CoCoMoRP

CoCoMoRP predicts mortality risk among confirmed COVID-19 patients using machine learning to inform community-level triage and resource allocation.

Key Features:

• Machine Learning Framework: Evaluates five algorithms—logistic regression, support vector machine, K nearest neighbor, random forest, and gradient boosting—for mortality risk prediction.
• Dataset: Uses publicly available South Korea COVID-19 surveillance data from January 20, 2020 to April 07, 2020 comprising 3,022 confirmed patients.
• Algorithm Performance: The gradient boosting model achieved area under the ROC curve (AUC)=0.966, Matthews Correlation Coefficient=0.656, Brier Score=0.013, and accuracy=0.987.

Scientific Applications:

• Prognostic Modeling: Provides a validated prognostic model for COVID-19 mortality risk using evaluated machine learning approaches and surveillance data.
• Clinical Decision Support: Identifies individuals at higher risk of mortality to inform patient management and targeted interventions.
• Public Health Policy: Informs community-level public health strategies and resource distribution for COVID-19 response.

Methodology:

Collected and analyzed data from 3,022 confirmed COVID-19 patients; applied five machine learning algorithms (logistic regression, support vector machine, K nearest neighbor, random forest, gradient boosting) to predict mortality; and performed comparative evaluation of algorithm performance using discrimination, calibration, and predictive ability metrics.