Mortality Prediction Utilizing Blood Biomarkers to Predict the Severity of COVID-19 Using Machine Learning Technique

Area of Science:

• Medical Informatics
• Clinical Prediction Models
• Public Health

Background:

• Accurate mortality prediction is crucial for resource allocation in severe illnesses like COVID-19.
• Existing models often lack external validation or focus solely on COVID-19 patients.

Purpose of the Study:

• To develop and validate an early prediction model for high mortality risk in both COVID-19 and non-COVID-19 patients.
• To identify key clinical parameters for predicting hospital death.

Main Methods:

• Retrospective analysis of two hospital datasets (Boston, USA and Tongji, China).
• Logistic regression model to identify predictors.
• Development of a nomogram-based scoring technique using Age, Lymphocyte count, D-dimer, CRP, and Creatinine (ALDCC).

Main Results:

• The ALDCC score identified key predictors for hospital mortality.
• The model achieved high performance with Area Under the Curve (AUC) values of 0.987 (development), 0.999 (internal validation), and 0.992 (external validation).
• Patients were categorized into Low, Moderate, and High mortality risk groups based on ALDCC score and probability.

Conclusions:

• The ALDCC prognostic model and score enable early identification of high-risk patients (COVID-19 and non-COVID-19).
• This tool can assist physicians in improving patient management and resource prioritization.
• The model demonstrates robust performance in external validation, suggesting generalizability.