Abstract 30: Prediction Scale Thresholds Selected to Rule Out Large Vessel Occlusion Stroke Result in Very Low Positive Predictive Values and Many False Positives

Abstract

Introduction: Many prior large vessel occlusion (LVO) prevalence and prediction scale accuracy studies have not had samples representative of a prehospital suspected stroke population. To address this, we studied emergency medical systems (EMS) identified prehospital suspected stroke patients brought to the Emergency Department (ED) at Zuckerberg San Francisco General Hospital from July 2017 to July 2018. Methods: Patients were eligible for the prevalence study if the EMS prehospital alert call included suspected stroke with a last known well time within 6 hours and a positive Cincinnati Prehospital Stroke Scale. LVO prediction scale scores were retrospectively calculated from arrival NIHSS subitems. We excluded patients missing NIHSS scores and scales requiring non-NIHSS data. LVO stroke included internal carotid, M1, M2, or basilar arteries. Diagnoses were determined by chart review. Prevalences, scale scores, and accuracy statistics were then calculated. We prespecified that negative results of scale thresholds must reduce the post-test probability to ≤5% to rule out LVO stroke and positive results must increase the post-test probability to ≥80% to rule in LVO stroke. Results: Of 220 EMS transported patients there were 30 LVO strokes (13.6%), 35 ICHs (15.9%), 45 non-LVO strokes (20.5%), and 110 mimics (50%). There were 184 patients eligible for the LVO prediction study. Table 1 shows the accuracy statistics of qualifying scale thresholds. False positive rates ranged from 58% to 80%. Only FAST-ED ≥7 resulted in a positive predictive value (PPV) of ≥80% but this missed 83% of LVO strokes. Conclusions: The prevalence of LVO stroke among EMS suspected acute stroke patients brought to our ED over one year was 13.6%. Prediction scale thresholds selected to rule out LVO stroke result in very low PPVs and many false positives. No scale achieved a PPV above 50% while maintaining a sensitivity above 50% suggesting limitations in the ability of scales to rule in LVO stroke.