Introduction Atrial fibrillation (AF) is closely associated with metabolic dysfunction. The uric acid–to–albumin ratio (UAR), integrating oxidative stress, inflammation, and nutritional status, reflect cardiometabolic burden, but evidence linking UAR to AF prognosis remains limited. Methods We analyzed clinical data from 1,908 AF patients at West China Hospital, with external validation from the MIMIC database (n=1,366). Associations were assessed using Kaplan–Meier analyses, restricted cubic splines, and multivariable Cox proportional hazards models. Incremental prognostic value beyond the CHA 2 DS 2 -VASc score was evaluated in both cohorts. Exploratory machine learning and SHAP analyses were employed to assess the variable importance of UAR. Subgroup and sensitivity analyses were performed in primary cohort, including additional cardiometabolic adjustment, analyses with cardiac mortality, competing risk models, and longer follow-up. Results Baseline characteristics differed across UAR quartiles, with high UARs associated with substantial burdens of metabolic comorbidities, heart failure, renal dysfunction, and elevated inflammatory and cardiac biomarkers. Mortality was higher in the highest UAR quartile (log-rank P<0.001). In the primary cohort, restricted cubic splines showed a J-shaped association between UAR and 1-year mortality (P for nonlinearity <0.001). In fully adjusted Cox models, UAR (per SD) predicted 1-year all-cause mortality in the primary cohort (HR 1.162, 95% CI 1.036–1.304) and in the MIMIC cohort (HR 1.137, 95% CI 1.092–1.185). Adding UAR to the CHA 2 DS 2 -VASc score improved discrimination (C-index 0.654 to 0.692; P = 0.001), reclassification (continuous NRI 0.178), calibration, and clinical net benefit, with consistent incremental performance in the MIMIC cohort. In both cohorts, SHAP analysis consistently identified UAR as one of the major contributors to mortality prediction. Findings were consistent across subgroups and sensitivity analyses. Conclusion UAR is an independent predictor of mortality in AF and captures cardiometabolic remodeling beyond conventional risk assessment. As a readily available biomarker, UAR may facilitate metabolically guided risk stratification and individualized management in AF populations.

BackgroundHeart failure mortality has risen sharply after years of decline, highlighting the limitations of current risk assessment tools in accuracy, complexity, and cost, and the need for improved predictive models. To address this gap, we developed and validated a deep learning model to improve short-term mortality prediction in heart failure patients.MethodsIn this retrospective study, we leveraged the Medical Information Mart for Intensive Care IV database to develop HF-ECGNet, combining an EfficientNet neural network and a Transformer architecture. We also developed a composite model integrating electrocardiogram-based (ECG) predictions and clinical features. We evaluated model performance using the area under the curve (AUC) and other metrics, with gradient-weighted class activation mapping (Grad-CAM) and Shapley additive explanations (SHAP) analyses for interpretability. We conducted comparisons with N-terminal pro-B-type natriuretic peptide and sequential organ failure assessment (SOFA) scores.ResultsWe analysed a total of 104 844 ECGs from 36 222 admissions. HF-ECGNet achieved an AUC of 0.664 for the first ECG during initial admission, improving to 0.721 for the last ECG. Incorporating three-day ECG data further enhanced performance, with AUCs of 0.691 (first admission) and 0.698 (last admission). HF-ECGNet outperformed NT-proBNP and SOFA. A composite model integrating ECG data and clinical features achieved the highest AUC of 0.725. Grad-CAM identified critical ECG patterns, while SHAP analysis highlighted ECG-derived features as the most influential predictors.ConclusionsHF-ECGNet demonstrates potential as a powerful tool for predicting short-term mortality in heart failure patients. Its innovative architecture and integration of clinical data enable more accurate and interpretable risk stratification. Future multi-centre validation is the critical step to fully ascertain its clinical utility and generalisability.

