Despite the proven superiority of the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) over the Cockcroft–Gault (CG) formula, current guidelines recommend the latter to assess renal function in patients treated with non-vitamin K antagonist oral anticoagulants (NOACs). To assess the relationship between the CG and the recommended CKD-EPI formulas, in a cohort of atrial fibrillation (AF) patients treated with NOACs, and the misclassifications introduced by the CG formula for renal function levels, we estimated renal function with three equations: CG, CKD-EPI with body surface adjustment (1.73 mL/m2, CKD-EPI) and without such adjustment (CKD-EPI_noBSA), in all consecutive AF patients discharged from NOACs from the Cardiology Division of a main city hospital between February 1st and May 31st 2018. We compared the different estimates of glomerular filtration rate and potential renal function class misclassifications. We reclassified 37/115 patients (32.1%) when switching from the CG to the CKD-EPI; and 24/115 (20.8%) switching from the CG to the CKD-EPI_noBSA formulas. Class reallocation was distributed across all levels of renal function, but mostly affected the “hyper-normal” function. In estimating consequences of such reallocation, a change in NOAC dosages would have occurred in 10/115 patients (8.7%) when switching from the CG to the CKD-EPI formula and in 10/115 patients when switching from the CG to the CKD-EPI_noBSA formula. Although the CG method has been traditionally used to calculate renal function in all NOAC studies, a renal dysfunction class reallocation occurs in a substantial fraction of hospital-admitted AF patients with the use of better estimates of renal function.