BMC Medical Informatics and Decision Making最新文献

{"title":"Power asymmetry and embarrassment in shared decision-making: predicting participation preference and decisional conflict.","authors":"Karin Antonia Scherer, Björn Büdenbender, Anja K Blum, Britta Grüne, Maximilian C Kriegmair, Maurice S Michel, Georg W Alpers","doi":"10.1186/s12911-025-02938-4","DOIUrl":"https://doi.org/10.1186/s12911-025-02938-4","url":null,"abstract":"<p><strong>Background: </strong>Shared decision-making (SDM) is the gold standard for patient-clinician interaction, yet many patients are not actively involved in medical consultations and hesitate to engage in decisions on their health. Despite considerable efforts to improve implementation, research on barriers to SDM within the patient-clinician relationship and interaction is scant. To identify potential barriers to urological patients' participation in decision-making, we developed two novel scales assessing power asymmetry (PA-ME) and embarrassment in medical encounters (EmMed). The present study validates both scales in a large sample comprising urological patients and non-clinical participants. It further examines the effects of both factors on participation preferences and decisional conflict among patients.</p><p><strong>Methods: </strong>Data were collected from 107 urological patients at a university hospital for Urology and Urosurgery in Germany. Patients completed self-report questionnaires before and after their clinical appointments. In addition, 250 non-clinical participants provided data via an online study. All participants rated perceived power asymmetry in the patient-clinician relationship and their experience of embarrassment in medical contexts using the PA-ME and EmMed scales. Urological patients further indicated their participation preference in decisions regarding both general and urological care prior to the consultation. Afterward, they assessed the level of perceived decisional conflict.</p><p><strong>Results: </strong>Factor analyses yielded power asymmetry and medical embarrassment as unidimensional constructs. Both questionnaires have good (PA-ME; α = 0.88), respectively excellent (EmMed; α = 0.95), internal consistency. Among urological patients, higher levels of perceived power asymmetry predicted lower generic participation preference (β = - 0.98, p <.001, adjusted R² = 0.14) and higher decisional conflict (β = 0.25, p <.01, adjusted R² = 0.07). While, in patients, embarrassment was not linked to generic participation preference before the consultation (p ≥.5), it resulted in higher decisional conflict after the consultation (β = 0.39, p <.001, adjusted R² = 0.14). Neither power asymmetry nor embarrassment were specifically associated with participation preference regarding urological care (p ≥.273).</p><p><strong>Conclusions: </strong>Given their promising psychometric properties, the new instruments are recommended for routine assessment of power asymmetry and embarrassment among patients. Addressing these factors may be helpful to reduce decisional conflict and increase participation preferences. Both factors are prerequisites for a successful SDM-process and active patient engagement in health-related decisions.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"120"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596347","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}

{"title":"An intelligent multi-attribute decision-making system for clinical assessment of spinal cord disorder using fuzzy hypersoft rough approximations.","authors":"Muhammad Abdullah, Khuram Ali Khan, Atiqe Ur Rahman","doi":"10.1186/s12911-025-02946-4","DOIUrl":"https://doi.org/10.1186/s12911-025-02946-4","url":null,"abstract":"<p><p>The data for diagnosing spinal cord disorder (SCD) are complex and often confusing, making it difficult for established diagnostic techniques to yield reliable results. This issue frequently necessitates expensive testing to get an accurate diagnosis. However, the diagnostic process can be enhanced by integrating theoretical frameworks that resemble fuzzy sets, which better manage complexity and uncertainty. This integration reduces the frequency of expensive diagnostic procedures, improving the effectiveness of decision-making. The goal of this work is to present lower and upper approximations for fuzzy hypersoft sets, which employ multi-argument-based parameters to improve the traditional lower and upper approximations of fuzzy sets and soft sets. An intelligent mechanism for decision assistance is established by proposing a robust algorithm, that is based on the proposed approximations. To validate the proposed algorithm, a prototype case study for the clinical diagnosis of SCD is discussed. The criteria are further refined by using pertinent sub-criteria, such as functional ability, imaging data, and neurological status criteria. Medical professionals would find the suggested approximations to be a very helpful tool as the results indicate that they could greatly improve diagnosis. This study contributes to the field of medical diagnostics by providing a sophisticated multi-criteria analytical tool that can manage the complexity and inherent ambiguity of SCD diagnosis.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"122"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596298","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}

