Age-Specific Prognostic Models for Sepsis-Associated Acute Kidney Injury: A Multicenter Cohort Study

BACKGROUND Sepsis-associated acute kidney injury (SA-AKI) exhibits distinct clinical outcomes across age groups, yet current prognostic methods seldom consider age-related pathophysiologic differences. This multicenter study explored age-specific prognostic models for patients with SA-AKI using real-world critical care data. MATERIAL AND METHODS We analyzed 3662 patients with SA-AKI from the MIMIC-IV and eICU databases, stratified into 3 age cohorts: under 65, 65-80, and over 80. For each cohort, we constructed clinical prediction models. Model performance was evaluated using receiver operating characteristic curve analysis, along with sensitivity and specificity at optimal thresholds. RESULTS Age-specific clinical models demonstrated superior predictive performance compared with conventional severity scores. For patients younger than 65 years, the optimal model - incorporating urinary infection, catheter-related infection, lactate, and norepinephrine use - achieved an area under the curve (AUC) of 0.753 (95% confidence intervals [CI], 0.721-0.785) with 67.0% sensitivity and 73.1% specificity. In the 65-80-year cohort, the optimal model - incorporating urinary infection, blood urea nitrogen, lactate, and vasopressor use - achieved an AUC of 0.769 (95% CI, 0.743-0.796) with 78.2% sensitivity. For patients older than 80 years, the optimal model - incorporating urinary infection, catheter-related infection, lactate, vasopressor use, and intensive care unit length of stay - achieved an AUC of 0.770 (95% CI, 0.737-0.803) with 79.7% sensitivity. Survival curves confirmed significant mortality risk stratification across all age groups. CONCLUSIONS Age-specific prognostic models incorporating clinically modifiable factors substantially improved mortality prediction in SA-AKI compared with conventional severity scores. These models facilitate personalized risk assessment and may guide age-tailored treatments for this high-risk population.