COVID-19 Racial Disparities in Testing, Infection, Hospitalization, and Death: Analysis of Epic Patient Data
The analysis is based on EHRN and KFF analysis of data from the Epic health record system, which includes data for patients from 53 health systems representing 399 hospitals across 21 states. Overall, the system includes data for roughly 55 million active patients. Active patients include those who have interacted with the health system in the past two years, as indicated by either a face-to-face visit or an order placed in their chart. The analysis was restricted to the 89% of active patients who had known race/ethnicity, resulting in a total of roughly 50 million active patients included in the analysis.
The analysis presents findings for Black, Hispanic, Asian, and White patients. Due to data limitations, we do not present findings for smaller population groups, including AIAN and NHOPI patients, or people who report multiple races. As availability of data for smaller population groups increase over time, it may allow for future analysis focused on the experiences of these populations.
We examined testing, infection, hospitalization, and death rates related to COVID-19 among active patients. In addition, we identified the level of care required at the time a patient tested positive for COVID-19 by race and ethnicity.
Further, we performed statistical analysis using data from 332,956 people who tested positive for COVID-19 to examine increased risk of hospitalization and death for Black, Hispanic, and Asian patients relative to White patients after controlling certain sociodemographic characteristics and health conditions known to increase risk of illness and death.
Specifically, we controlled for age, sex, and health conditions that a previous EHRN analysis had identified as being significantly associated with higher risk of hospitalization and death. These conditions included hypertension, diabetes, heart failure, chronic obstructive pulmonary disease (COPD), cerebrovascular disease or stroke, and obesity. The prior EHRN analysis also suggested a significant risk for patients who were immunocompromised. However, that condition was not included in the model due to continued refinements in the definition of an immunocompromised state. In addition, we controlled for social vulnerability based on where each person lives, using the CDC’s Social Vulnerability Index. The CDC’s Social Vulnerability Index identifies the level of social vulnerability associated with a census area based on 15 social factors, including poverty, income, employment, education, age, household composition, housing, transportation, and racial/ethnic distribution. It was developed to help public health officials and emergency response planners identify and map the communities that will most likely need support before, during, and after a hazardous event. Statistical controls were performed using Cox Proportional Hazards models using 95% confidence intervals.
Lily Rubin-Miller, MPH, Christopher Alban, MD, MBA and Sean Sullivan, MS, MPH are with the Epic Health Research Network. Samantha Artiga, MHSA is with KFF.