BMC Medical Informatics and Decision Making最新文献

{"title":"Optimizing stroke mortality prediction: a comparative study of sampling, cost-sensitive learning, and threshold adjustment.","authors":"Rahim Nikbakht-Fard, Leili Tapak, Mojtaba Khazei, Elaheh Talebi-Ghane","doi":"10.1186/s12911-026-03707-7","DOIUrl":"https://doi.org/10.1186/s12911-026-03707-7","url":null,"abstract":"<p><strong>Background: </strong>Stroke is a leading cause of mortality and disability worldwide, creating a critical need for accurate prediction tools to support early clinical decision making. However, mortality prediction is often hindered by class imbalance, as death events represent a minority of cases in most clinical datasets. Various machine learning (ML) approaches and imbalance-handling techniques, including sampling methods, cost-sensitive learning, and threshold adjustment, have been proposed to address this challenge. This study aimed to compare these strategies and identify an optimal ML framework for predicting in-hospital mortality among patients with stroke.</p><p><strong>Methods: </strong>Data from 2653 patients with stroke were analyzed. Five machine learning algorithms (XGBoost, RF, DNN, SVM, and LR) were evaluated using nested stratified 10-fold cross-validation. Class imbalance was addressed using threshold adjustment, cost-sensitive learning, and sampling methods (ENN, OSS, SMOTE, SMOTE-ENN, and SVM-SMOTE). Model performance was assessed using Accuracy, F1-score, G-mean, MCC, AUROC, and AUPRC. Statistical comparisons were performed using Friedman and post-hoc tests, and model interpretability was evaluated using SHAP.</p><p><strong>Results: </strong>Among the baseline models, XGBoost achieved the highest discrimination performance, with an AUROC of 0.898 ± 0.014 and an AUPRC of 0.656 ± 0.076. Support Vector Machine and Logistic Regression achieved the highest G-mean values, indicating a better balance between sensitivity and specificity. Threshold adjustment based on the G-mean criterion improved the G-mean of XGBoost from 0.661 to 0.804 and increased the F1-score from 0.551 to 0.624, while cost-sensitive learning achieved a comparable G-mean of 0.806. Among the sampling strategies, ENN produced the highest F1-score (0.615 ± 0.026), whereas SMOTE-ENN achieved the highest G-mean (0.799 ± 0.032). Although all sampling methods improved minority-class detection, statistical analysis showed no significant differences among sampling strategies (Friedman χ² = 9.44, p = 0.051). SHAP analysis identified consciousness status, awareness level, patient arrival condition, Glasgow Coma Scale category, respiratory complications, length of hospital stay, and age as the most influential predictors of in-hospital mortality.</p><p><strong>Conclusion: </strong>XGBoost achieved the best overall performance for predicting in-hospital mortality among patients with stroke. Threshold adjustment and cost-sensitive learning substantially improved minority-class detection, whereas ENN and SMOTE-ENN provided the most favorable sampling results. Overall, algorithm selection and threshold optimization had a greater impact on predictive performance than the specific sampling strategy. These findings support the potential use of machine-learning models for early risk stratification in stroke care.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148444578","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":"Red blood cell distribution width-to-albumin ratio as a novel predictor for mortality in breast cancer patients admitted to ICU: a retrospective analysis using MIMIC-IV 3.1.","authors":"Chenyan Hong, Ke Yin, Shenchao Guo, Jin Luo","doi":"10.1186/s12911-026-03711-x","DOIUrl":"https://doi.org/10.1186/s12911-026-03711-x","url":null,"abstract":"<p><strong>Background: </strong>The red blood cell distribution width-to-albumin ratio (RAR), an emerging biomarker integrating inflammation and nutritional status, has not been systematically evaluated for its prognostic value in breast cancer patients admitted to intensive care units (ICU).</p><p><strong>Methods: </strong>We conducted a retrospective cohort study using data from the MIMIC-IV 3.1 database, including 881 adult breast cancer patients admitted to the ICU. Patients were stratified into high- and low-RAR groups based on maximally selected rank statistics. Kaplan-Meier analysis, multivariable Cox proportional hazards models, restricted cubic splines, subgroup analysis, time-dependent concordance index (C-index) curves and Boruta feature selection (iterative random forest with shadow feature comparison) were applied to assess the association between RAR and 1-year all-cause mortality. Robustness was examined using E-value and propensity score weighting methods.