Real-time video analysis allows the identification of large vessel occlusion in patients with suspected stroke: feasibility trial of a "telestroke" pathway in Northwestern Switzerland.

BACKGROUND AND AIM: Loss of time is a major obstacle to efficient stroke treatment. Our telestroke path intends to optimize prehospital triage using a video link connecting ambulance personnel and a stroke physician. The objectives were as follows: (1) To identify patients suffering a stroke and (2) in particular large vessel occlusion (LVO) strokes as candidates for endovascular treatment. We have chosen the Rapid Arterial Occlusion Evaluation (RACE) scale for this purpose.

METHODS: This analysis aimed to verify the feasibility of prehospital stroke identification by video assessment. In this prospective telestroke cohort study, we included 97 subjects, in which the RACE score (items: facial palsy, arm and leg motor function, head and gaze deviation, and aphasia or agnosia) was applied, and the assessment videotaped by a trained member of the Emergency Medical Services (EMS) in the field using a mobile device. Each recorded patient video was independently assessed by three experienced stroke physicians from a certified stroke center and compared to the neuroimaging gold standard. Within this feasibility study, the stroke code was not altered by the outcome of the RACE assessment, and all patients underwent the standard procedures within the emergency unit.

RESULTS: We analyzed 97 patients (median age 78 years, 53% women), of whom 51 (52.6%) suffered an acute stroke, 12 (23.5%) of which were due to an LVO and 46 patients had symptoms mimicking a stroke. The sensitivity of stroke identification was 77.8%, and specificity was 53.6%. In regard to the identification of an LVO, sensitivity was 69.4% and specificity was 84.3%. The inter-rater agreement in the RACE-score assessment was ICC = 0.82 (intraclass-correlation coefficient).

CONCLUSION: These results confirm our hypothesis that the local telestroke concept is feasible. It allows correct (i) stroke and (ii) LVO identification in the majority of the cases and thus has the potential to assist in efficient prehospital triage.