A score to predict the stroke recurrence of patients with embolic stroke of undetermined source

Abstract

Aim: We aimed to develop a score and validate it in a prospective cohort to identify the patients with ESUS at high risk for stroke recurrence.

Methods: We assessed the stroke recurrence in ESUS patients of the Third China National Stroke Registry. We performed multivariable logistic regression analysis to identify predictors of stroke recurrence in the derivation cohort. Based on the coefficient of each covariate of the fitted multivariable model, we generated an integer-based point scoring system. We validated the score in the validation cohort assessing its discrimination and calibration.

Results: 2415 patients were included: 1611 in the derivation and 804 in the validation sample. We developed a scoring system (0-15 points) by assigning 2 points for hypertension, 3 points for diabetes mellitus, 4 points for multiple stage infarction, 2 points for watershed involved infarction, 1 points for left atrial diameter index (per increasing 2.5 mm/m2) and 3 points for without statin at discharge. The rate of stroke recurrence was 5.9% per year (95% CI 4.2-7.6%) in patients with low risk(a score of 0-5), 9.4% (7.3-11.5%) in patients with intermediate risk (6-10), and 26.8% (16.5-37.1%) in patients with high risk (11-15). The AUC (area under curve of receiver operator characteristic curve) of the score in the derivation cohort and validation cohort was, respectively, 0.60 (0.55-0.65) and 0.63 (0.56-0.70). The score was well calibrated both in the derivation cohort (p = 0.36) and validation cohort (p = 0.26) with the Hosmer-Lemeshow test.

Conclusion: The developed score can improve risk stratification after ESUS in secondary care settings.

Keywords: ESUS; Prediction; Prognosis.