Severity Prediction for COVID-19 Patients via Recurrent Neural Networks

  1. Junghwan Lee
  2. Jae Hyun Kim
  3. Casey Ta
  4. Cong Liu
  5. Chunhua Weng

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

The novel coronavirus disease-2019 (COVID-19) pandemic has threatened the health of tens of millions of people worldwide and posed enormous burden on the global healthcare systems. Many prediction models have been proposed to fight against the pandemic. In this paper, we propose a model to predict whether a patient infected with COVID-19 will develop severe outcomes based only on the patient’s historical electronic health records (EHR) using recurrent neural networks (RNN). The predicted severity risk score represents the probability for a person to progress into severe status (mechanical ventilation, tracheostomy, or death) after being infected with COVID-19. While many of the existing models use features obtained after diagnosis of COVID-19, our proposed model only utilizes a patient’s historical EHR so that it can enable proactive risk management before or at the time of hospital admission.

Article activity feed

  1. ScreenIT

    SciScore for 10.1101/2020.08.28.20184200: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    Institutional Review Board StatementIRB: This work received institutional review board approval (AAAR3954) with a waiver for informed consent.
    Consent: This work received institutional review board approval (AAAR3954) with a waiver for informed consent.
    Randomizationnot detected.
    Blindingnot detected.
    Power Analysisnot detected.
    Sex as a biological variableA patient’s demographic information vector is a simple concatenation of one-hot encoded sex (i.e. [1, 0] for male and [0, 1] for female) and min-max normalized age of the patient.

    Table 2: Resources

    No key resources detected.

    Results from OddPub: Thank you for sharing your code.

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    We believe obtaining a larger size of data across different institutions and nations or using other disease cohorts as proxy cohorts will resolve this limitation. One advantage of our approach is that our analysis used a standardized clinical data format, the OMOP Common Data Model. The source code for this analysis can be easily shared with others who have similarly formatted clinical data for evidence aggregation. While higher accuracies (0.73-0.99) were reported in other studies, the intended use of these models were often not clearly described7. The RNN model we propose is intended to aid decision making at the time of or before hospital admission due to COVID-19, since only historical EHR data were needed in the model. In addition, the RNN model can be applied to the general population that is not confirmed COVID-19 positive to identify people at high risk of developing potential severe outcomes if infected by COVID-19. The RNN model can readily be applied to the situations above with much larger datasets or in a real-time setting since it can compute the risk score of the patient in a small amount of time. We demonstrated the effectiveness of the risk score predicted by the RNN model by analyzing it with basic characteristics (i.e., age and total historical visit count) of the patients. From Figure 3a, we can confirm that the risk score is correlated with the patient developing severe outcomes from COVID-19. We found that there exists a statistically significant positiv...

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We found bar graphs of continuous data. We recommend replacing bar graphs with more informative graphics, as many different datasets can lead to the same bar graph. The actual data may suggest different conclusions from the summary statistics. For more information, please see Weissgerber et al (2015).

    Results from JetFighter: We did not find any issues relating to colormaps.

    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.

  2. Version published to 10.1101/2020.08.28.20184200v2 on medRxiv
  3. Version published to 10.1101/2020.08.28.20184200v1 on medRxiv