Hemodynamic instability and uncontrolled hemorrhage remain leading causes of preventable morbidity and mortality in trauma and perioperative critical care. This review summarizes recent advances in machine learning-based approaches for early detection before overt decompensation and for supporting time-critical hemorrhage management in trauma patients. Recent studies have explored machine learning across multiple stages of trauma care, including early warning systems, outcome and mortality prediction, prediction of massive transfusion needs, risk stratification, and bleeding monitoring. Outcome prediction - particularly mortality and complications such as sepsis - remains one of the most extensively studied domains. More recent work has increasingly favored neural network-based architectures, including deep and hybrid models, reflecting their capacity to model complex, high-dimensional, and temporal physiologic data, while ensemble methods such as extreme gradient boosting remain widely used due to their robustness to missing data and class imbalance. Although many models outperform traditional clinical scores in retrospective analyses, performance frequently declines during external validation, and few systems have demonstrated clinical impact in prospective or workflow-integrated settings. Machine learning-based predictive analytics show promise for anticipating hemodynamic instability and guiding hemorrhage management before conventional vital-sign thresholds are crossed. However, clinical adoption remains constrained by data quality, generalizability, interpretability, and integration into time-critical workflows. Future progress will depend on incremental performance gains and physiology-informed model design, rigorous external validation, and careful positioning of machine learning tools as decision-support systems that augment - rather than replace - clinician judgment.

Current pretransplant clinical scoring systems fail to accurately estimate post-liver transplant (LT) survival, making recipient selection challenging. Persistent immune dysfunction contributes to early LT recipient mortality but is not captured by existing models. To identify plasma biomarkers of pretransplant immune dysfunction and to develop a biomarker-based pre-LT risk stratification tool to predict post-LT mortality. Prospective biomarker analysis of consecutively enrolled adult LT recipients was conducted. Healthy controls were included for baseline comparison. Patients undergoing LT at Houston Methodist Hospital (October 1, 2013, to December 31, 2017) and Rutgers/University Hospital (January 1, 2019, to March 31, 2021) were enrolled under an institutional review board-approved protocol, and data were censored on March 31, 2023, and analyzed. Patients with cirrhosis older than 18 years undergoing deceased donor LT were included. Exclusion criteria were age greater than 70 years, cancer other than hepatocellular carcinoma, retransplant, status 1A, intraoperative mortality, multivisceral transplant (except liver-kidney), and sample availability. Plasma cytokine, chemokine, and immune exhaustion biomarkers were quantified using multiplex Luminex assays at the time of transplant. Clinical, demographic, and laboratory data were extracted from institutional databases. The primary outcome was all-cause mortality within 1 year of LT. Secondary outcomes included graft survival, infections, rejection, readmissions, and 24-month survival. A total of 779 adult LTs were performed between 2007 and 2017, with prospective biomarker analysis of 279 consecutively enrolled LT recipients between 2018 and 2022. Median (IQR) participant age was 56.7 (48.2-62.5) years, and 110 of 279 patients (39.4%) were female. Pre-LT plasma levels of B-cell activating factor, C-C motif chemokine ligand 1, eotaxin, fractalkine, interleukin 1β (IL-1β), sIL-6Rβ, metalloproteinase (MMP) 2, and MMP3 were significantly associated with 1-year post-LT mortality. Multivariable Cox proportional hazards modeling identified fractalkine and MMP3 as independent predictors of early post-LT mortality. These were used to develop the Liver Immune Frailty Index (LIFI), stratifying patients into low, moderate, and high risk. One-year mortality was 1.9%, 10.3%, and 63.6% for LIFI-low, -moderate, and -high, respectively. Relative risk of death within 1 year was 5.43 (95% CI, 1.59-18.60; P < .001) for LIFI-moderate and 33.41 (95% CI, 11.48-97.25; P < .001) for LIFI-high compared to LIFI-low. LIFI demonstrated strong discrimination (C statistic, 0.83) and was associated with post-LT infections, longer hospital stays, and reduced patient survival. Per the results of this diagnostic/prognostic study, LIFI identifies LT candidates with severe immune dysfunction at high risk for early post-LT mortality, offering a preoperative, objective tool to refine transplant candidacy, guide perioperative management, and improve LT outcomes.

