BMC Medical Informatics and Decision Making最新文献

{"title":"Outcomes of acute kidney injury patients with and without chronic kidney disease in intensive care units: a retrospective contrast analysis.","authors":"Yan Liang, Ji Zhang, Tianhao Weng, Wenxian Qiu","doi":"10.1186/s12911-026-03532-y","DOIUrl":"https://doi.org/10.1186/s12911-026-03532-y","url":null,"abstract":"<p><strong>Background: </strong>Chronic kidney disease (CKD) is a common complication among patients in the intensive care unit (ICU), however, the impact of pre-existing CKD on patient who were admitted to the ICU and diagnosed with acute kidney injury (AKI) remains controversial.</p><p><strong>Methods: </strong>Critically ill patients (18-90 years old) with AKI were enrolled as study participants from the Medical Information Mart for Intensive Care-III database. Patients with repeated ICU admissions and a length of stay less than 48 h were excluded. In-ICU mortality was considered the main endpoint, and multivariable Cox regression analyses were performed. Moreover, propensity score matching (PSM) was employed to adjust potential interference factors, and the three-year survival rate was analyzed using the Kaplan-Meier method.</p><p><strong>Results: </strong>This study included 20440 patients, divided into the pure AKI group (18441 patients) and the acute-on-chronic kidney disease (ACKD) group (1999 patients). Multivariable Cox regression analyses revealed lower in-ICU mortality in ACKD group than pure AKI group (7.0% versus 7.7%, hazard ratio (HR) 0.83, 95% confidence interval (CI) 0.69-0.99, P = 0.047), as well as after PSM (7.0% versus 9.9%, HR 0.79, 95% CI 0.63-0.99, P = 0.049). Statistically significant differences persisted when patients were stratified by AKI stages. The results demonstrated that preexisting CKD was associated with reduced in-ICU mortality among AKI stage 3 patients, regardless of the classification criteria used. However, the Kaplan-Meier method indicated a lower three-year survival in the ACKD group (P = 0.017).</p><p><strong>Conclusion: </strong>Among AKI patients in ICU, the ACKD group had lower ICU mortality compared to pure AKI group but experienced worse long-term survival.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147856017","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":"Forecasting hospital bed occupancy: a time series approach with prophet.","authors":"Mohammad Fattouh, L Lyssenko, F Heilmeyer, T Ball, Ch Haverkamp","doi":"10.1186/s12911-026-03542-w","DOIUrl":"https://doi.org/10.1186/s12911-026-03542-w","url":null,"abstract":"<p><strong>Background: </strong>Accurate hospital bed occupancy forecasting is essential for effective resource planning and patient flow management. While complex machine learning models have gained popularity in healthcare forecasting, their operational utility often falls short due to high maintenance costs and limited interpretability. This study evaluates the performance and practicality of Prophet, a parsimonious time-series model, for mid-term hospital bed occupancy forecasting.</p><p><strong>Methods: </strong>We applied the Prophet model to daily bed occupancy data from the Medical Center - University of Freiburg (2010-2023), incorporating public holidays and a COVID-19 pandemic indicator as exogenous regressors. Prophet decomposes time series into trend, seasonality, and holiday effects, offering interpretable components. Forecast accuracy was assessed via rolling cross-validation over 2022-2023 for horizons of 30, 60, 90, and 180 days. A production-ready forecasting pipeline and dashboard were also implemented using cloud-native tools.</p><p><strong>Results: </strong>Prophet achieved low MAPE values across all horizons (3.21%-3.53%) with coverage above 80%, demonstrating reliable accuracy comparable to or better than more complex models that often require higher computational resources and operational costs, such as deep neural networks. Component analysis revealed patterns aligned with hospital operations; weekly and yearly cycles, and holiday effects, highlighting the model's interpretability.</p><p><strong>Conclusions: </strong>This study shows that mid-term hospital bed occupancy can be accurately forecasted using a simple, interpretable model like Prophet. In contrast to more complex architectures, Prophet offers robust performance with minimal tuning, faster deployment, and clearer insights that are critical in operational settings. These findings reinforce the argument that, for structured forecasting tasks like bed occupancy, simple models can rival complex ones, not only in accuracy, but also in reproducibility, scalability, and operational value.