Using machine learning techniques to predict COVID-19 patient outcomes

Abstract EN

The recent outbreak of coronavirus disease 2019 (COVID-19) has affected human life to a great extent.

The pandemic has led to severe global socioeconomic disruption, causing the postponement or cancellation of major events.

COVID-19 is a novel and accelerating outbreak.

The virus spreads very quickly and has reached almost every part of the world with more than 100 million reported cases, and critical healthcare resources are limited.

Therefore, it is very important to predict which patients are most likely to develop severe illness and face the greatest risks of complications, including death.

In this study, we compare the performances of different machine learning algorithms in predicting COVID-19 patient outcomes based on combinations of patient risk factors.

The bestperforming algorithm is the random forest classifier, which achieves an F-score of 0.788 and an accuracy rate of 0.789.

The proposed model in this study is able to predict the outcome (i.e., dead, discharged, or stable) for any patient diagnosed with COVID-19 by using the same set of risk factors, namely, gender, country, symptoms, and chronic diseases.

The findings of this study can supplement clinical skills and assist doctors in predicting unexpected patterns to identify mild cases among diagnosed patients and the few cases that will progress to severe illness