OBJECTIVES: Previous research has identified risk factors that may affect the risk of bleeding when individuals are exposed to oral anticoagulants (OAC). It is unclear if this risk continues to exist with newer OACs and more intense monitoring of warfarin therapy. The purpose of this study was to assess the risk of bleeding events in patients on OAC medications including warfarin, apixaban, rivaroxaban, dabigatran or edoxaban based on known risk factors including demographics, medical conditions and concomitant medications.

METHODS: This study was a retrospective analysis using electronic health record (EHR) data from individuals receiving an OAC from 2015 up to 2020. The primary outcome of interest was bleeding events recorded in the EHR coded using ICD-9 and ICD-10. Known risk factors were used to predict bleeding using multi-variate logistic regression.

RESULTS: There was a total of 8,527 individuals on OACs. Among them, 35.6% (n= 3,035) were receiving warfarin. Less than half the study population were female (3,562, 41.8%). There were 1,255 (14.7%) bleeding events, excluding intracranial hemorrhage (ICH) (2, 0.02%) and gastrointestinal (GI) (510, 6.0%). Significant risk factors of GI bleeding included: aspirin [OR 1.21, CI95% 1.01-1.47]; clopidogrel [OR 1.61, CI95% 1.22-2.13]; and previous GI bleeding episodes [OR 5.23, CI95% 3.90-7.01]. For other bleeding outcomes, excluding GI bleeding and ICH, exposure to aspirin [OR 1.17, CI95% 1.03-1.32], clopidogrel [OR 1.40, CI95% 1.14-1.71], selective serotonin reuptake inhibitors (SSRI) [OR 1.19, CI95% 1.02-1.39], corticosteroids [OR 1.43, CI95% 1.20-1.70] and previous GI bleeding [OR 1.71, CI95% 1.28-2.29] were significantly predictors.

CONCLUSIONS: Individuals on OACs are at significant increased risk of bleeding, especially if they have known risk factors. This risk increases for individuals on concomitantly medications such as aspirin, clopidogrel, or SSRI or previous bleeding episodes. Optimal risk assessment for bleeding should accounting for different patient’s specific attributes.