Predicting the negative conversion time of nonsevere COVID-19 patients using machine learning methods

Area of Science:

• Medical Informatics
• Infectious Diseases
• Machine Learning

Background:

• Non-severe coronavirus disease 2019 (COVID-19) poses a significant public health challenge.
• Predicting negative conversion time is crucial for managing patient flow and preventing transmission, especially during the Omicron variant surge.
• Existing prediction methods may not fully leverage complex clinical and laboratory data.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting the negative conversion time of non-severe COVID-19 patients.
• To identify key clinical and laboratory features that influence COVID-19 negative conversion time.
• To compare the predictive performance of various machine learning algorithms.

Main Methods:

• Retrospective analysis of 376 non-severe COVID-19 patients.
• Feature selection using the least absolute shrinkage and selection operator (LASSO).
• Development and comparison of six machine learning models: MLR, KNNR, RFR, SVR, XGBR, and MLPR.
• Evaluation of model performance on a dedicated test set.

Main Results:

• Seven predictive features were identified: age, gender, vaccination status, IgG, lymphocyte ratio, monocyte ratio, and lymphocyte count.
• The Multilayer Perceptron Regression (MLPR) model demonstrated the strongest predictive performance and generalization ability.
• Vaccination status, IgG, lymphocyte count, and lymphocyte ratio were identified as protective factors, while male gender, age, and monocyte ratio were risk factors.

Conclusions:

• Machine learning, particularly MLPR, offers an effective approach for predicting COVID-19 negative conversion time in non-severe cases.
• Key predictors include vaccination status, gender, and IgG levels.
• Accurate prediction can optimize medical resource allocation and enhance disease control strategies.