BMC Medical Informatics and Decision Making最新文献

{"title":"Development and external validation of a machine-learning risk model for colorectal cancer triage in Sweden's fast-track pathway.","authors":"Michiel A van Nieuwenhoven, Yang Cao","doi":"10.1186/s12911-026-03813-6","DOIUrl":"https://doi.org/10.1186/s12911-026-03813-6","url":null,"abstract":"<p><strong>Background: </strong>Fast-track colonoscopy pathways detect colorectal cancer (CRC) but strain endoscopy capacity. We developed and externally validated multivariable risk models combining faecal immunochemical test (FIT) with clinical data to triage Swedish fast-track referrals.</p><p><strong>Methods: </strong>We analysed 2,539 fast-track colonoscopies (2016-2020) and 723 as a validation cohort (2021-2022). Predictors were age, sex, eight symptoms, binary FIT, haemoglobin, and suspicious imaging. In development, missing predictors were imputed within training/test splits; validation used complete cases. Class imbalance was handled with oversampling to balance classes. Logistic regression and CatBoost models were built. External validation assessed AUC, calibration, and calibration slope. Variable influence was examined with feature contribution analysis. Decision-curve analysis (DCA) used the entry rule as a treat-all baseline.</p><p><strong>Results: </strong>CRC prevalence was 16% in both cohorts. Development AUCs were 0.81 (95% CI 0.78-0.83) for CatBoost and 0.77 (0.74-0.80) for logistic regression. In external validation, AUCs were 0.67 (0.61-0.74) and 0.66 (0.58-0.75), respectively. FIT status, suspicious imaging, and age were the main predictors. Calibration drift was observed. The entry rule functions as treat-all (specificity 0%). CatBoost DCA curves were near or above treat-all at common thresholds, but 95% bands overlapped; advantage over treat-all was not confirmed.</p><p><strong>Conclusions: </strong>A multivariable approach using FIT, simple laboratory data, and clinical findings can support triage in a fast-track pathway. In validation, discrimination was modest, and superiority over the entry rule could not be shown because all referred patients underwent colonoscopy. External testing across regions, recording quantitative FIT, and routine recalibration are needed before implementation.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148905105","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":"Systematic review of machine learning approaches for predicting sickle cell crisis and mortality risk at the climate-health nexus.","authors":"Mary Ofuru Kama, Ifeyinwa Angela Ajah, Anayo Chukwu Ikegwu, Adaora Angela Obayi, Felix Oshiorenoya Uloko, Immaculate Chidimma Ogbuagu","doi":"10.1186/s12911-026-03796-4","DOIUrl":"10.1186/s12911-026-03796-4","url":null,"abstract":"<p><strong>Background: </strong>Sickle cell anemia (SCA) is a severe genetic blood disorder characterized by recurrent vaso-occlusive crises and increased mortality, with the greatest burden occurring in low- and middle-income countries. Climatic and environmental conditions, including temperature variability, humidity, rainfall, air pollution, and seasonal changes, have been associated with disease exacerbation. However, the extent to which these factors have been incorporated into predictive models remains unclear. This study systematically reviews the application of machine learning (ML) models for predicting SCA crises and mortality in relation to climate and environmental factors.</p><p><strong>Methodology: </strong>The PRISMA guidelines were used, and 34 peer-reviewed studies published between 2005 and 2026 were analyzed to identify the climate variables, ML approaches employed, and predictive performance. The reviewed studies applied a range of ML techniques, including artificial neural networks, random forests, support vector machines, decision trees, logistic regression, and deep learning models. Temperature, humidity, rainfall, wind speed, air quality indicators, and seasonal patterns were the most frequently examined environmental variables.</p><p><strong>Results: </strong>The findings indicate that most existing models rely predominantly on clinical and demographic data, with limited integration of climate information and inadequate representation of high-burden regions, especially Sub-Saharan Africa. Studies incorporating environmental variables reported improved predictive performance and highlighted the potential of climate-informed early warning systems for SCA management.</p><p><strong>Conclusion: </strong>The review recommends development of interdisciplinary, climate-aware ML frameworks, expansion of longitudinal environmental datasets, and increased research in underrepresented regions to support climate-resilient and patient-centered SCA care.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13548623/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148905015","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":"Machine learning-based prediction model for predicting the impact of insulin resistance on the risk of ischemic cardiomyopathy.","authors":"Tuersunjiang Naman, Hui Cheng, Xiao-Lin Yu, Zi-Tong Guo","doi":"10.1186/s12911-026-03693-w","DOIUrl":"10.1186/s12911-026-03693-w","url":null,"abstract":"<p><strong>Background: </strong>The triglyceride-to-high-density lipoprotein cholesterol (TG/HDL-C) ratio and triglyceride glucose-body mass (TyG-BMI) index are reliable indicators of insulin resistance (IR). This study investigated their association with ischemic cardiomyopathy (ICM) and developed a machine learning-based model for ICM risk prediction.