</p><p><strong>Results: </strong>Patients with high RAR (> 4.96) had significantly higher 1-year mortality compared to those with low RAR (log-rank P < 0.001). In adjusted models, high RAR independently predicted mortality (HR = 1.65, 95% CI: 1.33-2.06). Across the four models, E-values for the point estimates ranged from 2.18 (fully adjusted model, HR 1.65) to 2.82 (unadjusted model, HR 2.19); the E-value for the lower confidence limit of the primary model was 1.73. Each standard deviation increase in RAR was associated with a 17% higher mortality risk (HR = 1.17, 95% CI: 1.09-1.25). Restricted cubic spline analysis demonstrated a linear dose-response relationship (P for nonlinearity > 0.05). Incorporating RAR into SOFA and APS III scores improved prognostic performance (P < 0.001). Feature importance ranking further highlighted RAR as a major predictor. Sensitivity analyses using overlap-weighting (HR = 1.56, 95% CI: 1.25-1.96) and matching-weighting (HR = 1.53, 95% CI: 1.22-1.93) confirmed robustness.</p><p><strong>Conclusion: </strong>RAR is an easily obtainable biomarker derived from routine laboratory testing that may enhance risk stratification for ICU-admitted breast cancer patients. Its integration into prognostic models may facilitate early decision support.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148435275","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":"Toward tailored implementation of an artificial intelligence-based healthy dietary chatbot: differences in technology acceptance pathways by health consciousness and user characteristics.","authors":"Chiu-Liang Liu, Chien-Ta Ho, Tzu-Chi Wu","doi":"10.1186/s12911-026-03712-w","DOIUrl":"https://doi.org/10.1186/s12911-026-03712-w","url":null,"abstract":"<p><strong>Background: </strong>Artificial intelligence (AI) health chatbots are increasingly used for health education and health promotion, yet the mechanisms underlying their adoption and the heterogeneity across user groups remain insufficiently understood.</p><p><strong>Methods: </strong>Drawing on the Technology Acceptance Model (TAM), this study examined behavioral intention to use an artificial intelligence-based healthy dietary chatbot, incorporating interactivity and entertainment as key system features, and testing whether health consciousness and sociodemographic subgroups moderate technology acceptance pathways. An anonymous questionnaire survey was conducted after participants interacted with a LINE-based AI dietary chatbot. Structural equation modeling and subgroup analysis were applied to evaluate TAM relationships, moderation by health consciousness, and boundary conditions across demographic groups.</p><p><strong>Results: </strong>Perceived ease of use and perceived usefulness significantly predicted behavioral intention. Interactivity was primarily associated with intention indirectly through perceived ease of use and perceived usefulness, whereas entertainment showed a stronger direct association with intention. Health consciousness moderated the interactivity-usefulness and entertainment-intention relationships. Multi-group analyses further indicated pathway heterogeneity: interaction-related effects were stronger among younger, female, and more highly educated users, whereas older users placed greater weight on enjoyment and ease of use, and males exhibited a stronger usefulness-intention link.</p><p><strong>Conclusions: </strong>Overall, the findings support the applicability of TAM to artificial intelligence-based health chatbot adoption and suggest the coexistence of cognitive, interactivity-driven and affective, entertainment-driven acceptance routes. The moderating roles of health consciousness and user characteristics underscore the importance of segment-specific design and dissemination strategies to enhance adoption.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148435287","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":"Enhancing stratification for survival analyses across standardized data sources.","authors":"Mikhail Shubov, Mareile Beernink, Jasmin Carus, Alexander Johannes Wiederhold, Christopher Gundler","doi":"10.1186/s12911-026-03666-z","DOIUrl":"https://doi.org/10.1186/s12911-026-03666-z","url":null,"abstract":"<p><strong>Background: </strong>Patient stratification is crucial for advancing personalized medicine yet is complicated due to the fragmented and variable nature of healthcare data. The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) addresses the challenges regarding the data by offering a standardized framework for data integration across diverse sources and configurations. Recent advancements in machine learning, particularly transformer-based models such as Bidirectional Encoder Representations from Transformers (BERT), have demonstrated significant potential in extracting deep patient representations from electronic health records.