Developing neural network models for early detection of cardiac arrest in emergency department

Dong Hyun Jang, Joonghee Kim, You Hwan Jo, Jae Hyuk Lee, Ji Eun Hwang, Seung Min Park, Dong Keon Lee, Inwon Park, Doyun Kim, Hyunglan Chang

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Background: Automated surveillance for cardiac arrests would be useful in overcrowded emergency departments. The purpose of this study is to develop and test artificial neural network (ANN) classifiers for early detection of patients at risk of cardiac arrest in emergency departments. Methods: This is a single-center electronic health record (EHR)-based study. The primary outcome was the development of cardiac arrest within 24 h after prediction. Three ANN models were trained: multilayer perceptron (MLP), long-short-term memory (LSTM), and hybrid. These were compared to other classifiers including the modified early warning score (MEWS), logistic regression, and random forest. We used AUROC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the comparison. Results: During the study period, there were a total of 374,605 ED visits and 2,910,321 patient status updates. The ANN models (MLP, LSTM, and hybrid) achieved higher AUROC (AUROC: 0.929, 0.933, and 0.936; 95% confidential interval: 0.926–0.932, 0.930–0.936, and 0.933–0.939, respectively) compared to the non-ANN models, and the hybrid model exhibited the best performance. The ANN classifiers displayed higher performance in most of the test characteristics when the threshold levels of the classifiers were fixed to display the same positive result as those at the three MEWS thresholds (score ≥ 3, ≥4, and ≥5), and when compared with each other. Conclusions: The ANN improves upon MEWS and conventional machine learning algorithms for the prediction of cardiac arrests in emergency departments. The hybrid ANN model utilizing both baseline and sequence information achieved the best performance.

Original languageEnglish
Pages (from-to)43-49
Number of pages7
JournalAmerican Journal of Emergency Medicine
Volume38
Issue number1
DOIs
StatePublished - Jan 2020

Fingerprint

Neural Networks (Computer)
Heart Arrest
Hospital Emergency Service
Long-Term Memory
Short-Term Memory
Electronic Health Records
Logistic Models
Sensitivity and Specificity

Keywords

  • Cardiac arrest
  • Deep learning
  • Early warning system
  • Emergency department
  • Monitoring

Cite this

@article{ea1d99884f774d91b56987484d7c7937,
title = "Developing neural network models for early detection of cardiac arrest in emergency department",
abstract = "Background: Automated surveillance for cardiac arrests would be useful in overcrowded emergency departments. The purpose of this study is to develop and test artificial neural network (ANN) classifiers for early detection of patients at risk of cardiac arrest in emergency departments. Methods: This is a single-center electronic health record (EHR)-based study. The primary outcome was the development of cardiac arrest within 24 h after prediction. Three ANN models were trained: multilayer perceptron (MLP), long-short-term memory (LSTM), and hybrid. These were compared to other classifiers including the modified early warning score (MEWS), logistic regression, and random forest. We used AUROC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the comparison. Results: During the study period, there were a total of 374,605 ED visits and 2,910,321 patient status updates. The ANN models (MLP, LSTM, and hybrid) achieved higher AUROC (AUROC: 0.929, 0.933, and 0.936; 95{\%} confidential interval: 0.926–0.932, 0.930–0.936, and 0.933–0.939, respectively) compared to the non-ANN models, and the hybrid model exhibited the best performance. The ANN classifiers displayed higher performance in most of the test characteristics when the threshold levels of the classifiers were fixed to display the same positive result as those at the three MEWS thresholds (score ≥ 3, ≥4, and ≥5), and when compared with each other. Conclusions: The ANN improves upon MEWS and conventional machine learning algorithms for the prediction of cardiac arrests in emergency departments. The hybrid ANN model utilizing both baseline and sequence information achieved the best performance.",
keywords = "Cardiac arrest, Deep learning, Early warning system, Emergency department, Monitoring",
author = "Jang, {Dong Hyun} and Joonghee Kim and Jo, {You Hwan} and Lee, {Jae Hyuk} and Hwang, {Ji Eun} and Park, {Seung Min} and Lee, {Dong Keon} and Inwon Park and Doyun Kim and Hyunglan Chang",
year = "2020",
month = "1",
doi = "10.1016/j.ajem.2019.04.006",
language = "English",
volume = "38",
pages = "43--49",
journal = "American Journal of Emergency Medicine",
issn = "0735-6757",
publisher = "W.B. Saunders Ltd",
number = "1",

}

Developing neural network models for early detection of cardiac arrest in emergency department. / Jang, Dong Hyun; Kim, Joonghee; Jo, You Hwan; Lee, Jae Hyuk; Hwang, Ji Eun; Park, Seung Min; Lee, Dong Keon; Park, Inwon; Kim, Doyun; Chang, Hyunglan.

