Atrial fibrillation and prediction of mortality by conventional clinical score systems according to the setting of care

Background: Atrial fibrillation (AF) is associated with high morbidity and mortality, also among anticoagulated patients. Our aim was to evaluate the predictive role for long-term mortality of a series of risk stratification scores associated with cardiovascular or thromboembolic outcomes (CHADS2, CHA2DS2-VASc, ATRIA, TIMI-AF), and bleeding complications (HAS-BLED) in an unselected population of patients with AF. Methods: Single center, observational, prospective registry of consecutive patients with AF, undergoing clinical/echocardiographic evaluation in a University Hospital, as either in-patients or out-patients. We assessed the role of each single score as predictors of long-term survival according to clinical setting. Results: We enrolled 1051 patients, mean age 72 ± 12 years, who were followed for 797 ± 298 days. All the tested scores showed a good performance in prediction of mortality, together with several clinical factors (older age, chronic heart failure, diabetes, renal impairment, previous transient ischemic attack, left ventricular ejection fraction). The values at C-statistics ranged between modest (0.608–0.684) of inpatients to good (0.708–0.751) in outpatients without any statistical difference between the scores, excepted a lower performance of HAD-BLED. Conclusions: Risk scores currently adopted for decision making on starting oral anticoagulation provide good prediction of long-term survival in unselected AF patients, especially in the outpatient setting.