{"title":"Development of a generic decision guide for patients in oncology: a qualitative interview study.","authors":"Lia Schilling, Jana Kaden, Isabel Bán, Birte Berger-Höger","doi":"10.1186/s12911-025-02960-6","DOIUrl":"https://doi.org/10.1186/s12911-025-02960-6","url":null,"abstract":"<p><strong>Background: </strong>Many patients with cancer want to be involved in healthcare decisions. For adequate participation, awareness of one's own desires and preferences and sufficient knowledge about medical measures are indispensable. In order to support patient participation, a decision guide for patients with cancer was developed as part of a larger project called TARGET, which specifically aims to improve the care of patients with rare cancer.</p><p><strong>Methods: </strong>The development of the decision guide took place from 08.2022 to 03.2023. The decision guide is a single component of a complex intervention that aims to facilitate decision support in cancer care for patients. For the development, existing development and evaluation studies of Question Prompt Lists (QPLs) were identified through systematic literature searches in the MEDLINE via PubMed, PsycInfo, and CINAHL databases. The decision guide was pre-tested for feasibility, usability, completeness and acceptance with the target groups through guided individual interviews. Sociodemographic data were collected anonymously. An expert review was conducted. The verbatim transcribed interviews were analysed using content analysis according to Kuckartz with MAXQDA. The guide has been iteratively optimized based on the results.</p><p><strong>Results: </strong>A generic decision guide for patients with cancer for diagnostic or treatment decisions was developed in both PDF web-based formats, based on the Ottawa Personal Decision Guide. It was supplemented with decision-related questions from QPLs for patients with cancer. The pre-test comprised seven expert reviews of (psych)oncologists and experts in evidence-based health information and ten interviews with cancer patients (n = 7), family relatives (n = 2), and one caregiver. The results were coded into nine main categories. The results indicated a good feasibility, usability and acceptability of the guide. The tool was perceived as comprehensive and appropriate. Individual elements were identified as modifiable for better comprehensibility. The target audience appreciated the decision guide as a good support option.</p><p><strong>Conclusion: </strong>The decision guide is potentially a useful support option for patients with cancer facing medical decisions in their further course of treatment. In the TARGET project, it will be made available to patients and can be supplemented with decision coaching. Further steps for implementation into healthcare structures are necessary.</p><p><strong>Clinical trial number: </strong>Not applicable.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"125"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596341","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}

{"title":"Comparing conventional and Bayesian workflows for clinical outcome prediction modelling with an exemplar cohort study of severe COVID-19 infection incorporating clinical biomarker test results.","authors":"Brian Sullivan, Edward Barker, Louis MacGregor, Leo Gorman, Philip Williams, Ranjeet Bhamber, Matt Thomas, Stefan Gurney, Catherine Hyams, Alastair Whiteway, Jennifer A Cooper, Chris McWilliams, Katy Turner, Andrew W Dowsey, Mahableshwar Albur","doi":"10.1186/s12911-025-02955-3","DOIUrl":"https://doi.org/10.1186/s12911-025-02955-3","url":null,"abstract":"<p><strong>Purpose: </strong>Assessing risk factors and creating prediction models from real-world medical data is challenging, requiring numerous modelling decisions with clinical guidance. Logistic regression is a common model for such studies, for which we advocate the use of Bayesian methods that can jointly deliver probabilistic risk factor inference and prediction. As an exemplar, we compare Bayesian logistic regression with horseshoe priors and Projective Prediction variable selection with the established frequentist LASSO approach, to predict severe COVID-19 outcomes (death or ICU admittance) from demographic and laboratory biomarker data. Our study serves as guidance on data curation, variable selection, and performance assessment with cross-validation.</p><p><strong>Methods: </strong>Our source data is based on a retrospective observational cohort design with records from three National Health Service (NHS) Trusts in southwest England, UK. Models were fit to predict severe outcomes within 28 days after admission to hospital (or a positive PCR result if already admitted) using demographic data and the first result from 30 biomarker tests collected within 3 days after admission (or testing positive if already admitted).</p><p><strong>Results: </strong>Patients included hospitalized adults positive for COVID-19 from March to October 2020, 756 total patients: Mean age 71, 45% female, 31% (n=234) had a severe outcome, of whom 88% (n=206) died. Patients were split into training (n=534) and external validation groups (n=222). Using our Bayesian pipeline, we show a reduced variable model using Age, Urea, Prothrombin time (PT) C-reactive protein (CRP), and Neutrophil-Lymphocyte ratio (NLR) has better predictive performance (median external AUC: 0.71, 95% Quantile [0.7, 0.72]) relative to a GLM using all variables (external AUC: 0.67 [0.63, 0.71]).</p><p><strong>Conclusion: </strong>Urea, PT, CRP, and NLR have been highlighted by other studies, and respectively suggest that hypovolemia, derangement of circulation via clotting, and inflammation are strong predictive risk factors of severity. This study provides guidance on conventional and Bayesian regression and prediction modelling with complex clinical data.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"123"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596335","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}

