P.136 Evaluating outcome prediction models in endovascular treatment for acute ischemic stroke using baseline, treatment and post-treatment variables. (November 2021)

• Record Type: Journal Article
• Authors: Ospel, JM; Ganesh, A; Kappelhof, M; McDonough, R; Nogueira, R; McTaggart, R; Menon, B; Demchuk, A; Poppe, A; Tymianski, M; Hill, M; Goyal, M
• Is Part Of: Canadian journal of neurological sciences. Volume 48(2021)Supplement S3
• Page Start: S58
• Page End: S58
• Publication Date: 2021-11
• Subjects: Neurology -- Periodicals; Nervous system -- Surgery -- Periodicals; Electronic journals; 616.8
• DOI: 10.1017/cjn.2021.412
• Languages: English
• ISSNs: 0317-1671

Abstract

Abstract : Background: Predicting outcomes after endovascular treatment (EVT) for acute ischemic stroke with baseline variables remains a challenge. We assessed the performance of stroke outcome prediction models for EVT in acute ischemic stroke in an iterative fashion using baseline, treatment-related and post-treatment variables. Methods: Data from the ESCAPE-NA1 trial were used to build 4 outcome prediction models using multi-variable logistic regression: Model 1 included baseline variables only that are available prior to treatment decision-making, model 2 included additional treatment-related variables, model 3 additional early post-treatment variables, and model 4 additional late post-treatment variables. The primary outcome was 90-day modified Rankin Scale score 0-2. Model performance was compared using the area under the curve (AUC). Results: Among 1, 105 patients, good outcome was achieved by 666 (60.3%). When using baseline variables only (model 1), the AUC was 0.74 (95%CI:0.71-0.77); this iteratively improved when treatment and post-treatment variables were added to the models (model 2: AUC 0.77, 95%CI: 0.74-0.80, model 3: AUC 0.80, 95%CI:0.77-0.83, model 4: AUC 0.82, 95%CI:0.79-0.85). Conclusions: Predicting EVT outcomes using baseline variables alone is inaccurate in one in four patients, and may be inappropriate for patient selection. Even the most comprehensive models with treatment-related and post-treatment factors involve considerable uncertainty.