</p><p><strong>Methods: </strong>This study assesses the efficacy of BERT-based patient representation learning using OMOP CDM data for patient stratification. We harmonize originally incompatible datasets, including the established MIMIC-IV-2.2 and lung cancer data from the German cancer registry Schleswig-Holstein, within the OMOP CDM framework. BERT is pre-trained on the MIMIC-IV-2.2 dataset, and the derived representations are utilized to generate patient embeddings from the cancer registry (test) data. We employ k-means clustering on the embeddings to stratify patient subgroups. To evaluate whether the embeddings are useful for clustering, we divide the original cancer registry data into the corresponding groups and conduct survival analyses on selected columns for each cluster. The clustering method's effectiveness is assessed by comparing survival models trained on these clusters with naïve clusters derived only from the original dataset. In addition, we included a clinical expert review, in which a physician assessed the resulting cluster assignments for clinical plausibility and interpretability.</p><p><strong>Results: </strong>Our approach effectively identifies patient similarities across datasets and allowed for efficient patient stratification. Survival analyses show varied performance depending on the model and cluster characteristics, with up to a 11% improvement over naïve k-means clusters, demonstrating the benefits of transfer learning. For some groups of patients, the corresponding accuracy of the survival analysis increased by up to 7%, emphasizing the value of stratifying homogeneous subgroups.</p><p><strong>Conclusion: </strong>Utilizing standardized data and transformer-based foundation models to generate patient embeddings demonstrates effective knowledge transfer between two vastly different datasets and enables the identification of groups in which the accuracy of survival analysis can be significantly increased.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148434761","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":"Machine learning driven modeling of synergistic perinatal risk profiles in early onset pediatric cerebral palsy.","authors":"Foysal Ahammad, Munira Aden, Ayesha Banu, Muhammad Marwan, Sadam Hussain, Safa Salim, Adnan Mohammed, Tanvir Alam, Lisa Thornton, Farhan Mohammad","doi":"10.1186/s12911-026-03710-y","DOIUrl":"https://doi.org/10.1186/s12911-026-03710-y","url":null,"abstract":"<p><strong>Aim: </strong>To develop and evaluate machine learning (ML) models for early cerebral palsy (CP) prediction and identify synergistic perinatal risk factors in a pediatric population.</p><p><strong>Method: </strong>We conducted a retrospective case-control study using demographic, perinatal, and postnatal clinical data collected at Sidra Medicine, Qatar. Four ML models- Random Forest (RF), XGBoost, Support Vector Machine (SVM), and a feedforward neural network (FFN) were trained using clinically relevant features. Model performance was assessed using precision, recall, area under the curve (AUC), F1-score, and SHAP-based interpretability. A multidimensional interaction framework was used to evaluate cumulative risk across 16 subgroups.</p><p><strong>Results: </strong>All ML models exhibited high predictive accuracy (ROC-AUC: 0.98-0.99, and PR-AUC: 0.97-0.98), with four key factors: low birth weight (LBW), premature birth, neonatal intensive care unit (NICU) admission, and multiple pregnancies. Infants exposed to all four factors demonstrated a 93.15% incidence of CP (OR = 1382.67; p < 0.0001). A clear dose-response gradient was observed across exposure subgroups. SHAP analysis confirmed consistent cross-model importance of LBW, very preterm birth, NICU admission, and multigravidity. Cross-validation confirmed model robustness, and severity analysis identified NICU admission and birth weight as independent predictors of higher GMFCS classification.</p><p><strong>Conclusion: </strong>Four ML models achieved high predictive accuracy (ROC-AUC 0.98-0.99) for CP risk stratification in a Middle Eastern pediatric cohort, with LBW, very preterm birth, NICU admission, and multigravidity as the most consistent cross-model predictors. The synergistic interaction of these exposures - evidenced by a 93.15% CP incidence in the highest-risk subgroup - supports a paradigm shift from single-factor screening to exposure-weighted, ML-driven neonatal surveillance. These findings provide a data-driven foundation for early risk stratification and targeted intervention planning in high-risk neonatal population.