Using EHR data to identify coronavirus infections in hospitalized patients: Impact of case definitions on disease surveillance

Purpose: To evaluate the number, characteristics, and outcomes of patients identified hospitalized with coronavirus disease 2019 (COVID-19) using two different case definitions. Procedures: Electronic Health Record data were evaluated from patients hospitalized with COVID-19 through May 2020 at 52 health systems across the United States. Characteristics of inpatients with positive laboratory tests for SARS-CoV-2 were compared with those with clinical diagnosis of COVID-19 but without a confirmatory lab result. Findings: Of 14,371 inpatients with COVID-19, 6623 (46.1 %) had a positive laboratory result, and n = 7748 (52.9 %) had only a clinical diagnosis of COVID-19. Compared with clinically diagnosed cases, those with laboratory-confirmed COVID were similar in age and sex, but differed by race, ethnicity, and insurance status. Laboratory-confirmed cases were more likely to receive certain COVID-19 therapies including hydroxychloroquine, anti-IL6 agents and antivirals (p < 0.001). Those with laboratory–confirmed COVID-19 had lower rates of most complications such as myocardial infarction, but higher overall mortality (p < 0.001). Conclusion: We observed a two–fold difference in the number of patients hospitalized with COVID-19 depending on whether the case definition required laboratory confirmation. Variations in case definitions also led to differences in cohort characteristics, treatments, and outcomes.