An observational cohort study was conducted to compare the performance of the RIFLE (risk, injury, failure, loss and end-stage kidney disease), AKIN (acute kidney injury network) and conventional graded criteria to identify acute kidney injury (AKI) following SCT and to predict long-term mortality in 141 myeloablative allogeneic SCT (m-allo), 60 non-myeloablative allogeneic SCT (nm-allo) and 48 autologous SCT (auto) cases. The AKIN criteria had less ability to identify patients as having the lowest category, stage 1 (analogous to RIFLE risk): 33% (37%) in m-allo, 23% (32%) in nm-allo and 8.3% (16.7%) in auto. Cox regression showed that categories higher than the intermediate stage were independent predictors of mortality in all three definitions. The areas under receiver operating characteristic curves showed that both definition systems had similar and significant ability to predict mortality (0.643-0.649 in m-allo and 0.734-0.766 in nm-allo, respectively). These abilities of the conventional graded criteria were comparable with those of the RIFLE criteria. The RIFLE criteria have greater sensitivity than the AKIN criteria to identify patients with AKI and therefore are more favorable as a uniform definition system for post-SCT AKI. However, the RIFLE criteria do not improve on the clinical relevance of the conventional graded criteria.