Covid-19 vaccination priorities defined on machine learning.

Area of Science:

• Epidemiology
• Health Informatics
• Machine Learning Applications

Background:

• Defining priority vaccination groups is crucial for reducing mortality.
• COVID-19 vaccination strategies require effective risk stratification.

Purpose of the Study:

• To identify priority population groups for COVID-19 vaccination using machine learning (ML).
• To base vaccination priorities on in-hospital risk of death.

Main Methods:

• Retrospective cohort study of 49,197 RT-PCR-confirmed COVID-19 patients in 336 Brazilian hospitals.
• Extreme Gradient Boosting ML algorithm used to predict in-hospital death risk.
• Variables included age, sex, and 179 categories of chronic health conditions.

Main Results:

• The ML model for in-hospital death prediction achieved an AUC-ROC of 0.80.
• Mean patient age was 60.5 years, with a 17.9% in-hospital mortality rate.
• Patients were categorized into eleven distinct risk groups based on ML-identified variables.

Conclusions:

• Machine learning effectively defines population priorities for vaccination based on in-hospital death risk.
• This ML-driven approach is easily applicable by health system managers for resource allocation.