Background and purpose The purpose of this work was to develop a robust and user-friendly tool-kit to aid decision-making regarding Mechanical Thrombectomy (MT), based on easily-available patient variables that could accurately predict long-term functional outcome following MT. The aim of this study was to test the accuracy of four models developed to predict four types of post-treatment outcome.

Methods Data from patients with anterior circulation stroke who underwent MT between October 2009 and January 2018 (n=240) were identified from our Mechanical Thrombectomy Database. Four models were developed from the data to predict the four outcomes of interest. The patient explanatory variables were age, sex, initial NIHSS, ASPECTS, Collateral Score and Glasgow Coma Scale score.

Model 1: Prediction of survival: mRS score of 0–5 (alive) or 6 (dead).

Model 2: Prediction of good/poor outcome: mRS score of 0–3 (good), or 4–6 (poor).

Model 3: Prediction of good/poor outcome: mRS score of 0–2 (good), or 3–6 (poor).

Model 4: Prediction of mRS category: mRS score of 0–2 (no disability), 3 (minor disability), 4–5 (severe disability) or 6 (dead).

The accuracy and discriminative power of each predictive model were tested.

Results Prediction of survival was 87% accurate (area under the curve 0.87). Prediction of good/poor outcome was 91% accurate (area under the curve 0.94) for Model 2 and 95% accurate (area under the curve 0.98) for Model 3. Prediction of mRS category was 74% accurate, and increased to 95% using the ‘one-score-out rule’.

Conclusion This novel tool-kit is easy to use and provided accurate estimations of outcome. It can be used to aid decision making and consent for mechanical thrombectomy.

Disclosures S. Nayak: None. H. Wright: None. A. Bazarova: None. M. Raseta: None.