Objective: Multidrug-resistant organisms (MDROs) complicate hospital-acquired and ventilator-associated pneumonia (HAP/VAP). We aimed to develop and validate a machine learning (ML) model to predict MDR risk in pneumonia patients and assess its utility for clinical decision support. Patients and Methods: We developed multiple ML models using data from the MIMIC-IV database (n=802). Feature selection was performed using LASSO regression and chi-square tests. Four models—Logistic Regression, Random Forest, XGBoost, and LightGBM—were trained, tuned, and calibrated. Model performance was evaluated using ROC curves and calibration plots. The final model was selected based on discrimination, calibration, and interpretability (assessed via SHAP). External validation on an independent cohort from a Chinese tertiary hospital (n=213) demonstrated the reproducibility and generalizability of its performance. Results: Among the evaluated ML models, Logistic Regression demonstrated the best overall performance. On the MIMIC-IV internal test set, it achieved an area under the curve (AUC) of 0.798 (95% CI: 0.718– 0.872) with an accuracy of 0.807. External validation on an independent Chinese cohort confirmed the model’s robust generalizability, achieving an AUC of 0.845. SHAP analysis identified key predictive features consistently across both cohorts, including the systemic immune-inflammation index (SII), albumin level, C-reactive protein-to-albumin ratio (CAR), number of antibiotic classes, white blood cell count (WBC), and absolute lymphocyte count (LYM abs). All of these features were significantly associated with MDR risk. Conclusion: Multiple ML models effectively predicted MDR infections in pneumonia patients, with Logistic Regression exhibiting particularly strong overall performance. Model reliability was enhanced through feature selection and probability calibration, while interpretability was improved by SHAP analysis. External validation confirmed the generalizability of our approach, supporting its potential application in clinical infection control. Future studies should focus on validation and the integration of more diverse clinical data sources.

Oxidative stress is closely associated with the estimated glomerular filtration rate (eGFR) levels. A comprehensive indicator for evaluating dietary and lifestyle exposures to oxidative stress is the Oxidative Balance Score (OBS), with higher OBS suggesting more substantial antioxidant exposures. The aim of this study is to explore the relationship between OBS and eGFR levels. A total of 20,285 subjects were selected from the National Health and Nutrition Examination Survey (NHANES) from 2007 to 2018. One-way and multifactorial linear regression analyses were used to assess the association of OBS with eGFR. The relationship between OBS and eGFR levels was also assessed using restricted cubic splines (RCS) analysis. OBS levels were significantly and positively correlated with eGFR levels, and eGFR values were more susceptible to OBS in males than in females. Our findings suggest that higher OBS scores may be associated with higher eGFR levels, highlighting the importance of adhering to an antioxidant diet and lifestyle, which can help prevent kidney disease.

Objective: This study aimed to develop and validate an interpretable machine learning model for early prediction of carbapenem-resistant Klebsiella pneumoniae (CRKP) infection using routine peripheral blood indicators, providing a rapid risk assessment tool before antimicrobial susceptibility results. Methods: A retrospective study was conducted on 285 patients with culture-confirmed Klebsiella pneumoniae infection (242 carbapenem-susceptible Klebsiella pneumoniae (CSKP), 43 CRKP), with all isolates obtained from sputum specimens, from June 2024 to June 2025. Given the significant class imbalance between the CSKP (n = 242) and CRKP (n = 43) groups, key features were screened by LASSO regression and Boruta algorithm. Six machine learning models were constructed, with synthetic minority oversampling technique (SMOTE) for class imbalance processing and stratified 5-fold cross-validation for optimization. Model performance was evaluated by area under the curve (AUC), accuracy, calibration curve, decision curve analysis (DCA), and interpreted by SHapley Additive exPlanations (SHAP). Results: Seven core predictors (HGB, LYMPH_p, NEUT_p, MONO, BASO, PCT, AISI) were identified. The random forest model achieved the best performance, with an AUC of 0.863 (95% CI: 0.672– 0.980), accuracy of 0.821, sensitivity of 0.750 (95% CI: 0.571– 1.000), specificity of 0.834 (95% CI: 0.643– 0.841), and Brier score of 0.077. DCA showed clear net benefit in the threshold probability range of 0.1– 0.8. SHAP analysis confirmed aggregate index of systemic inflammation (AISI) as the most important predictive factor. Conclusion: The interpretable prediction model based on routine peripheral blood indicators demonstrated good discrimination, calibration, and clinical utility in this single-center retrospective study. It has the potential to assist early risk stratification of CRKP infection and support clinical decision-making. External multicenter prospective validation is warranted before clinical implementation.

This study aimed to investigate the association between the blood urea nitrogen-to-albumin ratio (BAR) and both short- and long-term all-cause mortality in