Real-World Bleeding Outcomes Among Hospitalized Patients Receiving Apixaban or Rivaroxaban: A Single-Center Retrospective Cohort Study.

Background: Direct oral anticoagulants (DOACs) are widely used among hospitalized patients, yet real-world inpatient bleeding data remains relatively sparse. Most evidence derives from randomized trials enrolling ambulatory, clinically stable patients, or those being treated for a single specific indication such as acute venous thromboembolism (VTE), which may not reflect the risk profile of acutely ill inpatients with multimorbidity, procedural exposures, and fluctuating renal function. This study describes the incidence and types of bleeding events in a real-world inpatient cohort receiving apixaban or rivaroxaban at a single tertiary care center. Methods: We conducted a single-center, retrospective cohort study of 867 hospital encounters involving apixaban (n = 722, 83.3%) or rivaroxaban (n = 145, 16.7%) between 2019 and 2021. The primary outcome was a non-adjudicated, non-standardized composite of any documented bleeding (gastrointestinal bleeding, hematuria, intracranial hemorrhage, vaginal bleeding, or epistaxis) during the index hospitalization; International Society on Thrombosis and Haemostasis (ISTH) major bleeding or clinically relevant non-major (CRNM) bleeding definitions were not applied. Secondary outcomes included individual bleeding subtypes, hemoglobin drop ≥2 g/dL, transfusion requirement, ICU/CCU admission, vasopressor use, and bleeding-related mortality. All comparisons are unadjusted and descriptive. Between-group comparisons used Fisher's exact test and the Mann-Whitney U test. Results: Bleeding was documented in 13 encounters (1.5% overall; apixaban 1.39%, rivaroxaban 2.07%); given reliance on clinical documentation rather than systematic adjudication, true incidence may be higher. The composite outcome was driven predominantly by gastrointestinal bleeding (n = 10, 76.9% of events). No hemoglobin drop ≥2 g/dL, transfusion, ICU/CCU admission, vasopressor use, or bleeding-related mortality occurred in either group. No statistically significant between-agent difference was observed (p = 0.47); all comparisons are unadjusted and descriptive. Median length of stay (LOS) differed between drug groups (8.6 vs. 6.2 days; p = 0.006), an exploratory finding confounded by patient class imbalance. Conclusions: In this single-center, retrospective cohort, bleeding was documented in 1.5% of encounters; ascertainment limitations mean true incidence may be higher. Non-adjudicated, non-standardized outcome definitions and the small number of events limit comparability with the literature and preclude definitive inference. The absence of a statistically significant between-agent difference should not be interpreted as evidence of equivalence. These findings are hypothesis-generating and should not be used for comparative inference. Prospective, multicenter studies using ISTH-standardized outcomes, indication-level data, and formal adjudication are needed.