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147856105","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":"Actual use, intention to use, and user attitudes toward mobile health applications for coronary heart disease self-management: a systematic review and meta-analysis.","authors":"Tesema Etefa Birhanu, Elena Vlahu-Gjorgievska, Khin Than Win","doi":"10.1186/s12911-026-03554-6","DOIUrl":"https://doi.org/10.1186/s12911-026-03554-6","url":null,"abstract":"<p><strong>Background: </strong>Mobile Health applications (mHealth) have become a promising approach to support self-management of coronary heart disease (CHD). No previous studies have examined user acceptance constructs, and the results remain inconsistent. The aim of this study is to synthesize and quantify the pooled prevalence of current use, intention to use, perceived usefulness, and positive user attitudes toward mHealth apps among patients with coronary heart disease.</p><p><strong>Methods: </strong>A systematic review and meta-analysis were conducted in accordance with PRISMA guidelines and registered in PROSPERO (CRD420251018916). Five electronic databases (PubMed, Web of Science, Cochrane Library, CINAHL, and SCOPUS) were searched for studies published in English between January 2015 and April 2025. Primary studies reporting at least one acceptance-related construct: actual use, intention to use, perceived usefulness, and positive user attitude among patients with CHD or cardiac events were included. Random-effects models (metaprop, REML) were used to estimate pooled prevalence. Heterogeneity, sensitivity test, subgroup analysis, and publication bias assessments were performed.</p><p><strong>Results: </strong>A total of 6113 participants in 17 studies. The pooled prevalence of actual use was [39% (95% CI: 24%-54%)], intention to use was [61% (95% CI: 53%-69%)], perceived usefulness was [69% (95% CI: 49%-88%)], and positive user attitude was [80% (95% CI: 69%-91%)]. Substantial heterogeneity was observed across studies. Sensitivity analysis indicated no influential outliers. The funnel plot and Egger's test indicate no statistically significant publication bias. However, the findings should be interpreted with caution due to substantial heterogeneity and a small number of studies.</p><p><strong>Conclusion: </strong>According to our findings, despite strong behavioural intention, high perceived usefulness, and positive user attitudes, actual usage remains relatively low, highlighting a gap between acceptance and implementation. The findings provide a potential quantitative basis for guiding the design and development of mHealth applications and for emphasizing user-centered interfaces to translate acceptance into sustained engagement in CHD self-management. These findings reveal significant untapped potential for mHealth-supported CHD self-care and the closely related cardiovascular population with similar self-management needs.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147856028","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":"Agentic GPT-5.0 system outperforms standard large language models and human experts in critical care clinical decision-making: a simulation study.","authors":"Evren Ekingen, Mete Ucdal","doi":"10.1186/s12911-026-03555-5","DOIUrl":"https://doi.org/10.1186/s12911-026-03555-5","url":null,"abstract":"<p><strong>Background: </strong>Agentic artificial intelligence (AI) systems employing multi-model architectures with iterative reasoning may surpass standard single-model large language models (LLMs) in complex clinical decision-making. Comprehensive comparisons of agentic versus standard LLM deployment against human specialists in critical care remain limited.</p><p><strong>Objective: </strong>This simulation study compared the performance of an agentic system combining GPT-5.0 and Gemini 2.0 Flash against two standard LLMs (Gemini 2.0 Flash and GPT-4o) and human specialists in acid-base disorder interpretation and sepsis management using text-based clinical vignettes.</p><p><strong>Methods: </strong>Forty-five clinical vignettes (20 acid-base, 25 sepsis) developed by an independent expert panel were evaluated by: (1) Gemini 2.0 Flash (standard single-turn); (2) GPT-4o (standard single-turn); (3) an agentic system combining GPT-5.0 and Gemini 2.0 Flash with multi-step reasoning and cross-verification; and (4) 20 board-certified physicians. Responses were anonymized and assessed by two blinded graders against pre-established gold standards using an explicit scoring rubric.