Neuroimaging in Headache Patients: The Sensitivity of Computerized Tomography (CT) in Missed Stroke Diagnoses

Abstract

Background: Stroke is the leading cause of disability in the US, costing $34 billion a year and affecting 800,000 patients. Early detection and treatment is the best way to improve outcomes. Yet, 12.5% of strokes are discharged from the ED within the prior 30 days, with headache the most common diagnosis. Neuroimaging, ideally, would catch an impending stroke, but head CT has variable sensitivity based on onset of symptoms and there are both Federal and provider-led (including Choosing Wisely and the American College of Radiology Appropriateness Criteria (ACR-AC) initiatives to reduce overuse of imaging. Purpose: This study examined variation in ED treatment for patients presenting with a headache, particularly focusing on use of neuroimaging. Potential missed strokes were identified to determine if CT or MRI could have captured stroke. Methods: Using HCUP 2013 Maryland State Emergency Department Dataset, and State Inpatient Data, patients who were seen in the ED within 30 days of a stroke with a complaint of headache were identified. Generalized linear mixed modeling determined if neuroimaging predicted stroke bounce back while controlling for patient and hospital variables. Results: Of the 63,942 headache visits in Maryland EDs, 337 patients presented with a stroke within 30 days of ED discharge. Half (54%) were seen in the ED the day of their stroke and 72% were seen within 7 days. A large majority of the stroke patients (82%) underwent CT for their ED headache visit. Patients who underwent CT for their headache were 2.5 times more likely to return with ischemic stroke, and 7.7 times more likely to return with hemorrhagic. Patients who underwent MRI were 1.7 times more likely to return with any stroke, and 2.8 times more likely to return with ischemic stroke. Conclusions: Providers were concerned about pathology, given the large percentage of patients imaged; however, imaging did not catch active ischemia or bleeding. The negative predictive value of imaging for headache may need to be reconsidered. Patients with high suspicion of pathology should be placed in observation and have appropriate follow up testing. The ACR-AC should be incorporated into diagnostic pathways to optimize use.