Characterizing COVID-19 Patients through a Community Health Information Exchange and EHR databases

Project Summary/AbstractThe COVID-19 pandemic is a significant public health problem that will require novel approaches formanagement and intervention. Knowledge of the disease's transmission, symptomatology, clinicalcourse, treatment and outcomes is rapidly evolving based on many sources. An important source foradvancing this knowledge are data from electronic health records (EHR) and health informationexchanges (HIE) because they can provide a real-time, unvarnished view of the disease. However,the initially "invisible" nature of the disease makes clear that clinicians and public health personnelwere at a significant disadvantage in discovering and quantifying the pandemic. There is an urgentneed to learn rapidly from EHR and other data to improve discovery and monitoring of patientsinfected by the coronavirus. The evolving dynamic and understanding of the incidence and course ofCOVID-19 requires that we develop new methods for discovery from data. The long-term goal of ourresearch is to develop collaborative filtering algorithms to facilitate access to and analysis of clinicaldata. The goal of this application is to characterize COVID-19 patients through data in a communityHIE, specifically the Indiana Network for Patient Care (INPC) within Indiana's HIE (IHIE), andunderstand how that characterization differs from that within the EHRs of individual health systems.Understanding how COVID-19 patients are represented in HIEs and EHRs will build an importantfoundation for downstream computational activities, such as real-time discovery, public healthsurveillance, intervention management and contact tracing. The two specific aims of this project areto (1) extract a cohort of patients suffering from COVID-19 and similar diseases from IHIE and (2)characterize patients according to several dimensions, such as demographics, signs and symptoms,and disease course using both the INPC as well as separate EHR data sets. The data, going back to1/1/2015, will be extracted from the INPC, and the clinical data warehouses at IU Health andEskenazi Health, two of our major health system partners. As of this writing, 230,749 individuals inIndiana (3.4 percent of the population of 6.73m) have been tested for the coronavirus, of whom32,078 (13.9 percent) have tested positive. We will apply computational phenotyping approachesusing both HIE and individual EHR data in order to help us evaluate to what degree data fromindividual EHRs can help approximate characterizations based on HIE data. This proposal issignificant because it will help us understand how HIE and EHR data can be used to characterizeboth COVID-19 and non-COVID-19 patients. It is innovative because it leverages multiplecomputational phenotyping methods on both individual organizations' EHR, as well as HIE, data togenerate a comprehensive characterization of COVID-19 and non-COVID-19 patients.