Korean J Transplant.  2023 Nov;37(Suppl 1):S61. 10.4285/ATW2023.F-6380.

Prediction of postdonation renal function using machine learning techniques and conventional regression models in living kidney donors

• Jeon J
• Song Y
• Yu JY
• Jung W
• Lee K
• Lee JE
• Huh W
• Cha WC
• Jang HR

• KMID: 2551648
• DOI: http://doi.org/10.4285/ATW2023.F-6380

Abstract

Background
The accurate prediction of renal function following a kidney donation and a careful selection of living donors are essential for living-kidney donation programs. We aimed to develop a prediction model for postdonation renal function follow-ing a living kidney donation using machine learning.
Methods
This retrospective cohort study using electronic medical records was conducted with 823 living kidney donors be-tween 2009 and 2020. The entire dataset was randomly split into a training set (80%) and a test set (20%). The main outcome was the accurate prediction of postdonation estimated glomerular filtration rate (eGFR) 12 months after nephrectomy. We com-pared the performance of various machine learning techniques as well as traditional regression models. The best-performing model was selected based on the mean absolute error (MAE) and root mean square error (RMSE).
Results
The mean age of the participants was 45.2±12.3 years, and 48.4% were males. The mean predonation and postdonation eGFRs were 101.3 and 68.8±12.7 mL/min/1.73 m ² , respectively. The XGBoost model with feature importance, including the eGFR, age, serum creatinine, 24-hour urine creatinine, 24-hour urine sodium, creatinine clearance, cystatin C, cystatin C-based eGFR, computed tomography volume of the remaining kidney/body weight, normalized GFR of the remaining kidney measured through a acid scan, and sex, showed the best performance with an MAE of 6.23 and RMSE of 8.06. The proportion of predicted eGFR values within 5% or 10% of the actual eGFR value were 0.39 and 0.58, respectively. We developed a web application titled Kidney Donation with Nephrologic Intelligence (KDNI) for ease of use in clinical practice.
Conclusions
The machine learning technique using XGBoost accurately predicted the postdonation eGFR after living kidney donation. This model can be easily applied in clinical practice using KDNI, the developed web application.