Objective  To compare the accuracy of three acute kidney injury (AKI) criteria of RIFLE, Acute Kidney Injury Network (AKIN) and Kidney Disease: Improving Global Outcomes (KDIGO) in predicting the early (30 d) postoperative death of liver transplant recipients.

  Methods  Clinical data of 173 liver transplant recipients were retrospectively analyzed. The incidence of postoperative AKI was calculated according to the three criteria of RIFLE, AKIN and KDIGO. The all-cause fatality rate and cause of death at postoperative 30 d were analyzed. The risk factors of death within 30 d after operation were analyzed by binary Logistic regression. The prediction accuracy of three criteria for death within 30 d after operation was compared by the receiver operating characteristic (ROC) curve.

  Results  According to the RIFLE, AKIN and KDIGO criteria, the incidences of postoperative AKI were 48.0%, 51.4% and 53.8%, respectively. Thirteen patients died within 30 d after operation and the fatality rate was 7.5%. RIFLE, AKIN and KDIGO stages were the independent risk factors for death within 30 d after operation (all P < 0.05). The area under the curve (AUC) of RIFLE, AKIN and KDIGO stages to predict death within 30 d after liver transplantation were 0.828, 0.766 and 0.844, respectively. There was a statistically significant difference between AKIN and KDIGO (P < 0.05).

  Conclusions  KDIGO criterion is better for predicting early death after liver transplantation. However, as a tool, the comparative selection among these three criteria still needs the evidence support from a large multicenter sample.