Background Currently, the coronavirus disease 2019 (COVID-19) has become pandemic globally. 10-20% of the cases are severe and more than 397,000 deaths have occurred. The risk factors for the mortality of critically ill COVID-19 patients remain to be elucidated. Conclusions Survived severe and non-survived COVID-19 patients had distinct clinical and laboratory characteristics, which were separated by principle component analysis. Logistic regression revealed several risk factors such as elder age, greater affected lobe numbers and higher level of serum CRP for the mortality of severe COVID-19 patients. Longitudinal changes of laboratory findings indicate the advancement of the disease and may be helpful in predicting the progression of severe patients.