Special circumstances may require the measurement of the anticoagulant effect of dabigatran etexilate. No data currently link any given coagulation test to bleeding outcomes in patients receiving dabigatran etexilate for atrial fibrillation. Nonvalvular atrial fibrillation patients receiving dabigatran etexilate of 110 mg (DE110) or 150 mg (DE150) were consecutively enrolled. The hemoclot thrombin inhibitor (HTI) assay, prothrombin time, and activated partial thromboplastin time (APTT) measurements were correlated with bleeding events during a prospective follow-up. There were 17 bleeding events (8.2%) in 208 patients (74.7 ± 10.3 years old, 67.9% male, median follow-up: 364 days), whereas 15 patients with bleeding events used DE110. Compared with DE110, the patients receiving DE150 were younger and more often male and had lower HAS-BLED and CHA2DS2VASc scores and better renal function. Patients’ HTI levels were very variable (DE110, 10–90th percentile: 20.5–223.9 ng/ml). A receiver-operator characteristic curve gave a median cutoff HTI level of 117.7 ng/ml to predict bleeding events (C-statistics: 0.65; P = 0.036), but no cutoff could be determined for prothrombin time or APTT. Based on the Kaplan–Meier analysis, a dabigatran etexilate level greater than 117.7 ng/ml was associated with a higher bleeding rate (15.4% vs. 4.9%, P = 0.01). After multivariate Cox regression analysis, HTI levels, history of stroke, and male sex were independent risk factors for bleeding events. Dabigatran etexilate-HTI levels were independently associated with bleeding in patients receiving routine clinical care. Blood sampling at multiple time points might be needed to increase reliability because of high variation of dabigatran etexilate-HTI levels.