To build a risk score (RS) model of fatal intracranial hemorrhage (FICH) in patients with acute leukemia, we retrospectively assessed risk factors in 792 patients newly diagnosed with acute leukemia, 41 of whom had analyzable FICH. We found that female gender (relative risk (RR) ¼ 5.234, Po0.001), acute promyelocytic leukemia (RR ¼ 4.057, P ¼ 0.003), leukocytosis (RR ¼ 3.301, P ¼ 0.004), thrombocytopenia (RR ¼ 3.283, P ¼ 0.005) and prolonged prothrombin time (RR ¼ 3.291, P ¼ 0.016) were significantly associated with occurrence of FICH in multivariate analysis. To calculate RS for FICH, one point was assigned for each risk factor, making the RS between 0 and 5. The RS model segregated patients into three prognostic groups: a low-risk group (LRG) for RS of 0 or 1; an intermediate-risk group (IRG) for RS of 2 or 3; and a high-risk group (HRG) for RS of 4 or 5. Expectation of FICH was well correlated with risk groups (all P-values o0.001). Overall survival was significantly shorter in the HRG compared with the LRG. The RS model we constructed to predict the occurrence of FICH will be verified through prospective studies.