Abstract: Objective: To examine the impact of SARS-CoV-2 on long-term glycemia. Research desing and methods: We conducted a retrospective inception cohort study using Veterans Health Administration data (March 1, 2020-May 31, 2022) among individuals with ≥ 1 positive nasal swab for SARS-CoV-2 and individuals with ≥ 1 laboratory test of any type but no positive swab. Two incident diabetes cohorts were defined based on: 1) a computable phenotype using a combination of diagnosis codes, laboratory tests, and receipt of glucose-lowering medications (n = 17,754); and 2) the presence of ≥ 2 HbA1c results ≥ 6.5% (n = 4,768). We fit log-binomial models examining associations of SARS-CoV-2 with diabetes remission, defined as ≥ 2 HbA1c measurements < 6.5% ≥ 90 days apart after cessation of any glucose-lowering medications. To help equalize laboratory surveillance of glycemia, we conducted a subgroup analysis among non-hospitalized participants. Results: In cohorts 1 and 2 respectively, 25% and 29% had ≥ 1 positive test for SARS-CoV-2 prior to enrollment, and 21% and 11% had remission. SARS-CoV-2 was associated with a higher chance of remission by both definitions (1: RR 1.22 [95%CI 1.14-1.29]; 2: RR 1.27 [95%CI 1.07-1.50]) over an average 503 (±202) and 494 (±184) days. The association was attenuated among non-hospitalized participants (1: RR 1.11 [1.04-1.20]; 2: R: 1.17 [95%CI 0.97-1.42]). Conclusions: Diabetes remission was more common in Veterans with new-onset diabetes after SARS-CoV-2. In non-hospitalized participants, who were likely to have more similar laboratory surveillance, the association was diminished. Differences in surveillance or transient hyperglycemia may explain the observed association.