Development and validation of a risk score to predict 30-day mortality in patients with atrial fibrillation-related stroke: GPS-GF score

Objective: Stroke due to atrial fibrillation (AF) is common and frequently devastating. However, there is no specific tool to accurately estimate the risk of mortality. This study aims to develop and validate a comprehensive risk score for predicting 30-day mortality in the patients with AF-related stroke. Methods: A retrospective multi-center clinical study was performed based on the data from the project of secondary prevention of stroke in patients with nonvalvular AF in Shaanxi province, China. A total of 1077 consecutive patients were randomly classified into derivation (66.7%, n=718) and internal validation cohort (33.3%, n=359). Independent predictors of 30-day mortality were obtained using univariate and multivariable analyses. The area under the receiver operating characteristic curve (AUROC) and the Hosmer-Lemeshow test were used to assess model discrimination and calibration, respectively. Results: Two hundred patients (18.6%) of 1077 participants died within 30days. An 8-point score was generated from the five independent predictors for 30-day mortality including Glasgow Coma Scale, pneumonia, midline shift on brain images, blood glucose, and female sex, which was named GPS-GF. The resulting score showed good discrimination (AUROC) and well calibrated (Hosmer-Lemeshow test) in the derivation (0.909; p=0.102) and internal validation cohort (0.922; p=0.153). Compared with iScore, the GPS-GF score exhibited remarkably better discriminative power and predictive accuracy regarding the 30-day mortality in patients with AF-related stroke. Conclusion: The GPS-GF score is a simple and valid tool for predicting 30-day mortality in patients with AF-related stroke.