{"title":"Advancing AI-driven thematic analysis in qualitative research: a comparative study of nine generative models on Cutaneous Leishmaniasis data.","authors":"Issam Bennis, Safwane Mouwafaq","doi":"10.1186/s12911-025-02961-5","DOIUrl":"https://doi.org/10.1186/s12911-025-02961-5","url":null,"abstract":"<p><strong>Background: </strong>As part of qualitative research, the thematic analysis is time-consuming and technical. The rise of generative artificial intelligence (A.I.), especially large language models, has brought hope in enhancing and partly automating thematic analysis.</p><p><strong>Methods: </strong>The study assessed the relative efficacy of conventional against AI-assisted thematic analysis when investigating the psychosocial impact of cutaneous leishmaniasis (CL) scars. Four hundred forty-eight participant responses from a core study were analysed comparing nine A.I. generative models: Llama 3.1 405B, Claude 3.5 Sonnet, NotebookLM, Gemini 1.5 Advanced Ultra, ChatGPT o1-Pro, ChatGPT o1, GrokV2, DeepSeekV3, Gemini 2.0 Advanced with manual expert analysis. Jamovi software maintained methodological rigour through Cohen's Kappa coefficient calculations for concordance assessment and similarity measurement via Python using Jaccard index computations.</p><p><strong>Results: </strong>Advanced A.I. models showed impressive congruence with reference standards; some even had perfect concordance (Jaccard index = 1.00). Gender-specific analyses demonstrated consistent performance across subgroups, allowing a nuanced understanding of psychosocial consequences. The grounded theory process developed the framework for the fragile circle of vulnerabilities that incorporated new insights into CL-related psychosocial complexity while establishing novel dimensions.</p><p><strong>Conclusions: </strong>This study shows how A.I. can be incorporated in qualitative research methodology, particularly in complex psychosocial analysis. Consequently, the A.I. deep learning models proved to be highly efficient and accurate. These findings imply that the future directions for qualitative research methodology should focus on maintaining analytical rigour through the utilisation of technology using a combination of A.I. capabilities and human expertise following standardised future checklist of reporting full process transparency.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"124"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596291","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}

{"title":"Analyzing the medical record homepages quality in a Chinese EMR system.","authors":"Dandan Ge, Yong Xia, Zhonghua Zhang","doi":"10.1186/s12911-025-02949-1","DOIUrl":"https://doi.org/10.1186/s12911-025-02949-1","url":null,"abstract":"<p><strong>Background: </strong>The medical record homepage represents the core and quintessential distillation of the entire medical record. This study aims to investigate the problems with the medical record homepages data quality after the upgrade of the electronic medical record system, while simultaneously proposing practical and feasible measures to catalyze substantive improvements in data quality standards.</p><p><strong>Methods: </strong>A retrospective analysis of data extracted from the medical record homepage system was conducted at a Chinese tertiary hospital affiliated with a medical university between January and December 2021. Analysis of Moment Structures (AMOS) was used to construct a structural equation model, with the aim of elucidating the influence of individual variables on dependent variables. Furthermore, a fish bone diagram analysis was utilized to systematically analyze the underlying causes of quality defects.</p><p><strong>Results: </strong>Among the 2,731 medical record homepages subjected to scrutiny, a substantial proportion of 1,531 records (56.1%) exhibited quality issues. The structural equation model revealed that patient demographic information exerted the most profound influence on data quality, as evidenced by the greatest value of the standardized total effects (β = -0.729), followed by surgery (β = -0.606) and diagnosis information (β = -0.363). Moreover, the fish-bone diagram analysis was employed to systematically dissect the underlying causes of quality defects in the medical record homepages, encompassing human factors, surroundings, regulatory system, and machinery.</p><p><strong>Conclusions: </strong>The predominant factor contributing to the poor data on the medical record homepage was inaccuracies in demographic information, closely followed by errors in surgical and diagnosis information. It is helpful to improve the data quality of the medical record homepages by establishing a coder qualification certification system, strengthening the construction of medical informatization, and adding data validation and prompt functions.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"121"},"PeriodicalIF":3.3,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143596321","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}

