Introduction: The effectiveness of endovascular thrombectomy for acute cerebral large vessel occlusion (LVO) was proved, but many patients did not received such interventions because capable operators were not placed at all hospitals. If the type of stroke [large vessel occlusion, subarachnoid hemorrhage (SAH), intracranial hemorrhage (ICH), other cerebral infarction (CI)] was predicted at prehospital, better access to appropriate interventions were capable. We, thus, developed the clinical prediction rules to classify the type of stroke who were suspected to suffer acute stroke at prehospital, and validated them.
Methods: We analyzed consecutive 1,229 patients who were suspected to suffer acute stroke from June, 2015 to March 2016. We obtained the history and physical signs at prehospital from paramedics and final diagnosis from hospital transferred. We constructed multivariate logistic regression models for 1) LVO, 2) SAH, 3) ICH, 4) CI, and developed the clinical prediction rules for each type. We prospectively validated the rules with another consecutive patients from August 2016 to July 2017 using mobile application.
Results: In the derivation cohort, 104 LVO, 57 SAH, 169 ICH, and 183 CI were observed. The area under the receiver operating curve (AUC) of the rules were 0.90 for LVO, 0.90 for SAH, 0.85 for ICH, and 0.65 for CI. The validation cohort of 932 patients showed the sensitivity and specificity of the rules were 0.53 and 0.95 for LVO, 0.73 and 0.96 for SAH, 0.52 and 0.85 for ICH, 0.63 and 0.70 for CI. The AUCs of LVO, SAH, ICH, and CI were 0.85, 0.96, 0.77, and 0.67, respectively.
Conclusions: The clinical prediction rule calculated by paramedics at prehospital can easily classify the patients who suspected to have stroke into LVO, SAH, ICH, and CI with excellent performance. By applying the rules, more patients would receive appropriate interventions without unnecessary delay.