Abstract W P322: Prediction of Intracerebral Hemorrhage Expansion With Clinical, Laboratory, Pharmacologic, and Noncontrast Radiographic Variables
Introduction: Hematoma expansion (HE) is a cause of excess mortality in intracerebral hemorrhage (ICH), and HE is potentially preventable if at-risk patients can be identified. Contrast extravasation on initial computed tomographic angiography (CTA) strongly predicts HE, yet most centers do not integrate CTA into early ICH management. We therefore asked whether other variables available at presentation might be used to predict HE. Methods: We searched the electronic medical records of a large integrated healthcare delivery system to identify patients with a hospitalization discharge diagnosis of ICH during a three year period (2008-2010). HE was defined as ICH volume increase by 1/3 or by 12.5mL, as determined by ABC/2 estimation, within 48 hours of presentation. Pre-specified patient variables including age, gender, medical comorbidities, medication use, and in-hospital vital signs were extracted. Stepwise multivariable logistic regression was performed to model HE in patients with at least two head CT scans (HE group), and HE or death by 48 hours among patients with only one head CT (HE+D group). The area under the receiver operating characteristic curve (AUROC) and pseudo r-squared (pseudo-r2) statistics were used to assess model goodness-of-fit. Results: For the 2 study cohorts, 91/257 (35%) had HE and 163/343 (48%) had HE+D. In stepwise multivariable logistic regression, antithrombotic use (OR = 1.9, P=0.04) and initial mNIHSS (OR = 1.06, P=0.001) were significant predictors in the HE model (AUROC = 0.6712, pseudo-R2 = 0.0641). In the HE+D model, age (OR = 1.02, P=0.02), initial mNIHSS (OR = 1.07, P<0.001), and initial hematoma volume (OR = 1.01, P=0.03) were significant predictors (AUROC = 0.7579, pseudo-R2 = 0.1722). Conclusion: Our model predicting HE+D was better fitted than the HE model. The resemblance to known predictors of ICH mortality in the HE+D model suggests that the improved fit may be driven by mortality rather than HE. CTA contrast extravasation remains the strongest predictor of HE and merits consideration as a standard in early ICH care.