Background Major bleeding (MB) and clinically relevant bleeding (CRB) are the most common seen complications associated with anticoagulation treatment for pulmonary embolism (PE) patients. A bleeding risk score (BRS) may help to accurately determine the risk of bleeding and make better decisions for patients in clinical practice. Methods Patients diagnosed as acute PE and who met the inclusion criteria in Beijing Chao-Yang Hospital from January 2009 to September 2013 were consecutively enrolled. Baseline data were collected. Four BRS (Kuijer score, RIETE score, Kearon score and Nieuwenhuis score) were assessed and compared using the area under the receiver operating characteristic curve (AUC). Results A total number of 563 patients were included in the study. Of which, 16 had MB and 89 had CRB within three months of anticoagulation treatment. Three months' cumulative incidence for MB and CRB events was 0.03 (95% CI 0.01-0.05) and 0.17 (95% CI 0.12-0.21), respectively. In our study population, the AUCs for Kuijer, RIETE, Kearon and Nieuwenhuis scores were 0.57 (95% CI 0.44 similar to 0.68), 0.56 (95% CI 0.45-0.71), 0.75 (95% CI 0.60 similar to 0.89) and 0.59 (95% CI 0.41 similar to 0.74), respectively. In addition, the AUCs of four BRS for CRB were all poor, with values less than 0.65. Conclusions Only the Kearon score appeared to have more accuracy in predicting the risk of MB. Further large prospective studies are needed to externally validate a BRS for CRB.