aimed to explore the technological feasibility of leveraging a business automation workflow (BAW) platform to connect the digital (eReferral) and nondigital referral platform (fax), enabling eReferral senders to send referrals to fax receivers, thereby improving the clinician experience.</p><p><strong>Methods: </strong>An eReferral via eFax solution was developed and evaluated on the BAW platform to connect the eReferral platform and the clinicians using fax. A selected number of fax receivers were identified and enabled on the eReferral platform as eFax receivers. Sending clinicians initiated eFaxes through the familiar eReferral workflow, with eFaxes transmitted to BAW and delivered to the target receiver via fax. Retry and reminder logic were built to improve the user experience. If the eFax failed after all retries, a message was sent to the sending clinician through the eReferral platform explaining the failure reason. The appointment information was entered into the eReferral platform by the sending clinicians to trigger patient email notifications. Surveys and focused interviews were conducted to collect clinicians' feedback.</p><p><strong>Results: </strong>From May 2022 to December 2023, 224 eFax receivers were enabled on the platform, processing 4504 eFaxes for 4132 unique patients and 843 unique senders across the province. Nearly 70% (3137/4504) of patients consented and received email notifications; 19% (875/4504) received appointment details after manual entry in the eReferral platform. On average, eFax referrals contained 5.6 pages, with a minimal 0.7% exceeding 30 pages. Initially, fax service retries were disabled to observe delivery error rates. This resulted in a 37.7% (1023/2712) fax failure. However, after implementing new retry logic in March 2023, the failure rate dropped significantly to 9.9% (304/3082), and 98.7% (2770/2806) of eFaxes were successfully delivered after automatic retries. Clinician feedback revealed a positive impact on sending clinicians' experience, maintaining their familiar workflow while accommodating fax-reliant receivers who can gradually transition to eReferral at their own pace.</p><p><strong>Conclusions: </strong>This project demonstrates the potential of the BAW platform to bridge the gap between fax and eReferral systems. It minimizes disruption for sending clinicians while allowing fax rec","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e62983"},"PeriodicalIF":3.1,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12234396/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144531319","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Caroline Bönisch, Christian Schmidt, Dorothea Kesztyüs, Hans A Kestler, Tibor Kesztyüs
{"title":"Proposal for Using AI to Assess Clinical Data Integrity and Generate Metadata: Algorithm Development and Validation.","authors":"Caroline Bönisch, Christian Schmidt, Dorothea Kesztyüs, Hans A Kestler, Tibor Kesztyüs","doi":"10.2196/60204","DOIUrl":"10.2196/60204","url":null,"abstract":"<p><strong>Background: </strong>Evidence-based medicine combines scientific research, clinical expertise, and patient preferences to enhance the patient outcomes and improve health care quality. Clinical data are crucial in aligning medical decisions with evidence-based practices, whether derived from systematic research or real-world data sources. Quality assurance of clinical data, mainly through predictive quality algorithms and machine learning, is essential to mitigate risks such as misdiagnosis, inappropriate treatment, bias, and compromised patient safety. Furthermore, excellent quality of clinical data is a prerequisite for the replication of research results in order to gain insights from practice and real-world evidence.</p><p><strong>Objective: </strong>This study aims to demonstrate the varying quality of medical data in primary clinical source systems at a maximum care university hospital and provide researchers with insights into data reliability through predictive quality algorithms using machine learning techniques.</p><p><strong>Methods: </strong>A literature review was conducted to evaluate existing approaches to automated quality prediction. In addition, embedded in the process of integrating care data into a medical data integration center (MeDIC), metadata relevant to this clinical data was stored, considering factors such as data granularity and quality metrics. Completed patient cases with echocardiographic and laboratory findings as well as medication histories were selected from 2001 to 2023. Two authors manually reviewed the datasets and assigned a quality score for each entry, with 0 indicating unsatisfactory and 1 satisfactory quality. Since quality control was considered a binary problem, corresponding classifiers were used for the quality prediction. Logistic regression, k-nearest neighbors, a naive bayes classifier, a decision tree classifier, a random forest classifier, extreme gradient boosting (XGB), and support vector machines (SVM) were selected as machine learning algorithms. Based on preprocessing the dataset, training machine learning algorithms on echocardiographic, laboratory, and medication data, and assessing various prediction models, the most effective algorithms for quality classification were to be identified. The performance of the predictive quality algorithms was assessed based on accuracy, precision, recall, and scoring.</p><p><strong>Results: </strong>There were 450 patient cases with complete information extracted from the MeDIC data pool. The laboratory and medication datasets had to be limited to 4000 data entries each to enable manual review; the echocardiographic datasets comprised 750 examinations. XGB demonstrated the highest performance for the echocardiographic dataset with an area under the receiver operating characteristic curve (AUC-ROC) of 84.6%. For laboratory data, SVM achieved an AUC-ROC score of 89.8%, demonstrating superior discrimination performance. Finally, regarding the med","PeriodicalId":56334,"journal":{"name":"JMIR Medical Informatics","volume":"13 ","pages":"e60204"},"PeriodicalIF":3.1,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12234397/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144531320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}