Aplicación de modelos de aprendizaje automático para la predicción de eventos adversos secundarios a vacunas COVID-19 y detección de factores importantes

Abstract

COVID-19 disease caused by the SARS-CoV-2 virus was the third leading cause of death in the United States for 2022, claiming more than six million lives worldwide since the outbreak began in 2019. COVID-19 vaccines have been available for more than two years, yet vaccine hesitancy still prevails to this day. One of the factors to vaccine hesitancy is the concern for vaccine safety and adverse reactions. The main objective of this work is to apply machine learning algorithms and develop a model to predict if an individual will have a serious adverse reaction or death based on patient information, medical history, and vaccine information. Additionally, through the application of feature importance techniques this study aims to identify potential risk factors for serious adverse reactions. Six machine learning algorithms were chosen for this study: Logistic Regression, Decision trees, Support Vector Machines (SVM), Naïve Bayes (NB), Random Forest (RF), K-Nearest Neighbors (kNN). The best result was achieved through Random Forest to predict a lethal adverse event post vaccination with an accuracy of 91.37%. Decision Tree provides an accuracy of 64.83% when predicting a severe adverse event. Age, gender, vaccine manufacturer and vaccine dose contribute the most to serious adverse reactions while age, gender and symptoms contribute the most towards patient death.