A Recurrent Neural Network Model for Predicting Activated Partial Thromboplastin Time After Treatment With Heparin: Retrospective Study.
ICU
activated partial thromboplastin time (aPTT)
critical care
deep learning
health care
heparin
machine learning
recurrent neural network
Journal
JMIR medical informatics
ISSN: 2291-9694
Titre abrégé: JMIR Med Inform
Pays: Canada
ID NLM: 101645109
Informations de publication
Date de publication:
13 Oct 2022
13 Oct 2022
Historique:
received:
02
05
2022
accepted:
11
08
2022
revised:
17
07
2022
entrez:
13
10
2022
pubmed:
14
10
2022
medline:
14
10
2022
Statut:
epublish
Résumé
Anticoagulation therapy with heparin is a frequent treatment in intensive care units and is monitored by activated partial thromboplastin clotting time (aPTT). It has been demonstrated that reaching an established anticoagulation target within 24 hours is associated with favorable outcomes. However, patients respond to heparin differently and reaching the anticoagulation target can be challenging. Machine learning algorithms may potentially support clinicians with improved dosing recommendations. This study evaluates a range of machine learning algorithms on their capability of predicting the patients' response to heparin treatment. In this analysis, we apply, for the first time, a model that considers time series. We extracted patient demographics, laboratory values, dialysis and extracorporeal membrane oxygenation treatments, and scores from the hospital information system. We predicted the numerical values of aPTT laboratory values 24 hours after continuous heparin infusion and evaluated 7 different machine learning models. The best-performing model was compared to recently published models on a classification task. We considered all data before and within the first 12 hours of continuous heparin infusion as features and predicted the aPTT value after 24 hours. The distribution of aPTT in our cohort of 5926 hospital admissions was highly skewed. Most patients showed aPTT values below 75 s, while some outliers showed much higher aPTT values. A recurrent neural network that consumes a time series of features showed the highest performance on the test set. A recurrent neural network that uses time series of features instead of only static and aggregated features showed the highest performance in predicting aPTT after heparin treatment.
Sections du résumé
BACKGROUND
BACKGROUND
Anticoagulation therapy with heparin is a frequent treatment in intensive care units and is monitored by activated partial thromboplastin clotting time (aPTT). It has been demonstrated that reaching an established anticoagulation target within 24 hours is associated with favorable outcomes. However, patients respond to heparin differently and reaching the anticoagulation target can be challenging. Machine learning algorithms may potentially support clinicians with improved dosing recommendations.
OBJECTIVE
OBJECTIVE
This study evaluates a range of machine learning algorithms on their capability of predicting the patients' response to heparin treatment. In this analysis, we apply, for the first time, a model that considers time series.
METHODS
METHODS
We extracted patient demographics, laboratory values, dialysis and extracorporeal membrane oxygenation treatments, and scores from the hospital information system. We predicted the numerical values of aPTT laboratory values 24 hours after continuous heparin infusion and evaluated 7 different machine learning models. The best-performing model was compared to recently published models on a classification task. We considered all data before and within the first 12 hours of continuous heparin infusion as features and predicted the aPTT value after 24 hours.
RESULTS
RESULTS
The distribution of aPTT in our cohort of 5926 hospital admissions was highly skewed. Most patients showed aPTT values below 75 s, while some outliers showed much higher aPTT values. A recurrent neural network that consumes a time series of features showed the highest performance on the test set.
CONCLUSIONS
CONCLUSIONS
A recurrent neural network that uses time series of features instead of only static and aggregated features showed the highest performance in predicting aPTT after heparin treatment.
Identifiants
pubmed: 36227653
pii: v10i10e39187
doi: 10.2196/39187
pmc: PMC9614623
doi:
Types de publication
Journal Article
Langues
eng
Pagination
e39187Informations de copyright
©Sebastian Daniel Boie, Lilian Jo Engelhardt, Nicolas Coenen, Niklas Giesa, Kerstin Rubarth, Mario Menk, Felix Balzer. Originally published in JMIR Medical Informatics (https://medinform.jmir.org), 13.10.2022.
Références
Kardiol Pol. 2014;72(11):997-1053
pubmed: 25524375
Br J Clin Pharmacol. 2021 Nov;87(11):4124-4139
pubmed: 33835524
J Thromb Haemost. 2021 Aug;19(8):2002-2006
pubmed: 33555096
Chest. 1995 Oct;108(4 Suppl):276S-290S
pubmed: 7555182
NPJ Digit Med. 2021 Oct 11;4(1):147
pubmed: 34635760
Res Pract Thromb Haemost. 2020 Jul 06;4(6):1004-1012
pubmed: 32864551
Chest. 2012 Feb;141(2 Suppl):e24S-e43S
pubmed: 22315264
J Am Heart Assoc. 2020 Sep;9(17):e016784
pubmed: 32809909
Eur Respir J. 2019 Oct 9;54(3):
pubmed: 31473594
Sci Rep. 2018 Apr 17;8(1):6085
pubmed: 29666385
Artif Organs. 1998 Sep;22(9):731-9
pubmed: 9754457
N Engl J Med. 1972 Aug 17;287(7):324-7
pubmed: 5041701
JMIR Med Inform. 2020 Jun 22;8(6):e17648
pubmed: 32568089
J Med Internet Res. 2021 May 20;23(5):e27118
pubmed: 34014171
JAMA Netw Open. 2022 Mar 1;5(3):e224205
pubmed: 35333359
Mayo Clin Proc. 2020 Dec;95(12):2775-2798
pubmed: 33276846
Intensive Care Med. 2014 Sep;40(9):1332-9
pubmed: 25091788
BMJ. 2020 Mar 20;368:l6927
pubmed: 32198138
Front Neuroinform. 2014 Feb 21;8:14
pubmed: 24600388
Res Pract Thromb Haemost. 2019 Sep 08;3(4):749-757
pubmed: 31624795
Thromb Haemost. 2006 Nov;96(5):547-52
pubmed: 17080209
Chest. 2010 Jun;137(6):1382-90
pubmed: 20081101
Arch Intern Med. 2000 Mar 27;160(6):761-8
pubmed: 10737275
Crit Care. 2015 Aug 18;19:287
pubmed: 26283414
Singapore Med J. 2005 Sep;46(9):450-6
pubmed: 16123828