The hemoglobin-to-red cell distribution width ratio (HRR), which has demonstrated better predictive ability than the red blood cell distribution width (RDW) and hemoglobin (Hb) level, has not been used to predict orthopedic surgical outcomes and may be a novel prognostic parameter for mortality. In this single-center cohort study, data of 363 patients (aged ≥60 years) who underwent surgery for fragility hip fracture at our institution between January 2016 and December 2018 were retrospectively analyzed. Multivariable Cox proportional hazards and Kaplan-Meier survival curve analyses were performed to compare the high and low HRR and RDW groups, divided based on cutoff values. The power of mortality prediction over time was assessed by comparing Harrell's concordance index using the bootstrapping method. Among 363 patients, the overall mortality was 48.48% (176/363), with a mean±standard deviation of 4.31±2.09 (0.02-7.44) years. HRR was significantly associated with all-cause mortality after hip fracture surgery (hazard ratio: 0.989; 95% confidence interval: 0.978-0.999; p=0.044). Moreover, during the follow-up period after 1 year, HRR demonstrated the second-highest predictive ability for mortality among all laboratory parameters and indicators reflecting general condition, and it remained unaffected by anemia status for up to 4 years. HRR is proposed as a novel prognostic indicator for mid- to long-term survival after hip fracture surgery in older patients.

Sarcopenia is a prevalent issue among older patients with hip fracture and is a risk factor for poor clinical outcomes. Computed tomography (CT)-derived muscle density and areawere reported to predict the prognosis of fracture patients. This study aimed to assess the efficacy of these CT-based muscle parameters in predicting 1-year mortality in the oldest-old patients with hip fracture (≥ 80 years). A retrospective study was conducted on 324 hip fracture patients aged ≥ 80 years from 2018 to 2022. The cross-sectional area (CSA) and Hounsfield units (HU) of periarticular hip muscles were measured from CT images. The primary outcome was 1-year mortality. A multivariate logistic regression model was constructed, and its performance was assessed using ROC analysis, calibration curves, and Hosmer-Lemeshow testing. A nomogram was developed for model visualization and early clinical application. The 1-year mortality rate in this cohort was 13.0% (42/324). Survivors and non-survivors significantly differed in age, red blood cells (RBC), platelets, albumin, urea, and gluteal muscle parameters (all P < 0.05). Multivariate analysis identified four mortality predictors: older age (P = 0.048), lower albumin (P = 0.025), reduced gluteal density (P = 0.031), and smaller muscle area (P = 0.045). Gluteus maximus density and area independently predicted 1-year mortality (P < 0.05) in oldest-old hip fracture patients. Our predictive model incorporating age, muscle density, albumin, and muscle area showed a moderate predictive value (AUC = 0.741). This CT-based method offers a practical alternative to traditional sarcopenia assessments, facilitating early risk identification in this hip fracture population.

Sepsis secondary to pneumonia is a prominent cause of intensive care unit (ICU)admissions and mortality among older adults, yet early bedside risk stratification poses significant challenges. This study aimed to evaluate the predictive value of the Acute Physiology and Chronic Health Evaluation II (APACHE II)and the National Early Warning Score (NEWS), both individually and in combination, alongside admission serum lactate levels, for predicting mortality in geriatric ICU patients with pneumonia-related sepsis. In this single-center retrospective cohort study, we analyzed patients aged 65 years and older who were admitted between January 1, 2020, and July 1, 2025. Sepsis was defined according to Sepsis-3 criteria; APACHE II (using the worst values within the first 24 hours) and NEWS (measured at ICU admission) were recorded, along with the first lactate and other biomarkers obtained within the first 24 hours. We assessed mortality predictors using logistic regression and evaluated model discrimination through receiver operating characteristic (ROC)analysis. Among the 179 patients (median age 80), the ICU mortality rate was 64.8%. Non-survivors exhibited significantly higher APACHE II and NEWS scores, as well as elevated lactate and inflammatory markers (all p<0.001). In multivariable analysis, APACHE II (OR 1.130; p<0.001), NEWS (OR 1.239; p=0.003), and a history of stroke (OR 2.856; p=0.041) were identified as independent predictors of mortality, whereas lactate did not demonstrate independent predictive capability. Although lactate improved the discrimination of a baseline clinical-laboratory model (AUC increased from 0.67 to 0.75), it offered no incremental benefit when APACHE II and NEWS were included; the combined APACHE II+NEWS model achieved the highest AUC of 0.85. Exploratory cut-offs identified very high-risk subgroups (APACHE II >21 with NEWS >8 or lactate >2 mmol/L), with mortality rates approximating 86-87%. In conclusion, APACHE II and NEWS are robust early predictors of mortality in geriatric patients with pneumonia-related sepsis, while lactate may assist in early risk stratification but provides limited prognostic value beyond these scoring systems.

