Risk assessment in COVID-19 patients: A multiclass classification approach.

Area of Science:

• Biomedicine
• Public Health
• Data Science

Background:

• Understanding SARS-CoV-2 infection is crucial for identifying severe COVID-19 risk factors.
• Artificial Intelligence (AI) and Machine Learning (ML) show promise in biomedical applications.
• Pandemic research benefits from advanced computational approaches.

Purpose of the Study:

• To apply ML algorithms for multiclass classification of COVID-19 patient risks.
• To identify key factors associated with different COVID-19 disease severities.
• To develop predictive models for COVID-19 complications.

Main Methods:

• Utilized a dataset from the Mexican Ministry of Health.
• Implemented three ML algorithms: Random Forest, GBM, and XGBoost.
• Performed multiclass classification for risk detection.

Main Results:

• Achieved high accuracy with Random Forest (89.86%), GBM (89.37%), and XGBoost (89.97%).
• Identified significant risk factors: sex, age, symptom duration, dyspnea, polypnea, diabetes, and hypertension.
• Established a framework for predicting individual and group risk.

Conclusions:

• ML models can effectively predict COVID-19 severity and associated risks.
• Demographic, clinical, and comorbidity factors are critical predictors.
• Findings support targeted interventions and personalized treatment strategies.