BackgroundDiabetes mellitus (DM) is not associated with increased mortality in critically ill patients, a phenomenon known as the “diabetes paradox”. However, DM is a risk factor for increased mortality in patients with COVID-19. This study aims to investigate the association of DM and stress-induced hyperglycemia at intensive care unit (ICU) with mortality in this population. MethodsThis is a retrospective study. Electronic medical records from patients admitted from March 2020 to September 2020 were reviewed. Primary outcome was mortality. Secondary outcomes were ICU and hospital mortality and stay, and need for mechanical ventilation and renal replacement therapy. Results187 patients were included. Overall mortality was 43.2%, higher in patients with DM (55.7% vs. 34%; p = 0.007), even after adjustment for age, hypertension, and disease severity. When patients were separated into groups, named normoglycemia (without DM and glycemia ≤140 mg/dL), stress-induced hyperglycemia (without DM and glycemia >140 mg/dL), and DM (previous diagnosis or HbA1c ≥ 6.5%), the mortality rate was 25.8%, 37.3%, and 55.7%, respectively (p = 0.021). Mortality was higher in patients with higher glycemic variability. No statistical difference related to secondary outcomes was observed. ConclusionsDM, hyperglycemia, and glycemic variability associated with increased mortality in critically ill patients with severe COVID-19, but did not increase the rates of other clinical outcomes. More than stress-induced hyperglycemia, DM was associated with mortality.