Abstract Number: Esoc2026a1377 Real-World Agreement of Nihss Scores Between Paramedics and Stroke Physicians in Suspected Stroke: Secondary Analysis From the Paranaspp Trial

Abstract Background and aims The National Institutes of Health Stroke Scale (NIHSS) is widely used in-hospital but rarely applied prehospital. The ParaNASPP trial introduced paramedic-performed NIHSS, supported by structured training and a digital tool, for suspected stroke cases. This secondary analysis evaluated real-world agreement on NIHSS scores between paramedics and stroke physicians. Methods We analyzed patients from the stepped-wedge cluster-randomized ParaNASPP trial conducted in Oslo, Norway (2019–2021) who had both a prehospital and a NIHSS score at hospital arrival. Agreement in total scores was assessed using Bland–Altman analysis, Cohen’s kappa for dichotomized NIHSS (0–5 vs. ≥6), and weighted kappa for individual items. Limits of agreement (LoA) were interpreted using a predefined four-grade scale (A–D), where grades A–C are considered acceptable. Sensitivity analyses examined whether agreement varied by time interval (≤30, 31–60, 60 minutes. Results 443 patients were included for analysis (median age 74 years, 45% female). Mean difference in total scores was 0.76 points; LoA ranged from –5.90 to +7.42 (Grade B). Very good interrater reliability was observed for dichotomized categories (κ = 0.87). Weighted kappa for individual items ranged from 0.90 to 0.98, strongest for consciousness, facial palsy, and motor function, weakest for ataxia, sensory, and dysarthria. No systematic differences were observed across time intervals. Conclusions Paramedics trained and supported by a digital tool achieved acceptable agreement with stroke physicians in NIHSS scoring in a real-life setting. Although differences likely reflect interrater variability rather than elapsed time, they remained within clinically acceptable thresholds for prehospital decision-making. Conflict of interest Helge F. Bugge: nothing to disclose, Maren R. Hov: nothing to disclose, Else C. Sandset: nothing to disclose, Mona Guterud:nothing to disclose.