An Explainable Machine Learning Model to Predict Acute Kidney Injury After Cardiac Surgery: A Retrospective Cohort Study.
SHAP
acute kidney injury
cardiac surgery
machine learning
prediction model
shapley additive explanations
Journal
Clinical epidemiology
ISSN: 1179-1349
Titre abrégé: Clin Epidemiol
Pays: New Zealand
ID NLM: 101531700
Informations de publication
Date de publication:
2023
2023
Historique:
received:
25
01
2023
accepted:
27
09
2023
medline:
11
12
2023
pubmed:
11
12
2023
entrez:
11
12
2023
Statut:
epublish
Résumé
To derive and validate a machine learning (ML) prediction model of acute kidney injury (AKI) that could be used for AKI surveillance and management to improve clinical outcomes. This retrospective cohort study was conducted in Fuwai Hospital, including patients aged 18 years and above undergoing cardiac surgery admitted between January 1, 2017, and December 31, 2018. Seventy percent of the observations were randomly selected for training and the remaining 30% for testing. The demographics, comorbidities, laboratory examination parameters, and operation details were used to construct a prediction model for AKI by logistic regression and eXtreme gradient boosting (Xgboost). The discrimination of each model was assessed on the test cohort by the area under the receiver operator characteristic (AUROC) curve, while calibration was performed by the calibration plot. A total of 15,880 patients were enrolled in this study, and 4845 (30.5%) had developed AKI. Xgboost model had the higher discriminative ability compared with logistic regression (AUROC, 0.849 [95% CI, 0.837-0.861] vs 0.803[95% CI 0.790-0.817], ML models can provide clinical decision support to determine which patients should focus on perioperative preventive treatment to preemptively reduce acute kidney injury by predicting which patients are not at risk.
Sections du résumé
Background
UNASSIGNED
To derive and validate a machine learning (ML) prediction model of acute kidney injury (AKI) that could be used for AKI surveillance and management to improve clinical outcomes.
Methods
UNASSIGNED
This retrospective cohort study was conducted in Fuwai Hospital, including patients aged 18 years and above undergoing cardiac surgery admitted between January 1, 2017, and December 31, 2018. Seventy percent of the observations were randomly selected for training and the remaining 30% for testing. The demographics, comorbidities, laboratory examination parameters, and operation details were used to construct a prediction model for AKI by logistic regression and eXtreme gradient boosting (Xgboost). The discrimination of each model was assessed on the test cohort by the area under the receiver operator characteristic (AUROC) curve, while calibration was performed by the calibration plot.
Results
UNASSIGNED
A total of 15,880 patients were enrolled in this study, and 4845 (30.5%) had developed AKI. Xgboost model had the higher discriminative ability compared with logistic regression (AUROC, 0.849 [95% CI, 0.837-0.861] vs 0.803[95% CI 0.790-0.817],
Conclusion
UNASSIGNED
ML models can provide clinical decision support to determine which patients should focus on perioperative preventive treatment to preemptively reduce acute kidney injury by predicting which patients are not at risk.
Identifiants
pubmed: 38076638
doi: 10.2147/CLEP.S404580
pii: 404580
pmc: PMC10706584
doi:
Types de publication
Journal Article
Langues
eng
Pagination
1145-1157Informations de copyright
© 2023 Gao et al.
Déclaration de conflit d'intérêts
The authors report no conflicts of interest in this work.
Références
J Am Soc Nephrol. 2005 Jan;16(1):162-8
pubmed: 15563569
J Med Internet Res. 2023 Jan 5;25:e41142
pubmed: 36603200
Clin Kidney J. 2022 Aug 02;15(12):2266-2280
pubmed: 36381375
J Cardiothorac Vasc Anesth. 2021 Mar;35(3):866-873
pubmed: 32713734
J Card Surg. 2022 Nov;37(11):3838-3845
pubmed: 36001761
J Med Chem. 2020 Aug 27;63(16):8761-8777
pubmed: 31512867
Ann Thorac Surg. 2012 Jan;93(1):337-47
pubmed: 22186469
Ann Intern Med. 2009 May 5;150(9):604-12
pubmed: 19414839
Crit Care. 2020 Jul 31;24(1):478
pubmed: 32736589
JAMA Netw Open. 2021 Mar 1;4(3):e212240
pubmed: 33783520
Comput Methods Programs Biomed. 2019 Mar;170:1-9
pubmed: 30712598
J Clin Med. 2018 Oct 03;7(10):
pubmed: 30282956
J Nephrol. 2019 Dec;32(6):937-945
pubmed: 31243735
Lancet Digit Health. 2020 Apr;2(4):e179-e191
pubmed: 33328078
Crit Care Med. 2018 Jul;46(7):1070-1077
pubmed: 29596073
J Thorac Cardiovasc Surg. 2016 Jul;152(1):245-251.e4
pubmed: 27045042
J Clin Epidemiol. 2019 Jun;110:12-22
pubmed: 30763612
Crit Rev Clin Lab Sci. 2021 Aug;58(5):354-368
pubmed: 33556265
Crit Care. 2018 Aug 18;22(1):197
pubmed: 30119691
Anesth Analg. 2021 Sep 1;133(3):570-577
pubmed: 34153017
Kidney Int. 2019 Mar;95(3):590-610
pubmed: 30709662
Int J Cardiol. 2022 Jan 15;347:21-27
pubmed: 34774886
Curr Opin Anaesthesiol. 2017 Feb;30(1):60-65
pubmed: 27820742
N Engl J Med. 2015 Oct 8;373(15):1408-17
pubmed: 26436207
Kidney Int. 2007 Sep;72(5):624-31
pubmed: 17622275
Clin Chem Lab Med. 2017 Jul 26;55(8):1074-1089
pubmed: 28076311
Clin J Am Soc Nephrol. 2008 Jul;3(4):962-7
pubmed: 18354074
Intensive Care Med. 2020 Mar;46(3):383-400
pubmed: 31965266
JAMA. 2016 Mar 1;315(9):877-88
pubmed: 26906014
BMC Med Inform Decis Mak. 2022 May 18;22(1):137
pubmed: 35585624
Eur J Cardiothorac Surg. 2022 Oct 4;62(5):
pubmed: 35521994