Prognostic Value of Scoring Systems in Elderly Sepsis-Associated AKI Patients: A Multicenter Retrospective Cohort Study

Background Sepsis-associated acute kidney injury (AKI) in elderly patients carries high mortality risks, yet the comparative performance of severity scores in predicting outcomes remains unclear. This study evaluated the prognostic accuracy of SAPS II, LODS, SAPS III, MELD and Charlson Comorbidity Index for 28- and 90-day mortality in this vulnerable population. Methods A retrospective multicenter cohort study analyzed 2455 elderly (≥65 years) sepsis-AKI patients from MIMIC-IV and eICU databases. Participants were stratified into survival (n = 1693) and non-survival (n = 762) groups. Receiver operating characteristic (ROC) curves and decision curve analysis(DCA) assessed discriminative power of each predictor for mortality endpoints. Results For 28-day mortality, SAPS II demonstrated superior predictive performance (AUC 0.803, 95% CI 0.782-0.824), followed by LODS (AUC 0.772) and SAPS III (AUC 0.759), all p < 0.001. MELD (AUC 0.626), and Charlson (AUC 0.585) showed modest discrimination. Similarly, SAPS II maintained the highest AUC (0.789) for 90-day mortality, with SAPS III (0.757) ranking second. Charlson again showed limited predictive value (AUCs 0.580-0.575). Decision curve analysis demonstrated the superior predictive utility of SAPS II for both 28-day and 90-day mortality compared to other clinical scoring systems. Observationally, all scores showed higher specificity values (>0.65) than sensitivity values in our dataset. Furthermore, while urinary tract infection was associated with increased mortality in elderly AKI patients, our subsequent analysis demonstrated its limited predictive capacity for 28-day and 90-day mortality outcomes. Conclusion In our cohort of elderly sepsis-AKI patients, SAPS II demonstrated the strongest observed prognostic performance for both 28- and 90-day mortality compared to MELD, LODS, and Charlson scores. These findings suggest SAPS II may be valuable for risk stratification in clinical decision-making for this high-risk population, though further validation is warranted.