Objective To analyze the prediction efficiency of scoring models at home and abroad on delayed graft function (DGF) after renal transplantation in China.
Methods The clinical data of 112 donors and 220 recipients undergoing renal transplantation were prospectively analyzed. The DGF predicted by KDRI model, Jeldres model, and model of our center was compared with actual DGF incidence of renal transplant recipients. The prediction efficiency of each model was analyzed. The predictive accuracy was compared by the area under curve (AUC) of receiver operating characteristic (ROC) curve.
Results The DGF incidence of 220 renal transplant recipients was 14.1% (31/220). DGF prediction using KDRI model showed that 41 cases were high risk donors, the AUC was 0.57, the sensitivity was 0.37, the specificity was 0.66, and the positive predictive value was 22%. DGF prediction using Jedres model showed that 22 cases were high risk recipients, the AUC was 0.56, the sensitivity was 0.13, the specificity was 0.92 and the positive predictive value was 20%. DGF prediction using the model of our center showed that 25 cases were high risk donors, the AUC was 0.80, the sensitivity was 0.53, the specificity was 0.84, the positive predictive value was 40%.
Conclusions Compared with the KDRI and Jedres models, the prediction model of our center has higher AUC and sensitivity with a better prediction efficiency on DGF. Therefore, it is a suitable evaluation system of donors from donation after citizen's death in Chinese.
DownLoad: