The efficacy of corticosteroids in the treatment of patients with severe COVID-19 remains unknown. We evaluated the impact of corticosteroids on clinical improvement among severe COVID-19 patients. In this retrospective, two-centered, cohort study, we enrolled 101 patients with severe COVID-19: with 39 patients in the steroid group and 63 patients in the non-steroid group. The primary endpoint was Time to Clinical Improvement (TTCI) by up to 28 days after the treatment. Secondary endpoints included the rate of CAT scan improvement, the percentage of negative SARS-Cov-2 RT-PCR tests by Day 28, and the time to discharge. We found that patients in the steroid group did not have significant differences of TTCI from patients in the non-steroid group by 28 days after the treatment (median, 19 days vs. 20 days; hazard ratio, 1.07; p=0.797). The CAT scan improvement rate was not statistically different between the two groups by Day 28 (87.2% vs. 79.0%, p=0.170). The negative test of SARS-CoV2 RT-PCR by Day 28 was 68.4% in the steroid group, 87.1% in the non-steroid group (p= 0.060). Time to discharge was significantly longer in the steroid group than the non-steroid group (35 days vs 21 days, p=0.005). Our findings indicated the short-term corticosteroid at a low to moderate dose did not improve the clinical outcomes for patients with severe COVID-19. Further randomized clinical trials are needed to confirm the findings.