- Very Large and Varied Population:
- Over 600,000 discrete patients with COVID case enrolled in 2020
- 21 different health systems from across the nation with diverse populations
- The Potential of Longitudinal Data: Because each patient maintains a unique identifier and health system data are updated regularly, we can:
- Evaluate the frequency and types of re-treatment (inpatient or outpatient) for both COVID and non-COVID illnesses after the index COVID diagnosis.
- Assess the evolution of the disease and its treatments on outcomes (e.g., changes in death rates within and across health systems).
- Limited Data Set: In calendar year 2020, data were entirely de-identified, HIPPA compliant. In calendar year 2021, the data set will be transformed to a limited data set allowing for collection of dates of service and five-digit zip code.
- Extensive Array of Data Elements: Approximately 195 discrete EHR data elements (in 2020) provide the capacity to examine relations between these variables and outcomes.
- Regular Data Pulls Retrospective to February 1, 2020: Through 2020, health system data was extracted monthly (will move to quarterly in 2021). These repeated data extractions allow for the addition and refinement of data elements over time to:
- Add data elements of interest (e.g., Remdesivir, neurologic symptoms).
- Improve accuracy of data elements collected (e.g., COVID diagnostic criteria, death definition).
- COVID Cases Only: Since the cohort is limited to individuals diagnosed with COVID, we are unable to examine risk of developing COVID or compare findings of cohort members with individuals without a COVID diagnosis.
- Discrete Data Elements Only: The data pulled is almost exclusively limited to discrete EHR data elements. It typically does not include data from text fields within the EHR.
- Missing Data: For many EHR data elements (e.g., vaping), a large proportion of patients have no data reported.
- Harmonizing EHR Data across Health Systems: Merging data from different health systems highlights the idiosyncratic nature of how each system records EHR data, even if on the same EHR platform (e.g., Epic)
- Patient Data is limited to Health System Where COVID Was Diagnosed: Thus, if patient dies or subsequently receives care at alternative sites, we will not capture those events.