To the Editor In a recent Original Investigation in JAMA Internal Medicine, Graham and colleagues1 compared outcomes among patients with nonvalvular atrial fibrillation (NVAF) treated with dabigatran or rivaroxaban. Using Medicare claims data, they found that rivaroxaban, compared with dabigatran, was associated with a trend toward less thromboembolic stroke and more intracranial and extracranial bleeding. In the accompanying Editor’s Note, Parks and Redberg state, “The additional information should lead us to prescribe dabigatran over rivaroxaban for patients with atrial fibrillation.”2(p1672) The methodological features of this study do not allow such strong conclusions, and these statements overstate the findings of this analysis, specifically related to drawing causal inferences.