Development of a prognostic model for mortality in COVID-19 infection using machine learning

Area of Science:

• Infectious Diseases
• Medical Informatics
• Biochemistry

Background:

• Coronavirus disease 2019 (COVID-19), caused by SARS-CoV-2, rapidly became a global pandemic in 2020.
• Hospitals faced challenges managing COVID-19 cases due to a lack of effective treatments, vaccines, and clinical guidelines.
• Urgent need for actionable knowledge and predictive tools for patient management during the pandemic.

Purpose of the Study:

• To identify prognostic serum biomarkers for predicting mortality in patients with SARS-CoV-2 infection.
• To develop a machine learning model for early identification of high-risk COVID-19 patients.
• To aid in clinical decision-making and resource allocation for severe COVID-19 cases.

Main Methods:

• Retrospective study evaluating laboratory data and mortality from 398 patients with confirmed SARS-CoV-2 infection.
• Development of a machine learning model (support vector machine) using five serum chemistry parameters: c-reactive protein, blood urea nitrogen, serum calcium, serum albumin, and lactic acid.
• Model trained to predict patient expiration status up to 48 hours prior to death.

Main Results:

• The machine learning model demonstrated high predictive performance with 91% sensitivity and 91% specificity (AUC 0.93) for predicting mortality.
• Analysis identified key serum chemistry parameters and their combinations that significantly impact outcomes in SARS-CoV-2 infection.
• The model successfully predicted patient expiration status on independent testing data.

Conclusions:

• Serum chemistry parameters can serve as valuable prognostic biomarkers for COVID-19 mortality.
• Machine learning models can effectively predict mortality risk in SARS-CoV-2 infected patients using readily available laboratory data.
• These findings support the integration of predictive models into clinical practice for improved COVID-19 patient management.