Background: Clinical trial data suggest greater likelihood of gastrointestinal (GI) bleeding for patients taking dabigatran 150 mg compared to warfarin, and similar likelihood for patients taking dabigatran 75 mg. Dabigatran-associated GI bleeding is concerning because no antidote exists to reverse the effects of dabigatran, possibly increasing the likelihood of death. Objective: Examine mortality after GI bleeding for patients taking warfarin, dabigatran 150mg, or dabigatran 75mg. Method: We used Medicare inpatient, carrier, and pharmacy claims for beneficiaries age 65 or older with new atrial fibrillation from 11/2010 through 12/2011 who received dabigatran 150mg, dabigatran 75mg, or warfarin. GI bleeds were identified by the primary diagnosis on inpatient claims, and death dates were determined from the CMS enrollment file. Patient demographics, comorbid conditions, and prior health service use were determined from claims incurred during the 12 months prior to the GI bleed. Each patient taking dabigatran 150 mg or 75 mg was matched to a single patient taking warfarin based on patient characteristics at the time of the bleed using an optimal propensity matching algorithm. Results: Of 2,889 GI bleeding events, 2326 (81%), 465 (16%), and 98 (3%) occurred in patients taking warfarin, dabigatran 150 mg, and dabigatran 75 mg, respectively. Compared to patients taking warfarin, patients on dabigatran 150mg were less likely to have renal disease, heart failure, or previous GI bleed, and had fewer previous inpatient days of care. In contrast, patients on dabigatran 75 mg were older and more likely to have renal disease, heart failure, or previous GI bleed compared to patients taking warfarin. The unadjusted 30-day mortality rate was 5.3%, 6.5%, and 8.1% for patients taking warfarin, dabigatran 150 mg, and dabigatran 75 mg and did not differ significantly by anticoagulant type. In the propensity-matched samples, mortality was similar for patients taking dabigatran 150 mg compared to warfarin (6.6% vs. 5.5%; p=0.49), and dabigatran 75 mg compared to warfarin (8.2% vs. 5.1%; p=0.39). Conclusions: Death after GI bleeding does not differ for patients taking dabigatran or warfarin, despite the lack of antidote for dabigatran.