BACKGROUND In people with suspected stroke the first assessment occurs under time pressure in the emergency department and is based solely on clinical information. Working hypotheses include mechanism and localization of the clinical deficit on imaging. To assess the performance of neurologists in such situation, we investigated the accuracy with which board‐certified neurologists make the correct diagnosis based solely on clinical information. METHODS In this prospective diagnostic accuracy study done at an emergency department of a university hospital, neurologists had to commit themselves to a diagnosis in people with suspected acute stroke. The main analysis was the accuracy with which they distinguished vascular from nonvascular causes using the discharge diagnosis as a reference. Secondary analyses included the distinction of ischemic from hemorrhagic strokes, and the accuracy with which the lesion location and site of vessel occlusion were identified. The performance of neurologists was also compared to residents and medical students. RESULTS Of 800 people with suspected stroke, 567 (71%) had a vascular (508 ischemic stroke or transient ischemic attack and 59 hemorrhagic stroke) and 233 (29%) had a nonvascular disorder (72 seizures, 33 migraine auras, 12 functional neurological disorders, and 116 other diseases). Vessel occlusion was found in 227 of 410 people with ischemic stroke. Neurologists identified vascular origin with an accuracy of 0.86 (95% CI: 0.83–0.89), a sensitivity of 0.93 (0.90–0.95), and a specificity of 0.66 (0.58–0.73). The accuracy to identify ischemia compared with hemorrhage was 0.91 (0.87–0.93). Neurologists’ accuracy to predict the presence of vessel occlusion was 0.66 (0.61–0.71), of exact lesion location was 42%, and of the affected blood vessel 57%. CONCLUSION In people with acute neurological deficits, the accuracy with which neurologists identify vascular origin is high and depends on neurological education. Experienced physicians should be involved early in the management of people with “code stroke.”
Read full abstract