Designing COVID-19 Mortality Predictions to Advance Clinical Outcomes: Evidence from the Department of Veterans Affairs

Abstract

Using administrative data on all veterans who enter Department of Veterans Affairs (VA) medical centers throughout the United States, this paper uses machine learning methods to predict mortality rates for COVID-19 patients between March and August 2020. First, using comprehensive data on over 10,000 veterans' medical history, demographics, and lab results, we estimate five AI models. Our XGBoost model performs the best, producing an AUROC and AUPRC of 0.87 and 0.41, respectively. Second, through a unique collaboration with the Washington D.C. VA medical center, we develop a dashboard that incorporates these risk factors and the contributing sources of risk, which we deploy across local VA medical centers.

Note: Funding: This was work was completed in part through the authors' roles in the Department of Veterans Affairs.

Declaration of Interests: There are no conflicts of interest.

Keywords: Artificial Intelligence, Clinical Practice, Coronavirus, Machine Learning, Socio-Demographics, Veterans, Department of Veterans Affairs