Abstract
Background: To determine whether early dynamic changes in the systemic immune-inflammation index (SII) improve prediction of acute kidney injury (AKI) and 1-year mortality in critically ill patients. Methods: In this retrospective cohort study of 17,491 ICU admissions from the MIMIC-IV database, we calculated three SII metrics within the first 24 h of ICU stay: the 24-h SII_slope and the extreme values (SII_min, SII_max). LASSO-selected multivariable logistic regression was used to predict AKI, and Cox proportional hazards models assessed associations with 1-year mortality. A prognostic nomogram integrating SOFA score, APS III score, and log-transformed SII_min and SII_max was developed using the rms package in R. Model performance was evaluated by AUC of ROC curves, calibration plots, decision curve analysis (DCA), and Kaplan-Meier survival curves stratified by SII quartiles. Results: The LASSO-based logistic model identified a steeper 24-h SII_slope as an independent predictor of AKI (AUC 0.739; patients who developed AKI had significantly higher predicted risk than those who did not). Higher SII_min and SII_max were each associated with reduced 1-year survival (log-rank p=0.047 for SII_min quartiles). The nomogram for 1-year mortality demonstrated excellent discrimination (AUC 0.823) and good calibration, and DCA confirmed its clinical utility. Conclusions: Early dynamic changes in SII-especially the 24-h slope-and the first-day SII extremes independently predict AKI and long-term mortality in ICU patients. A nomogram combining SII metrics with standard severity scores may facilitate individualized risk stratification in critical care.