Use of Different Inclusion Criteria to Compare COVID-19 Hospital Admission Rates by Race and Ethnicity: A Cohort...

Background/Objectives: Reports documenting racial disparities in COVID-19 hospitalization rates from electronic medical record data have used different sample selection methods. Studies including all individuals with a positive COVID-19 test may be vulnerable to misclassification bias if hospitalization status is not captured for all individuals (i.e., if they went to a non-study hospital). A few studies have used only patients who tested positive in the ED and have found different results. In this study, we explore the implications of using different sets of inclusion criteria for analyses that compare COVID-19 hospital admissions by race and ethnicity. Methods: Two separate data sets were created by applying the two different COVID-19 testing inclusion criteria to medical records data from a single academic health system. We used logistic regression to compare the odds of COVID-19 hospitalization across race and ethnicity for each data set and compared our results with previous studies. Results: We found that using all positive COVID-19 tests as the study sample resulted in higher odds of hospitalization for Black and Hispanic patients relative to White patients. In contrast, using only patients who tested positive in the ED resulted in higher odds of hospitalization for White patients. These findings matched the findings of other studies. Conclusions: Patient inclusion criteria should be considered carefully when comparing results from studies of COVID-19 hospitalization.