An interpretable mortality prediction model for COVID-19 patients – alternative approach

  1. Peter Gemmar

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

The pandemic spread of coronavirus leads to increased burden on healthcare services worldwide. Experience shows that required medical treatment can reach limits at local clinics and fast and secure clinical assessment of the disease severity becomes vital. In [1] a model is presented for predicting the mortality of COVID-19 patients from their biomarkers. Three biomarkers have been selected by ranking with a supervised Multi-tree XGBoost classifier. The prediction model is built up as a binary decision tree with depth three and achieves AUC scores of up to 97.84±0.37 and 95.06± 2.21 for training and external test data sets, resp.

In human assessment and decision making influencing parameters usually aren’t considered as sharp numbers but rather as Fuzzy terms [2], and inferencing primarily yields Fuzzy terms or continuous grades rather than binary decisions. Therefore, I examined a Sugenotype Fuzzy classifier [3] for disease assessment and decision support. In addition, I used an artificial neural network ( SOM , [4]) for selecting the biomarkers. Modelling and validation was done with the identical data base provided by [1]. With the complete training and test data sets, the Fuzzy prediction model achieves improved AUC scores of up to 98.59 or 95.12 The improvements with the Fuzzy classifier obviously become clear as physicians can interpret output grades to belong to positive or negative class more or less strongly. An extension of the Fuzzy model, which takes into account the trend in key features over time, provides excellent results with the training data, which, however, could not be finally verified due to the lack of suitable test data. The generation and training of the Fuzzy models was fully automatic and without additional adjustment with the help of ANFIS from Matlab©.

Article activity feed

  1. ScreenIT

    SciScore for 10.1101/2020.06.14.20130732: (What is this?)

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

    Table 1: Rigor

    Institutional Review Board Statementnot detected.
    Randomizationnot detected.
    Blindingnot detected.
    Power Analysisnot detected.
    Sex as a biological variablenot detected.

    Table 2: Resources

    No key resources detected.

    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: An explicit section about the limitations of the techniques employed in this study was not found. We encourage authors to address study limitations.

    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.
    • 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 10.1101/2020.06.14.20130732v1 on medRxiv