External validation of ABCD series scores for predicting early stroke events following transient ischemic attack in a large nationwide registry

In the context of modern guideline-based strategies, new validations of prognostic scores for predicting early stroke risk are needed. We aimed to compare the validity of the ABCD series scores and assess the incremental values of risk components for predicting in-hospital stroke events in patients with transient ischemic attack (TIA).We abstracted data from the Chinese Stroke Center Alliance (CSCA), a nationwide registry with 68,433 TIA patients admitted within 7 days of symptom onset from 1476 hospitals. TIA was defined by time-based criteria according to the World Health Organization (WHO). The discrimination of ABCD, ABCD2, ABCD2-I, and ABCD3 scores for predicting in-hospital stroke events was assessed by the area under the receiver-operating characteristics curves (AUC). The incremental predictive values of added risk predictor were determined by net reclassification improvement (NRI) and integrated discrimination improvement (IDI).A total of 29,286 TIA patients were included, of whom 1466 (5.0%) had in-hospital stroke events. Compared with ABCD2-I score (AUC 0.79, 95% confidence interval [CI] 0.77-0.80), ABCD (AUC 0.58, 95% CI 0.57-0.60), ABCD2 (AUC 0.58, 95% CI 0.56-0.59), and ABCD3 (AUC 0.58, 95% CI 0.56-0.60) had lower predictive utility. An incremental value was observed when adding infarction on DWI (IDI = 0.0597, NRI = 1.1036) into ABCD2 score to be ABCD2-I.The traditional scales utilizing medical history (ABCD, ABCD2, and ABCD3 scores) show fair ability for predicting in-hospital stroke events after TIA, but the ABCD2-I score, which adds infarction on DWI, improves the predictive ability.