</p><p><strong>Results: </strong>For acid-base disorders, the agentic system achieved 91.0% overall accuracy (95% CI 85.2-96.8%), significantly outperforming GPT-4o (78.0%, P = .002), Gemini (74.5%, P < .001), and human specialists (83.0%, P = .038). SSC hour-1 bundle compliance was 96.8% for the agentic system versus 82.4% for GPT-4o, 79.2% for Gemini, and 90.4% for humans (all P < .05). ROC analysis demonstrated superior discrimination for the agentic system (AUC = 0.932) compared to humans (0.856), GPT-4o (0.814), and Gemini (0.786). Subgroup findings in complex case categories are exploratory given small case numbers.</p><p><strong>Conclusions: </strong>In this simulation study using text-based clinical vignettes, an agentic AI system combining GPT-5.0 and Gemini 2.0 Flash demonstrated significantly higher performance than standard LLM implementations and human medical specialists in structured tasks of acid-base interpretation and sepsis bundle compliance. These simulation-based findings suggest that agentic architectures may represent a promising direction for structured clinical decision support; prospective validation in real clinical environments with actual patient data is essential before implementation.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147856040","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":"Research on depression diagnosis method based on multi-scale analysis of frontal lead EEG.","authors":"Huigang Wang, Weilin Kuang","doi":"10.1186/s12911-026-03521-1","DOIUrl":"https://doi.org/10.1186/s12911-026-03521-1","url":null,"abstract":"<p><strong>Background: </strong>Depression is one of the most prevalent mental disorders globally, severely affecting individuals' emotional, cognitive, and physical functions while imposing profound socioeconomic impacts. Traditional diagnostic approaches primarily rely on clinical judgment and self-assessment scales; however, these methods carry inherent risks of misdiagnosis and missed diagnosis, necessitating more precise and efficient diagnostic tools.</p><p><strong>Aims: </strong>This study employs a two-channel frontal EEG system for depression detection, aiming to simplify data acquisition processes and reduce costs while ensuring high classification accuracy.</p><p><strong>Method: </strong>Electroencephalography (EEG), as a non-invasive biosignal monitoring technique, enables real-time recording of brain electrical activity. By extracting multiple features including relative power, fuzzy entropy, and mutual information, combined with multi-scale analysis techniques, the detection accuracy for depression was further enhanced.</p><p><strong>Results: </strong>The study compared three traditional machine learning models with three deep learning models, among which the Gated Recurrent Unit (GRU) model demonstrated superior performance, achieving a classification accuracy of 91.02% and exhibiting strong robustness.</p><p><strong>Conclusions: </strong>The aforementioned approach provides preliminary technical support for the application of EEG signals in depression detection, and represents a proof-of-concept for multi-scale feature-enhanced automated depression screening. Further validation in larger, clinically representative, and externally verified cohorts is necessary before practical deployment can be considered.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147834040","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":"Strategies to support implementation of infectious disease decision support systems: a scoping review.","authors":"Kaia M Nielsen, Sara R Packull-McCormick, Lauren E Grant","doi":"10.1186/s12911-026-03515-z","DOIUrl":"https://doi.org/10.1186/s12911-026-03515-z","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Background: &lt;/strong&gt;Decision support systems (DSSs) are computerized tools that analyse data to guide actions and inform decision-making. Although DSSs are increasingly used in infectious disease contexts, their adoption remains inconsistent. Effective implementation is essential for integration of these tools into practice. Implementation strategies that promote sustainable uptake have not been comprehensively mapped for infectious disease DSSs. This scoping review aimed to identify and describe implementation strategies, associated outcomes, and use of theories, models, and frameworks (TMFs) in the implementation of infectious disease DSSs designed for early warning, detection, or prevention.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;A scoping review was conducted following the Joanna Briggs Institute methodology and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. A limited search in MEDLINE was used to develop the search strategy based on relevant publications. The final search was applied across MEDLINE, CAB Direct, AGRICOLA, and Web of Science using a time limiter (2000-2025). Studies were screened against eligibility criteria by two independent reviewers. Data were charted on implementation strategies, outcomes, and TMFs. Strategies were mapped to the Expert Recommendations for Implementing Change (ERIC) taxonomy, and outcomes were mapped to Proctor's implementation outcomes taxonomy. Data were synthesized by grouping DSSs according to their primary level of public health action: clinical/individual, population/program, or system/governance. Reported or codable strategies, outcomes, and TMFs were summarized within each level of action.