Background Copeptin, the C-terminal fragment of provasopressin, has emerged as a potential prognostic biomarker in sepsis. However, its predictive accuracy for mortality in adult patients with sepsis remains uncertain. We conducted a systematic review and meta-analysis to evaluate the diagnostic performance of elevated blood copeptin levels for mortality prediction in this population. Methods We systematically searched PubMed, Embase, Web of Science, Wanfang Data, and CNKI from inception to 22 May 2025, for observational studies assessing copeptin levels at admission or within 48 h in adults with sepsis. Pooled sensitivity, specificity, likelihood ratios, diagnostic odds ratio (DOR), and area under the summary receiver operating characteristic curve (AUC) were calculated using a random-effects model. Study quality was assessed using QUADAS-2. Results Ten prospective studies involving 1,637 patients were included. Pooled sensitivity and specificity of elevated copeptin for predicting mortality were 0.77 (95% CI: 0.70–0.83; I 2 = 52%) and 0.76 (95% CI: 0.67–0.83; I 2 = 86%), respectively. The pooled positive and negative likelihood ratios were 3.16 (95% CI: 2.33–4.29) and 0.30 (95% CI: 0.23–0.40), with a DOR of 10.40 (95% CI: 6.62–16.33). The summary AUC was 0.83 (95% CI: 0.79–0.86), indicating good overall prognostic accuracy. Subgroup analysis according to the cutoffs of copeptin did not significantly affect the results. No significant publication bias was detected ( p = 0.58). Conclusion Elevated blood copeptin levels within 48 h of sepsis diagnosis show good prognostic accuracy for short-term mortality in adult patients with sepsis. These findings support the potential clinical utility of copeptin as a risk stratification tool in sepsis management. Systematic review registration https://www.crd.york.ac.uk/prospero/ , identifier CRD42024587540.

Objectives: The incidence of stroke is higher among type 2 diabetes mellitus (T2DM) patients with a higher mortality rate. Prognostic scores for stroke patients can assist with treatment planning and counseling. The objective of this study was to create a machine-learning-based prognostic score to estimate in-hospital mortality in acute stroke with T2DM. Materials and Methods: This study used data from claims-based diabetes registry at Dr. Sardjito General Hospital, Yogyakarta, Indonesia, to identify patients diagnosed with acute stroke and T2DM between January 2016 and December 2020. Four machine learning algorithms were trained and evaluated based on standard performance metrics. Important features were selected from the best-performing model and implemented in a web-based in-hospital mortality prediction scoring system. Results: Of the 18,652 patients in the registry, the final analytic dataset comprised 749 patients (557 survivors and 192 non-survivors). The random forest showed superiority compared to other models. The six most important features were length of stay, sepsis, pneumonia, age, dyslipidemia, and hemiplegia. Using these features, the web-based system estimates the probability of in-hospital death for an individual patient. Conclusion: Machine learning analysis may support an in-hospital mortality prediction score for patients with acute stroke and T2DM patients by leveraging the key features identified by the random forest model.

Elderly patients with acute kidney injury (AKI) face a significantly increased mortality risk. Recent advances in machine learning technology have made it possible to predict the risk of death in patients at an early stage, which help to enable timely clinical intervention, optimize treatment strategies, and allocate hospital resources reasonably. We conducted a retrospective analysis of elderly patients admitted to the People's Liberation Army General Hospital (PLAGH) between 2008 and 2018. This study included data on demographic characteristics, comorbidities, and laboratory test results. We employed five machine learning algorithms, including L2-regularized logistic regression (L2-logistic), Least Absolute Shrinkage and Selection Operator (LASSO), eXtreme Gradient Boosting (XGBoost), Random Forest (RF), and Multi-layer Perceptron (MLP). To address the class imbalance issue , we employed oversampling techniques. Model performance was primarily evaluated using the area under the receiver operating characteristic curve (AUC), and SHapley Additive exPlanations (SHAP) values were introduced to enhance the interpretability of the prediction models. A total of 1290 AKI patients were enrolled in the study, with a 28-day mortality rate of 25.43%. Through data oversampling, the XGBoost model with random oversampling was identified as the optimal predictive model. The model achieved an AUC of 0.8659 in the validation cohort. Furthermore, external validation was performed using the eICU Collaborative Research Database (eICU-CRD), yielding an AUC of 0.6317. SHAP analysis revealed that Mechanical Ventilation, Peak serum creatinine within 7 days, Stage of AKI, urine protein and sreum albumin levels were the top five predictive factors for 28-day mortality in elderly patients with AKI. This comprehensive approach demonstrates how predictive healthcare analytics can enhance clinical decision-making and ultimately improve patient outcomes.

