Validation of renal-risk models for the prediction of non-renal replacement therapy cardiac surgery-associated acute kidney injury

Background Cardiac surgery-associated acute kidney injury (AKI) is a serious complication of cardiac surgery, even when renal replacement therapy (RRT) is not required. The existing risk models for cardiac surgery associated AKI are designed to predict AKI requiring RRT (RRT-AKI). The aim of this study is to validate three risk models for the prediction of RRT-dependent and non-RRT AKI after cardiac surgery. Methods Retrospective analysis on 7675 consecutive adult patients undergoing cardiac surgery with cardiopulmonary bypass. AKI was defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria for stage 1 and 2. RRT AKI and non-RRT AKI were defined according to the need for RRT. Three risk models were validated separately for RRT and non-RRT AKI: the Cleveland Risk Score, the Bedside Risk Score, and the Simplified Renal Index Scoring Scheme. Discrimination power was assessed with Receiver Operating Characteristics analysis and c-statistics. Results There were 502 (6.5%) non-RRT AKI events, 128 (1.7%) RRT-AKI events, and 7045 (91.8%) no-events. The three models performed well for predicting RRT-AKI (c-statistics 0.75–0.79) and poorly for predicting non-RRT AKI (c-statistics 0.54–0.59). The models had an excellent calibration for RRT-AKI but not for non-RRT AKI. Preoperative serum creatinine and estimated glomerular filtration rate were associated with RRT AKI but not with non-RRT AKI. Mortality was 12.2% in non-RRT AKI and 46.9% in RRT-AKI, significantly (P = 0.001) higher than in patients without AKI (1.3%). Conclusions The existing risk models are inadequate for predicting non-RRT AKI following cardiac surgery, both in terms of discrimination and calibration.