Identification of clinically-relevant phenoclusters among severe COVID ARDS patients
Over the course of the COVID-19 pandemic, hospital systems have been overwhelmed by the number of patients, resulting in a global healthcare crisis. Patients with COVID-19 have a wide variety of clinical outcomes, and risk factors are still not well understood. Among those with severe COVID-19, a subgroup of patients further develop COVID-19-related Acute Respiratory Distress Syndrome (COVID ARDS), leading to a high burden on hospital ICUs as well as high mortality rates. Furthermore, there is ongoing debate as to whether tailored treatments are required for COVID ARDS patients compared to ARDS patients without COVID-19. In order to improve treatments and standard of care, it is vital to understand the clinical and genetic landscape of patients with COVID-19.
In previous work, our group has developed machine learning techniques to identify subgroups within COVID ARDS based on electronic health records from a New York City healthcare system. We showed that COVID ARDS patients can be grouped into distinct groups based on their clinical features and those distinct groups have differences in outcomes. We wish to validate and improve our pipeline using genomic features. By doing so, we hope to identify novel biomarkers for COVID ARDS that can guide tailored treatments. In this proposal, we will use data from the UK Biobank to profile COVID-19 patients into subgroups using both clinical and genetic information, confirming the presence of these patient clusters in the UK population. Our objectives are to characterize these sub-groups based on clinical outcome, identifying potential sources of genetic differences within these groups, and by doing so, better understand the clinical and genetic factors that distinguish severity groups of COVID-19 patients. We will use these results to identify whether tailored treatments could be developed to improve the outcomes of COVID ARDS patients.