Triage of Patients with Intracerebral Hemorrhage to Comprehensive Versus Primary Stroke Centers.

OBJECTIVES: The management of patients admitted with intracerebral hemorrhage (ICH) mostly occurs in an ICU. While guidelines recommend initial treatment of these patients in a neurocritical care or stroke unit, there is limited data on which patients would benefit most from transfer to a comprehensive stroke center where on-site neurosurgical coverage is available 24/7. As neurocritical units become more common in primary stroke centers, it is important to determine which patients are most likely to require neurosurgical intervention and transfer to comprehensive stroke centers. MATERIALS AND METHODS: This is a retrospective observational cohort study conducted at an academic comprehensive stroke center in the United States. Four-hundred-fifty-nine consecutive patients transferred or directly admitted to the neurocritical care unit from 2016-2018 with the primary diagnosis of ICH were included. Univariate statistics and multivariate regression were used to identify clinical characteristics associated with neurosurgical intervention, defined as undergoing craniotomy, ventriculostomy, or endovascular embolization of an arteriovenous malformation (AVM). RESULTS: The following variables were associated with neurosurgical intervention in multivariate analysis: age (OR 0.38, 95% CI 0.27-0.55), admission Glasgow Coma Scale (OR 0.29, 95% CI 0.18-0.48), the presence of intraventricular hemorrhage (OR 2.82, CI 1.71-4.65), infratentorial location of ICH (OR 2.28, 95% CI 1.20-4.31), previous antiplatelet use (OR 2.04, 95% CI 1.24-3.34), and an AVM indicated on CT Angiogram (OR 2.59, 95% CI 1.19-5.63) were independently associated with the need for neurosurgical intervention. This was translated into a scoring system to help make quick triage decisions, with high sensitivity (99%, 95% CI 97-99%) and negative predictive value (98%, 95% CI 89-99%). CONCLUSIONS: Using previously well described predictors of severity in ICH patients, we were able to develop a scoring system to predict the need for neurosurgical intervention with high sensitivity and negative predictive value.