Abstract WP112: A Clinical Score To Predict Reduced Ejection Fraction In Acute Ischemic Stroke

Introduction: Transthoracic echocardiography (TTE) is valuable in evaluating stroke etiology. A left ventricular (LV) thrombus, the identification of which can immediately impact clinical management, is rarely seen unless the ejection fraction (EF) is less than 50%. A clinical prediction score to identify stroke patients with EF less than 50% may help guide the timing of TTE after stroke. Methods: The CAESAR registry includes all patients with ischemic stroke at our medical center. We derived a clinical prediction score using CAESAR data from calendar years 2011-2016. We included all patients who underwent TTE and had a quantitative EF measurement. We selected clinical factors, laboratory values, and vital signs based on biological plausibility and the results of multiple logistical regression with backward selection. Results: Of 2,116 patients with ischemic stroke from 2011-2016, 1,045 patients had an EF measurement. The mean age was 63 (SD, 15), 49% were women, and the mean EF was 63% (SD, 14%). Reduced EF was identified in 171 patients (16%). Our final model comprised historical variables (coronary disease, heart failure, and chronic kidney disease) and clinical parameters from the time of admission (NIHSS score, heart rate, serum potassium, and serum creatinine). The model AUC was 0.80 (95% CI, 0.76-0.85) and model calibration was good (Figure). At a predicted probability threshold of 0.1, the score’s sensitivity for reduced EF was 80% with a specificity of 62%. In a sensitivity analysis excluding patients with atrial fibrillation, the AUC was 0.77 (95% CI 0.70-0.83) and calibration remained good. Conclusions: We derived a clinical score with good performance for predicting reduced EF in acute ischemic stroke patients. If externally validated, such a score may help identify which patients are most likely to benefit from an expedited inpatient TTE.