Objective: We evaluated the reliability of eight clinical prediction models for symptomatic intracerebral hemorrhage (sICH) and long-term functional outcome in stroke patients treated with thrombolytics according to clinical practice.
Methods: In a cohort of 169 patients, 60 patients (35.5%) received IV rtPA according to the European license criteria. The remaining patients received off-label IV rtPA and/or were treated with intra-arterial thrombolysis. We used receiver operator characteristic curves to analyze the discriminative capacity of the MSS score, the HAT score, the SITS SICH score, the SEDAN score and the GRASPS score for sICH according to the NINDS and the ECASSII criteria. Similarly, the discriminative capacity of the s-TPI, the iScore and the DRAGON score were assessed for the modified Rankin Scale (mRS) score at 3 months poststroke. An area under the curve (c-statistic) >0.8 was considered to reflect good discriminative capacity. The reliability of the best performing prediction model was further examined with calibration curves. Separate analyses were performed for patients meeting the European license criteria for IV rtPA and patients outside these criteria.
Results: For prediction of sICH c-statistics were 0.66-0.86 and the MMS yielded the best results. For functional outcome c-statistics ranged from 0.72 to 0.86 with s-TPI as best performer. The s-TPI had the lowest absolute error on the calibration curve for predicting excellent outcome (mRS 0-1) and catastrophic outcome (mRS 5-6).
Conclusions: All eight clinical models for outcome prediction after thrombolysis for acute ischemic stroke showed fair predictive value in patients treated according daily practice. The s-TPI had the best discriminatory ability and was well calibrated in our study population.
Keywords: Prognosis; Stroke, acute; Stroke, outcome; Thrombolysis.
Copyright © 2014 Elsevier B.V. All rights reserved.