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Of the 18,708 records identified, 26 studies reported in 27 publications met the inclusion criteria. In total, 20 unique implementation strategies were identified through reviewer coding of narrative descriptions, 7 implementation outcomes were inferred from descriptive indicators, and 3 TMFs were reported in the literature. Clinical/individual-level DSSs included 13 strategies, 7 outcomes, and 2 TMFs; population/program-level DSSs included 13 strategies, 7 outcomes, and 1 TMF; and system/governance-level DSSs included 7 strategies, 3 outcomes, and no TMFs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusions: &lt;/strong&gt;Implementation of infectious disease DSSs most often involved activities that mapped to educational and stakeholder engagement strategies, with limited reported use of guiding theories or frameworks. Although outcome reporting was relatively common, outcome definitions and depth of outcome reporting varied widely. More deliberate use of TMFs and systematic outcome evaluation could strengthen the evidence base for DSS implementation. In policy and management contexts, better alignment between strategy selection, system design, and public health objectives may enhance sustainable DSS adoption and impact.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Registration: &lt;/strong&gt;Not","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147834013","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 auxiliary diagnostic system for early osteoporosis screening using stacking ensemble learning.","authors":"Qian Wang, Jun Liu, Min Zhu, Hong-Ying Ye, Yan Shang","doi":"10.1186/s12911-026-03350-2","DOIUrl":"https://doi.org/10.1186/s12911-026-03350-2","url":null,"abstract":"","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":"26 1","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13147576/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147834027","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 of ventilator therapeutic pressure for optimized CPAP titration.","authors":"Jingjing Huang, Yuxin Huang, Chengyu Zhang, Jun Shao, Cheng Zhang","doi":"10.1186/s12911-026-03535-9","DOIUrl":"https://doi.org/10.1186/s12911-026-03535-9","url":null,"abstract":"<p><strong>Background: </strong>Accurate prediction of therapeutic pressure for Continuous Positive Airway Pressure (CPAP) therapy is essential for effective treatment of Obstructive Sleep Apnea (OSA). Existing methods often rely on complex sleep-related parameters and small sample sizes, limiting their generalizability. This study aims to develop a more accessible, data-driven model using readily available demographic and physiological variables to predict CPAP pressure, improving both accuracy and scalability.</p><p><strong>Methods: </strong>We employed a machine learning approach, integrating decision trees, gradient boosting algorithms (LightGBM, XGBoost, CatBoost), and neural networks to predict therapeutic pressure. Forward selection based on the Akaike Information Criterion (AIC) was used to identify the most relevant variables. The model was trained on a dataset of 2,092 patients, with model performance assessed using mean absolute error (MAE).</p><p><strong>Results: </strong>The most influential variables identified were BMI, neck circumference, and waist-to-hip ratio. Among the algorithms, LightGBM achieved the highest predictive accuracy, with the lowest MAE. Ensemble methods, such as voting, did not improve performance beyond LightGBM alone. Subsample analyses revealed that prediction accuracy varied across BMI ranges and ventilator brands.</p><p><strong>Conclusions: </strong>The study demonstrates that BMI and other physical parameters play a pivotal role in determining CPAP pressure, offering a simplified yet effective prediction model. This approach has significant potential for clinical applications, particularly in resource-limited settings, where access to complex sleep studies may be restricted. Future research could enhance the model by incorporating real-time physiological data and expanding data collection to diverse populations.