Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments

Amin Naemi; Thomas Schmidt; Marjan Mansourvar; Uffe Kock Wiil

doi:10.3233/SHTI200713

Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments

Amin Naemi, Thomas Schmidt, Marjan Mansourvar, Uffe Kock Wiil

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

97 Downloads (Pure)

Abstract

Early detection of deterioration at hospitals could be beneficial in terms of reducing mortality and morbidity rates and costs. In this paper, we present a model based on Long Short-Term Memory (LSTM) neural network used in deep learning to predict the illness severity of patients in advance. Hence, by predicting health severity, this model can be used to identify deteriorating patients. Our proposed model utilizes continuous monitored vital signs, including heart rate, respiratory rate, oxygen saturation, and blood pressure automatically collected from patients during hospitalization. In this study, a short-time prediction using a sliding window approach is applied. The performance of the proposed model was compared with the Multi-Layer Perceptron (MLP) neural network, a feedforward class of neural network, based on R2 score and Root Mean Square Error (RMSE) metrics. The results showed that the LSTM has a better performance and could predict the illness severity of patients more accurately.

Original language	English
Title of host publication	Integrated Citizen Centered Digital Health and Social Care
Editors	Alpo Värri, Jaime Delgado, Parisis Gallos, Maria Hägglund, Kristiina Häyrinen, Ulla-Mari Kinnunen, Louise B. Pape-Haugaard, Laura-Maria Peltonen, Kaija Saranto, Philip Scott
Publisher	IOS Press
Publication date	2020
Pages	152-156
ISBN (Print)	978-1-64368-144-3
ISBN (Electronic)	978-1-64368-145-0
DOIs	https://doi.org/10.3233/SHTI200713
Publication status	Published - 2020
Event	2020 Special Topic Conference of the European Federation for Medical Informatics - Duration: 26. Nov 2020 → 27. Nov 2020

Conference

Conference	2020 Special Topic Conference of the European Federation for Medical Informatics
Period	26/11/2020 → 27/11/2020

Series	Studies in Health Technology and Informatics
Volume	275
ISSN	0926-9630

Keywords

clinical deterioration
emergency department
health informatics
LSTM
machine learning algorithms
recurrent neural network
time series
Neural Networks, Computer
Emergency Service, Hospital
Humans
Clinical Deterioration
Early Diagnosis

Documents & Links

10.3233/SHTI200713Licence: CC BY-NC

Open Access VersionFinal published version, 190 KBLicence: CC BY-NC

1 Ph.D. thesis

Predictive Models for Identifying Clinical Deterioration
Naemi, A., 9. May 2022, Syddansk Universitet. Det Tekniske Fakultet. 248 p.
Research output: Thesis › Ph.D. thesis

Cite this

Naemi, A., Schmidt, T., Mansourvar, M., & Wiil, U. K. (2020). Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments. In A. Värri, J. Delgado, P. Gallos, M. Hägglund, K. Häyrinen, U.-M. Kinnunen, L. B. Pape-Haugaard, L.-M. Peltonen, K. Saranto, & P. Scott (Eds.), Integrated Citizen Centered Digital Health and Social Care (pp. 152-156). IOS Press. https://doi.org/10.3233/SHTI200713

Naemi, Amin ; Schmidt, Thomas ; Mansourvar, Marjan et al. / Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments. Integrated Citizen Centered Digital Health and Social Care. editor / Alpo Värri ; Jaime Delgado ; Parisis Gallos ; Maria Hägglund ; Kristiina Häyrinen ; Ulla-Mari Kinnunen ; Louise B. Pape-Haugaard ; Laura-Maria Peltonen ; Kaija Saranto ; Philip Scott. IOS Press, 2020. pp. 152-156 (Studies in Health Technology and Informatics, Vol. 275).

@inproceedings{9f531faf02ed453b89570adc0a76009e,

title = "Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments",

abstract = "Early detection of deterioration at hospitals could be beneficial in terms of reducing mortality and morbidity rates and costs. In this paper, we present a model based on Long Short-Term Memory (LSTM) neural network used in deep learning to predict the illness severity of patients in advance. Hence, by predicting health severity, this model can be used to identify deteriorating patients. Our proposed model utilizes continuous monitored vital signs, including heart rate, respiratory rate, oxygen saturation, and blood pressure automatically collected from patients during hospitalization. In this study, a short-time prediction using a sliding window approach is applied. The performance of the proposed model was compared with the Multi-Layer Perceptron (MLP) neural network, a feedforward class of neural network, based on R2 score and Root Mean Square Error (RMSE) metrics. The results showed that the LSTM has a better performance and could predict the illness severity of patients more accurately.",

keywords = "clinical deterioration, emergency department, health informatics, LSTM, machine learning algorithms, recurrent neural network, time series, Neural Networks, Computer, Emergency Service, Hospital, Humans, Clinical Deterioration, Early Diagnosis",

author = "Amin Naemi and Thomas Schmidt and Marjan Mansourvar and Wiil, {Uffe Kock}",

year = "2020",

doi = "10.3233/SHTI200713",

language = "English",

isbn = "978-1-64368-144-3",

series = "Studies in Health Technology and Informatics",

publisher = "IOS Press",

pages = "152--156",

editor = "Alpo V{\"a}rri and Jaime Delgado and Parisis Gallos and H{\"a}gglund, {Maria } and Kristiina H{\"a}yrinen and Ulla-Mari Kinnunen and Pape-Haugaard, {Louise B.} and Laura-Maria Peltonen and Kaija Saranto and Philip Scott",

booktitle = "Integrated Citizen Centered Digital Health and Social Care",

address = "Netherlands",

note = "2020 Special Topic Conference of the European Federation for Medical Informatics, EFMI STC 2020 ; Conference date: 26-11-2020 Through 27-11-2020",

