Abstract WP54: Patient Location Impacts Accuracy Of Inpatient Stroke Alerts

Abstract

Background: Strokes that occur during inpatient hospitalization require rapid identification and stroke-alert activation from hospital staff in heterogeneous settings, often resulting in overutilization of stroke team resources on stroke mimics. An understanding of the factors that impact accuracy of identification of inpatient strokes would assist in improving care in this population. Methods: This was a retrospective analysis from a single Joint-Commission certified Comprehensive Stroke Center. All patients for whom an inpatient stroke alert was called between 2/1/2019 - 4/30/2021 were included. Metrics captured in our local stroke-alert registry were analyzed, with comparisons made between patients who were determined to have a true stroke versus stroke mimic. Results: Of 705 inpatient alerts, mean patient age was 66 years, 50% were male, 62% occurred in the daytime, and mean NIHSS was 12. The majority of alerts were stroke mimics (76%); most common non-stroke diagnoses were encephalopathy, medical decompensation, and seizure. Among patients with true stroke, 86% were ischemic, with 13 patients receiving thrombolytics and 21 patients undergoing mechanical thrombectomy. One stroke mimic received thrombolytics. Although most alerts originated from medicine units (38%) and floor beds (49%), true stroke was most often diagnosed on cardiology (39%) and hematology/oncology (38%) units, and in the ICU or intermediate care setting (54%). Logistic regression adjusted for age and time of day showed true strokes were significantly associated with male sex, location (stratified by specialty), shorter time to CT completion, and higher initial NIHSS (p<0.05). Conclusions: A majority of inpatient alerts are stroke mimics, particularly on medicine floors. Stroke education could be tailored to staff to incorporate these findings to ensure that stroke patients are not missed, while stroke team resources are not overutilized on mimics.