Machine learning identifies ICU outcome predictors in a multicenter COVID-19 cohort.
ARDS
COVID-19
Critical care
Outcome
Prognostic models
Journal
Critical care (London, England)
ISSN: 1466-609X
Titre abrégé: Crit Care
Pays: England
ID NLM: 9801902
Informations de publication
Date de publication:
17 Aug 2021
17 Aug 2021
Historique:
received:
17
06
2021
accepted:
01
08
2021
entrez:
18
8
2021
pubmed:
19
8
2021
medline:
27
8
2021
Statut:
epublish
Résumé
Intensive Care Resources are heavily utilized during the COVID-19 pandemic. However, risk stratification and prediction of SARS-CoV-2 patient clinical outcomes upon ICU admission remain inadequate. This study aimed to develop a machine learning model, based on retrospective & prospective clinical data, to stratify patient risk and predict ICU survival and outcomes. A Germany-wide electronic registry was established to pseudonymously collect admission, therapeutic and discharge information of SARS-CoV-2 ICU patients retrospectively and prospectively. Machine learning approaches were evaluated for the accuracy and interpretability of predictions. The Explainable Boosting Machine approach was selected as the most suitable method. Individual, non-linear shape functions for predictive parameters and parameter interactions are reported. 1039 patients were included in the Explainable Boosting Machine model, 596 patients retrospectively collected, and 443 patients prospectively collected. The model for prediction of general ICU outcome was shown to be more reliable to predict "survival". Age, inflammatory and thrombotic activity, and severity of ARDS at ICU admission were shown to be predictive of ICU survival. Patients' age, pulmonary dysfunction and transfer from an external institution were predictors for ECMO therapy. The interaction of patient age with D-dimer levels on admission and creatinine levels with SOFA score without GCS were predictors for renal replacement therapy. Using Explainable Boosting Machine analysis, we confirmed and weighed previously reported and identified novel predictors for outcome in critically ill COVID-19 patients. Using this strategy, predictive modeling of COVID-19 ICU patient outcomes can be performed overcoming the limitations of linear regression models. Trial registration "ClinicalTrials" (clinicaltrials.gov) under NCT04455451.
Sections du résumé
BACKGROUND
BACKGROUND
Intensive Care Resources are heavily utilized during the COVID-19 pandemic. However, risk stratification and prediction of SARS-CoV-2 patient clinical outcomes upon ICU admission remain inadequate. This study aimed to develop a machine learning model, based on retrospective & prospective clinical data, to stratify patient risk and predict ICU survival and outcomes.
METHODS
METHODS
A Germany-wide electronic registry was established to pseudonymously collect admission, therapeutic and discharge information of SARS-CoV-2 ICU patients retrospectively and prospectively. Machine learning approaches were evaluated for the accuracy and interpretability of predictions. The Explainable Boosting Machine approach was selected as the most suitable method. Individual, non-linear shape functions for predictive parameters and parameter interactions are reported.
RESULTS
RESULTS
1039 patients were included in the Explainable Boosting Machine model, 596 patients retrospectively collected, and 443 patients prospectively collected. The model for prediction of general ICU outcome was shown to be more reliable to predict "survival". Age, inflammatory and thrombotic activity, and severity of ARDS at ICU admission were shown to be predictive of ICU survival. Patients' age, pulmonary dysfunction and transfer from an external institution were predictors for ECMO therapy. The interaction of patient age with D-dimer levels on admission and creatinine levels with SOFA score without GCS were predictors for renal replacement therapy.
CONCLUSIONS
CONCLUSIONS
Using Explainable Boosting Machine analysis, we confirmed and weighed previously reported and identified novel predictors for outcome in critically ill COVID-19 patients. Using this strategy, predictive modeling of COVID-19 ICU patient outcomes can be performed overcoming the limitations of linear regression models. Trial registration "ClinicalTrials" (clinicaltrials.gov) under NCT04455451.
Identifiants
pubmed: 34404458
doi: 10.1186/s13054-021-03720-4
pii: 10.1186/s13054-021-03720-4
pmc: PMC8370055
doi:
Banques de données
ClinicalTrials.gov
['NCT04455451']
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
295Subventions
Organisme : deutsche forschungsgemeinschaft
ID : DFG RO 3671/8-1
Organisme : bundesministerium für bildung, wissenschaft, forschung und technologie
ID : DIFUTURE, 01ZZ1804D
Informations de copyright
© 2021. The Author(s).
Références
Crit Care Resusc. 2020 Sep;22(3):200-211
pubmed: 32900326
J Zhejiang Univ Sci B. 2020 May;21(5):378-387
pubmed: 32425003
JAMA. 2020 Apr 28;323(16):1574-1581
pubmed: 32250385
Stat Methods Med Res. 1995 Sep;4(3):187-96
pubmed: 8548102
N Engl J Med. 2020 Jun 11;382(24):2372-2374
pubmed: 32302078
Arterioscler Thromb Vasc Biol. 2014 Sep;34(9):1977-84
pubmed: 25012129
BMJ. 2020 May 22;369:m1966
pubmed: 32444366
JAMA. 2012 Jun 20;307(23):2526-33
pubmed: 22797452
Crit Care. 2016 Jul 22;20(1):226
pubmed: 27448995
Age Ageing. 2021 Feb 26;50(2):326-334
pubmed: 33201181
N Engl J Med. 2020 May 21;382(21):2012-2022
pubmed: 32227758
Lancet Respir Med. 2020 May;8(5):475-481
pubmed: 32105632
Am J Respir Crit Care Med. 2020 Jun 1;201(11):1430-1434
pubmed: 32267160
BMC Infect Dis. 2021 Jan 9;21(1):39
pubmed: 33421989
J Biomed Inform. 2009 Apr;42(2):377-81
pubmed: 18929686
Crit Care Resusc. 2020 Apr 06;22(2):119-125
pubmed: 32248675
Dan Med J. 2020 Apr 27;67(5):
pubmed: 32351196
Anaesth Crit Care Pain Med. 2020 Oct;39(5):553-561
pubmed: 32278670
Am Rev Respir Dis. 1988 Sep;138(3):720-3
pubmed: 3202424
J Thromb Haemost. 2018 Feb;16(2):231-241
pubmed: 29193703
Lancet. 2020 Oct 10;396(10257):1071-1078
pubmed: 32987008
Intensive Care Med. 2021 Jan;47(1):60-73
pubmed: 33211135
Haematologica. 2021 Aug 01;106(8):2170-2179
pubmed: 34011137
Eur Heart J. 2020 May 7;41(18):1747-1756
pubmed: 31923316
J Biomed Inform. 2019 Jul;95:103208
pubmed: 31078660
Lancet. 2020 Jun 6;395(10239):1763-1770
pubmed: 32442528
N Engl J Med. 2021 Jun 3;384(22):2092-2101
pubmed: 33835769
Z Gesundh Wiss. 2022;30(2):361-364
pubmed: 32318325
Lancet. 2020 May 16;395(10236):1579-1586
pubmed: 32416782
Intensive Care Med. 1996 Jul;22(7):707-10
pubmed: 8844239
Eur Heart J. 2017 Feb 14;38(7):500-507
pubmed: 27252451