Background: The FAST-ED scale is used prehospital to identify ischemic strokes due to large vessel occlusion (LVO). Its real-world accuracy when applied by EMS is uncertain. Mapping NIHSS elements to FAST-ED provides a standardized reference. We aimed to quantify discrepancies between EMS and NIHSS-derived FAST-ED, evaluate documentation patterns, and test whether incorporating vascular risk factors improves discrimination. Methods: We retrospectively reviewed 58 ischemic stroke patients transported by EMS with documented FAST-ED scores. NIHSS elements from the initial hospital exam were mapped to FAST-ED to create a reference standard. Underscored (EMS NIHSS) cases were identified, and EMS narratives analyzed for documentation patterns. Discrimination for LVO was assessed using AUC, sensitivity, specificity, PPV, and NPV at prespecified cut-offs. Logistic regression models tested the added predictive value of atrial fibrillation (AFib), diabetes (DM), and anticoagulation (AC). Results: Fifteen patients had LVO. EMS underscored 12 and overscored 27 relative to NIHSS-derived FAST-ED. Narratives emphasized “weakness” and “slurred speech” but rarely described cortical signs; neglect was never documented despite present in 9 NIHSS-derived cases. At FAST-ED ≥4, EMS sensitivity was 47% and specificity 83% (PPV 50%, NPV 80%), while NIHSS-derived achieved 53% and 90% (PPV 67%, NPV 84%). Alternative thresholds improved sensitivity: EMS ≥3 reached 67%/58%, NIHSS ≥3 reached 73%/78%. AUCs were 0.655 for EMS FAST-ED and 0.788 for NIHSS-derived. Adding risk factors improved discrimination: EMS +AFib 0.672, +DM 0.763, +AC 0.730, all three 0.773. NIHSS-derived FAST-ED +DM reached 0.858, all three 0.864. Conclusions: EMS FAST-ED often under-recognized cortical features and substituted non-protocol elements, reducing sensitivity. Incorporating structured cortical assessments and readily available risk factors improved discrimination and represents a practical opportunity to strengthen prehospital LVO triage.
D'Souza et al. (2026) studied this question.