Related Experiment Video
Updated: Aug 16, 2026
Basics of Multivariate Analysis in Neuroimaging Data
Published on: July 25, 2010
Prospective optimization of patient selection for emergency cranial computed tomography: univariate and multivariate
W R Reinus1, F L Zwemer, J R Fornoff
1Mallinckrodt Institute of Radiology, Jewish Hospital, St. Louis, MO, 63110, USA.
Rationale And Objectives:
To determine if the clinical variables that are important for selecting patients for emergency cranial computed tomography (CT) are population dependent.
Methods:
Prior to obtaining scans, physicians working in an emergency department in a level II trauma center completed a form describing the indication for the CT examination. These data were matched to the CT scan results and analyzed statistically using univariate and multivariate methods. These results were compared with a prior study at a level I trauma center.
Results:
Of 551 patients having cranial CT, neurologic examination was positive in 340 and CT scan was positive in 122. The neurologic examination correlated strongly with the results of the CT scan (P < 0.00001). In this patient population, the most important clinical predictors of 17 abnormal CT scans from the 211 patients without positive neurologic examinations were seizure and a history of neoplasm. These high-yield variables differ from our prior retrospective study in which intoxication and amnesia were the important predictors in patients with negative examinations. The difference in predictors between the populations most likely results from different prevalences of trauma and ischemic disease.
Conclusions:
Abnormal neurologic examination is the most important criterion available to select patients for emergency cranial CT. Other variables (eg, seizure, amnesia, intoxication, and history of neoplasm) that help select patients without neurologic findings appear to be population dependent.
