An ensemble prediction model for COVID-19 mortality risk

Area of Science:

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

Background:

• Early identification of high-risk coronavirus disease (COVID-19) patients is essential for timely intervention and improved outcomes.
• Existing machine learning models have shown limitations in predicting mortality across diverse patient cohorts.
• There is a need for robust predictive models to assess COVID-19 patient mortality risk in independent cohorts.

Purpose of the Study:

• To develop and validate a machine learning model for accurate early prediction of death risk in COVID-19 patients.
• To identify key clinical features that are predictive of mortality in SARS-CoV-2 infected individuals.
• To ensure the model's generalizability and utility in independent patient cohorts.

Main Methods:

• Utilized a cohort of 4711 COVID-19 patients, analyzing clinical and laboratory features.
• Developed a novel data preprocessing technique for cleaning clinical data.
• Employed an ensemble machine learning approach to identify critical predictive features.

Main Results:

• Identified 14 key clinical features demonstrating strong predictive performance for mortality.
• Achieved an area under the receiver operating characteristic curve of 0.907.
• Successfully validated the predictive utility of these 14 features in a large independent cohort of 15,790 patients.

Conclusions:

• The 14 identified clinical features are robust predictors of death risk in COVID-19 patients.
• This model offers a valuable tool for early risk stratification in clinical settings.
• The findings support the use of this machine learning approach to aid in critical clinical decision-making for SARS-CoV-2 patients.