Abstract Introduction: Perforation peritonitis is a life-threatening emergency associated with high morbidity and mortality. Early risk stratification is vital, particularly in resource-limited settings. The Mannheim peritonitis index (MPI) is a widely used prognostic tool, but its performance in Vietnamese patients remains underexplored. The objective of this study was to evaluate the prognostic accuracy of MPI in predicting inhospital mortality among patients undergoing surgery for secondary peritonitis due to hollow viscus perforation. Methods: This prospective observational study included 176 patients treated at a tertiary care center in southern Vietnam from April 2023 to May 2024. Clinical data were collected to calculate MPI scores. Mortality outcomes were analyzed using the receiver-operating characteristic curve and multivariate logistic regression. Results: The mean patient age was 64.9 ± 15.7 years; 70.5% were male. Generalized peritonitis occurred in 96%, and feculent contamination in 5.1%. The most frequent perforation sites were the stomach (43.8%) and duodenum (21%). Overall mortality was 21%. MPI demonstrated excellent predictive performance (area under the curve = 0.944; 95% confidence interval: 0.905–0.988; P &lt; 0.001). An MPI score ≥25 yielded 91.9% sensitivity and 79.9% specificity. Feculent peritonitis (odds ratio [OR] =26.5; P = 0.020) and preoperative organ failure (OR = 51.4; P &lt; 0.001) were independent mortality predictors. Conclusion: MPI is a reliable and efficient tool for early mortality prediction in secondary peritonitis. A threshold score ≥25 effectively identifies high-risk patients and should be integrated into clinical decision-making in emergency surgical settings.

Artificial intelligence (AI) is transforming health care by enhancing diagnostics, improving patient outcomes, and reducing administrative burdens through advanced algorithms, with applications in medical imaging, virtual care, and automated data analysis. However, its role in palliative and hospice care remains underexplored. This review synthesizes research on AI applications in palliative and hospice care, examining its technological and clinical contributions to inform future research and guide clinical implementation. An integrative literature review, guided by Whittemore and Knafl's framework, analyzed qualitative, quantitative, and mixed-method studies. Registered with PROSPERO. A comprehensive search across 11 databases: Academic Search Complete, CINAHL, Cochrane Library, PubMed, Medline, Web of Science, Scopus, PsycINFO, ProQuest Dissertations & Theses Global, ACM Digital Library, and IEEE Xplore, identified English-language studies published from 2010 to 2024. Studies on AI applications in clinical settings, model validation, and key findings were included, with quality assessed using the Mixed Methods Appraisal Tool. Seventy studies (2018-2024) were included, primarily quantitative analyses of retrospective clinical and administrative data. AI applications supported mortality prediction, symptom monitoring, patient needs identification, communication facilitation, care planning, and resource allocation. Early tools included rule-based and structured-data models, while more recent approaches integrate unstructured clinical notes, wearable devices, and multimodal data for individualized prognostication and timely interventions. Key barriers included reliance on retrospective or single-center datasets, limited generalizability, ethical and equity concerns, and challenges in integrating AI into clinical workflows. AI holds potential in enhancing timely, patient-centered palliative and hospice care, supporting prognostication, symptom management, and decision-making. Successful integration requires attention to clinician trust, workflow alignment, equity, and ethical considerations. To maximize its impact on underutilization, future research should focus on multicenter validation, representative datasets, ethical deployment, and seamless integration into clinical practice.

