Skip to main content

ORIGINAL RESEARCH article

Front. Med.
Sec. Nephrology
Volume 11 - 2024 | doi: 10.3389/fmed.2024.1428073
This article is part of the Research Topic Artificial Intelligence and Big Data for Value-Based Care - Volume III View all 4 articles

Dynamic survival prediction of end-stage kidney disease using random survival forest for competing risk analysis

Provisionally accepted
Daniel Christiadi Daniel Christiadi 1,2*Kevin Chai Kevin Chai 3Aaron Chuah Aaron Chuah 1Bronwyn Loong Bronwyn Loong 4Dan Andrews Dan Andrews 1Aron Chakera Aron Chakera 5Giles Desmond Walters Giles Desmond Walters 2,6Simon Hee-Tang Jiang Simon Hee-Tang Jiang 1,2
  • 1 Department of Immunology and Infectious Disease, John Curtin School of Medical Research, College of Health and Medicine, Australian National University, Canberra, Australian Capital Territory, Australia
  • 2 Department of Nephrology, The Canberra Hospital, Canberrra, Australia
  • 3 Curtin School of Population Health, Faculty of Health Sciences, Curtin University, Perth, Australia
  • 4 Research School of Finance, Actuarial Studies and Statistics, College of Business & Economics, Australian National University, Canberra, Australian Capital Territory, Australia
  • 5 Department of Renal Medicine, Sir Charles Gairdner Hospital, Nedlands, WA, Australia
  • 6 School of Medicine , Australian National University, Garran, ACT, Australia

The final, formatted version of the article will be published soon.

    A static predictive model that uses only baseline clinicopathological data cannot capture the heterogeneity of predictor trajectories in chronic kidney disease (CKD) progression. Using longitudinal clinicopathological data, we developed and validated a dynamic survival prediction model to predict end-stage kidney disease (ESKD) with death as a competing risk.A sequence of random survival forests was trained with a landmarking approach and optimised with a pre-specified prediction horizon of five years. The predicted cumulative incidence function (CIF) values were used to create a personalised dynamic prediction plot.Baseline demographics and thirteen longitudinal clinicopathological data items from 4,950 patients were used to develop the models. Analysing the variable importance for ESKD and death, a sequence of reduced models utilising age, serum measurement of albumin, bicarbonate, chloride, eGFR, and haemoglobin was created with a median concordance index for ESKD 84.84% and death 84.1%; median integrated Brier score ESKD 0.03 and death 0.038 across all landmark times, which were validated in an external cohort comprised of 8,729 patients.We developed and validated a dynamic survival prediction model to predict ESKD with death as competing risks using common longitudinal, clinicopathological data to assist clinicians in dialysis planning in patients with chronic kidney disease.

    Keywords: Dynamic prediction model, end-stage kidney disease, Landmarking, Random survival forest, competing risk

    Received: 05 May 2024; Accepted: 28 Nov 2024.

    Copyright: © 2024 Christiadi, Chai, Chuah, Loong, Andrews, Chakera, Walters and Jiang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

    * Correspondence: Daniel Christiadi, Department of Immunology and Infectious Disease, John Curtin School of Medical Research, College of Health and Medicine, Australian National University, Canberra, ACT 2601, Australian Capital Territory, Australia

    Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.