Abstract TP126: Discrimination of Ischemic and Hemorrhagic Stroke by Clinical Stroke Risk Prediction Scores

Introduction: We have recently published a scoring system for predicting reduced left atrial appendage emptying velocity, CHIRP 3 M -1 [coronary artery disease (1), congestive heart failure (1), increased left atrial volume index ≥42 mL/m 2 (1), rhythm atrial fibrillation (AF, 1), paroxysmal AF (2), persistent AF (3), longstanding persistent/permanent AF (4), and >moderate mitral regurgitation (-1)]. Clinical stroke prediction scores guide anticoagulation decisions to prevent ischemic stroke in AF. We sought to evaluate comparative performance of CHIRP 3 M -1 , CHA 2 DS 2 -VASc, CHADS 2 , R 2 CHADS 2 , and ATRIA scores in differentiating ischemic from hemorrhagic stroke. Methods: Eligible study participants were identified from The University of Kansas Medical Center stroke registry and included those ≥18 years old who presented in calendar year 2021, with a neurologist-confirmed diagnosis of either ischemic or hemorrhagic stroke. Additional data was collected with retrospective chart review. Differences between clinical risk scores were examined using the Mann-Whitney U (non-parametric) test. Receiver operating characteristic (ROC) curves were evaluated for predictive accuracy. Differences between area under ROC curves (AUCs) were compared using DeLong’s test. Results: Among the 904 stroke patients (age 65.8 ± 14.6 years, 47.7% female, 72.3% White) included in our study, 678 (75.0%) were diagnosed with ischemic stroke and 226 (25.0%) with hemorrhagic stroke. For all stroke risk scoring systems, the average score was higher for ischemic compared to hemorrhagic stroke ( Table 1 , Figure 1 ); CHA 2 DS 2 -VASc (3.1 ± 1.6 vs. 2.7 ± 1.4, p = 0.0007), CHIRP 3 M -1 (1.7 ± 2.7 vs. 1.0 ± 2.1, p < 0.0001) and R 2 CHADS 2 scores (2.3 ± 1.5 vs. 1.8 ± 1.3, p < 0.0001). For discriminating ischemic from hemorrhagic stroke, the CHA 2 DS 2 -VASc score demonstrated the weakest AUC (0.574). The R 2 CHADS 2 score showed modest prediction accuracy (AUC 0.609), while the CHIRP 3 M -1 score achieved a comparable AUC of 0.593 ( Table 1 , Figure 2 ). The DeLong’s test showed AUC for CHIRP 3 M -1 was not statistically different from AUC of R 2 CHADS 2 ( p = 0.3973) or CHA 2 DS 2 -VASc ( p = 0.4546). Conclusions: CHIRP 3 M -1 was not statistically superior to other risk prediction scores to discriminate hemorrhagic from ischemic stroke. As all clinical classification systems have limited ability to make this discrimination, bleeding risk needs to be evaluated separate from the stroke risk for aiding anticoagulation decisions.