Predicting the COVID-19 infection with fourteen clinical features using machine learning classification algorithms.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Infectious Disease Diagnostics

Background:

• Reverse Transcription Polymerase Chain Reaction (RT-PCR) is the standard for COVID-19 confirmation but faces limitations like reagent shortages, time constraints, and specialized laboratory requirements.
• Previous alternatives, including Chest CT and X-ray imaging with deep learning, present challenges such as radiation exposure, high costs, and limited device availability.
• A need exists for rapid, cost-effective diagnostic tools for COVID-19, particularly in resource-limited settings.

Purpose of the Study:

• To develop and evaluate predictive models for COVID-19 diagnosis using clinical features.
• To identify the most accurate machine learning classifier for distinguishing between positive and negative COVID-19 cases.
• To provide an accessible diagnostic alternative when RT-PCR testing is insufficient.

Main Methods:

• Retrospective analysis of 114 COVID-19 cases from Taizhou Hospital, Zhejiang Province, China.
• Development of six predictive models utilizing distinct classifiers: BayesNet, Logistic, IBk, CR, PART, and J48.
• Models were trained and validated based on 14 selected clinical features.

Main Results:

• The CR (Classification and Regression) meta-classifier demonstrated the highest accuracy at 84.21% in predicting COVID-19 status.
• All developed models utilized 14 clinical features for prediction.
• The study identified CR as the most effective classifier among the six evaluated.

Conclusions:

• The developed CR-based predictive model offers a promising, accurate, and efficient method for COVID-19 diagnosis.
• This approach can significantly aid in the early detection of COVID-19, especially in scenarios with limited RT-PCR availability.
• The findings are particularly relevant for developing countries facing shortages of testing kits and specialized diagnostic facilities.