Bleeding risk represents a major concern in anticoagulated patients with atrial fibrillation (AF). Several bleeding prediction scores have been described: HAS-BLED, ATRIA, HEMORR2HAGES and ORBIT. Of these, only HAS-BLED considers quality of anticoagulation control amongst vitamin K antagonist (VKA) users. We hypothesised that predictive value of bleeding risk scores other than HAS-BLED could be improved incorporating time in therapeutic range (TTR) in warfarin-treated patients. Of the 127 adjudicated major bleeding events, 21.3% of events occurred in ‘low-risk’ HAS-BLED category (1.8 per 100 patient-years), compared to higher proportions (≥50% of events; ~2.5 per 100 patient-years) in ‘low-risk’ categories for other scores. Only the ‘low-risk’ HAS-BLED category was associated with the absence of investigator-defined major bleeding events (OR: 1.46;95% CI: 1.00–2.15). ‘High’ or ‘medium/high’ risk categories for the HAS-BLED (p = 0.023) or ORBIT (p = 0.022) scores, respectively, conferred significant risk for adjudicated major bleeding events. On Cox regression analysis, adjudicated major bleeding was associated only with HAS-BLED (HR: 1.62;95% CI: 1.06–2.48) and ORBIT (HR: 1.83;95% CI: 1.08–3.09) ‘high-risk’ categories. Adding ‘labile INR’ (TTR < 65%) to ORBIT, ATRIA and HEMORR2HAGES significantly improved their reclassification and discriminatory performances. In conclusion, HAS-BLED categorised adjudicated major bleeding events in low-risk and high-risk patients appropriately, whilst ORBIT and ATRIA categorised most major bleeds into their ‘low-risk’ patient categories. Adding TTR to ORBIT, ATRIA and HEMORR2HAGES led to improved predictive performance for major bleeding.