Time Windows Voting Classifier for COVID-19 Mortality Prediction

  1. Tiong-Thye Goh
  2. MengJun Liu

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

The ability to predict COVID-19 patients’ level of severity (death or survival) enables clinicians to prioritise treatment. Recently, using three blood biomarkers, an interpretable machine learning model was developed to predict the mortality of COVID-19 patients. The method was reported to be suffering from performance stability because the identified biomarkers are not consistent predictors over an extended duration.

Methods

To sustain performance, the proposed method partitioned data into three different time windows. For each window, an end-classifier, a mid-classifier and a front-classifier were designed respectively using the XGboost single tree approach. These time window classifiers were integrated into a majority vote classifier and tested with an isolated test data set.

Results

The voting classifier strengthens the overall performance of 90% cumulative accuracy from a 14 days window to a 21 days prediction window.

Conclusions

An additional 7 days of prediction window can have a considerable impact on a patient’s chance of survival. This study validated the feasibility of the time window voting classifier and further support the selection of biomarkers features set for the early prognosis of patients with a higher risk of mortality.

Article activity feed

  1. ScreenIT

    SciScore for 10.1101/2021.07.02.21259934: (What is this?)

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

    Table 1: Rigor

    Ethicsnot detected.
    Sex as a biological variablenot detected.
    Randomizationnot detected.
    Blindingnot detected.
    Power Analysisnot detected.

    Table 2: Resources

    Software and Algorithms
    SentencesResources
    Statistical analyses were computed using SPSS (version 24).
    SPSS
    suggested: (SPSS, RRID:SCR_002865)

    Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    Our study comes with certain limitations. First, the samples are retrospective and consist of 375 cases with 201 (53.6%) who survived and 174 (46.4%) who died. While the small sample size limits generalisation, it does provide a direction for further analysis when larger data sets become available. Second, a more comprehensive analysis is needed to establish the stability of the algorithm. Our method may have reduced the instability, but we do need further evidence in support. Third, the data distribution with respect to mortality would affect the performance of the algorithm which requires further validation that takes hospital context into consideration (Wang et al., 2020).

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

    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.
    • No funding statement was detected.
    • No protocol registration statement was detected.

    Results from scite Reference Check: We found no unreliable references.

    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 10.1101/2021.07.02.21259934v1 on medRxiv