Personalized Predictions and Prescriptions for COVID-19 Patients: A Machine Learning Approach


The COVID-19 pandemic has created unprecedented challenges worldwide. Healthcare providers make difficult decisions on patient triage, treatment and care management on a daily basis. We design analytical tools to support these decisions and combat the pandemic. Leveraging electronic health records and registry data from institutions spanning Europe and the United States, we develop and validate a novel mortality risk calculator for hospitalized COVID-19 patients. Subsequently, we extend our machine learning framework to not only yield predictive insights but also offer actionable prescriptions. In particular, we present a methodology for personalizing the use of ACE inhibitors and ARBs for COVID-19 treatment. Our approach assesses the potential benefit from the use of these drugs at the individual level, providing novel insights into predictors of treatment effectiveness. The proposed models validate known clinical risk factors and uncover the importance of individual clinical characteristics that could guide patient triage and care planning. Our tools are openly available for clinicians and are currently used in hospitals in Spain and Italy.

University of Cambridge Public Health Seminar