}

Naemi, A , Schmidt, T, Mansourvar, M & Wiil, UK 2020, Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments. in A Värri, J Delgado, P Gallos, M Hägglund, K Häyrinen, U-M Kinnunen, LB Pape-Haugaard, L-M Peltonen, K Saranto & P Scott (eds), Integrated Citizen Centered Digital Health and Social Care. IOS Press, Studies in Health Technology and Informatics, vol. 275, pp. 152-156, 2020 Special Topic Conference of the European Federation for Medical Informatics, 26/11/2020. https://doi.org/10.3233/SHTI200713

Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments. / Naemi, Amin ; Schmidt, Thomas; Mansourvar, Marjan et al.
Integrated Citizen Centered Digital Health and Social Care. ed. / Alpo Värri; Jaime Delgado; Parisis Gallos; Maria Hägglund; Kristiina Häyrinen; Ulla-Mari Kinnunen; Louise B. Pape-Haugaard; Laura-Maria Peltonen; Kaija Saranto; Philip Scott. IOS Press, 2020. p. 152-156 (Studies in Health Technology and Informatics, Vol. 275).

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments

AU - Naemi, Amin

AU - Schmidt, Thomas

AU - Mansourvar, Marjan

AU - Wiil, Uffe Kock

PY - 2020

Y1 - 2020

N2 - Early detection of deterioration at hospitals could be beneficial in terms of reducing mortality and morbidity rates and costs. In this paper, we present a model based on Long Short-Term Memory (LSTM) neural network used in deep learning to predict the illness severity of patients in advance. Hence, by predicting health severity, this model can be used to identify deteriorating patients. Our proposed model utilizes continuous monitored vital signs, including heart rate, respiratory rate, oxygen saturation, and blood pressure automatically collected from patients during hospitalization. In this study, a short-time prediction using a sliding window approach is applied. The performance of the proposed model was compared with the Multi-Layer Perceptron (MLP) neural network, a feedforward class of neural network, based on R2 score and Root Mean Square Error (RMSE) metrics. The results showed that the LSTM has a better performance and could predict the illness severity of patients more accurately.

AB - Early detection of deterioration at hospitals could be beneficial in terms of reducing mortality and morbidity rates and costs. In this paper, we present a model based on Long Short-Term Memory (LSTM) neural network used in deep learning to predict the illness severity of patients in advance. Hence, by predicting health severity, this model can be used to identify deteriorating patients. Our proposed model utilizes continuous monitored vital signs, including heart rate, respiratory rate, oxygen saturation, and blood pressure automatically collected from patients during hospitalization. In this study, a short-time prediction using a sliding window approach is applied. The performance of the proposed model was compared with the Multi-Layer Perceptron (MLP) neural network, a feedforward class of neural network, based on R2 score and Root Mean Square Error (RMSE) metrics. The results showed that the LSTM has a better performance and could predict the illness severity of patients more accurately.

KW - clinical deterioration

KW - emergency department

KW - health informatics

KW - LSTM

KW - machine learning algorithms

KW - recurrent neural network

KW - time series

KW - Neural Networks, Computer

KW - Emergency Service, Hospital

KW - Humans

KW - Clinical Deterioration

KW - Early Diagnosis

U2 - 10.3233/SHTI200713

DO - 10.3233/SHTI200713

M3 - Article in proceedings

C2 - 33227759

AN - SCOPUS:85096738139

SN - 978-1-64368-144-3

T3 - Studies in Health Technology and Informatics

SP - 152

EP - 156

BT - Integrated Citizen Centered Digital Health and Social Care

A2 - Värri, Alpo

A2 - Delgado, Jaime

A2 - Gallos, Parisis

A2 - Hägglund, Maria

A2 - Häyrinen, Kristiina

A2 - Kinnunen, Ulla-Mari

A2 - Pape-Haugaard, Louise B.

A2 - Peltonen, Laura-Maria

A2 - Saranto, Kaija

A2 - Scott, Philip

PB - IOS Press

T2 - 2020 Special Topic Conference of the European Federation for Medical Informatics

Y2 - 26 November 2020 through 27 November 2020

ER -

Naemi A , Schmidt T, Mansourvar M, Wiil UK. Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments. In Värri A, Delgado J, Gallos P, Hägglund M, Häyrinen K, Kinnunen UM, Pape-Haugaard LB, Peltonen LM, Saranto K, Scott P, editors, Integrated Citizen Centered Digital Health and Social Care. IOS Press. 2020. p. 152-156. (Studies in Health Technology and Informatics, Vol. 275). doi: 10.3233/SHTI200713

Personalized Predictive Models for Identifying Clinical Deterioration Using LSTM in Emergency Departments

Abstract

Conference

Keywords

Documents & Links

Fingerprint

Related research output

Predictive Models for Identifying Clinical Deterioration

Cite this