</p><p><strong>Methods: </strong>In total, 1,603 subjects participated in this study. Univariable logistic regression analysis was conducted, and variables with P < 0.05 were selected for multivariable logistic regression to identify independent risk factors for ICM. Variables meeting this criterion were adopted to create eight machine learning models, from which the optimal model was selected. Using this best-performing model, SHAP values were visualized, and an online calculator was developed. The model was validated via a calibration plot and DCA.</p><p><strong>Results: </strong>Univariate and multivariate logistic regression analyses revealed that TyG-BMI, age, ejection fraction, TC/HDL-C, sex, HDL-C, TC, BMI, hemoglobin, diabetes, and hypertension were independent risk factors for ICM (P < 0.05). Based on these factors, SHAP visualization and an online calculator were developed. The calibration plot indicated strong alignment between the model's predicted and actual values, whereas the DCA demonstrated the model's clinical utility.</p><p><strong>Conclusion: </strong>The TyG-BMI and TC/HDL-C ratio independently predict ICM risk, with the XGB model identified as the most effective for ICM risk prediction, indicating substantial clinical applicability.</p><p><strong>Clinical trial registration number: </strong>Not applicable.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13523236/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148849859","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":"Decentralized rare disease studies in Germany: first results and hurdles of secondary use of patient data.","authors":"Michele Zoch, Christian Gierschner, Jens Weidner, Martin Sedlmayr, Gabriele Müller, Daniela Choukair, Georg F Hoffmann, Nicole Toepfner, Reinhard Berner, Fabian Prasser, Josef Schepers, Helge Hebestreit","doi":"10.1186/s12911-026-03688-7","DOIUrl":"10.1186/s12911-026-03688-7","url":null,"abstract":"<p><strong>Background: </strong>The research challenges associated with rare diseases is characterized by a scarcity of information as well as reliable data due to their low prevalence. The problem of the \"underpowered studies\" is stemmed from a small research community, limited study participants and scarce data. The German project \"Collaboration on Rare Diseases - Medical Informatics\" (CORD-MI), tackles these problems by improving research opportunities and patient care by employing innovative IT solutions for collaborative data use across 20 German university hospitals. One possibility was to conduct decentralized studies based on secondary data. Three studies based on four rare diseases served as examples: (1) Cystic Fibrosis (CF), (2) Phenylketonuria (PKU), and (3) Kawasaki Disease and Multisystem Inflammatory Syndrome in Children (MIS-C).</p><p><strong>Methods: </strong>All three decentralized studies were conducted using routine inpatient data from German university hospitals. For each use case, an interdisciplinary team defined research questions, created analysis scripts based on the Core Data Set of the Medical Informatics Initiative (MII), executed these locally at participating sites, and aggregated anonymized results for descriptive statistical analysis. A frequency threshold rule was applied to protect patient privacy. Study protocols and analysis scripts for data extraction and evaluation are publicly available, and the studies were reported in detail in accordance with the Reporting of Studies Conducted Using Observational Routinely-Collected Health Data (RECORD) Statement.</p><p><strong>Results: </strong>Results from up to 17 German university hospitals were achieved for all three decentralized studies. The challenges in the areas of health care process bias, inaccurate coding of rare diseases, difficult verification of study results, and loss of information due to masking of small study case numbers as well as imprecise definition of the cohorts are discussed.</p><p><strong>Conclusions: </strong>Naming the hurdles enables the identification of areas for improvements, which will be the base for the development of new approaches or adaptations of existing tools and methodologies for the future. Although adaptation would make an impact, the initial results already show that decentralized analyses based on secondary use of patient data can improve research and thus also the care for people with rare diseases.</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":"2026-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13471175/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148719574","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":"Interpretable machine learning model for early prediction of deep vein thrombosis in elderly ICU patients: multicenter development and external validation.","authors":"Jiabao Yang, Xiangrong Zhao, Juan Ren","doi":"10.1186/s12911-026-03747-z","DOIUrl":"10.1186/s12911-026-03747-z","url":null,"abstract":"<p><strong>Background: </strong>Deep vein thrombosis (DVT) is a frequent complication in elderly intensive care unit (ICU) patients, while reliable prediction tools for this population remain limited. This study aimed to develop and externally validate an interpretable machine learning model for early prediction of DVT in critically ill elderly patients.