Differences among Clinical Classification Schemes for Predicting Stroke in Atrial Fibrillation: Implications for Therapy in Daily Practice

Abstract

Several clinical classification schemes (CCSs) for predicting stroke in nonvalvular atrial fibrillation (NVAF) have been developed to help identify patients eligible for anticoagulation. To estimate the agreement in predicting the risk of stroke among four widespread CCSs, and to determine their implications for thromboprophylaxis in clinical practice. The authors conducted a prospective, multicenter, observational study of adults with NVAF in 12 emergency departments (EDs) in July 2000 and February 2001. The proportions of patients classified as having high, moderate, and low risk of stroke among the following CCSs were compared: the Atrial Fibrillation Investigators (AFI), the Stroke Prevention in Atrial Fibrillation (SPAF), the CHADS(2) (an acronym for congestive heart failure, hypertension, age >75 years, diabetes mellitus, and prior stroke or transient ischemic attack), and the American College of Chest Physicians (ACCP). One thousand two hundred twenty patients were included. The proportions of patients stratified as having high/moderate/low risk of stroke according to each CCS were: 70%/22%/8% (AFI), 38%/41%/21% (SPAF), 13%/45%/42% (CHADS(2)), and 86%/7%/7% (ACCP). The agreement was medium between AFI and ACCP (kappa = 0.52) and poor among the rest of them (AFI/SPAF, kappa = 0.01; AFI/CHADS(2), kappa = 0.02; SPAF/CHADS(2), kappa = 0.18; SPAF/ACCP, kappa = 0.11; CHADS(2)/ACCP, kappa = 0.03). The agreements in selecting patients as eligible for antiplatelet therapy or anticoagulation were: AFI/SPAF, kappa = 0.45; AFI/CHADS(2), kappa = 0.22; AFI/ACCP, kappa = 0.91; SPAF/CHADS(2), kappa = 0.47; SPAF/ACCP, kappa = 0.11; CHADS(2)/ACCP, kappa = 0.03. In the ED population studied, these CCSs showed relevant differences in the risk of stroke stratification and, therefore, in the identification of patients with NVAF eligible for anticoagulation.