The study objective was to identify the demographic, physiologic, and anatomic variables associated with outcomes of patients with intracranial hemorrhages. We performed a prospective study of all patients with known or suspected intracranial hemorrhages transferred from a community hospital to the neurosurgical service of tertiary care centers in Toronto. Outcomes measured were patient disposition (admitted v discharged immediately), management (neurosurgical interventions v observation) and survival until discharge v death. Patient variables recorded pre- and post-transfer were analyzed using the statistical programs C.H.A.I.D. and C.A.R.T.(Answer-Tree Software, SPSS Inc.Headquarters, Chicago, IL). Two linear regression trees were constructed to reveal associations with each outcome. Sixty-seven patients between the ages of 24 and 89 were included in the study. Of these, 1 was referred for an epidural bleed, 20 for subdural bleeds, 21 for subarachnoid hemorrhages, and 25 for intracerebral hemorrhages. Thirteen of the patients in the study died or remained with permanent, severe neurological deficits. Of these, none had epidural hemorrhages, 4 had subdural hemorrhages, 5 had subarachnoid hemorrhages, and 4 had intracerebral hemorrhages. The linear regression tree constructed with C.A.R.T. revealed that the most important predictor of outcome was the GCS score of the patient on arrival at the neurosurgical center. Specifically, 71.43% of individuals arriving with a GCS score less than or equal to 7 died or remained with a severe neurological deficit, whereas only 5.66% of individuals arriving with a GCS score greater than or equal to 8 had such an outcome (P 62 with a GCS 62 negated the possibility of normal functional outcome (not statistically significant). These findings may have wide-ranging implications regarding the transfer and treatment of patients with intracranial hemorrhages, use of resources and counseling of families.