Dabigatran is approved for stroke risk reduction in patients with nonvalvular atrial fibrillation (NVAF). Data from diverse clinical practice settings will help establish whether the risk:benefit ratio seen in clinical trials is comparable with routine clinical care. This study aimed to compare the safety and effectiveness of dabigatran and warfarin in clinical practice. We undertook a propensity score-matched (PSM) cohort study (N=12,793 per group; mean age 74) comparing treatment with dabigatran or warfarin in the US Department of Defense claims database, October 2009 to July 2013. Treatment-naïve patients with first prescription claim for dabigatran (either FDA-approved dose) or warfarin between October 2010 and July 2012 (index) and a diagnosis of NVAF during the 12 months before index date were included. Primary outcomes were stroke and major bleeding. Secondary outcomes included ischaemic and haemorrhagic stroke, major gastrointestinal (GI), urogenital or other bleeding, myocardial infarction (MI) and death. Time-to-event was investigated using Kaplan-Meier survival analyses. Outcomes comparisons were made utilising Cox-proportional hazards models of PSM groups. Dabigatran users experienced fewer strokes (adjusted hazard ratio [95 % confidence intervals] 0.73 [0.55–0.97]), major intracranial (0.49 [0.30–0.79]), urogenital (0.36 [0.18–0.74]) and other (0.38 [0.22–0.66]) bleeding, MI (0.65 [0.45–0.95]) and deaths (0.64 [0.55–0.74]) than the warfarin group. Major bleeding (0.87 [0.74–1.03]) and major GI bleeding (1.13 [0.94–1.37]) was similar between groups and major lower GI bleeding events were more frequent (1.30 [1.04–1.62]) with dabigatran. In conclusion, compared with warfarin, dabigatran treatment was associated with a lower risk of stroke and most outcomes measured, but increased incidence of major lower GI bleeding.