A Machine Learning-Based Web Tool for the Severity Prediction of COVID-19

Area of Science:

• Medical Informatics
• Public Health
• Epidemiology

Background:

• The COVID-19 pandemic highlighted significant variations in patient outcomes.
• Predictive modeling offers a method to understand these outcome disparities.
• Identifying risk factors is crucial for effective disease management.

Purpose of the Study:

• To correlate demographic and clinical patient data with COVID-19 severity.
• To demonstrate the utility of machine learning (ML) in predicting COVID-19 prognosis.
• To develop a web tool for predicting disease outcomes.

Main Methods:

• Enrolled 344 unvaccinated patients with confirmed SARS-CoV-2 infection.
• Integrated data from questionnaires and medical records.
• Applied various classification machine learning algorithms to identify predictive features.

Main Results:

• Identified age, sex, hypertension, obesity, and cancer as significant predictors of severe COVID-19.
• Selected the optimal machine learning algorithm and hyperparameters for prediction.
• Developed a prognostic tool based on 111 independent features.

Conclusions:

• The developed prognostic tool can guide personalized therapeutic strategies for COVID-19 patients.
• The tool may encourage vaccination among vulnerable populations by illustrating potential risks.
• Machine learning is a valuable approach for disease prognosis and personalized medicine.