The combined impact of the systemic immune-inflammation index (SII) and the triglyceride-glucose index (TyG) on short- and long-term outcomes in patients with coronary heart disease (CHD) remains unclear. This study aimed to investigate the association between the SII combined with the TyG index and mortality risk in CHD patients. This study included 970 CHD patients from the MIMIC-IV database. Participants were divided into four quartiles based on median SII and TyG values: T1 (low SII and low TyG), T2 (low SII and high TyG), T3 (high SII and low TyG), and T4 (high SII and high TyG). The combined impact of these indices on CHD patient prognosis was explored using Cox proportional hazards models, Kaplan-Meier survival curves, receiver operating characteristic (ROC) curves, and subgroup analyses. Compared with the T1 group, the T4 group showed a significantly increased risk of mortality. Multivariate Cox regression analysis revealed the highest 30-day (HR = 2.446, 95% CI 1.366-4.382, p = 0.003) and 1-year (HR = 1.586, 95% CI 1.054-2.387, p = 0.027) mortality risks in the T4 group. ROC curves further confirmed that the combined SII-TyG index demonstrated superior predictive capability for both short-term (AUC = 0.728, 95% CI 0.677-0.778) and long-term (AUC = 0.742, 95% CI 0.703-0.780) outcomes in CHD patients compared to the SII or TyG index alone. Subgroup analyses indicated that the predictive value of the combined index demonstrated considerable universality across different clinical strata. Furthermore, incorporating the combined index into baseline models moderately improved mortality prediction performance. Similarly, a cohort of 1,621 CHD patients from the eICU2.0 database yielded comparable findings: elevated SII and TyG indices correlated with increased in-hospital mortality risk (T4 HR = 3.378, 95% CI 1.716-6.649, p < 0.001), and ROC curve analysis confirmed that combined assessment demonstrated the highest predictive efficacy (AUC = 0.800, 95% CI 0.760-0.840). In summary, the combined assessment of SII and TyG indices assists clinicians in identifying high-risk populations and improving outcomes for patients with CHD.

Interpretable Machine Learning Model based on SHAP Explanations for Mortality Prediction in Patients Requiring Extra-corporeal Membrane Oxygenation: A Retrospective Study

Machine learning prediction of 1-year mortality in older patients with heart failure: a nationwide, multicenter, prospective cohort study

AimsArtificial intelligence (AI) has emerged as a promising tool for echocardiographic image analysis, potentially improving efficiency and reducing inter-observer variability. Real-world comparisons between AI-based analysis and human expert interpretation, and their correlation to clinical outcomes, remain limited. This study aimed to evaluate the correlation between AI-based echocardiographic analysis and human expert interpretation and to compare their association with one-year mortality in hospitalized patients.Methods and resultsWe conducted a retrospective analysis of 889 consecutive hospitalized patients who underwent a clinically indicated echocardiographic exam. All studies were read and analysed by both human echocardiographic experts and by commercially available AI software (Us2.ai). We performed correlation analysis of common echocardiographic variables obtained by human vs. AI and compared their performance in the prediction of 1-year mortality. Of the 889 patients, 731 (82%) patients (mean age 68 ± 16, 46% Females) had sufficient echocardiographic data to be included in the analysis. Most parameters exhibited a strong correlation between human and AI-derived measurements. AI-derived LVEF values were significantly higher than human estimates (mean difference 5.8%, P < 0.001). In a multivariable model, AI- and human-based mortality prediction were comparable (AUC 0.67 vs. 0.66, P = 0.86). When including AI-obtained automated left ventricular strain analysis, the AI-based model was superior to humans in predicting 1-year mortality (AUC 0.73 vs. 0.66, P = 0.048).ConclusionAI-based echocardiographic analysis shows excellent correlation with human-derived measurements. Incorporating automated strain analysis resulted stronger association with mortality of the AI analysis compared to standard human analysis.

Explainable AI for critical care: a systematic review of interpretable models for sepsis and ICU mortality prediction.

Development of advanced lung cancer inflammation index-based machine learning models for predicting stroke and mortality: A comparative and interpretable study.

Effect of artificial intelligence in extracorporeal membrane oxygenation: a systematic review and meta-analysis.

Variability in the urinary albumin-to-creatinine ratio predicts mortality in patients with type 2 diabetes.