Background: Inappropriate dosing of direct oral anticoagulants (DOACs) has been associated with clinical safety and efficacy; however, little is known about clinical data associated with an inappropriate DOAC dosing in Japan. In addition, there is no report in which the appropriateness of DOAC dosing between prescription for inpatients and for outpatients was examined. In this study, we aimed to investigate the prevalence and factors associated in the inappropriate dosing of DOACs in patients with atrial fibrillation (AF).Methods: The retrospective cohort study was conducted at a single Japanese university hospital. Both inpatients and outpatients, who were diagnosed with AF and for whom treatment with either dabigatran, rivaroxaban, apixaban, or edoxaban was initiated between April 1, 2014 and March 31, 2018, were enrolled in the study. Appropriateness of DOAC dosing was assessed according to the manufacturer's labeling recommendations (dose reduction criteria) of each DOAC. Inappropriate reduced dose, namely, underdosing, was defined as prescription of a reduced dose of DOAC despite the patient not meeting the dose reduction criteria. Inappropriate standard dose, namely, overdosing, was defined as prescription of a standard dose of DOAC despite the patient meeting the dose reduction criteria. Inappropriate DOAC dosing was defined as a deviation of the recommended dose (both underdosing and overdosing). Results: A total of 316 patients (dabigatran, 28; rivaroxaban, 107; apixaban, 116; and edoxaban, 65) were included, with a median (interquartile range) age of 75 (66-81) years and 62.3% male. DOACs were prescribed at an appropriate standard dose in 39.2% of patients, an appropriate reduced dose in 36.7%, an inappropriate standard dose in 2.5%, and an inappropriate reduced dose in 19.3%. Multivariate analysis revealed that the inappropriate dosing of DOACs was significantly associated with prescriptions for outpatients (vs. inpatients; odds ratio [OR] 2.87, 95% confidence interval [CI] 1.53-5.62, p < 0.001) and those with higher HAS-BLED scores (OR 1.87, 95% CI 1.42-2.51, p < 0.001).