COVID-19 is a new coronavirus infectious disease. We aimed to study the characteristics of thyroid hormone levels in patients with COVID-19 and to explore whether thyroid hormone predicts all-cause mortality of severely or critically ill patients. The clinical data of 100 patients with COVID-19, who were admitted to Wuhan Tongji Hospital from February 8 to March 8, 2020, were analyzed in this retrospective study. The patients were followed up for 6-41 days. Patients were grouped into non-severe illness and severe or critical illness, which included survivors and non-survivors. Multivariate Cox proportional hazards analysis was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for all-cause mortality in association with continuous and the lower two quartiles of thyroid hormone concentrations in severely or critically ill patients. The means of free T3 (FT3) were 4.40, 3.73 and 2.76 pmol/L in non-severely ill patients, survivors and non-survivors, respectively. The lower (versus upper) two quartiles of FT3 was associated with all-cause mortality HR (95% CI) of 9.23 (2.01, 42.28). The HR (95% CI) for all-cause mortality in association with continuous FT3 concentration was 0.41 (0.21, 0.81). In the multivariate-adjusted models, free T4 (FT4), TSH and FT3/FT4 were not significantly related to all-cause mortality. Patients with FT3 less than 3.10 pmol/L had increased all-cause mortality. FT3 concentration was significantly lower in patients with severe COVID-19 than in non-severely ill patients. Reduced FT3 independently predicted all-cause mortality of patients with severe COVID-19.