New treatments for acute stroke require a rapid triage system, which minimizes treatment delays and maximizes selection of eligible patients. Our aim was to create a score for assessing the probability of brain hemorrhage among patients with acute stroke based upon clinical information. Of 1805 patients in the Stroke Data Bank, 1273 had infarction (INF) and 237 had parenchymatous hemorrhage (HEM) verified by CT. INF and HEM discriminators were determined by logistic regression and used to create a score. ROC curve was used to choose the cut-point for predicting HEM (score </= 2), with sensitivity of 76% and specificity of 83%. External validation was done using the NOMASS cohort. Although the use of a practical score by emergency personnel cannot replace the gold-standard brain image differentiation of HEM from INF for thrombolytic therapy, this score can help to select patients for stroke trials and pre-hospital treatments, alert CT scan technicians, and warn stroke teams of incoming patients to reduce treatment delays.