I ABSTRACTObjectives: To examine the pattern of nontrauma cranial CT use in an urban ED, to identify the rate of significant CT abnormalities in this setting, and to develop criteria for restricting the ordering of CT scans. Methods: A prospective, observational study of a case series of adults who underwent cranial CT scanning for nontraumatic cases was performed at the EDs of an urban teaching hospital and an affiliated community hospital with a combined annual census of 110,000. Clinically significant CT scans were defined as: 1) acute stroke, 2) CNS malignancy, 3 ) acute hydrocephalus, 4) intracranial bleeding, or 5) intracranial infection. x2recursive partitioning was used to derive a decision rule to restrict ordering of CT scans. Results: Only 61 (8%) of 806 CT scans revealed clinically significant abnormalities. The presence of any of the following: age 260 years, focal neurologic deficit, headache with vomiting, or altered mental status, was 100% sensitive (95% CI: 94-100%) and 31% specific (95% CI: 28-33%) in detecting clinically significant CT scans. This set of features had positive and negative predictive values of 11% (95% CI: 8-13%) and 100% (95% C1: 98-loo%), respectively. If these criteria had been used to restrict cranial CT use, 229 fewer patients (28%) would have had CT scans obtained and no clinically significant abnormalities would have been missed. Conclusion: Clinically significant CT abnormalities were uncommon in this study population, suggesting that current criteria for ordering nontrauma cranial CT scans may be too liberal. In this study, a set of clinical criteria was derived that may be useful at separating patients into high-and low-risk categories for clinically significant cranial CT abnormalities. Before these results are applied clinically, these criteria should be validated in larger, prospective studies.