Real-time infectious disease endurance indicator system for scientific decisions using machine learning and rapid...

Area of Science:

• Medical Informatics
• Computational Biology
• Epidemiology

Background:

• COVID-19, a global pandemic caused by SARS-CoV-2, necessitates accurate mortality prediction for effective patient management.
• Basic blood tests offer a widely accessible alternative to complex imaging for predicting patient outcomes.
• Identifying key predictive factors in blood work is crucial for improving therapeutic strategies.

Purpose of the Study:

• To develop and validate a machine learning model for predicting COVID-19 mortality using routine blood test data.
• To identify the most significant blood biomarkers and demographic factors associated with COVID-19 mortality.
• To assess the model's predictive accuracy and timeliness for clinical application.

Main Methods:

• Utilized machine learning (ML) methodologies, specifically XGBoost feature importance and neural network classification.
• Analyzed routine blood test results, including lactate dehydrogenase (LDH), lymphocytes, neutrophils, and high-sensitivity C-reactive protein (hs-CRP), along with patient age.
• Validated the model's predictive performance using three distinct datasets based on days to outcome.

Main Results:

• A combination of five factors (age, LDH, lymphocytes, neutrophils, hs-CRP) accurately predicted mortality in 96% of cases.
• The optimal ML model achieved high accuracy and a 90% precision rate up to 16 days prior to the mortality event.
• Model performance was consistently demonstrated across different timeframes to outcome.

Conclusions:

• Machine learning models leveraging accessible blood test data can effectively predict COVID-19 mortality.
• The identified biomarkers provide valuable insights for early risk stratification and targeted treatment strategies.
• This approach offers a practical, timely, and reliable tool to support clinical decision-making in infectious disease management.