Comprehensive analysis of clinical data for COVID-19 outcome estimation with machine learning models

Abstract

COVID-19 is a global threat for the healthcare systems due to the rapid spread of the pathogen that causes it. In such situation, the clinicians must take important decisions, in an environment where medical resources can be insufficient. In this task, the computer-aided diagnosis systems can be very useful not only in the task of supporting the clinical decisions but also to perform relevant analyses, allowing them to understand better the disease and the factors that can identify the high risk patients. For those purposes, in this work, we use several machine learning algorithms to estimate the outcome of COVID-19 patients given their clinical information. Particularly, we perform 2 different studies: the first one estimates whether the patient is at low or at high risk of death whereas the second estimates if the patient needs hospitalization or not. The results of the analyses of this work show the most relevant features for each studied scenario, as well as the classification performance of the considered machine learning models. In particular, the XGBoost algorithm is able to estimate the need for hospitalization of a patient with an AUC-ROC of $0.8415\pm 0.0217$ while it can also estimate the risk of death with an AUC-ROC of $0.7992\pm 0.0104$ . Results have demonstrated the great potential of the proposal to determine those patients that need a greater amount of medical resources for being at a higher risk. This provides the healthcare services with a tool to better manage their resources.

Keywords: COVID-19; Classification; Clinical data; Feature selection; Machine learning.

Graphical abstract

Fig. 1

Overall description of the pipeline of the proposed methodology.

Fig. 2

Description of the method that was used to address the problem of imbalancing that exists in the original dataset.

Fig. 3

Representative graphical results obtained in the experiment I. (a) Representative confusion matrix obtained with the most appropriate classifier for this experiment (XGBoost) training with all the features. (b) ROC curves of the classifiers used in this experiment, training with the whole set of features.

Fig. 4

Ranking of features according to the score given by each feature selection method for the experiment I (estimation of Non-Hospitalization/Hospitalization). The $x$ axes are displayed in logarithmic scale to improve the visualization of the differences among the scores. Those features with a negligible score were removed from the chart. (a) Fisher Scoring. (b) Mutual Information. (c) Variance-based Ranking.

Fig. 5

Evolution of the AUC-ROC values for the problem of classifying Non-Hospitalized/Hospitalized (experiment I) given the number of features using the XGBoost algorithm.

Fig. 6

Remarkable graphical results of the experiment II. (a) Confusion matrix representative of the most appropriate classifier for this experiment (decision tree) trained with the whole set of features. (b) ROC curves of the classifiers used in this experiment after training with all the features.

Fig. 7

Ranking of features according to the score given by each feature selection method for the experiment II (estimation of Survival/Death). The $x$ axes are shown in logarithmic scale to improve the visualization of the differences among the scores. Those features with a negligible score were removed from the chart. (a) Fisher Scoring. (b) Mutual Information. (c) Variance-based Ranking.

Fig. 8

Evolution of the AUC-ROC values for the problem of classifying Survival/Death (experiment II) given the number of features using the XGBoost algorithm.