{"title":"AI-Driven decision-making for personalized elderly care: a fuzzy MCDM-based framework for enhancing treatment recommendations.","authors":"Abeer Aljohani","doi":"10.1186/s12911-025-02953-5","DOIUrl":"10.1186/s12911-025-02953-5","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Global healthcare systems face enormous challenges due to the ageing population, demanding novel measures to assure long-term efficacy and viability. The expanding senior population, which requires specialised and efficient healthcare solutions, emphasises the importance of improving healthcare sustainability. Recognising the importance of personalised healthcare recommendations in improving patient outcomes as well as facility sustainability, this study tackles the crucial need for targeted treatments to help the elderly navigate the complicated healthcare landscape.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objectives: &lt;/strong&gt;Through the integration of automation with the Fuzzy VIKOR approach as well as Electronic Health Record (EHR) data, this work seeks to create an automated decision-making mechanism that improves personalised healthcare suggestions for the elderly. By using automated data-driven observations, Fuzzy VIKOR to handle decision-making uncertainty as well as the clinical depth of EHR data, the primary objective is to increase the efficacy and accuracy of treatment choices. In order to guarantee that treatment recommendations are not only medically beneficial but also in line with each patient's needs and preferences, this research aims to close the gap between automated intelligence as well as patient-centered care.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Method: &lt;/strong&gt;The Fuzzy VIKOR approach is used with Electronic Health Record (EHR) data to establish a strong framework for personalised healthcare recommendations. AI techniques are employed to enhance data processing, while Fuzzy VIKOR is used to control uncertainty in decision-making, whereas EHR data gives comprehensive clinical insights. The combination of these aspects enables the creation of a system that compensates for uncertainties in medical knowledge and patient preferences, culminating in a ranked array of treatment alternatives customised to the difficulties of healthcare decision-making for the aged.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The study shows how the proposed methodology improves therapy selection for senior populations. By combining AI-powered analysis, Fuzzy VIKOR, and EHR data, the study provides a refined and personalised approach to healthcare recommendations, providing ranked treatment alternatives based on individual characteristics and preferences. The findings demonstrate the potential of this strategy to handle healthcare complexity and contribute to the developing era of precision medicine.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;Finally this study makes an important contribution to the continuing discussion about the sustainability of healthcare for the elderly. The combination of AI-driven methodologies, the Fuzzy VIKOR technique and EHR data offers a promising approach to improving therapy selection in the setting of precision medicine. By accepting personalised healthcare recommendations, this study anticipates a future in which elderly people's unique characteristi","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"119"},"PeriodicalIF":3.3,"publicationDate":"2025-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11889780/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143584837","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":"Retinal vein occlusion risk prediction without fundus examination using a no-code machine learning tool for tabular data: a nationwide cross-sectional study from South Korea.","authors":"Na Hyeon Yu, Daeun Shin, Ik Hee Ryu, Tae Keun Yoo, Kyungmin Koh","doi":"10.1186/s12911-025-02950-8","DOIUrl":"10.1186/s12911-025-02950-8","url":null,"abstract":"<p><strong>Background: </strong>Retinal vein occlusion (RVO) is a leading cause of vision loss globally. Routine health check-up data-including demographic information, medical history, and laboratory test results-are commonly utilized in clinical settings for disease risk assessment. This study aimed to develop a machine learning model to predict RVO risk in the general population using such tabular health data, without requiring coding expertise or retinal imaging.</p><p><strong>Methods: </strong>We utilized data from the Korea National Health and Nutrition Examination Surveys (KNHANES) collected between 2017 and 2020 to develop the RVO prediction model, with external validation performed using independent data from KNHANES 2021. Model construction was conducted using Orange Data Mining, an open-source, code-free, component-based tool with a user-friendly interface, and Google Vertex AI. An easy-to-use oversampling function was employed to address class imbalance, enhancing the usability of the workflow. Various machine learning algorithms were trained by incorporating all features from the health check-up data in the development set. The primary outcome was the area under the receiver operating characteristic curve (AUC) for identifying RVO.</p><p><strong>Results: </strong>All machine learning training was completed without the need for coding experience. An artificial neural network (ANN) with a ReLU activation function, developed using Orange Data Mining, demonstrated superior performance, achieving an AUC of 0.856 (95% confidence interval [CI], 0.835-0.875) in internal validation and 0.784 (95% CI, 0.763-0.803) in external validation. The ANN outperformed logistic regression and Google Vertex AI models, though differences were not statistically significant in internal validation. In external validation, the ANN showed a marginally significant improvement over logistic regression (P = 0.044), with no significant difference compared to Google Vertex AI. Key predictive variables included age, household income, and blood pressure-related factors.