Prediction and Classification of COVID-19 Admissions to Intensive Care Units (ICU) Using Weighted Radial Kernel SVM...

Area of Science:

• Computational biology
• Medical informatics
• Machine learning applications in healthcare

Background:

• COVID-19 pandemic has strained healthcare systems globally.
• Accurate prediction of ICU admissions is crucial for resource allocation.
• Machine learning offers potential solutions for clinical decision support.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting ICU admissions in COVID-19 patients.
• To identify key clinical and laboratory features that distinguish ICU from non-ICU patients.
• To assess the performance of a weighted radial kernel SVM with Recursive Feature Elimination (RFE).

Main Methods:

• Retrospective analysis of clinical and laboratory data from 100 COVID-19 patients.
• Application of weighted radial kernel Support Vector Machine (SVM) coupled with Recursive Feature Elimination (RFE).
• Comparison with Linear Discriminant Analysis (LDA) and other SVM variants (linear, polynomial kernels).

Main Results:

• The weighted radial kernel SVM with RFE demonstrated superior performance in discriminating between ICU and non-ICU admissions.
• RFE identified significant predictors including patient weight, PCR Ct Value, CCL19, INF-β, BLC, INR, PT, PTT, CKMB, HB, platelets, RBC, urea, creatinine, and albumin.
• The model successfully distinguished ICU from non-ICU COVID-19 patients based on these features.

Conclusions:

• Weighted radial kernel SVM with RFE is a promising tool for predicting COVID-19 ICU admissions.
• This ML approach can aid hospital decision-makers in optimizing resource allocation.
• Further prospective validation is recommended for real-world application.