Machine Learning Prediction of End-stage Kidney Disease for Clinical Decision Support

Dan Andrews⁰, Aaron Chuah¹
(0) JCSMR
(1) Australian National University

Find me on Wed Nov 25th, 1:30-2:50pm AEDT in Remo, table 88

Abstract
Chronic kidney disease (CKD) is a major source of morbidity and mortality globally. Whilst CKD may ultimately culminate in end stage kidney disease (ESKD), CKD may progress at highly variable rates or not progress at all. ESKD is associated with a marked increase in mortality and morbidity and without renal replacement therapy (RRT) in the form of haemodialysis, peritoneal dialysis, or kidney transplantation is a terminal condition.

As ESKD approaches, clinicians are required to make difficult decisions. RRT requires either the formation of permanent dialysis access or evaluation of suitability for transplantation. Preparation for RRT is associated with significant cost and side effects such as post-operative infection and bleeding. The capacity of physicians to correctly identify which, and when, patients will require RRT is poor. Therefore, any method which will improve the ability to correctly identify patients who will require RRT is highly desirable.

Data-driven predictive modelling is a rapidly advancing field and has been employed in a range of clinical scenarios. Recent advances have demonstrated the capacity of predictive modelling as an adjunct in clinical care in kidney disease.

In this work, we will present a machine learning algorithm capable of predicting which patients will progress to ESKD and the time when they will reach ESKD. This data-driven approach, based on thousands of patient records from the Canberra Hospital, collected over nearly a decade, will assist treating physicians in planning if and when an individual patient should be prepared for RRT.