Development and external validation of a prognostic model for ischaemic stroke after surgery

Background: There is an under-recognised patient cohort at elevated risk of postoperative ischaemic stroke. We aimed to develop and validate a prognostic model for the identification of such patients at high risk of ischaemic stroke within 1 yr after noncardiac surgery. Methods: This was a hospital registry study of adult patients undergoing noncardiac surgery between 2005 and 2017 at two independent healthcare networks in Massachusetts, USA without a preoperative indication for therapeutic anticoagulation. Logistic regression was used to fit a model from a priori defined candidate predictors for the outcome 1 yr postoperative ischaemic stroke. To enhance clinical applicability, the model was simplified to a scoring system and externally validated. Results: In the development (n=107 756) and validation (n=141 724) cohorts, 1.4% and 0.5% of patients had an ischaemic stroke up to 1 yr postoperatively. The final model included 13 variables (patient characteristics, comorbidities, procedural factors), considering sub-models conditional on a previous history of ischaemic stroke. Areas under the curve were 0.89 (95% confidence interval 0.89-0.90) and 0.88 (95% confidence interval 0.86-0.89) in the development and validation cohorts. Decision curve analysis indicated positive net benefits superior to other prediction instruments. Conclusions: Stroke after surgery (STRAS) screening can reliably identify patients with a high risk for ischaemic stroke during the first year after surgery. A STRAS-guided risk stratification may inform the recruitment to future randomised trials testing the efficacy of treatments for the prevention of postoperative ischaemic stroke.