In: American Journal of Emergency Medicine, Vol. 38, No. 1, 01.2020, p. 43-49.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Developing neural network models for early detection of cardiac arrest in emergency department

AU - Jang, Dong Hyun

AU - Kim, Joonghee

AU - Jo, You Hwan

AU - Lee, Jae Hyuk

AU - Hwang, Ji Eun

AU - Park, Seung Min

AU - Lee, Dong Keon

AU - Park, Inwon

AU - Kim, Doyun

AU - Chang, Hyunglan

PY - 2020/1

Y1 - 2020/1

N2 - Background: Automated surveillance for cardiac arrests would be useful in overcrowded emergency departments. The purpose of this study is to develop and test artificial neural network (ANN) classifiers for early detection of patients at risk of cardiac arrest in emergency departments. Methods: This is a single-center electronic health record (EHR)-based study. The primary outcome was the development of cardiac arrest within 24 h after prediction. Three ANN models were trained: multilayer perceptron (MLP), long-short-term memory (LSTM), and hybrid. These were compared to other classifiers including the modified early warning score (MEWS), logistic regression, and random forest. We used AUROC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the comparison. Results: During the study period, there were a total of 374,605 ED visits and 2,910,321 patient status updates. The ANN models (MLP, LSTM, and hybrid) achieved higher AUROC (AUROC: 0.929, 0.933, and 0.936; 95% confidential interval: 0.926–0.932, 0.930–0.936, and 0.933–0.939, respectively) compared to the non-ANN models, and the hybrid model exhibited the best performance. The ANN classifiers displayed higher performance in most of the test characteristics when the threshold levels of the classifiers were fixed to display the same positive result as those at the three MEWS thresholds (score ≥ 3, ≥4, and ≥5), and when compared with each other. Conclusions: The ANN improves upon MEWS and conventional machine learning algorithms for the prediction of cardiac arrests in emergency departments. The hybrid ANN model utilizing both baseline and sequence information achieved the best performance.

AB - Background: Automated surveillance for cardiac arrests would be useful in overcrowded emergency departments. The purpose of this study is to develop and test artificial neural network (ANN) classifiers for early detection of patients at risk of cardiac arrest in emergency departments. Methods: This is a single-center electronic health record (EHR)-based study. The primary outcome was the development of cardiac arrest within 24 h after prediction. Three ANN models were trained: multilayer perceptron (MLP), long-short-term memory (LSTM), and hybrid. These were compared to other classifiers including the modified early warning score (MEWS), logistic regression, and random forest. We used AUROC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for the comparison. Results: During the study period, there were a total of 374,605 ED visits and 2,910,321 patient status updates. The ANN models (MLP, LSTM, and hybrid) achieved higher AUROC (AUROC: 0.929, 0.933, and 0.936; 95% confidential interval: 0.926–0.932, 0.930–0.936, and 0.933–0.939, respectively) compared to the non-ANN models, and the hybrid model exhibited the best performance. The ANN classifiers displayed higher performance in most of the test characteristics when the threshold levels of the classifiers were fixed to display the same positive result as those at the three MEWS thresholds (score ≥ 3, ≥4, and ≥5), and when compared with each other. Conclusions: The ANN improves upon MEWS and conventional machine learning algorithms for the prediction of cardiac arrests in emergency departments. The hybrid ANN model utilizing both baseline and sequence information achieved the best performance.

KW - Cardiac arrest

KW - Deep learning

KW - Early warning system

KW - Emergency department

KW - Monitoring

UR - http://www.scopus.com/inward/record.url?scp=85064040562&partnerID=8YFLogxK

U2 - 10.1016/j.ajem.2019.04.006

DO - 10.1016/j.ajem.2019.04.006

M3 - Article

C2 - 30982559

AN - SCOPUS:85064040562

VL - 38

SP - 43

EP - 49

JO - American Journal of Emergency Medicine

JF - American Journal of Emergency Medicine

SN - 0735-6757

IS - 1

ER -