Background/Aims: To review the currently available severity scores to predict outcome of acute kidney injury (AKI) patients, to discuss the problems with such scores, and to provide information for the development of more accurate AKI severity scores in the future. Methods: Literature review and multivariate analysis using a large international database for AKI. Results: Although general severity scores have good discrimination and calibration abilities to predict outcome of critically ill patients, the accuracy of these systems for AKI patients has been questioned. To improve prediction ability, multiple AKI severity scores have been published in the literature. However, most of these scores were developed and tested in a single center, or if multicentric, they were confined to a single country. Seven variables (mechanical ventilation, bilirubin, age, oliguria, hypotension, sepsis and platelet count) are often found as common risk factors in these severity scores and should be included in future AKI severity scores. Although several studies have consistently reported that both low creatinine and high urea at the start of RRT are related to worse outcome in AKI patients, they might not improve prediction ability. Conclusion: Using available information and a large database collected internationally, a more accurate score for AKI is likely to be developed.