方山 真朱

BACKGROUND: Clinically important gastrointestinal bleeding (CIGIB) is a serious complication in critically ill patients, contributing to prolonged ICU stays and increased mortality. Despite efforts to identify high-risk patients, no previous studies have employed machine learning models to predict CIGIB during ICU stay or identify key predictors in this context. METHODS: This single-center retrospective study included ICU patients aged 18 years or older admitted between 2017 and 2024. Patients with ICU stays of less than 24 hours or GIB within 24 hours of admission were excluded. Machine learning models, including XGBoost, Random Forest, and L1-regularized logistic regression, were trained using patient data from the first 24 hours of ICU admission. Model performance was assessed using AUROC, precision, recall, and F1 scores. Shapley Additive Explanations (SHAP) were employed to evaluate key predictors. RESULTS: A total of 7,357 ICU patients were included, of whom 171 (2.3%) experienced CIGIB. The XGBoost model demonstrated the highest predictive performance with an AUROC of 0.84. Key predictors included APACHE III scores, hematocrit levels, APTT, creatinine and respiratory rate, while invasive mechanical ventilation and stress ulcer prophylaxis within the first 24 hours of ICU admission did not rank among the top 20 predictors based on SHAP values. CONCLUSIONS: This study represents the first application of machine learning for predicting CIGIB in ICU patients, providing valuable insights into risk stratification. The model demonstrated high predictive accuracy and interpretability, highlighting its potential to guide early intervention and prophylaxis. Further multi-center studies and interventional trials are needed to validate these findings and refine clinical risk prediction strategies.

BACKGROUND: Reduced or delayed access to medical resources on weekends could lead to worsening outcomes, in critically ill infected patients requiring intensive care unit (ICU) admission. OBJECTIVE: To investigate the "weekend effect," on critically ill infected patients in Japanese ICUs for the first time. DESIGN: Multicenter retrospective cohort study. METHODS: We examined data from Japanese ICU patients participating in the DIANA study, a multicenter international observational cohort study. This prospective investigation enrolled critically ill patients with infections admitted to the ICU. The primary endpoint was successful discharge from the ICU within 28 days of admission. Outcome measures were evaluated through both univariate and covariate Cox regression analyses, providing hazard ratios (HRs) along with estimated 95% confidence intervals (CIs). RESULTS: Out of the 276 patients enrolled in the DIANA study across 31 facilities, 208 patients (75.4%) meeting the inclusion criteria were included in the analysis. The weekday ICU admission group comprised 156 patients (75.0%), while the weekend ICU admission group comprised 52 patients (25.0%). In the multivariate Cox regression analysis, there were no statistically significant differences observed in the rates of ICU discharge alive within 28 days and 14 days (28 days, HR: 0.94, 95% CI: 0.63-1.40; 14 days, HR: 0.97, 95% CI: 0.64-1.48). Furthermore, the overall ICU mortality rates at 28 days and 14 days after ICU admission did not show statistical significance between patients admitted on weekends and those admitted on weekdays (ICU mortality, 28 days: 13.5% vs 11.5%, p = 0.806; 14 days: 7.7% vs 10.9%, p = 0.604). CONCLUSION: The rates of ICU discharge alive within 28 days after ICU admission did not differ significantly between weekday and weekend admissions, both in the unadjusted and adjusted analyses. Moreover, further well-designed studies are warranted to thoroughly assess this effect.