Abstract 086: Design and validation of a scale for the prediction of large vessel occlusion in stroke patients in the Latin American population Escala Predictiva en Ictus de Oclusion EsPIO

Introduction There are multiple limitations in the approach and management of ischemic stroke in the Latin American people, particularly, in the management of patients requiring thrombectomy. Although there are numerous scales to predict large vessel occlusion (LVO), they have not been validated in this population; therefore, our objective was to design and validate a scale to predict large vessel occlusion in this populace, in order to provide predictive tools for the decision making in the stroke patients. Methods A multivariate logistic regression model was constructed from the National Institutes of Health Stroke Scale (NIHSS) items, using the backward stepwise methodology to select the items that best predicted the model, eliminating items with p ≥ 0.05. The significant items comprised the EsPIO ( Escala Precdictiva en Ictus de Oclusión, i.e. Occlusion Prediction Scale in Stroke). For its creation, we use a validation cohort with the 66% of patients, in whom neuroimaging was performed to assess cervical and cerebral vessels. Sensitivity, specificity, ROC curves, positive predictive value (PPV) and negative predictive value (NPV) were evaluated. The remaining 33% of the cohort was used for a validation cohort. Finally, the results of derivation and the validation cohort in the EsPIO scale were compared. Results We included 668 patients; LVO was identified in 15.5%. In the validation cohort, the four significant items were gaze disturbance, arm motor function, language, and extinction or inattention, with which the EsPIO scale was constructed, ranging from 0 to 11 points. With >=3 points, it had a sensitivity (S) of 89% and specificity (E) of 46%, with >=4 points it had S 85% and E 60%, choosing this point, PPV 19%, and NPV 97% when the frequency was 10%. The ROC curve was 0.831 (95%CI 0.771‐0.891). The derivation and validation cohort were compared, with no significant differences. Conclusion The results obtained with the creation of the EsPIO scale show very high negative predictive values. This scale may aid in determining which patients may benefit from being translated to thrombectomy‐capable centers.