Optimal Prognostic Accuracy: Machine Learning Approaches for COVID-19 Prognosis with Biomarkers and Demographic Information

The global emergence of the unprecedented COVID-19 pandemic in late 2019 has led to millions of infections and thousands of fatalities, profoundly affecting various aspects of life. The continual genetic evolution of the virus highlights the critical need for accurate prognostic tools. In response, this study introduces a machine learning-based approach that leverages the capabilities of Quantum Neural Networks (QNN) to predict the lethal risk of COVID-19 patients, aiming to support physicians and healthcare administrators with insightful data for informed decision-making. Leveraging demographic information and biomarker-based dataset, the Relief method and Matrix factorization feature selection approaches were employed for the feature engineering. In addition, our in-depth descriptive analysis revealed specific demographic factors and comorbidities that intricately contribute to an elevated risk of mortality among COVID-19 patients, emphasizing the nuanced dynamics influencing fatal outcomes. A comprehensive comparative analysis compared QNN with a range of machine learning and deep learning algorithms. The findings demonstrate the potential of our prototype prognostic model in stratifying the mortality risk among COVID-19 patients, offering valuable insights for healthcare professionals to make informed decisions in clinical practice.