In the current study, a receiver operating characteristic (ROC) curve was applied to explore predictors for severity and prognosis of acute respiratory distress syndrome (ARDS). Clinical data including mechanical ventilation, clinical manifestation, performance of APACHE II score, and PaO2/FiO(2) ratio and AaDO(2) were obtained within 24-hour of admission to the hospital. Protein levels of surfactant, pulmonary-associated protein A and D (SP-A and -D), Clara cell protein 16 (CC-16), neutrophil elastase (NE) and peptidase inhibitor 3 (PI3) in the patients' serum were determined by ELISA. Mortality of the hospitalized patients was 30.7% and there were significant differences in APACHE II score, PaO2/FiO(2) ratio and AaDO(2), SP-A and -D, CC-16 and NE between survivors and non-survivors. By comparing the areas under the ROC curve of logistic regression model, it was found that APACHE II score and serum SP-D level displayed larger ROC areas with the left shift of the curves. The findings of the current study suggested that, in addition to APACHE II, serum SP-D level could also be used as a biomarker of monitoring mortality rate in ARDS.