</p><p><strong>Methods: </strong>In this multicenter retrospective study, ICU patients aged ≥ 75 years from two tertiary medical centers in China were enrolled. Patients with pre-existing DVT were excluded. A development cohort (Suzhou, n = 1,717) and an external validation cohort (Xi'an, n = 533) were established. Clinical variables collected within 24 h after ICU admission were used to predict incident DVT during ICU stay. Least absolute shrinkage and selection operator regression was applied for feature selection, and four machine learning models were developed and compared. Model discrimination, calibration, and interpretability were evaluated using area under the curve (AUC), calibration analysis, and SHapley Additive exPlanations (SHAP).</p><p><strong>Results: </strong>DVT occurred in 15.2% and 21.0% of patients in the development and external validation cohorts, respectively. Among the evaluated models, XGBoost achieved the best performance, with AUCs of 0.843 in internal validation and 0.811 in external validation. Calibration analysis demonstrated good agreement between predicted and observed risks. Key predictors included D-dimer, C-reactive protein, platelet-to-lymphocyte ratio, procalcitonin, mechanical ventilation, and mechanical thromboprophylaxis. SHAP analysis quantified the contribution of individual predictors.</p><p><strong>Conclusions: </strong>We developed and externally validated an interpretable machine learning model for early DVT prediction in elderly ICU patients. The model may support risk stratification and clinical decision-making in critical care settings.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13673899/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149191126","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":"The word and the way: strategies for domain-specific BERT pre-training in German medical NLP.","authors":"Henry He, Johann Frei, Raphael Schmitt","doi":"10.1186/s12911-026-03728-2","DOIUrl":"10.1186/s12911-026-03728-2","url":null,"abstract":"<p><strong>Background: </strong>Digital healthcare generates vast amounts of clinical texts that hold potential for AI-assisted applications. However, existing German biomedical language models either rely on older architectures or are trained on limited data, which may hinder their performance in real-world settings.</p><p><strong>Methods: </strong>To explore the impact of domain adaptation strategies in German clinical NLP, we developed a family of domain-specific RoBERTa-based language models, collectively referred to as ChristBERT (Clinical- and Healthcare-Related Issues and Subjects Tuned BERT). To address the lack of large-scale German clinical corpora, we curated a 13.5 GB dataset consisting of scientific publications, clinical texts, and health-related web content. Additionally, we employed data augmentation via translation of English clinical corpora. Three domain adaptation strategies were explored: continued pre-training, pre-training from scratch, and pre-training with domain-specific vocabulary adaptation.</p><p><strong>Results: </strong>The resulting models were evaluated on three medical named entity recognition and two text classification tasks. Our models consistently outperformed four existing general-purpose and medical German models on four out of five tasks. The results demonstrate that the choice of domain adaptation strategy significantly influences downstream task performance. Based on the empirical results, pre-training from scratch is effective for highly specialized clinical texts, whereas continued pre-training is suited for more commonly written medical texts.</p><p><strong>Conclusions: </strong>ChristBERT establishes a new state-of-the-art for German clinical language modeling. Our findings indicate that the optimal domain adaptation strategy is task-dependent and remains crucial, as adapted models consistently outperformed general-purpose language models in our experiments. To support further research and application in German medical NLP, all developed models are publicly released.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13449853/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148683448","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":"GERBEHRT: a BERT-based model tailored for German electronic health records - potential in chronic kidney disease prediction.","authors":"Anja Seidel, Edgar Steiger, Friedrich Alexander von Samson-Himmelstjerna, Lars Eric Kroll","doi":"10.1186/s12911-026-03733-5","DOIUrl":"10.1186/s12911-026-03733-5","url":null,"abstract":"<p><strong>Background: </strong>Chronic kidney disease (CKD) is a critical, progressive condition associated with high mortality and substantial healthcare costs. Early detection is essential, as it can slow disease progression and improve patient outcomes. With the increasing availability of large-scale electronic health records (EHRs), the question arises to what extent these data, when combined with machine-learning algorithms specifically tailored to EHR characteristics, can enhance personalized CKD risk prediction.