ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

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

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: SA-PO332

Machine Learning Approach for Hemodialysis Prescription: Model Development and Validation Study

Session Information

Category: Dialysis

  • 701 Dialysis: Hemodialysis and Frequent Dialysis

Authors

  • Bian, Xueqin, Second Affiliated Hospital of Nanjing Medical University, Nanjing, China
  • Zhou, Yang, Second Affiliated Hospital of Nanjing Medical University, Nanjing, China
  • Ye, Hong, Second Affiliated Hospital of Nanjing Medical University, Nanjing, China
  • Yang, Junwei, Second Affiliated Hospital of Nanjing Medical University, Nanjing, China
Background

The prediction model of hemodialysis prescription is established based on machine learning to achieve precise hemodialysis treatment.

Methods

We obtained 108,638 hemodialysis sessions in 965 independent maintenance hemodialysis(MHD)patients in our hospital from October 1, 2020 to June 31, 2021 using random sampling. The sessions were randomly divided into training (70%), calibration (10%), and testing (20%) sets. Apply XGBoost to analyze and extract effective feature data. XGBoost, Random Forest Regression, K-Nearest Neighbor, Support Vector Regression and Linear Regression were used to develop the prescription model and model fusion. Training and update the model using reinforcement learning. The area under the receiver operating characteristic curves, the area under the precision-recall curves, F1 scores and MSE obtained to assess model stability and accuracy.

Results

There were 108,638 dialysis records in 965 patients, of whom 62.2 were male, the average age was 59.3±13.2 years, the median dialysis age was 102.3 months, BMI was 23.8±3.9, and the primary disease was diabetic nephropathy (38.6%). There are 13 labels in the hemodialysis prescription, of which 6 labels are continuous variables and the regression model is used, and 7 labels are discontinuous variables. The average accuracy is greater than 0.88, and the mean square error is less than 0.067. The average compliance rate of dialysis prescriptions was 86%.

Conclusion

The artificial intelligence dialysis prescription established based on machine learning has better stability and accuracy.

Funding

  • Government Support – Non-U.S.