Validation of the prehospital stroke scales as a tool for in-hospital large vessel occlusion stroke: whether we satisfied?

Abstract

Background

Prehospital stroke severity scales have been widely used to identify whether community stroke patients presented with large vessel occlusion (LVO) or not. However, whether these scales are also applicable to in-hospital stroke patients remains unknown.

Purpose

We aim to validate and compare the predictive capability of these scales for these patients.

Material and methods

From January 2016 to October 2020, a total of 243 patients who activated in-hospital stroke alerts, were included in this study. The area under the curve (AUC) was used to assess the predictive ability of five scales (Field Assessment Stroke Triage for Emergency Destination [FAST-ED], Rapid Arterial Occlusion Evaluation [RACE], Los Angeles Motor Scale [LAMS], Cincinnati Prehospital Stroke Severity Scale [CPSSS], and Prehospital Acute Stroke Severity scale [PASS]) for LVO. In addition, multivariable logistic analysis was adopted to determine the predictors of LVO in our patients cohort.

Results

Finally, 94 (38.7%) patients were confirmed presence of persistent LVO. The AUC for the FAST-ED, RACE, LAMS, CPSSS, and PASS scales to predict the presence of LVO in patients activating in-hospital stroke alerts were 0.82, 0.89, 0.86, 0.81, and 0.79, respectively. After multivariable analysis, baseline NIHSS (adjusted odds ratio [OR] = 1.160, 95% confidence interval [CI] = 1.110–1.212; P < 0.001) atrial fibrillation (adjusted OR = 2.940, 95% CI = 1.387–6.230; P = 0.005) and cardiac/pulmonary procedure (adjusted OR = 6.861, 95% CI = 2.437–19.315; P < 0.001) remained independent predictors of LVO.

Conclusion

The prehospital stroke scales also showed good predictive capabilities in discriminating LVO among inpatients who activated stroke alerts. However, given that inpatients' history is more readily available, a specifically designed in-hospital stroke scale that combines stroke severity and history is warranted.