Abstract P165: Characteristics and Predictors of Mimics and Confirmed Strokes in a Stroke Alerts System at a Large Urban Medical Center

Background: Stroke alert systems provide rapid evaluation of suspected strokes to aid timely thrombolysis and mechanical thrombectomy. Given consequences of delayed or missed diagnosis, these systems are intentionally more sensitive than specific for confirmed stroke. However, substantial physician and hospital resources are involved. We sought to evaluate characteristics and predictors of stroke mimics and confirmed strokes in a large stroke alert program. Methods: The study is a retrospective investigation of the stroke alert system at the University of Colorado Hospital. Variables included patient, clinical, stroke and stroke alert characteristics, and outcome. Variables were compared for patients with confirmed stroke, and mimics. Results: Stroke alerts were evaluated in 125 patients over a one-month period in 2019. Median age was 60 years (IQR 47-72), 52% were female, 51% were white, and 24% black or African American. Median initial NIHSS score was 4 (IQR 1-11), and onset was outside the hospital in 70%. One third of stroke alerts were confirmed stroke or TIA, and 66% were mimics. Most stroke alerts were called by the departments of Emergency Medicine (78%), Medicine (10%) and Surgery (9%). The most common of more than 40 stroke alert symptoms and combinations were weakness (45%), aphasia (20%), altered mental status (16%) and facial droop or numbness (each 14%). Stroke mimics were more likely to be female (58% female vs. 41% male, OR 2.206, 95% CI 1.025-4.745, p=0.041), had better initial NIHSS scores [3 (IQR 1-9) mimics vs. 6 (IQR 2-15) stroke, p=0.015), and were more often discharged to home (77% mimics vs. 51% strokes, OR 2.051, 95% CI 1.269-3.316, p=0.004). Stroke onset location and service initiating stroke alert were not associated with confirmed stroke. Weakness was the only symptom associated with confirmed stroke (58% stroke vs. 37% mimics, OR 2.447, 95% CI 1.137-5.268, p=0.021). Conclusions: Of stroke alerts in our institution, two thirds were stroke mimics, and these were more likely to be female patients. Weakness as a stroke alert symptom, alone or with other symptoms, was associated with confirmed stroke. Additional study is warranted to improve specificity and optimize utilization of physician and hospital resources in stroke alert programs.