A CHUAH 1, D CHRISTIADI 1, G WALTERS 2, K KARPE 2, A KENNARD 2, R SINGER 2, G TALAULIKAR 2, K MCKEON 2, W GE 3, H SUOMINEN 3, T ANDREWS 1, S JIANG 1
1John Curtin School of Medical Research, The Australian National University, Canberra, Australia, 2Department of Renal Medicine, The Canberra Hospital, Garran, Australia, 3School of Computing, The Australian National University, Canberra, Australia
Aim: To investigate the utility of machine-learning (ML) built with the aid of large datasets in the prediction of end-stage kidney disease (ESKD).
Background: 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.
Results: A total of 12,371 patients were included, with 2,388 were found to have adequate density (four eGFR data points in the first 2 years) for subsequent analysis. Patients were divided into 80%/20% ratio for training and testing datasets. Of 2,388 patients, 263(11%) developed ESKD in the period of observation.
ML model has superior performance than nephrologist in predicting ESKD within 2 years with 94% accuracy, 60% sensitivity, 98% specificity, 75% positive predictive value, and 95% negative predictive value. eGFR and glucose were found to be highly contributing to the ESKD prediction performance.
Conclusions: The computational predictions had higher accuracy and precision which indicates the potential integration into clinical workflows for decision support.
Daniel Christiadi is a PhD Student at John Curtin School of Medical Research, studying Artificial Intelligence in the Prediction and Reclassification of Kidney Disease.