We recommend that all patients with AF or AFL (paroxysmal, persistent, or permanent), should be stratified using a predictive index for stroke risk (eg, CHADS<sub>2</sub>) and for the risk of bleeding (eg, HAS-BLED), and that most patients should receive either an OAC or ASA. (Recommendation Strong, Quality High)<br/><br/>We recommend that most patients at intermediate risk of stroke (CHADS<sub>2</sub> = 1) should receive OAC therapy. (Recommendation Strong, Quality High)<br/><br/> We suggest, based on individual risk/benefit considerations, that ASA is a reasonable alternative for some. (Recommendation Conditional, Quality Moderate)<br/><br/><b>Values and preferences: </b>This recommendation places a relatively high value on comparisons with warfarin showing that dabigatran and apixaban have greater efficacy and rivaroxaban has similar efficacy for stroke prevention; dabigatran and rivaroxaban have no more major bleeding and apixaban has less; all three new OACs have less intracranial hemorrhage and are much simpler to use. The recommendation places less value on the following features of warfarin: long experience with clinical use, availability of a specific antidote, and a simple and standardized test for intensity of anticoagulant effect. The preference for one of the new OACs over warfarin is less marked among patients already receiving warfarin with stable INRs and no bleeding complications.