Using Machine Learning to Predict Development of Heart Failure, during Post-Acute COVID-19, by Race and Ethnicity

Abstract

Roughly 6 million Americans have Heart Failure (HF), and this number could increase to 8 million by 2030 [1]. As of early 2022, about 76 million Americans have been diagnosed with novel coronavirus (COVID-19) and of those, around 900,000 have subsequently died [2]. Our goal for this paper is two-fold: 1) use machine learning (ML) algorithms to predict the development of HF during the post-acute COVID-19 period, with emphasis on race and ethnicity, and 2) determine how feature importance differs across the race and ethnicity groups. We apply Logistic Regression, Random Forest Classifier [3], and XGBoost Classifier [4] to predict the development of HF in patients of various races and ethnicities during the post-COVID period. These models show promising results for the use of ML algorithms to predict the development of HF in patients post-COVID.