Elevated serum lactate dehydrogenase (LDH) was commonly reported in COVID-19 patients. However, the relationship between LDH and the incidence of severe cases has not been characterized in those patients.We retrospectively analyzed the characteristics of patients from a designated isolation medical center for COVID-19 patients diagnosed from February 6 to March 1. Variables accessed within 48 hours on admission were compared between patients with and without the severe disease. Logistic model analyses were performed to examine the prognostic value of LDH for predicting severe disease.52 (28.6%) patients later developed severe disease. Comparing to non-severe cases, severe cases had a higher level of serum LDH (321.85 ± 186.24 vs 647.35 ± 424.26, P < .001), neutrophils (5.42 ± 3.26 vs 9.19 ± 6.33, P < .001), and C-reactive protein (38.63 ± 43.14 vs 83.20 ± 51.01, P < .001). The patients with severe disease tended to be male (44.6% vs 80.8%, P < .001), lower level of serum albumin (31.41 ± 6.20 vs 27.18 ± 5.74, P < .001), and SpO2 (96.30 ± 2.75 vs 92.37 ± 8.29, P < .001). In the multivariate analysis model, LDH and sex remained independent risk factors for severe disease. The serum LDH predicted severe cases with an area under the curve (AUC) of 0.7999. A combination of serum LDH and sex predicted severe cases with an AUC of 0.849. A combination of serum LDH accessed on admission and sex had a better predictive performance than the serum LDH (P = .0238).Serum LDH on admission combined with sex is independently associated with severe disease in COVID-19.