Machine learning to predict venous thrombosis in acutely ill medical patients.
acute medically ill
machine learning
personalized medicine
super learner
venous thromboembolism
Journal
Research and practice in thrombosis and haemostasis
ISSN: 2475-0379
Titre abrégé: Res Pract Thromb Haemost
Pays: United States
ID NLM: 101703775
Informations de publication
Date de publication:
Feb 2020
Feb 2020
Historique:
received:
29
08
2019
revised:
02
10
2019
accepted:
06
10
2019
entrez:
29
2
2020
pubmed:
29
2
2020
medline:
29
2
2020
Statut:
epublish
Résumé
The identification of acutely ill patients at high risk for venous thromboembolism (VTE) may be determined clinically or by use of integer-based scoring systems. These scores demonstrated modest performance in external data sets. To evaluate the performance of machine learning models compared to the IMPROVE score. The APEX trial randomized 7513 acutely medically ill patients to extended duration betrixaban vs. enoxaparin. Including 68 variables, a super learner model (ML) was built to predict VTE by combining estimates from 5 families of candidate models. A "reduced" model (rML) was also developed using 16 variables that were thought, a priori, to be associated with VTE. The IMPROVE score was calculated for each patient. Model performance was assessed by discrimination and calibration to predict a composite VTE end point. The frequency of predicted risks of VTE were plotted and divided into tertiles. VTE risks were compared across tertiles. The ML and rML algorithms outperformed the IMPROVE score in predicting VTE (c-statistic: 0.69, 0.68 and 0.59, respectively). The Hosmer-Lemeshow goodness-of-fit The super learner algorithms improved discrimination and calibration compared to the IMPROVE score for predicting VTE in acute medically ill patients.
Sections du résumé
BACKGROUND
BACKGROUND
The identification of acutely ill patients at high risk for venous thromboembolism (VTE) may be determined clinically or by use of integer-based scoring systems. These scores demonstrated modest performance in external data sets.
OBJECTIVES
OBJECTIVE
To evaluate the performance of machine learning models compared to the IMPROVE score.
METHODS
METHODS
The APEX trial randomized 7513 acutely medically ill patients to extended duration betrixaban vs. enoxaparin. Including 68 variables, a super learner model (ML) was built to predict VTE by combining estimates from 5 families of candidate models. A "reduced" model (rML) was also developed using 16 variables that were thought, a priori, to be associated with VTE. The IMPROVE score was calculated for each patient. Model performance was assessed by discrimination and calibration to predict a composite VTE end point. The frequency of predicted risks of VTE were plotted and divided into tertiles. VTE risks were compared across tertiles.
RESULTS
RESULTS
The ML and rML algorithms outperformed the IMPROVE score in predicting VTE (c-statistic: 0.69, 0.68 and 0.59, respectively). The Hosmer-Lemeshow goodness-of-fit
CONCLUSION
CONCLUSIONS
The super learner algorithms improved discrimination and calibration compared to the IMPROVE score for predicting VTE in acute medically ill patients.
Identifiants
pubmed: 32110753
doi: 10.1002/rth2.12292
pii: S2475-0379(22)01969-0
pmc: PMC7040551
doi:
Types de publication
Journal Article
Langues
eng
Pagination
230-237Informations de copyright
© 2020 The Authors. Research and Practice in Thrombosis and Haemostasis published by Wiley Periodicals, Inc on behalf of International Society on Thrombosis and Haemostasis.
Références
JAMA. 2016 Dec 13;316(22):2402-2410
pubmed: 27898976
Public Health. 2015 Mar;129(3):189-90
pubmed: 25747566
Chest. 2007 Sep;132(3):936-45
pubmed: 17573514
Stat Appl Genet Mol Biol. 2007;6:Article25
pubmed: 17910531
Thromb Haemost. 2017 Apr 3;117(4):801-808
pubmed: 28150851
J Thromb Haemost. 2013 Mar;11(3):467-73
pubmed: 23279085
Chest. 2011 Sep;140(3):706-714
pubmed: 21436241
N Engl J Med. 2011 Dec 8;365(23):2167-77
pubmed: 22077144
J Hosp Med. 2012 Mar;7(3):231-8
pubmed: 22190427
Thromb Haemost. 2005 Oct;94(4):750-9
pubmed: 16270626
N Engl J Med. 2013 Feb 7;368(6):513-23
pubmed: 23388003
Am Heart J. 2014 Mar;167(3):335-41
pubmed: 24576517
J Thromb Haemost. 2010 Nov;8(11):2450-7
pubmed: 20738765
Circulation. 2015 Nov 17;132(20):1920-30
pubmed: 26572668
J Am Coll Cardiol. 2018 Apr 17;71(15):1650-1660
pubmed: 29650121
J Am Heart Assoc. 2014 Nov 17;3(6):e001152
pubmed: 25404191
Lancet Respir Med. 2015 Jan;3(1):42-52
pubmed: 25466337
Am J Med. 2016 Sep;129(9):1001.e9-1001.e18
pubmed: 27107925
JAMA. 2017 Dec 12;318(22):2184-2186
pubmed: 29234791
JAMA. 2016 Apr 26;315(16):1735-49
pubmed: 27022822
N Engl J Med. 2018 Sep 20;379(12):1118-1127
pubmed: 30145946
N Engl J Med. 2016 Aug 11;375(6):534-44
pubmed: 27232649