</p><p><strong>Conclusion: </strong>This study demonstrates the feasibility of developing an accessible, cost-effective RVO risk prediction tool using health check-up data and no-code machine learning platforms. Such a tool has the potential to enhance early detection and preventive strategies in general healthcare settings, thereby improving patient outcomes.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"118"},"PeriodicalIF":3.3,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11889835/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143584845","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":"NLP modeling recommendations for restricted data availability in clinical settings.","authors":"Fabián Villena, Felipe Bravo-Marquez, Jocelyn Dunstan","doi":"10.1186/s12911-025-02948-2","DOIUrl":"10.1186/s12911-025-02948-2","url":null,"abstract":"<p><strong>Background: </strong>Clinical decision-making in healthcare often relies on unstructured text data, which can be challenging to analyze using traditional methods. Natural Language Processing (NLP) has emerged as a promising solution, but its application in clinical settings is hindered by restricted data availability and the need for domain-specific knowledge.</p><p><strong>Methods: </strong>We conducted an experimental analysis to evaluate the performance of various NLP modeling paradigms on multiple clinical NLP tasks in Spanish. These tasks included referral prioritization and referral specialty classification. We simulated three clinical settings with varying levels of data availability and evaluated the performance of four foundation models.</p><p><strong>Results: </strong>Clinical-specific pre-trained language models (PLMs) achieved the highest performance across tasks. For referral prioritization, Clinical PLMs attained an 88.85 % macro F1 score when fine-tuned. In referral specialty classification, the same models achieved a 53.79 % macro F1 score, surpassing domain-agnostic models. Continuing pre-training with environment-specific data improved model performance, but the gains were marginal compared to the computational resources required. Few-shot learning with large language models (LLMs) demonstrated lower performance but showed potential in data-scarce scenarios.</p><p><strong>Conclusions: </strong>Our study provides evidence-based recommendations for clinical NLP practitioners on selecting modeling paradigms based on data availability. We highlight the importance of considering data availability, task complexity, and institutional maturity when designing and training clinical NLP models. Our findings can inform the development of effective clinical NLP solutions in real-world settings.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"116"},"PeriodicalIF":3.3,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11889813/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143584839","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":"Diagnosis extraction from unstructured Dutch echocardiogram reports using span- and document-level characteristic classification.","authors":"Bauke Arends, Melle Vessies, Dirk van Osch, Arco Teske, Pim van der Harst, René van Es, Bram van Es","doi":"10.1186/s12911-025-02897-w","DOIUrl":"10.1186/s12911-025-02897-w","url":null,"abstract":"<p><strong>Background: </strong>Clinical machine learning research and artificial intelligence driven clinical decision support models rely on clinically accurate labels. Manually extracting these labels with the help of clinical specialists is often time-consuming and expensive. This study tests the feasibility of automatic span- and document-level diagnosis extraction from unstructured Dutch echocardiogram reports.</p><p><strong>Methods: </strong>We included 115,692 unstructured echocardiogram reports from the University Medical Center Utrecht, a large university hospital in the Netherlands. A randomly selected subset was manually annotated for the occurrence and severity of eleven commonly described cardiac characteristics. We developed and tested several automatic labelling techniques at both span and document levels, using weighted and macro F1-score, precision, and recall for performance evaluation. We compared the performance of span labelling against document labelling methods, which included both direct document classifiers and indirect document classifiers that rely on span classification results.</p><p><strong>Results: </strong>The SpanCategorizer and MedRoBERTa.nl models outperformed all other span and document classifiers, respectively. The weighted F1-score varied between characteristics, ranging from 0.60 to 0.93 in SpanCategorizer and 0.96 to 0.98 in MedRoBERTa.nl. Direct document classification was superior to indirect document classification using span classifiers. SetFit achieved competitive document classification performance using only 10% of the training data. Utilizing a reduced label set yielded near-perfect document classification results.</p><p><strong>Conclusion: </strong>We recommend using our published SpanCategorizer and MedRoBERTa.nl models for span- and document-level diagnosis extraction from Dutch echocardiography reports. For settings with limited training data, SetFit may be a promising alternative for document classification. Future research should be aimed at training a RoBERTa based span classifier and applying English based models on translated echocardiogram reports.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"25 1","pages":"115"},"PeriodicalIF":3.3,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11887187/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143572153","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}