Identifying risk of adverse outcomes in COVID-19 patients via artificial intelligence-powered analysis of 12-lead intake electrocardiogram.
12-lead ECG
Arrhythmia
Artificial intelligence
COVID-19
Deep learning
Heart failure prognosis
Mortality
Risk factors
Journal
Cardiovascular digital health journal
ISSN: 2666-6936
Titre abrégé: Cardiovasc Digit Health J
Pays: United States
ID NLM: 101771268
Informations de publication
Date de publication:
Apr 2022
Apr 2022
Historique:
pubmed:
11
1
2022
medline:
11
1
2022
entrez:
10
1
2022
Statut:
ppublish
Résumé
Adverse events in COVID-19 are difficult to predict. Risk stratification is encumbered by the need to protect healthcare workers. We hypothesize that artificial intelligence (AI) can help identify subtle signs of myocardial involvement in the 12-lead electrocardiogram (ECG), which could help predict complications. Use intake ECGs from COVID-19 patients to train AI models to predict risk of mortality or major adverse cardiovascular events (MACE). We studied intake ECGs from 1448 COVID-19 patients (60.5% male, aged 63.4 ± 16.9 years). Records were labeled by mortality (death vs discharge) or MACE (no events vs arrhythmic, heart failure [HF], or thromboembolic [TE] events), then used to train AI models; these were compared to conventional regression models developed using demographic and comorbidity data. A total of 245 (17.7%) patients died (67.3% male, aged 74.5 ± 14.4 years); 352 (24.4%) experienced at least 1 MACE (119 arrhythmic, 107 HF, 130 TE). AI models predicted mortality and MACE with area under the curve (AUC) values of 0.60 ± 0.05 and 0.55 ± 0.07, respectively; these were comparable to AUC values for conventional models (0.73 ± 0.07 and 0.65 ± 0.10). There were no prominent temporal trends in mortality rate or MACE incidence in our cohort; holdout testing with data from after a cutoff date (June 9, 2020) did not degrade model performance. Using intake ECGs alone, our AI models had limited ability to predict hospitalized COVID-19 patients' risk of mortality or MACE. Our models' accuracy was comparable to that of conventional models built using more in-depth information, but translation to clinical use would require higher sensitivity and positive predictive value. In the future, we hope that mixed-input AI models utilizing both ECG and clinical data may be developed to enhance predictive accuracy.
Sections du résumé
BACKGROUND
BACKGROUND
Adverse events in COVID-19 are difficult to predict. Risk stratification is encumbered by the need to protect healthcare workers. We hypothesize that artificial intelligence (AI) can help identify subtle signs of myocardial involvement in the 12-lead electrocardiogram (ECG), which could help predict complications.
OBJECTIVE
OBJECTIVE
Use intake ECGs from COVID-19 patients to train AI models to predict risk of mortality or major adverse cardiovascular events (MACE).
METHODS
METHODS
We studied intake ECGs from 1448 COVID-19 patients (60.5% male, aged 63.4 ± 16.9 years). Records were labeled by mortality (death vs discharge) or MACE (no events vs arrhythmic, heart failure [HF], or thromboembolic [TE] events), then used to train AI models; these were compared to conventional regression models developed using demographic and comorbidity data.
RESULTS
RESULTS
A total of 245 (17.7%) patients died (67.3% male, aged 74.5 ± 14.4 years); 352 (24.4%) experienced at least 1 MACE (119 arrhythmic, 107 HF, 130 TE). AI models predicted mortality and MACE with area under the curve (AUC) values of 0.60 ± 0.05 and 0.55 ± 0.07, respectively; these were comparable to AUC values for conventional models (0.73 ± 0.07 and 0.65 ± 0.10). There were no prominent temporal trends in mortality rate or MACE incidence in our cohort; holdout testing with data from after a cutoff date (June 9, 2020) did not degrade model performance.
CONCLUSION
CONCLUSIONS
Using intake ECGs alone, our AI models had limited ability to predict hospitalized COVID-19 patients' risk of mortality or MACE. Our models' accuracy was comparable to that of conventional models built using more in-depth information, but translation to clinical use would require higher sensitivity and positive predictive value. In the future, we hope that mixed-input AI models utilizing both ECG and clinical data may be developed to enhance predictive accuracy.
Identifiants
pubmed: 35005676
doi: 10.1016/j.cvdhj.2021.12.003
pii: S2666-6936(21)00146-8
pmc: PMC8719367
doi:
Types de publication
Journal Article
Langues
eng
Pagination
62-74Subventions
Organisme : NCATS NIH HHS
ID : UL1 TR002319
Pays : United States
Informations de copyright
© 2021 Heart Rhythm Society.
Références
JAMA. 2020 Aug 25;324(8):782-793
pubmed: 32648899
Nat Med. 2020 Jun;26(6):886-891
pubmed: 32393799
Nature. 2015 May 28;521(7553):436-44
pubmed: 26017442
Nat Med. 2019 Jan;25(1):70-74
pubmed: 30617318
Neural Comput. 1997 Nov 15;9(8):1735-80
pubmed: 9377276
Europace. 2021 Jan 27;23(1):123-129
pubmed: 33097933
N Engl J Med. 2020 Oct 8;383(15):1483-1488
pubmed: 32706956
Circ Arrhythm Electrophysiol. 2020 Aug;13(8):e008437
pubmed: 32986471
Stem Cell Reports. 2021 Mar 9;16(3):478-492
pubmed: 33657418
J Am Coll Cardiol. 2020 Feb 25;75(7):722-733
pubmed: 32081280
J Biomed Inform. 2019 Jul;95:103208
pubmed: 31078660
Nat Rev Cardiol. 2020 Sep;17(9):543-558
pubmed: 32690910
Lancet. 2021 Aug 14;398(10300):599-607
pubmed: 34332652
J Biomed Inform. 2009 Apr;42(2):377-81
pubmed: 18929686
JAMA Netw Open. 2020 Aug 3;3(8):e2017267
pubmed: 32749465
J Clin Med. 2020 Nov 12;9(11):
pubmed: 33198406
Scand J Trauma Resusc Emerg Med. 2020 Oct 6;28(1):98
pubmed: 33023615
Circ Arrhythm Electrophysiol. 2020 Dec;13(12):e009355
pubmed: 33185118
BMC Bioinformatics. 2011 Mar 17;12:77
pubmed: 21414208
JAMA Cardiol. 2020 Jul 1;5(7):802-810
pubmed: 32211816
Cardiol Rev. 2021 Jan/Feb;29(1):43-47
pubmed: 32947478
Am J Cardiol. 1994 May 15;73(13):937-49
pubmed: 8184849
J Am Coll Cardiol. 2020 Jun 9;75(22):2871-2872
pubmed: 32283124
Stroke. 2021 Jan;52(2):563-572
pubmed: 33430638
Lancet. 2019 Sep 7;394(10201):861-867
pubmed: 31378392
Sci Rep. 2021 May 26;11(1):10949
pubmed: 34040033
JAMA Psychiatry. 2020 May 1;77(5):534-540
pubmed: 31774490
N Engl J Med. 2021 Jan 28;384(4):353-361
pubmed: 33503344
Clin Infect Dis. 2022 Feb 11;74(3):387-394
pubmed: 33822023
Circulation. 2020 Nov 24;142(21):2080-2082
pubmed: 33054349
J Am Heart Assoc. 2020 Apr 7;9(7):e014717
pubmed: 32200712