Abstract Body: This study compares the accuracy of manual stroke scale calculations against electronically calculated scores using the Ultimate Stroke Scale (USS), a new software designed to automate large vessel occlusion (LVO) screening scales from a modified National Institute of Health Stroke Scale (NIHSS). The USS has the potential to streamline LVO screening with enhanced accuracy using multiple validated stroke scales simultaneously.

We prospectively applied eight stroke screening scales (NIHSS, BE-FAST, VAN, LAMS, FAST-ED, EMS RACE, 3-ISS, and PASS) to 199 stroke activations between January 2021 to December 2023. These values were recalculated with the USS calculator which incorporates inputs from a modified NIHSS score, including up to two additional points for hand grip strength. A Bland-Altman analysis was conducted to assess agreement between manual and USS-calculated scores.

The NIHSS showed a percentage error of -8.24% and a mean difference of -0.97 (LoA: -3.88 to 1.93). The BE-FAST scale exhibited a percentage error of -14.72% and a mean difference of -0.12 (LoA: -0.85 to 0.60). The VAN scale had a percentage error of -21.76% and a mean difference of -0.11 (LoA: -0.83 to 0.62). The LAMS scale had a percentage error of 6.59% and a mean difference of 0.15 (LoA: -1.20 to 1.50). The FAST-ED scale had a percentage error of -4.82% and a mean difference of -0.15 (LoA: -2.10 to 1.80). The EMS-RACE scale had a percentage error of -9.99% and a mean difference of -0.39 (LoA: -3.20 to 2.42). The 3-ISS scale exhibited the highest percentage error of -29.36% and a mean difference of -0.54 (LoA: -2.47 to 1.39). The PASS scale had the lowest percentage error at -2.86% and a mean difference of -0.04 (LoA: -0.66 to 0.58). The combined percentage error for all scales was -8.44%, increasing slightly to -8.61% when excluding the NIHSS score. Excluding both NIHSS and 3-ISS reduced the combined error to -5.44%.

Our findings demonstrate a general agreement between the manual and USS-calculated scores, with the strongest concordance observed in PASS, FAST-ED, and LAMS. Although some scales exhibited larger discrepancies, the moderate overall combined percentage error suggests that USS-calculated scores are generally consistent with manual calculations. These findings support the potential of the USS software to streamline LVO stroke screening, although further validation is necessary.