ObjectiveA tool to stratify the risk of stroke recurrence in patients with embolic stroke of undetermined source (ESUS) could be useful in research and clinical practice. We aimed to determine whether a score can be developed and externally validated for the identification of patients with ESUS at high risk for stroke recurrence.MethodsWe pooled the data of all consecutive patients with ESUS from 11 prospective stroke registries. We performed multivariable Cox regression analysis to identify predictors of stroke recurrence. Based on the coefficient of each covariate of the fitted multivariable model, we generated an integer-based point scoring system. We validated the score externally assessing its discrimination and calibration.ResultsIn 3 registries (884 patients) that were used as the derivation cohort, age, leukoaraiosis, and multiterritorial infarct were identified as independent predictors of stroke recurrence and were included in the final score, which assigns 1 point per every decade after 35 years of age, 2 points for leukoaraiosis, and 3 points for multiterritorial infarcts (acute or old nonlacunar). The rate of stroke recurrence was 2.1 per 100 patient-years (95% confidence interval [CI] 1.44–3.06) in patients with a score of 0–4 (low risk), 3.74 (95% CI 2.77–5.04) in patients with a score of 5–6 (intermediate risk), and 8.23 (95% CI 5.99–11.3) in patients with a score of 7–12 (high risk). Compared to low-risk patients, the risk of stroke recurrence was significantly higher in intermediate-risk (hazard ratio [HR] 1.78, 95% CI 1.1–2.88) and high-risk patients (HR 4.67, 95% CI 2.83–7.7). The score was well-calibrated in both derivation and external validation cohorts (8 registries, 820 patients) (Hosmer-Lemeshow test χ2: 12.1 [p = 0.357] and χ2: 21.7 [p = 0.753], respectively). The area under the curve of the score was 0.63 (95% CI 0.58–0.68) and 0.60 (95% CI 0.54–0.66), respectively.ConclusionsThe proposed score can assist in the identification of patients with ESUS at high risk for stroke recurrence.