Machine Learning Models for the Prediction of Renal Failure in Chronic Kidney Disease: A Retrospective Cohort Study

Abstract

This study assessed the feasibility of five separate machine learning (ML) classifiers for predicting disease progression in patients with pre-dialysis chronic kidney disease (CKD). The study enrolled 858 patients with CKD treated at a veteran's hospital in Taiwan. After classification into early and advanced stages, patient demographics and laboratory data were processed and used to predict progression to renal failure and important features for optimal prediction were identified. The random forest (RF) classifier with synthetic minority over-sampling technique (SMOTE) had the best predictive performances among patients with early-stage CKD who progressed within 3 and 5 years and among patients with advanced-stage CKD who progressed within 1 and 3 years. Important features identified for predicting progression from early- and advanced-stage CKD were urine creatinine and serum creatinine levels, respectively. The RF classifier demonstrated the optimal performance, with an area under the receiver operating characteristic curve values of 0.96 for predicting progression within 5 years in patients with early-stage CKD and 0.97 for predicting progression within 1 year in patients with advanced-stage CKD. The proposed method resulted in the optimal prediction of CKD progression, especially within 1 year of advanced-stage CKD. These results will be useful for predicting prognosis among patients with CKD.

Keywords: kidney disease; random forest; renal dialysis.

Figure 1

Flow chart for the selection of study subjects.

Figure 2

Flow chart of model training and performance evaluation.

Figure 3

Plots showing the area under the receiver operating characteristic curve (AUROC) for the ability of the random forest (RF) classifier trained using synthetic minority over-sampling technique (SMOTE) to predict the progression of early-stage chronic kidney disease (CKD) to end-stage renal disease (ESRD) within (a) 3 and (b) 5 years. Plots showing the AUROC for the ability of the RF classifier trained using to predict the progression of advanced-stage CKD to ESRD within (c) 1 and (d) 3 years.

Figure 4

Positive and negative impacts of 13 features on the prediction of progression from early-stage chronic kidney disease (CKD) to end-stage renal disease (ESRD) within (a) 3 and (b) 5 years. Positive and negative impacts of 24 features on the prediction of progression from advanced-stage CKD to ESRD within (c) 1 and (d) 3 years. Features are ranked in descending according to Shapley additive explanations (SHAP values), where the top feature represents the most informative feature. Each dot in the plot represents the value for an individual patient. The color represents the scale of the feature’s value, ranging from high (red) to low (blue), for the observation.

Figure 5

Plots showing the under the receiver operatic characteristic curve (AUROC) of five models, including six features from (a) original and (b) synthetic minority over-sampling technique (SMOTE) data for predicting end-stage renal disease progression within 5 years in patients with early-stage chronic kidney disease. Plots showing the AUROC of five models, including 10 features from (c) original and (d) SMOTE data for predicting end-stage renal disease progression within 1 year in patients with advanced-stage chronic kidney disease. GNB, Gaussian naïve Bayes; LR, linear regression; RF, random forest; SVM, support vector machine; XGBoost, extreme gradient boosting.