Stroke Triage Alpha Rescue Score (STARS) : Prehospital phone-based tool to identify cerebral large vessel occlusion

Introduction. - Prehospital identification of stroke patients with large vessel occlusion (LVO) is crucial to optimize transport to an endovascular thrombectomy (EVT)-capable center. Existing scores require medical or paramedical expertise and specific teachings. We aimed to validate a simple prehospital phone-based score for LVO identification. Methods. - First, the Stroke Triage Alpha Rescue Score (STARS) was designed among a retrospective cohort of 3452 stroke alerts from the Emergency Medical Dispatch Centers (EMDC) registry. Phone-based neurological assessment, characteristics, final diagnosis, and brain imaging were analyzed. The items weighting was based on odds ratios (ORs) of significant variables after multivariate analysis. Secondly, STARS was assessed by dispatchers for all suspected strokes in a prospective cohort. Receiver operating curve (ROC) and predictive performances were calculated for the external validation. Results. - In the retrospective cohort, the best balance between specificity and sensitivity was obtained with a combination of six items (presence or absence of: upper, lower limbs motor deficits, language impairment, facial palsy, consciousness disorder, smoking). Each item value was determined according to their respective OR to obtain a total STARS between 0 and 13, with the most favorable cut-off higher or equal to 7 points. Then, STARS was prospectively assessed in 799 stroke alerts between March 2021 and March 2022. LVO was finally detected in 63 of 653 patients (9.6%). The area under the ROC was 0.83 [95% CI: 0.78; 0.88]. STARS > 7 had 0.83 accuracy, 0.65 sensitivity, and 0.85 specificity for detecting LVO. Conclusion. - STARS is a simple and accurate phone-based score, without specific teaching, to detect LVO in prehospital network. STARS correctly classified 83% of stroke alerts and could be useful to identify patients with LVO eligible for EVT. # 2025 Elsevier Masson SAS. All rights are reserved, including those for text and data mining, AI training, and similar technologies.