Machine Learning Improves Upon Clinicians' Prediction of End Stage Kidney Disease.
XGBoost (Extreme Gradient Boosting)
chronic kidney disease
end stage kidney disease (ESKD)
machine learning (ML)
prediction model
Journal
Frontiers in medicine
ISSN: 2296-858X
Titre abrégé: Front Med (Lausanne)
Pays: Switzerland
ID NLM: 101648047
Informations de publication
Date de publication:
2022
2022
Historique:
received:
16
12
2021
accepted:
18
02
2022
entrez:
4
4
2022
pubmed:
5
4
2022
medline:
5
4
2022
Statut:
epublish
Résumé
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. 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). 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. The computational predictions had higher accuracy, specificity and positive predictive value, which indicates the potential integration into clinical workflows for decision support.
Sections du résumé
Background and Objectives
UNASSIGNED
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
UNASSIGNED
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
UNASSIGNED
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
UNASSIGNED
The computational predictions had higher accuracy, specificity and positive predictive value, which indicates the potential integration into clinical workflows for decision support.
Identifiants
pubmed: 35372378
doi: 10.3389/fmed.2022.837232
pmc: PMC8965763
doi:
Types de publication
Journal Article
Langues
eng
Pagination
837232Informations de copyright
Copyright © 2022 Chuah, Walters, Christiadi, Karpe, Kennard, Singer, Talaulikar, Ge, Suominen, Andrews and Jiang.
Déclaration de conflit d'intérêts
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Références
Kidney Int. 2006 Nov;70(10):1694-705
pubmed: 16969387
Nephrology (Carlton). 2013 Jan;18(1):22-5
pubmed: 23252801
Nat Rev Dis Primers. 2017 Nov 23;3:17088
pubmed: 29168475
PLoS One. 2016 Jul 06;11(7):e0158765
pubmed: 27383068
JAMA. 2016 Jan 12;315(2):164-74
pubmed: 26757465
Arch Intern Med. 2009 Feb 23;169(4):342-50
pubmed: 19237717
J Am Soc Nephrol. 2017 Jun;28(6):1839-1850
pubmed: 28031406
PLoS One. 2017 Jun 2;12(6):e0177678
pubmed: 28574989
Math Biosci. 2019 Apr;310:24-30
pubmed: 30768948
Kidney Int. 2020 Jun;97(6):1117-1129
pubmed: 32409237
Nature. 2019 Aug;572(7767):116-119
pubmed: 31367026
Nat Mach Intell. 2020 Jan;2(1):56-67
pubmed: 32607472
J Ren Care. 2015 Mar;41(1):62-75
pubmed: 25597792
Clin J Am Soc Nephrol. 2010 Jan;5(1):72-9
pubmed: 19965531
Medicine (Baltimore). 2016 Mar;95(11):e3013
pubmed: 26986114
PLoS One. 2018 Mar 21;13(3):e0192895
pubmed: 29561894
Nephrol Dial Transplant. 2014 Feb;29(2):333-41
pubmed: 24081862
Diabet Med. 2008 Oct;25(10):1151-6
pubmed: 19046192
J Am Med Inform Assoc. 2015 Jul;22(4):872-80
pubmed: 25896647
JAMA. 2011 Apr 20;305(15):1553-9
pubmed: 21482743
BMC Genomics. 2020 Jan 2;21(1):6
pubmed: 31898477
BMC Nephrol. 2021 May 24;22(1):194
pubmed: 34030639
J Med Syst. 2017 May;41(5):85
pubmed: 28401396
Comput Math Methods Med. 2016;2016:6080814
pubmed: 27022406
JAMA Netw Open. 2019 Mar 1;2(3):e190968
pubmed: 30901048
BMC Nephrol. 2015 Mar 28;16:37
pubmed: 25885460
Nephrol Dial Transplant. 2011 Jan;26(1):220-6
pubmed: 20610526
PLoS One. 2015 Mar 04;10(3):e0118432
pubmed: 25738806