Simple variables predict miserable outcome after intravenous thrombolysis

Abstract

Background and purpose: To test the predictability of miserable outcome amongst ischaemic stroke patients receiving intravenous thrombolysis (IVT) based on a simple variables model (SVM) and to compare the model's predictive performance with that of an existing score which includes imaging and laboratory parameters (DRAGON).

Methods: The SVM consists of the parameters age, independence before stroke, normal Glasgow coma verbal score, able to lift arms and able to walk. In a derivation cohort (n = 1346) and a validation cohort (n = 638) of consecutive IVT-treated stroke patients, the probability estimated by SVM and the observed occurrence of miserable 3-month outcome (modified Rankin score 5-6) were compared. The performances of SVM and the DRAGON score were compared. The area under the receiver operating curve (AUC) (95% confidence interval, CI) and the bootstrapping approach were used to compare the predictive performance.

Results: The AUCs to predict miserable outcome in the derivation cohort were 0.807 (95% CI 0.774-0.838) using the SVM and 0.822 (0.790-0.850) using the DRAGON score (P = 0.3). For the validation cohort, AUCs were 0.786 (0.742-0.829) for the SVM and 0.809 (0.774-0.845) for the DRAGON score (P = 0.23). Only one patient with an SVM probability of >70% for miserable outcome in either cohort had a good outcome whilst 83% had a miserable outcome. An online SVM calculator to estimate the probability of miserable outcome for individual patients is available under http://www.unispital-basel.ch/SVM-Tool.

Conclusion: The SVM was similar in accuracy to the DRAGON score for predicting miserable outcome after IVT. As these simple variables are available already at the pre-hospital stage, the SVM may facilitate and accelerate pre-hospital triage of patients at high risk for miserable outcome after IVT towards endovascular treatment.

Keywords: intravenous thrombolysis (IVT); outcome prediction; prognosis; stroke.