Age-Stratified Analysis of COVID-19 Outcome Using Machine Learning Predictive Models

Area of Science:

• Medical Informatics
• Computational Biology
• Epidemiology

Background:

• The COVID-19 pandemic overwhelmed global health systems.
• Clinical laboratory tests are crucial for assessing disease severity and mortality risk.

Purpose of the Study:

• To apply predictive models to COVID-19 outcome data.
• To stratify patients by age and analyze mortality risk factors.

Main Methods:

• Utilized three COVID-19 patient datasets.
• Employed the extreme gradient boosting (XGBoost) algorithm for prediction.
• Applied SHAP (Shapley additive explanations) for feature importance analysis.

Main Results:

• XGBoost achieved excellent predictive performance.
• Achieved an Area Under the Receiving Operator Characteristic Curve (AUROC) of 0.97 for patients up to 65 years old.
• Identified key features influencing COVID-19 outcomes.

Conclusions:

• Predictive modeling, particularly XGBoost, is effective for COVID-19 outcome prediction.
• Age stratification improves model accuracy for risk assessment.
• Feature importance analysis enhances understanding of disease progression.