Research of mortality risk prediction based on hospital admission data for COVID-19 patients

Math Biosci Eng. 2023 Jan 11;20(3):5333-5351. doi: 10.3934/mbe.2023247.

Abstract

As COVID-19 continues to spread across the world and causes hundreds of millions of infections and millions of deaths, medical institutions around the world keep facing a crisis of medical runs and shortages of medical resources. In order to study how to effectively predict whether there are risks of death in patients, a variety of machine learning models have been used to learn and predict the clinical demographics and physiological indicators of COVID-19 patients in the United States of America. The results show that the random forest model has the best performance in predicting the risk of death in hospitalized patients with COVID-19, as the COVID-19 patients' mean arterial pressures, ages, C-reactive protein tests' values, values of blood urea nitrogen and their clinical troponin values are the most important implications for their risk of death. Healthcare organizations can use the random forest model to predict the risks of death based on data from patients admitted to a hospital due to COVID-19, or to stratify patients admitted to a hospital due to COVID-19 based on the five key factors this can optimize the diagnosis and treatment process by appropriately arranging ventilators, the intensive care unit and doctors, thus promoting the efficient use of limited medical resources during the COVID-19 pandemic. Healthcare organizations can also establish databases of patient physiological indicators and use similar strategies to deal with other pandemics that may occur in the future, as well as save more lives threatened by infectious diseases. Governments and people also need to take action to prevent possible future pandemics.

Keywords: COVID-19; death risk feature; emergency triage strategy; ensemble learning; machine learning; tree model.

MeSH terms

  • COVID-19* / epidemiology
  • Hospitalization
  • Hospitals
  • Humans
  • Machine Learning
  • Pandemics