Atrial fibrillation (AF) is considered one of the main causes of ischemic stroke. The CHA2DS2VASC score can predict the stroke risk. Proper anticoagulation can significantly reduce such risk. Anticoagulation involves a risk of bleeding, which can be predicted by the HAS-BLED score. The non-documentation of both AF risk scores in the medical notes of patients presenting with acute or paroxysmal AF has alerted our team, for fear of missing administering the proper anticoagulation. A baseline audit showed that the CHA2DS2VASC score was documented in 27% of such patients, while the HAS-BLED score was documented in 5% of them. A quality improvement project was planned and included two PDSA cycles over a period of five months. The first cycle was based on raising the awareness of the junior doctors on risk assessment of AF patients and the importance of anticoagulation prescription. This involved educational posters, emails, and presentations. On the second cycle, the baseline audit showed a 69% improvement (from 27% to 46%) in the rate of CHA2DS2VASC score assessment and a 140% improvement (from 5% to 12%) in the rate of HAS-BLED scoring. However, the rate of anticoagulation prescription remained almost the same in both cycles (55% and 52%). A doctors’ survey showed that they are not quite comfortable prescribing long-term anticoagulation on acute care units because of the inadequate assessment of the bleeding risk, and the associated comorbidities that can disturb the action of the oral anticoagulants. They preferred to refer the patient to a specialised AF clinic for further assessment and prescription. On auditing the patient referral on discharge, it was found that 91% of the patient were seen at the AF clinic in 2 weeks. The remainders were referred to their GP for further assessment and prescription. That seemed to be a safer way of anticoagulation prescription for such patients.