Existing estimates of diabetes incidence after SARS-CoV-2 infection are derived from retrospective electronic health records (EHR) studies, which may lead to surveillance and misclassification bias given differences in receiving clinical care between the exposed and unexposed groups. We used data from the Veterans Health Administration (VHA)-based, longitudinal observational cohort study Epidemiology, Immunology and Clinical Characteristics of COVID-19 (EPIC3) and the VHA Corporate Data Warehouse to compare diabetes hazard rates over 12 months of follow-up (n=1245) in participants with and without a positive test for SARS-CoV-2. We fit Cox proportional hazard models, adjusted for age, sex, race/ethnicity, Charlson comorbidity index (CCI), and annual number of laboratory tests of any type except for SARS-CoV-2 related types.
Results: Compared to participants with a positive test for SARS-CoV-2, participants without a positive test were older and had a higher number of laboratory tests. They were more likely to have taken at least one dose of COVID19 vaccine, live in an urban area, being current smoker, and have a hospital stay within 30 days post enrollment. After adjustment, SARS-CoV-2 was not significantly associated with risk of incident diabetes overall (HR 0.71, 95%CI 0.33-1.51), in inpatients (HR 1.22, 95%CI 0.08-18.78), or in outpatients (HR 0.50, 95%CI 0.19-1.34).
L. Huang: None. J. Lee: None.
GR014814
