A large quantity of ischemic stroke events occur in patients hospitalized for non-stroke-related reason. No scale has been developed to identify the large vessel occlusion (LVO) among inpatient stroke alerts. We aimed to develop a novel evaluation scale to predict LVO from in-hospital stroke alerts. Data from consecutive in-hospital stroke alerts were analyzed at a single high volume stroke center between January 2016 and October 2020. We developed a predictive scale based on the first half of patients (training group) using multivariate logistic regression and evaluated it in the remaining half of patients (validation group) adopting receiver operating curve. Receiver operating characteristics of the scale were analyzed to evaluate its value for the detection of LVO. A total of 243 patients were enrolled for further study, among them, 94 (38.7%) had confirmed LVO. Three risk factors independently predicted the presence of LVO: recent cardiac or pulmonary procedure (1 point), neurological deficit scale (≥1: 2 points), and history of atrial fibrillation (1 point). The CAPS scale was generated based on predictive factors and demonstrated highly effective discrimination in identifying the presence of LVO in the training group (area under curve = 0.956) and the validation group (area under curve = 0.940). When the score ≥2, CAPS scale showed 97.9% sensitivity, 79.2% specificity, 74.8% positive predictive value, and 98.3% negative predictive value for discriminating LVO. CAPS scale was developed for identifying LVO among inpatient stroke alerts with high sensitivity and specificity, which may help to quickly prompt responses by appropriate stroke teams.
Read full abstract