External Validation of SAFE Score to Predict Atrial Fibrillation Diagnosis after Ischemic Stroke: A Retrospective Multicenter Study

Abstract

Introduction. The screening for atrial fibrillation (AF) scale (SAFE score) was recently developed to provide a prediction of the diagnosis of AF after an ischemic stroke. It includes 7 items: $age\ge 65$ years, bronchopathy, thyroid disease, cortical location of stroke, intracranial large vessel occlusion, NT-ProBNP ≥250 pg/mL, and left atrial enlargement. In the internal validation, a good performance was obtained, with an $AUC=0.88$ (95% CI 0.84-0.91) and sensitivity and specificity of 83% and 80%, respectively, for $scores\ge 5$ . The aim of this study is the external validation of the SAFE score in a multicenter cohort. Methods. A retrospective multicenter study, including consecutive patients with ischemic stroke or transient ischemic attack between 2020 and 2022 with at least 24 hours of cardiac monitoring. Patients with previous AF or AF diagnosed on admission ECG were excluded. Results. Overall, 395 patients were recruited for analysis. The SAFE score obtained an $AUC=0.822$ (95% CI 0.778-0.866) with a sensitivity of 87.2%, a specificity of 65.4%, a positive predictive value of 44.1%, and a negative predictive value of 94.3% for a SAFE $score\ge 5$ , with no significant gender differences. Calibration analysis in the external cohort showed an absence of significant differences between the observed values and those predicted by the model (Hosmer-Lemeshow’s test 0.089). Conclusions. The SAFE score showed adequate discriminative ability and calibration, so its external validation is justified. Further validations in other external cohorts or specific subpopulations of stroke patients might be required.