A deep-learning model to continuously predict severe acute kidney injury based on urine output changes in critically ill patients.
Acute kidney injury
Artificial intelligence
EAlert
KDIGO
Journal
Journal of nephrology
ISSN: 1724-6059
Titre abrégé: J Nephrol
Pays: Italy
ID NLM: 9012268
Informations de publication
Date de publication:
Dec 2021
Dec 2021
Historique:
received:
01
02
2021
accepted:
02
04
2021
pubmed:
27
4
2021
medline:
29
1
2022
entrez:
26
4
2021
Statut:
ppublish
Résumé
Acute Kidney Injury (AKI), a frequent complication of pateints in the Intensive Care Unit (ICU), is associated with a high mortality rate. Early prediction of AKI is essential in order to trigger the use of preventive care actions. The aim of this study was to ascertain the accuracy of two mathematical analysis models in obtaining a predictive score for AKI development. A deep learning model based on a urine output trends was compared with a logistic regression analysis for AKI prediction in stages 2 and 3 (defined as the simultaneous increase of serum creatinine and decrease of urine output, according to the Acute Kidney Injury Network (AKIN) guidelines). Two retrospective datasets including 35,573 ICU patients were analyzed. Urine output data were used to train and test the logistic regression and the deep learning model. The deep learning model defined an area under the curve (AUC) of 0.89 (± 0.01), sensitivity = 0.8 and specificity = 0.84, which was higher than the logistic regression analysis. The deep learning model was able to predict 88% of AKI cases more than 12 h before their onset: for every 6 patients identified as being at risk of AKI by the deep learning model, 5 experienced the event. On the contrary, for every 12 patients not considered to be at risk by the model, 2 developed AKI. In conclusion, by using urine output trends, deep learning analysis was able to predict AKI episodes more than 12 h in advance, and with a higher accuracy than the classical urine output thresholds. We suggest that this algorithm could be integrated in the ICU setting to better manage, and potentially prevent, AKI episodes.
Sections du résumé
BACKGROUND
BACKGROUND
Acute Kidney Injury (AKI), a frequent complication of pateints in the Intensive Care Unit (ICU), is associated with a high mortality rate. Early prediction of AKI is essential in order to trigger the use of preventive care actions.
METHODS
METHODS
The aim of this study was to ascertain the accuracy of two mathematical analysis models in obtaining a predictive score for AKI development. A deep learning model based on a urine output trends was compared with a logistic regression analysis for AKI prediction in stages 2 and 3 (defined as the simultaneous increase of serum creatinine and decrease of urine output, according to the Acute Kidney Injury Network (AKIN) guidelines). Two retrospective datasets including 35,573 ICU patients were analyzed. Urine output data were used to train and test the logistic regression and the deep learning model.
RESULTS
RESULTS
The deep learning model defined an area under the curve (AUC) of 0.89 (± 0.01), sensitivity = 0.8 and specificity = 0.84, which was higher than the logistic regression analysis. The deep learning model was able to predict 88% of AKI cases more than 12 h before their onset: for every 6 patients identified as being at risk of AKI by the deep learning model, 5 experienced the event. On the contrary, for every 12 patients not considered to be at risk by the model, 2 developed AKI.
CONCLUSION
CONCLUSIONS
In conclusion, by using urine output trends, deep learning analysis was able to predict AKI episodes more than 12 h in advance, and with a higher accuracy than the classical urine output thresholds. We suggest that this algorithm could be integrated in the ICU setting to better manage, and potentially prevent, AKI episodes.
Identifiants
pubmed: 33900581
doi: 10.1007/s40620-021-01046-6
pii: 10.1007/s40620-021-01046-6
pmc: PMC8610952
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
1875-1886Informations de copyright
© 2021. The Author(s).
Références
Intensive Care Med. 2017 Jun;43(6):764-773
pubmed: 28130688
Kidney Int. 2017 Aug;92(2):440-452
pubmed: 28416224
Crit Care Med. 2012 Apr;40(4):1164-70
pubmed: 22067631
N Engl J Med. 2008 Jul 3;359(1):7-20
pubmed: 18492867
Sci Data. 2018 Sep 11;5:180178
pubmed: 30204154
J Clin Med. 2020 Mar 03;9(3):
pubmed: 32138284
Crit Care. 2011 Jul 19;15(4):R172
pubmed: 21771324
Crit Care. 2007;11(2):R31
pubmed: 17331245
Kidney Int. 2021 May;99(5):1179-1188
pubmed: 32889014
Am J Clin Nutr. 2016 May;103(5):1197-203
pubmed: 27030535
Clin J Am Soc Nephrol. 2020 Mar 6;15(3):423-429
pubmed: 32075806
JAMA. 2005 Aug 17;294(7):813-8
pubmed: 16106006
Lancet. 2015 May 16;385(9981):1966-74
pubmed: 25726515
Lancet. 2015 Jun 27;385(9987):2616-43
pubmed: 25777661
Intensive Care Med. 2015 Aug;41(8):1411-23
pubmed: 26162677
Sci Rep. 2019 Dec 2;9(1):18141
pubmed: 31792326
Crit Care. 2013 Jun 20;17(3):R112
pubmed: 23787055
Ann Intensive Care. 2019 Jun 7;9(1):65
pubmed: 31175471
Adv Chronic Kidney Dis. 2017 Jul;24(4):194-204
pubmed: 28778358
BMC Nephrol. 2017 Feb 20;18(1):70
pubmed: 28219327
Curr Opin Nephrol Hypertens. 2013 Nov;22(6):637-42
pubmed: 24100217
Kidney Int. 2011 Oct;80(7):760-7
pubmed: 21716258
Int Urol Nephrol. 2018 Aug;50(8):1483-1488
pubmed: 29556903
Nephron Clin Pract. 2012;120(4):c179-84
pubmed: 22890468
Intensive Care Med. 2013 Mar;39(3):420-8
pubmed: 23291734
Crit Care. 2020 Apr 21;24(1):164
pubmed: 32316994
Am J Kidney Dis. 2018 Jan;71(1):9-19
pubmed: 28754457
J Am Soc Nephrol. 2013 Jan;24(1):37-42
pubmed: 23222124
BMJ Open. 2019 Jun 1;9(5):e025117
pubmed: 31154298
Sci Data. 2016 May 24;3:160035
pubmed: 27219127
Postgrad Med J. 2016 Jan;92(1083):9-13
pubmed: 26512125
Crit Care. 2020 Apr 23;24(1):171
pubmed: 32326981
BMC Med Inform Decis Mak. 2019 Jan 31;19(Suppl 1):16
pubmed: 30700291
Crit Care. 2012 Jul 12;16(4):R123
pubmed: 22789111
Int J Clin Pract. 2016 Apr;70(4):330-9
pubmed: 26799821
Kidney Int. 2015 Sep 09;:
pubmed: 26352301