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148418941","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 and validation of a machine learning model for identifying liver fibrosis in MAFLD: a retrospective model development study.","authors":"Chenyu Hu, Yu Qiu, Yuxin Zhong, Wei Zhou, Zhiqi Zhang, Yi Huang","doi":"10.1186/s12911-026-03638-3","DOIUrl":"https://doi.org/10.1186/s12911-026-03638-3","url":null,"abstract":"<p><strong>Background: </strong>Metabolic dysfunction-associated fatty liver disease (MAFLD), as the most prevalent chronic liver disease globally, is closely related to the prevalence of metabolic syndrome. Liver fibrosis is a core stage in the progression of this disease, significantly increasing the risk of liver cirrhosis, liver failure, and hepatocellular carcinoma. Early identification of high-risk patients is crucial for blocking disease progression. We developed and internally validated a machine learning model to identify MAFLD patients with VCTE-defined liver fibrosis using routinely collected clinical variables.</p><p><strong>Method: </strong>This study selected 2728 research subjects from January 2020 to October 2025. Through the least absolute shrinkage and selection operator (LASSO) regression, Boruta algorithm, and recursive feature elimination (RFE), eight key variables, namely ALT, AST, SBP, BMI, WHR, DBP, LDL-C, and GGT, were selected. Classification models employed logistic regression, decision trees, random forests, XGBoost, LightGBM, support vector machines (SVM), and artificial neural networks (ANN). Nomogram and SHapley additive exPlanations (SHAP) were used to explain the constructed model.</p><p><strong>Results: </strong>This study included 2728 patients with 1013 cases (38%) of liver fibrosis. XGBoost, LightGBM, and Random Forest demonstrated broadly comparable performance.</p><p><strong>Conclusion: </strong>The proposed machine learning model has clinical application potential in identifying liver fibrosis defined by VCTE in patients with MAFLD. External validation and prospective clinical evaluation are required before routine clinical implementation.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148418811","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":"A systematic review of patient decision aids for hypertension.","authors":"Jan Berghold, Lena Fischer, Alexander Pachanov, Leon Vincent Schewe, Dawid Pieper","doi":"10.1186/s12911-026-03700-0","DOIUrl":"10.1186/s12911-026-03700-0","url":null,"abstract":"<p><strong>Background: </strong>Hypertension is a major cause of premature death and a modifiable risk factor for cardiovascular disease. Effective management includes both pharmacological treatment strategies, including antihypertensive medications, and non-pharmacological treatment strategies, such as lifestyle changes. Shared decision making (SDM) between patients and healthcare professionals is essential to determine the most appropriate treatment for hypertension. Patient decision aids (PDAs) facilitate SDM by providing information on treatment options and helping patients to clarify their values and preferences. However, there is no comprehensive comparison of PDAs for hypertension.</p><p><strong>Objective: </strong>This systematic review aimed to identify and evaluate PDAs for arterial hypertension using the International Patient Decision Aid Standards (IPDAS) criteria.</p><p><strong>Methods: </strong>A comprehensive search of bibliographic databases (PubMed, Embase) and gray literature (Google Scholar, Google) was performed in October 2023 and updated in April 2026. Studies and tools were included if they presented a PDA for patients with arterial hypertension, met the IPDAS definition of a PDA and were in German or English. Two researchers screened the literature, extracted data and assessed the quality of the PDAs using the IPDAS Minimal Criteria. Data were synthesized narratively.</p><p><strong>Results: </strong>Of 1,874 records screened, five PDAs met the inclusion criteria. These PDAs were developed between 2019 and 2024 and originated from the United States, United Kingdom, and Germany. All PDAs addressed the two option areas of medication and lifestyle but differed in the number of individual treatment options presented (range: 2-12 options) and the level of detail in which the options were addressed. All five PDAs were accessible online. Common design features of the PDAs included an option grid format for presenting treatment options with associated benefits and adverse effects, visuals and graphics to enhance comprehension, and value clarification. Quality assessments showed varying levels of compliance with the IPDAS criteria, with overall scores ranging from 49/116 to 87/116.</p><p><strong>Conclusions: </strong>Few PDAs exist despite the prevalence of hypertension. PDAs varied in quality and format, highlighting gaps in the development process. Further research is needed to evaluate the effectiveness of PDAs for hypertension and to determine which formats are most effective in supporting SDM in different patient populations.</p><p><strong>Clinical trial number: </strong>Not applicable.