Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Accepted for/Published in: JMIR Formative Research

Date Submitted: Feb 17, 2021
Date Accepted: Aug 1, 2021
Date Submitted to PubMed: Aug 16, 2021

The final, peer-reviewed published version of this preprint can be found here:

Semisupervised Deep Learning Techniques for Predicting Acute Respiratory Distress Syndrome From Time-Series Clinical Data: Model Development and Validation Study

Lam C, Pellegrini E, Iqbal Z, Green-Saxena A, Hoffman J, Calvert J, Mao Q, Das R

Semisupervised Deep Learning Techniques for Predicting Acute Respiratory Distress Syndrome From Time-Series Clinical Data: Model Development and Validation Study

JMIR Form Res 2021;5(9):e28028

DOI: 10.2196/28028

PMID: 34398784

PMCID: 8447921

Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.

Accurate ARDS prediction from standard time series clinical data using gradient boosted trees and semi supervised deep learning

  • Carson Lam; 
  • Emily Pellegrini; 
  • Zohora Iqbal; 
  • Abigail Green-Saxena; 
  • Jana Hoffman; 
  • Jacob Calvert; 
  • Qingqing Mao; 
  • Ritankar Das

ABSTRACT

Background:

A high number of patients hospitalized with COVID-19 also develop Acute Respiratory Distress Syndrome (ARDS).

Objective:

In response to the need for clinical decision support tools to help manage the pandemic, we developed a semi-supervised machine learning (SSL) algorithm to predict ARDS in general and COVID-19 populations.

Methods:

SSL techniques were applied to 53,432 encounters from patients admitted to the intensive care unit at a large academic health center between 2001-2012. SSL techniques were also applied to 45,507 COVID-19 encounters from patients admitted to ten United States hospitals from 2019-2020. A baseline XGBoost (XGB) model and an evolving recurrent neural network (RNN) model trained using SSL were applied to the task of predicting ARDS at 12-, 24-, and 48- hours prior to onset. Model performance was assessed with regard to area under the receiver operating characteristic (AUROC).

Results:

Higher performance was achieved by the SSL as compared to the baseline XGB model in the general ICU population and much higher performance was achieved by the SSL in the COVID-19 positive population. The SSL demonstrated AUROCs of 0.96 for all time frames while baseline XGB demonstrated AUROCs of 0.75 in the same time frames. A cluster analysis demonstrated that ARDS in COVID-19 populations presented dissimilarly from ARDS in general populations.

Conclusions:

An SSL model outperformed an XGB model for early ARDS prediction in both general and COVID-19 patient populations. This study supports that ARDS in COVID-19 patients warrants further research as a distinct clinical event. Clinical Trial: N/A


 Citation

Please cite as:

Lam C, Pellegrini E, Iqbal Z, Green-Saxena A, Hoffman J, Calvert J, Mao Q, Das R

Semisupervised Deep Learning Techniques for Predicting Acute Respiratory Distress Syndrome From Time-Series Clinical Data: Model Development and Validation Study

JMIR Form Res 2021;5(9):e28028

DOI: 10.2196/28028

PMID: 34398784

PMCID: 8447921

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.