The purpose of this study was to assess the accuracy of the parameters used in conventional hemostatic testing and thromboelastography (TEG) in predicting bleeding risk in patients with hematologic diseases. Patients diagnosed with a hematologic disease were divided into bleeding (n = 125) and non-bleeding (n = 509) groups according to clinical signs and symptoms. Several parameters were measured in all patients via traditional hemostatic testing and TEG, including platelet counts (PLT) and maximum amplitude (MA). The sensitivity and specificity of each parameter for predicting bleeding risk were determined via receiver operating characteristic curves. PLT had a sensitivity of 81.1 % and a specificity of 74.4 %, and MA had a sensitivity of 74.7 % and a specificity of 72.0 %. Specificity was higher for both parameters together (77.6 %) than for either alone (P < 0.01). In a subgroup analysis of patients with PLT < 20 x 10(9)/L, sensitivity was higher for both parameters together (84.6 %) than for either alone (P = 0.003). Although all parameters evaluated predicted bleeding risk, PLT and MA were especially accurate. We suggest that the combination of PLT and MA better assesses bleeding risk than do other parameters and that the use of this metric may help to guide decisions regarding platelet transfusion in patients with thrombocytopenic hematologic diseases.