Endothelial activation and stress index associated with in-hospital mortality risk in patients with end-stage renal disease: a retrospective analysis based on the MIMIC database and machine learning model development.

BACKGROUND: End-stage renal disease (ESRD) is a severe chronic renal disorder with high mortality, requiring dialysis treatment or kidney transplantation. The endothelial activation and stress index (EASIX), reflecting inflammatory status and endothelial dysfunction, has demonstrated predictive value for outcomes in multiple diseases. However, its association with in-hospital mortality (IHM) risk in ESRD patients remains unclear, and machine learning prediction models remain unestablished. METHODS: Clinical data for ESRD patients were extracted from the MIMIC-IV (3.1) database. The outcome was IHM. The link between EASIX and IHM risk was explored using Cox proportional hazards regression, Kaplan-Meier survival curves, restricted cubic splines (RCS), and subgroup analysis. The Boruta algorithm and LASSO regression were employed to screen important features. Multiple models were established using machine learning algorithms, validated, and compared. SHAP analysis was applied to the optimal model. RESULTS: The study included 997 ESRD patients, with 194 in-hospital deaths, accounting for 19.5% of the total cohort. The Kaplan-Meier curve revealed the highest IHM rate among patients with the highest EASIX level (Q4). Elevated EASIX levels showed a significant positive link with an enhanced IHM risk in ESRD patients (HR [95% CI] = 2.086 [1.697, 2.564]). RCS showed an approximate linear relationship, with this association consistent across multiple subgroups. The gradient boosting machine was identified as the optimal model (AUC of training set = 0.822, AUC of validation set = 0.763). SHAP analysis identified EASIX as a key contributor to IHM risk. CONCLUSION: EASIX is significantly positively correlated with an elevated IHM risk in ESRD patients, making it a crucial predictor of IHM in ESRD. EASIX can serve as an early predictive tool to guide the identification, intervention, and prevention of adverse outcomes in ESRD. CLINICAL TRIAL NUMBER: Not applicable.