Abstract P542: Classifying COVID-19 Related Hospitalizations and Deaths in the Collaborative Cohort of Cohorts for COVID-19 Research (C4R) Study

Objectives: Administrative criteria are often used to define COVID-19 outcomes despite challenges in differentiating hospitalization “for” vs “with” COVID-19 and unknown reliability for identifying COVID-related cardiopulmonary complications. We implemented a protocol to adjudicate COVID-related events across 14 studies participating in C4R. Methods: C4R cohorts ascertained COVID-related hospitalizations and deaths via questionnaires or ongoing surveillance. Medical records were reviewed by C4R physicians using a standardized protocol to assign COVID-related diagnoses as definite or probable. C4R diagnoses were compared against ICD codes assigned to the events. This report includes data from events centrally adjudicated by C4R as of September 2022. Results: Among 144 events ascertained across 7 cohorts as potentially COVID-related, SARS-CoV-2 infection was confirmed in 107/117 (91%) non-fatal hospitalizations and 26/27 (96%) deaths. Of confirmed infections, COVID-19 illness was diagnosed as the cause of 101/107 (94%) hospitalizations and 24/26 (92%) deaths ( Table ). Of non-fatal hospitalizations with infection, 72% were diagnosed with severe or critical COVID-19, 78% with COVID-associated pneumonia, and 29% with renal failure; other complications were less common. C4R review did not validate diagnoses indicated by ICD codes for 1 COVID-19 infection, 5 pneumonias, 1 stroke, and 7 renal failure cases. C4R review did identify diagnoses that were not ICD-coded for 13 infections, 31 pneumonias, 6 myocardial infarctions, 5 venous thromboses, and 21 renal failure cases. Conclusions: Protocolized medical records review by C4R confirmed COVID-19 as the cause of hospitalization or death in 87% of events ascertained as potentially COVID-related and 94% of those in which SARS-CoV-2 infection was confirmed. Both false-positive and false-negative misclassification by ICD criteria was observed. Protocolized review may be useful to identify and validate COVID-related events for epidemiologic research.