A prediction model for large vessel occlusion in suspected stroke patients based on available prehospital information

Objective:Early and accurate identification of large vessel occlusion(LVO)acute ischemic stroke(AIS)patients is critically important for stroke management.Practicable scales with simple items can facilitate prehospital paramedics distinguishing LVO-AIS patients with high efficiency and help to avoid unnecessary and costly delays.The current study aims to develop a screening tool to predict AIS-LVO patients based on prehospital available data.Method:A total of 251 suspected stroke patients who were transported to the emergency department of our hospital via emergency medical services were consecutively enrolled from August,2020 to January,2022.Data including demographic information,medical history,clinical manifestations,and vital signs were collected.A multivariate logistic regression model was developed based on statistically significant variables selected from univariate analysis.Result:Forty-two patients(16.7%)were diagnosed as LVO-AIS based on imaging validation at admission.A comprehensive model was developed with past medical history factors such as atrial fibrillation and coronary heart disease,vital signs such as systolic blood pressure,and prominent symptoms and signs such as gaze palsy,facial paralysis,and dysarthria.The model showed better diagnostic performance in terms of area under the receiver operating characteristic curves(0.884,95%CI,0.830-0.939),which was higher than other common prehospital prediction scales such as the Face,Arm,Speech,Time test(FAST),the Field Assessment Stroke Triage for Emergency Destination(FAST-ED)scale,and the Gaze-Face-Arm-Speech-Time test(G-FAST).Calibration curve analysis,decision curve analysis,and clinical impact curve analysis further validated the reliability,net benefit,and potential clinical impact of the prediction model,respectively.Conclusion:We conducted a prediction model based on prehospital accessible factors including past history of atrial fibrillation and coronary heart disease,systolic blood pressure,and signs such as gaze palsy,facial palsy,and dysarthria.The prediction model showed good diagnostic power and accuracy for identification of the high-risk patients with LVO and may become an effective tool for the LVO recognition in prehospital settings.Future studies are warranted to refine and validate the model further in order to enhance the accuracy and objectivity of clinical judgments.