Comparing different machine learning techniques for predicting COVID-19 severity.

Area of Science:

• Medical Informatics
• Machine Learning in Medicine
• Computational Biology

Background:

• The ongoing global spread of Coronavirus disease 2019 (COVID-19) necessitates effective tools for patient management.
• Machine learning (ML) has shown promise in disease diagnosis and treatment outcome prediction.
• Predicting COVID-19 severity at admission is crucial for timely intervention and resource allocation.

Purpose of the Study:

• To evaluate the performance of different machine learning techniques in predicting COVID-19 severity upon admission.
• To identify key features that are important predictors of severe COVID-19.
• To compare the predictive capabilities of Random Forest (RF), Support Vector Machine (SVM), and Logistic Regression (LR) models.

Main Methods:

• A retrospective study was conducted using data from JinYinTan Hospital (January 26 - March 28, 2020).
• Eighty-six demographic, clinical, and laboratory features were selected using LassoCV, correlation analysis, expert opinions, and literature review.
• RF, SVM, and LR models were trained to predict severe COVID-19, with performance evaluated using the Area Under the Curve (AUC).

Main Results:

• A total of 287 patients were analyzed, with 36.6% classified as severe cases.
• The Random Forest model achieved the highest AUC (0.970), outperforming SVM (0.948) and LR (0.928).
• Key predictors for severe COVID-19 included chest CT findings, neutrophil-to-lymphocyte ratio, lactate dehydrogenase, and D-dimer levels.

Conclusions:

• The Random Forest model is a highly effective tool for predicting severe COVID-19 at admission.
• Accurate prediction of COVID-19 severity can facilitate optimized patient care and resource management.
• Identifying critical predictive features aids in understanding disease progression and potential therapeutic targets.