Li DH, Wald R, Blum D, et al. - Researchers performed this study with a development cohort and a validation cohort to construct prediction models for death at 90-day and 1-year after renal replacement therapy (RRT) start in critically ill patients with acute kidney injury (AKI). For the development of the models, all patients who started RRT in the ICU for AKI at a tertiary care hospital between 2007 and 2014 were used, and for external validation, data from the multicentre OPTIMAL-AKI study were used. To predict 90-day mortality, 11 variables were incorporated in the model (AUC:0.79, 95%CI:0.76–0.82). Reduced performance of the 90-day mortality model was seen upon validation in the OPTIMAL-AKI cohort (AUC:0.61, 95%CI:0.54–0.69) and modest calibration was demonstrated. Experts got similar findings for the mortality model at 1-year. Findings revealed the limited ability of routinely obtained variables at the time of RRT commencement for predicting mortality among critically ill patients with AKI who start RRT.