Machine Learning Improves Upon Clinicians' Prediction of End Stage Kidney Disease
Aaron Chuah1, Giles Walters2, Daniel Christiadi2
1Department of Immunology and Infectious Disease, John Curtin School of Medical Research, Australian National University (ANU), Canberra, ACT, Australia.
Insights
Machine learning accurately predicts End-Stage Kidney Disease (ESKD) risk, outperforming human experts and existing equations. This advance in chronic kidney disease management offers improved clinical decision support.
Area of Science:
- Nephrology
- Artificial Intelligence
- Predictive Analytics
Background:
- Chronic kidney disease (CKD) progression to End-Stage Kidney Disease (ESKD) significantly increases mortality and morbidity.
- Predicting CKD progression is challenging due to its variable nature.
Purpose of the Study:
- To develop and evaluate a machine learning (ML) model for predicting ESKD risk.
- To compare the ML model's performance against nephrologists and the Kidney Failure Risk Equation (KFRE).
Main Methods:
- An observational, retrospective study of 12,371 patients from The Canberra Hospital (1996-2018).
- Demographic, clinical, and laboratory data were extracted; time series data were featurized.
- An XGBoost ML model was trained and tested, with performance compared to nephrologists and KFRE models.
Main Results:
- The ML model achieved 93.9% accuracy, 60% sensitivity, 97.7% specificity, and 75% positive predictive value in predicting ESKD within 2 years.
- The ML model significantly outperformed six nephrologists and both 4- and 8-variable KFRE models across all metrics.
- Estimated Glomerular Filtration Rate (eGFR) and glucose levels were key predictors in the ML model.
Conclusions:
- The developed ML model demonstrates superior performance in predicting ESKD compared to current methods.
- The computational predictions show potential for integration into clinical workflows for enhanced decision support in CKD management.
Background And Objectives:
Chronic kidney disease progression to ESKD is associated with a marked increase in mortality and morbidity. Its progression is highly variable and difficult to predict.
Methods:
This is an observational, retrospective, single-centre study. The cohort was patients attending hospital and nephrology clinic at The Canberra Hospital from September 1996 to March 2018. Demographic data, vital signs, kidney function test, proteinuria, and serum glucose were extracted. The model was trained on the featurised time series data with XGBoost. Its performance was compared against six nephrologists and the Kidney Failure Risk Equation (KFRE).
Results:
A total of 12,371 patients were included, with 2,388 were found to have an adequate density (three eGFR data points in the first 2 years) for subsequent analysis. Patients were divided into 80%/20% ratio for training and testing datasets.ML model had superior performance than nephrologist in predicting ESKD within 2 years with 93.9% accuracy, 60% sensitivity, 97.7% specificity, 75% positive predictive value. The ML model was superior in all performance metrics to the KFRE 4- and 8-variable models.eGFR and glucose were found to be highly contributing to the ESKD prediction performance.
Conclusions:
The computational predictions had higher accuracy, specificity and positive predictive value, which indicates the potential integration into clinical workflows for decision support.
Related Concept Videos
01:28Chronic Kidney Disease III: Interprofessional Care
01:25Chronic Kidney Disease I: Introduction
01:18Chronic Kidney Disease IV: Nursing Management
01:30Acute Kidney Injury IV: Diagnostic Studies and Prevention
01:28Kidney Transplant I: Introduction
01:22Acute Kidney Injury I: Introduction