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13355325/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148418965","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":"Artificial intelligence for detection, prediction, and severity classification of diabetic peripheral neuropathy: a systematic review.","authors":"Tay Jia Ying, Ang Yee Gary","doi":"10.1186/s12911-026-03676-x","DOIUrl":"https://doi.org/10.1186/s12911-026-03676-x","url":null,"abstract":"<p><strong>Background: </strong>Diabetic peripheral neuropathy (DPN) is a common complication of diabetes and an important contributor to foot ulceration and lower limb amputation. Early detection remains challenging because conventional screening methods are often subjective, resource-intensive, and insensitive to subclinical disease. Artificial intelligence (AI) has increasingly been applied to support DPN assessment, but its applications across different clinical tasks have not been clearly synthesised. This systematic review evaluates AI applications for DPN detection, prognostic prediction, risk stratification, and severity classification.</p><p><strong>Methods: </strong>A systematic literature search was conducted across PubMed, Scopus, and IEEE Xplore for studies published from January 2021 to 1 September 2025. Studies developing or validating AI models for DPN assessment using patient datasets were included. Study selection was performed by one reviewer, with a second reviewer contributing to uncertain inclusion decisions. Data extraction and risk-of-bias assessment were performed by one reviewer and verified by another using the Prediction Model Risk of Bias Assessment Tool (PROBAST). Due to heterogeneity in clinical tasks, data types, outcome definitions, and validation approaches, findings were synthesised narratively.</p><p><strong>Results: </strong>Twenty-six studies were included and stratified by clinical task: diagnostic detection of existing DPN, prognostic prediction or risk stratification, and severity classification. Diagnostic detection studies mainly used corneal confocal microscopy, plantar pressure or gait-based assessments, nerve conduction studies, electromyography, and selected clinical variables. Prognostic prediction and risk stratification studies relied mainly on electronic medical record variables, including age, diabetes duration, glycated haemoglobin, renal markers, inflammatory markers, and cardiovascular risk factors. Severity classification studies used electrophysiological measures, structured clinical examination findings, and symptom-based scores. Across tasks, age, diabetes duration, and glycaemic control were recurrent predictors, while imaging, gait, plantar pressure, and electrophysiological features were more task-specific. Reported internal performance was often high, but most studies used small or single-centre datasets, heterogeneous reference standards, and limited reporting of predictor importance. External validation was uncommon, and risk of bias was frequently high, particularly in the analysis domain.</p><p><strong>Conclusions: </strong>AI approaches for DPN assessment show promise, but interpretation depends on task. Future studies should develop task-specific models, report predictor importance transparently, and undertake external and prospective validation before routine clinical implementation.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148419153","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":"Applications of digital twin technology in cardiometabolic disease management: a scoping review.","authors":"Meihua Yin, Ruzhen Luo, Shengying Hu, Hong Lin, Hongyu Sun, Yumei Sun","doi":"10.1186/s12911-026-03673-0","DOIUrl":"https://doi.org/10.1186/s12911-026-03673-0","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Cardiometabolic diseases are characterized by long disease courses, substantial inter-individual heterogeneity, and complex interactions among multiple risk factors, posing significant challenges to long-term management and precision interventions. Digital twin technology, which constructs individualized digital representations of patients, has demonstrated unique potential in precision disease management and clinical decision support. However, there remains a gap in the systematic synthesis of its applications in cardiometabolic disease management.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Objective: &lt;/strong&gt;This scoping review aimed to systematically outline the core application scenarios of digital twin technology in cardiometabolic disease management, summarize the integrated data sources and key technical approaches used to construct and operate digital twins, identify current technical, ethical, and clinical challenges, and outline future research directions.