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

Area of Science:

• Medical Informatics
• Computational Biology
• Public Health

Background:

• The COVID-19 pandemic poses a significant global health challenge, straining healthcare systems.
• Clinical decision-making during the pandemic is complex, often requiring resource management amidst scarcity.
• Computer-aided diagnosis systems offer potential for clinical support and disease analysis.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting COVID-19 patient outcomes.
• To identify clinical factors associated with high risk of death and hospitalization in COVID-19 patients.
• To support healthcare resource management by identifying patients requiring intensive medical care.

Main Methods:

• Utilized several machine learning algorithms to analyze clinical data from COVID-19 patients.
• Conducted two primary studies: predicting risk of death and predicting need for hospitalization.
• Evaluated model performance using metrics such as AUC-ROC.

Main Results:

• Identified key clinical features relevant to patient outcomes in both risk of death and hospitalization prediction.
• The XGBoost algorithm demonstrated strong performance in predicting hospitalization (AUC-ROC $$) and risk of death (AUC-ROC $$).
• Results highlight the potential for machine learning to accurately stratify patient risk.

Conclusions:

• Machine learning models show significant promise in predicting COVID-19 patient outcomes.
• Accurate prediction of patient risk can enhance clinical decision-making and optimize healthcare resource allocation.
• This approach provides a valuable tool for healthcare services to manage resources effectively during health crises.