</p><p><strong>Methods: </strong>We developed a transformer model adapted from BEHRT (Bidirectional Encoder Representations from Transformers for EHRs) and specifically tailored for German (GER) EHRs, which we refer to as GERBEHRT. GERBEHRT was pre-trained on outpatient claims data from more than 9 million statutorily insured patients and fine-tuned with nearly 1 million additional patients to predict CKD. The model incorporates EHR features not previously explored in BERT-based approaches and introduces an efficient method to represent multiple attributes per medical concept, such as diagnoses and medications. GERBEHRT was compared with more traditional models and predictions restricted to established risk factors, and the importance of its input features was assessed through an ablation study.</p><p><strong>Results: </strong>In a test cohort of 3.7 million patients with 1.5% CKD positives, GERBEHRT achieved an area under the receiver operating characteristic curve (AUROC) of 87.9% and an average precision (AVPR) of 11.4% for the three-year prediction of incident moderate-to-severe CKD, outperforming risk-factor-based models (AUROC/AVPR: 83.6/6.4%) and more traditional algorithms using the full EHR (AUROC/AVPR: 86.9/10.1%).</p><p><strong>Conclusions: </strong>Predicting moderate-to-severe CKD based on real-world EHRs remains challenging. However, our proposed architecture was able to make more accurate predictions than traditional approaches and feature sets, underscoring the importance of comprehensive EHR utilization and the potential of tailored deep learning models for personalized CKD risk prediction and targeted patient screening.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13445917/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148677398","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":"Dual-modality ultrasound radiomics of the medial gastrocnemius muscle for detecting diabetic peripheral neuropathy: a multicenter prospective study.","authors":"Wenqian Qiu, Erfeng Chen, Wanyan Li, Rong Xiao, Tingting Xie, Danqing He","doi":"10.1186/s12911-026-03726-4","DOIUrl":"10.1186/s12911-026-03726-4","url":null,"abstract":"<p><strong>Background: </strong>Diabetic peripheral neuropathy (DPN) affects 20%-50% of patients with type 2 diabetes mellitus (T2DM), yet early screening is limited by restricted access to nerve conduction studies (NCS). We aimed to develop and validate a dual-modality ultrasound radiomics model combining B-mode ultrasound (BMUS) and shear wave elastography (SWE) of the medial gastrocnemius (MG) - an accessible surrogate reflecting neurogenic muscle atrophy - for noninvasive DPN detection.</p><p><strong>Methods: </strong>A total of 190 T2DM patients (380 limbs) enrolled between June 2025 and February 2026 were allocated to training (158 limbs), internal test (68 limbs), and external validation (154 limbs) cohorts. DPN was defined as abnormal nerve conduction in at least two peripheral nerves, per ADA criteria. MG BMUS and SWE images were acquired bilaterally by blinded ultrasonographers. Sequential ICC (≥ 0.75), Mann-Whitney U (P < 0.05), Spearman (|r| > 0.7), and elastic-net filtering of 1,888 features selected 14 final features (9 BMUS, 5 SWE). An XGBoost MG-RadScore was derived via ten-fold cross-validation and integrated with clinical predictors in a final combined model. Performance was assessed by AUC, calibration, Hosmer-Lemeshow test, DCA, and SHAP.</p><p><strong>Results: </strong>Diabetes duration (OR 1.13, 95% CI: 1.07-1.19, P < 0.001) and right MG mean SWE modulus (RMG_SWEmean; OR 0.94, P = 0.022) were independent clinical predictors. The clinical model achieved AUCs of 0.741/0.780/0.783; the MG-RadScore model showed higher discrimination than it (0.833/0.788/0.804). The combined model achieved AUCs of 0.895 (95% CI: 0.839-0.940), 0.861 (95% CI: 0.745-0.949), and 0.854 (95% CI: 0.785-0.909). Calibration was satisfactory (Hosmer-Lemeshow P > 0.05); DCA confirmed favorable net benefit; SHAP identified diabetes duration as the top predictor, followed by MG-RadScore and RMG_SWEmean.</p><p><strong>Conclusion: </strong>A dual-modality MG ultrasound radiomics model integrating BMUS and SWE features with diabetes duration and RMG_SWEmean accurately detects DPN, achieving AUCs above 0.85 across multicenter prospective cohorts. This muscle-targeted, SHAP-interpretable approach offers a noninvasive alternative to nerve sonography and may facilitate early DPN screening where NCS is unavailable.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13660074/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149170080","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":"Application of knowledge graphs in health management for chronic disease patients: a scoping review.","authors":"Ziting Xu, Chunhao Dai, Haoyuan Li, Wen Li, Teng Yang, Luyao Yan, Lijuan Yang, Minmin Leng","doi":"10.1186/s12911-026-03704-w","DOIUrl":"10.1186/s12911-026-03704-w","url":null,"abstract":"<p><strong>Background: </strong>Chronic diseases impose a significant burden on global health. Knowledge graphs (KGs), which integrate and represent multisource information as interconnected networks, provide a promising approach for enabling personalized and dynamic health management. The aim of this scoping review is to systematically map the current landscape of KG applications in chronic disease health management.</p><p><strong>Methods: </strong>In accordance with the Arksey and O'Malley framework and the PRISMA-ScR guidelines, a systematic search was conducted in eight databases, namely, the Wanfang Database, CNKI, VIP, SinoMed, PubMed, Embase, Web of Science, and CINAHL, from the establishment of the databases to August 2025. Two researchers independently screened the literature and extracted data on the basis of the inclusion and exclusion criteria.