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A scoping review was conducted on November 29, 2025. Relevant studies on the application of digital twin technology in cardiometabolic disease management were systematically searched, screened, and selected. Data were extracted and synthesized with a focus on application scenarios, data sources, modeling approaches, implementation stages, and reported challenges.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Of the 5596 articles identified, 32 met inclusion criteria after screening. Digital twin applications in cardiometabolic diseases primarily focused on individualized risk prediction and stratification, optimization of treatment and intervention strategies, and clinical decision support. Integrated data sources mainly included physiological and metabolic data, electronic data of clinical documents, medical imaging, and behavioral and lifestyle-related data. Data-driven models were the most commonly used approach, followed by mechanistic models, hybrid models, multi-scale or multi-physics models, and simulation-based optimization methods. Most studies remained at the proof-of-concept or early implementation stage. Major challenges included limited data availability and integration, difficulties in model personalization and validation, ethical and privacy concerns, and insufficient integration into clinical and care workflows.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Digital twin technology shows significant promise for precision management of cardiometabolic diseases. Existing studies indicate that digital twins are primarily applied in individualized risk assessment, optimization of treatment strategies, disease progression simulation, and long-term management support. However, prevailing research remains predominantly data-driven and is not yet fully aligned with sustained disease management and routine care processes. Future efforts should emphasize mechanistic and hybrid modeling approaches, conduct multicenter prospective validations and randomized controlled t","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148419131","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":"Enhancing patient care with AI agents: integrating advanced chatbot technologies for improved healthcare delivery.","authors":"Yasmine Abu Adla, Ahmad El Hajj","doi":"10.1186/s12911-026-03694-9","DOIUrl":"https://doi.org/10.1186/s12911-026-03694-9","url":null,"abstract":"<p><p>The integration of artificial intelligence (AI) in healthcare presents significant opportunities to enhance patient care and streamline medical workflows. However, challenges related to accuracy, reliability, and accessibility continue to limit the widespread adoption of AI-driven medical assistants. To the best of our knowledge, this is the first study to integrate multiple advanced AI techniques into a unified system designed to provide real-time, context-aware medical insights while ensuring accuracy, engagement, and interpretability. This study aims to develop an AI-powered medical agent capable of assisting both patients and healthcare professionals by generating informed medical responses, automating healthcare-related tasks, and improving patient interaction through interactive and reliable assistance. The methods employed in this research include Retrieval-Augmented Generation (RAG) for contextualized medical responses, the Wikipedia API for real-time knowledge retrieval, and knowledge graphs for mapping symptoms to potential diseases. Additionally, a symptom checker tool enables preliminary diagnosis and personalized health recommendations using a prompt-based system. The system was evaluated through automated performance assessments and expert reviews. The results demonstrate that the knowledge graph tool achieved 95% accuracy in medical query responses, showcasing its reliability in symptom-disease mapping. Additionally, sentiment analysis of patient interactions reached 97% accuracy, reinforcing the system's ability to understand and respond empathetically. The Wikipedia-based retrieval system maintained an average response time of 10 seconds, ensuring real-time applicability. Furthermore, 100 medical experts rated the AI agent's responses 4.6 for comprehensiveness, 4.5 for engagement, and 4.5 for empathy and tone on a 5-point scale. These findings suggest that integrating advanced AI techniques enhances the accuracy, responsiveness, and contextual relevance of AI-driven medical agents. This scalable and reliable system presents a viable solution to reduce healthcare workload, enhance patient engagement, and democratize access to trusted medical information, reinforcing its potential as a transformative tool in modern healthcare.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148418850","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}