ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005


The Latest on Twitter

Kidney Week

Abstract: TH-PO849

Prediction of Peritoneal Membrane Function in Pre-Dialysis ESRD Patients

Session Information

  • Peritoneal Dialysis - I
    November 02, 2017 | Location: Hall H, Morial Convention Center
    Abstract Time: 10:00 AM - 10:00 AM

Category: Dialysis

  • 608 Peritoneal Dialysis


  • Xu, Hui, Xiangya Hospital, Central South University, Changsha, Hunan, China, Changsha, Hunan, China

Background: Peritoneal membrane function decides the efficiency of peritoneal dialysis (PD). However, an effective method is still lack to value the peritoneal membrane function discrimination for pre-dialysis patient is. As a valid classification and prediction tool, random forest using clmay serve efficiently for predicting pre-dialysis peritoneal membrane function in ESRD patients.


Firstly, random forest method was used to build discrimination model for predicting peritoneal membrane function. The clinical data of 247 patients was used as the training set and the other 50 was used as the validation set. Secondly, random forest-based algorithm was applied in training set for model development and validation set.


The discrimination model performed well for the primary objective. 10-fold cross validation was considered to be internal validation, the evaluation for this model showed that its accuracy rate, sensitivity and specificity respectively reached 0.862, 0.877, and 0.795. The coefficient of martensite (MCC) was turned out to be 0.60 and AUC (area under the receiver operating characteristics curve) was 0.840 (Fig.1a). For external validation, test was conducted on validation set. And the results showed that the accuracy rate, sensitivity, and specificity were respectively 0.78, 0.765, and 0.812. MCC was 0.546, AUC was 0.731(Figure.1b). Of the 28 variables considered, 7 were selected by the model for peritoneal membrane function prediction, among which thrombin time and urine volume reveal evident significance(Figure1c).


Conclusion: Random forest based model provides a robust tool to predict the peritoneal membrane function in non-dialysis ESRD patients.

Figure 1


  • Government Support - Non-U.S.