The National COVID Cohort Collaboration
The National COVID Cohort Collaborative (N3C, covid.cd2h.org) represents a partnership between the Clinical and Translational Science Awards (CTSA) program hubs (60 institutions), the National Center for Advancing Translational Science (NCATS), the Center for Data to Health (CD2H), and the community. As previously described17, N3C is a secure enclave of electronic health records (EHRs) that are harmonized across multiple health systems in the United States. Data domains are programmatically extracted from the EHR and normalized to the Observational Medical Outcomes Partnership (OMOP) common data model. The N3C includes patients with known or suspected COVID-19 infection, and by design, a control group without COVID-19 infection. COVID-19 was phenotyped using diagnostic codes, procedure codes, and laboratory codes, following guidelines from the Centers for Disease Control and Prevention. Pre-pandemic historical EHRs have been extracted since January 2018, with ongoing record collection continuing to the present day. All N3C activities have been approved by a central institutional review board at Johns Hopkins University (Reliance Protocol IRB00249128).
We performed a secondary analysis of the de-identified ‘Level 2’ N3C dataset, which removed 17 personal identifiers and date-shifted longitudinal data to protect patient privacy. Cases of COVID-19 were included in our analysis if the first date of diagnosis of COVID-19 occurred during hospitalization, including preadmission testing. Non-COVID-19 hospitalizations were identified by the absence of a diagnosis of COVID-19, including time intervals before hospitalization and during post-discharge follow-up. We limited our study population to patients surviving at hospital discharge and excluded all patients with a documented history of HF on or before the index hospitalization discharge date. For the purposes of our analysis, indexed hospitalizations were identified by admissions no later than March 1, 2020, and the end of surveillance was March 31, 2022.
Clinically diagnosed conditions were extracted from the N3C enclave if documented in the EHR, using Spark SQL version 3.0.2. HF diagnosed by a physician was used both as an exclusion factor (if documented on or before the index hospitalization discharge date) and as a clinical outcome after discharge. Available biomarkers of heart damage (B-type natriuretic peptide [BNP]N-terminal prohormone brain natriuretic peptide [NT-proBNP], cardiac troponin I and cardiac troponin T) were extracted from the index hospitalization, using the first laboratory value in case of serial testing. For the purposes of this analysis, natriuretic peptides were considered elevated by BNP > 100 pg/mL or NT-proBNP > 300 pg/mL. Medical history (eg, hypertension, obesity, coronary heart disease, diabetes, chronic kidney disease, and chronic lung disease) was extracted if documented by the index hospitalization discharge date, as well than the use of cardiovascular drugs (angiotensin converting enzyme inhibitor [ACEi]angiotensin II receptor blocker [ARB], beta-blockers and statins). Diagnosed conditions were considered present or absent if noted in the electronic health record. Demographics were extracted from the Hospitalization Index, with race/ethnicity categorized into the following groups: White, Black, Asian, or Other, with Other including multirace and Hispanic ethnicity. In most cases, documented race/ethnicity was based on self-report.
All statistical and data management analyzes were performed and documented in the secure N3C Enclave programming environment. Aggregate counts and summary data were queried using Spark SQL version 3.0.2. Descriptive characteristics were compared by chi-square, 2-sample tests you-tests, or Wilcoxon rank sum tests. Multivariate Cox regression models were analyzed using R version 3.5.1. The hazard ratios of incident HF for patients with COVID-19 post-recovery compared to those who were not hospitalized with COVID-19 were analyzed by Cox regression, adjusted for age , gender, race/ethnicity (white vs. non-white), hypertension, obesity, coronary heart disease, diabetes, chronic kidney disease and chronic lung disease, and use of cardiovascular medications (ACEi, ARA, beta -blockers or statins). Accumulated post-discharge follow-up time up to date of HF event, death, or end of surveillance, whichever comes first. The potential modification of the association between post-recovery COVID-19 and incident HF was assessed by stratification (age ≤65 vs. 65), (female vs. male), (white vs. non-white race/ethnicity) and (cardiovascular drug use versus no use). Racial/ethnic groups were categorized as white vs. non-white to analyze potential modification by racial or ethnic minority status. The magnitude of effect modification was assessed by testing the multiplicative interaction of the stratification variable with post-recovery COVID-19 status. Since elevated natriuretic peptides at index hospitalization may indicate an existing undocumented HF, we performed a sensitivity analysis excluding all patients identified with elevated BNP or NT-proBNP.
Summary of reports
Further information on the research design can be found in the summary of nature research reports linked to this article.