Machine learning approaches in Covid-19 severity risk prediction in Morocco.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• COVID-19 Research

Background:

• COVID-19 severity prediction is crucial for resource allocation.
• Limited research exists on machine learning for COVID-19 severity in Morocco.
• Combining biological and non-biological data offers a novel approach.

Purpose of the Study:

• To develop and validate machine learning models for predicting COVID-19 severity.
• To identify key clinical and biological indicators associated with severe illness.
• To aid healthcare facilities in prioritizing patient care and hospital admission.

Main Methods:

• Utilized COVID-19 patient data (337 cases) from Cheikh Zaid Hospital.
• Employed data reduction algorithms and machine learning classifiers (X_GBoost, AdaBoost, Random Forest, ExtraTrees).
• Implemented a novel feature engineering method using Uniform Manifold Approximation and Projection (UMAP).

Main Results:

• C-reactive protein, platelets, and D-dimers were identified as key predictors of severity.
• The UMAP-based feature engineering achieved 100% accuracy, specificity, and sensitivity.
• Machine learning models demonstrated high prognostic prediction capabilities.

Conclusions:

• The developed machine learning models show exceptional performance in predicting COVID-19 severity.
• This predictive tool can significantly enhance hospital resource management and patient triage.
• This study represents a pioneering effort in applying machine learning for COVID-19 severity assessment in Morocco.