</p><p><strong>Results: </strong>A total of 15 studies published between 2018 and 2025 were included. In these studies, the KG construction process generally involved five stages: data collection, information extraction, knowledge fusion, graph construction, and visualization. KGs were applied across a range of chronic diseases, including metabolic, cardiovascular, and respiratory diseases, as well as cancers. In terms of application scenarios, self-health management was the most common (9 studies, 60.0%), followed by clinical decision support (5 studies, 33.3%), intelligent question answering (4 studies, 26.7%), and health monitoring/early warning (1 study, 6.7%). With respect to system maturity, 12 studies (80.0%) were classified as prototype systems, 3 studies (20.0%) were classified as pilot implementations, and none were classified as clinically deployed systems. The reported outcomes were mainly technical or feasibility oriented, and 7 studies (46.7%) did not report any explicit evaluation metrics.</p><p><strong>Conclusion: </strong>Knowledge graphs show promise for integrating heterogeneous health data in chronic disease management. However, research has focused predominantly on system development and proof-of-concept validation, with limited high-quality evidence regarding clinical effectiveness. To support clinically meaningful implementation, in future work, rigorous real-world evaluation should be prioritized and challenges related to data quality, interoperability, and model interpretability should be addressed.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13651244/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149147351","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":"AI-powered risk prediction models for preventable maternal mortality in rural settings: a systematic review.","authors":"Joy Aifuobhokhan, Ayodeji Ogunjinmi, Chukwuemeka Abraham Agbarakwe, Deborah Oladunmolu Oduguwa, Annie Peter Essiet, Temitayo Osunkiyesi, Akinbogun Modesire","doi":"10.1186/s12911-026-03716-6","DOIUrl":"https://doi.org/10.1186/s12911-026-03716-6","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Maternal mortality remains disproportionately high in low- and middle-income countries, particularly in rural settings with limited access to skilled obstetric care. Artificial intelligence and machine learning models offer promise for early risk prediction, yet their methodological rigour, applicability, and deployment feasibility in resource-constrained rural contexts remain inadequately synthesised. This systematic review evaluated AI-powered risk prediction models for preventable maternal mortality, with an emphasis on their suitability for rural and low-resource settings.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A systematic literature search was conducted across PubMed, Scopus, Web of Science, IEEE Xplore, Google Scholar, and African Journals Online for studies published from January 2015 to August 2025. Studies employing AI or machine learning to predict maternal mortality or severe maternal outcomes were included. The Prediction model Risk Of Bias Assessment Tool (PROBAST) assessed methodological quality across four domains: participants, predictors, outcomes, and analysis. Data extraction captured study characteristics, model architectures, performance metrics, validation strategies, and rural implementation considerations. This review was registered with PROSPERO (CRD420251174343) and reported in accordance with PRISMA 2020 guidelines.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Twenty-eight studies met inclusion criteria, predominantly from sub-Saharan Africa (n = 12) and South Asia (n = 8). Dataset sizes ranged from 402 to over 31 million records from national surveys (n = 14), hospital registries (n = 9), and Internet of Things monitoring systems (n = 5). Random Forest (n = 14), ensemble methods (n = 11), and neural networks (n = 11) were most frequently employed. Reported area under the receiver operating characteristic curve values ranged from 0.70 to 0.95 (median 0.84), with sensitivity 70-92% and specificity 65-85%. PROBAST assessment revealed low risk of bias for participants (24/28), predictors (25/28), and outcomes (24/28), but substantial concerns in the analysis domain (14/28 low risk, 8/28 high risk). Key limitations included reliance on synthetic oversampling without external validation, inadequate reporting of calibration, and small sample sizes in IoT studies. Only 11 studies (39%) conducted external validation. Common predictors were maternal age, blood pressure, gestational age, parity, and antenatal care attendance. Rural implementation barriers included limited connectivity, data sparsity, workforce training needs, and the absence of explainability frameworks.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;AI-powered models demonstrate strong discrimination performance for maternal mortality prediction when trained on large, representative datasets. However, methodological weaknesses, particularly inadequate external validation and calibration assessment, limit confidence in generalizability. The underrepresentation ","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149223368","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}