Angelica Byrd1, Daniel Mestre1, Drashti Patel1, Fahad Khan1, Amita Singh1, Christina Wilson1
1University of Florida
Objective:
Evaluate predictors of inaccurate stroke alerts to improve utilization of resources and the quality of care provided.
Background:
Approximately 38% of stroke alerts activated in the emergency department (ED) annually at our academic comprehensive stroke center accurately identify a true stroke. A stroke alert activates a multidisciplinary team to allow for rapid triage and advanced imaging, but stroke mimics can result in high resource utilization, unnecessary exposure to radiation, delayed care of the underlying cause of symptoms, and burnout among providers.
Design/Methods:
A retrospective chart review was conducted of all stroke alerts activated in our ED between May 2022 and April 2023 to evaluate demographics and clinical characteristics. Accurate stroke was defined as a final diagnosis of either ischemic stroke, transient ischemic attack (TIA), or intracranial hemorrhage (ICH). Univariate and binomial logistic regression analyses were completed.
Results:
Among 1,537 patients (mean age 66 ± 17 years, 49% male), 41% were diagnosed with an accurate stroke (25% ischemic, 16% ICH, 1% TIA). Most common mimics were toxic/metabolic encephalopathy (33%), seizure (13%) and migraine (10%). Logistic regression showed predictors of stroke mimics were younger age, female gender, diabetes mellitus, prior stroke, prior or presenting seizure, sensory complaint, altered mental status, and dizziness. Stroke mimics were less likely to have atrial fibrillation, weakness, language deficit, intubation, systolic blood pressure over 160, or leukocytosis, and were more likely to present by personal vehicle than by emergency services (p<0.05). There were no differences in accuracy based on time of day or year.
Conclusions:
A majority of ED stroke mimics were associated with specific presenting symptoms, particularly altered mental status, prior seizure, and dizziness. Focused education or use of triage scores that incorporate these findings can be tailored towards staff to reduce the frequency of inaccurate stroke alerts while maintaining a high positive predictive value.