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147833961","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":"Interpretable machine learning for postoperative nausea and vomiting prediction in elderly orthopedic patients: a comparative study.","authors":"Li-Heng Li, Hao Guo, Hao Wang, Yu-Bo Xie","doi":"10.1186/s12911-026-03527-9","DOIUrl":"https://doi.org/10.1186/s12911-026-03527-9","url":null,"abstract":"<p><strong>Background: </strong>Postoperative nausea and vomiting (PONV) prolongs hospitalization and reduces patient satisfaction. Identifying high-risk elderly patients requires accurate absolute risk assessments, yet existing tools often lack probability calibration and transparency.</p><p><strong>Methods: </strong>We included 1216 elderly patients undergoing elective hip or knee surgery. To strictly prevent data leakage, the dataset was partitioned into training, validation, and independent test sets in a 7:1:2 ratio prior to any imputation or feature selection. Following the systematic hyperparameter optimization of 12 distinct machine learning algorithms, a StackNet meta-model was developed by fusing optimal base-learner probabilities with raw clinical features. Clinical utility was evaluated via Brier scores and Decision Curve Analysis (DCA), alongside SHapley Additive exPlanations (SHAP) interpretability.</p><p><strong>Results: </strong>Overall PONV incidence was 33%. The StackNet model achieved an AUC of 0.9338, significantly outperforming the conventional Logistic Regression baseline (AUC = 0.7564, p < 0.001) with superior calibration (Brier score = 0.102). On the independent test set, the StackNet model achieved an accuracy of 0.7860, sensitivity of 0.9250, specificity of 0.7178, and AUC of 0.9338, while the Logistic Regression baseline achieved an accuracy of 0.6584, sensitivity of 0.6750, specificity of 0.6503, and AUC of 0.7564. SHAP analysis identified preoperative frailty status and baseline hemoglobin levels as primary risk drivers.</p><p><strong>Conclusion: </strong>The StackNet framework offers highly calibrated absolute risk estimates for PONV in elderly orthopedic patients. Combined with SHAP transparency, it provides a clinically actionable tool to facilitate personalized antiemetic prophylaxis while avoiding unnecessary medical interventions due to overestimated risks.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147833963","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":"Optimized explainable AI and digital twin for patient flow improvement in ICU during respiratory epidemics.","authors":"Miguel Ortiz-Barrios, Sebastián Arias-Fonseca, Helder Jose Celani de Souza, Jose F Torres-Avila, Ana Maldonado-Olea, Isidro J Ángel-Gaviria, Tobías A Parodi-Camano, Omar Ayala Ruiz, Lorena Del Jesus Sarao-Cruz, Martha María Sánchez-Bolívar","doi":"10.1186/s12911-026-03523-z","DOIUrl":"https://doi.org/10.1186/s12911-026-03523-z","url":null,"abstract":"<p><strong>Background: </strong>Respiratory epidemics often place substantial pressure on intensive care units (ICU), which are continuously challenged to managing acute and life-threatening conditions under unpredictable workloads. During these periods, ICUs usually exhibit inefficient patient flows, treatment delays, and critical resource shortages. Proactive decision-making and precise interventions are therefore pivotal for patient survival and minimizing long-term sequelae.</p><p><strong>Methods: </strong>This paper proposes a robust approach combining Artificial Intelligence (AI), Bayesian Optimization, and Digital Twin (DT) to support ICU patient flow management. An eXtreme Gradient Boosting (XGBoost) algorithm is used to predict the patient transfer probability from the emergency department (ED) to the ICU within the next 24 h. Bayesian optimization is employed for efficient hyperparameter tuning of the XGBoost model. Then, the transfer predictions are inserted into a DT to verify ICU capacity for timely care and design interventions for process mismatches.</p><p><strong>Results: </strong>A case study from a European healthcare group validates the proposed approach. The specificity of the prediction XGBoost model was 94.90% (CI 95% 91.72% - 97.11%), whereas the sensitivity was 81.55% (CI 95% 72.70% - 88.51%). Finally, the median ICU bed waiting time decreased to between 66.74 and 69.38 h after implementing a patient transfer policy with a partner hospital having available ICU beds.</p><p><strong>Conclusions: </strong>This study demonstrates the effectiveness of AI-DT in predicting the probability of ICU transfers, assessing the operational response of emergency wards and intensive care units, and crafting practical scenarios for enhancing patient flow management.</p>","PeriodicalId":9340,"journal":{"name":"BMC Medical Informatics and Decision Making","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-05-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147811343","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}