ABSTRACT NUMBER: ESOC2026A2342 DIFFERENCES IN THE SENSITIVITY OF PREHOSPITAL STROKE SCALES FOR PREDICTION OF LARGE, MEDIUM AND DISTAL VESSEL OCCLUSIONS

Abstract Background and aims Several prehospital stroke scales have been developed for timely prehospital identification and triage of stroke patients selected for endovascular thrombectomy. We aim to compare the sensitivity of these scales in a large nationwide patient cohort. Methods Patients registered 01/01/2015-06/03/2023 in the two Swedish stroke registries were included. Patients without documented vessel occlusion or complete NIHSS were excluded. The following stroke scale scores were deducted from NIHSS using an automated algorithm: RACE, SAVE, C-STAT, CPSS, PASS, 3-ISS, FAST-ED, mNIHSS, PreHAST, A2L2, modNIHSSambulance, FAST, APSS, G-FAST, mG-FAST, VAN, VES, MPSS, CG-FAST, FAST-PLUS, EMSA, SNOW/POMONA, ACT-FAST and PreSS. Scores ≥1 were considered as positive for each scale, since established cut-off values only were available for LVO. For sensitivity analysis, occlusions were subdivided into Large Vessel Occlusion (LVO; ICA, M1, V4, Basilar) or Medium Vessel occlusion/Distal Vessel Occlusion (MeVO/DVO; A1-A3-,M2-M3-, P1-P3). Results Of 5809 patients selected for EVT, 1802 patients were available for further analysis. All scales except 3-ISS, A2L2, ACT-FAST, C-STAT AND FAST-PLUS demonstrated >90% sensitivity for LVO, MeVO/DMVO, anterior/posterior vessel occlusion, and LVO/MeVO/DMVO in the right or left hemisphere. The highest sensitivity (>99%) regardless of occlusion site was seen with mNIHSS, preHAST and modNIHSSambulance. Conclusions In patients triaged to EVT, there were significant differences in the sensitivity for vessel occlusion among the tested prehospital stroke scales. The highest sensitivity was seen with mNIHSS, preHAST and modNIHSSambulance. Conflict of interest Magnus Esbjörnsson: nothing to disclose; Johan Wasselius: nothing to disclose; Björn Hansen: nothing to disclose; Gunnar Andsberg: nothing to disclose; Teresa Ullberg: nothing to disclose.