Impact of diagnostic coding schemas on major bleeding risk assessment for oral anticoagulants in patients with atrial fibrillation using administrative claims data.

Abstract

Introduction: This study evaluated how different published diagnostic coding schemas impact the assessment of major bleeding risks associated with direct oral anticoagulants (DOACs) and warfarin.

Methods: This retrospective cohort study included patients with atrial fibrillation who initiated DOACs or warfarin between 2012 and 2019, using Taiwan's national claims database. Major bleeding events, including gastrointestinal bleeding (GIB), intracranial hemorrhage (ICH), and other major bleeding events, were identified using coding schemas from Cunningham et al, the Mini-Sentinel protocol, and Yao et al. Propensity score matching was performed to ensure covariate balance. Incidence rates and hazard ratios (HRs) were estimated to evaluate the bleeding risks.

Results: After matching, each cohort comprised 20,704 patients. The number of reported major bleeding events was influenced by the strictness of the coding schema, with Cunningham's yielding the most events, followed by the Mini-Sentinel and Yao's schemas. DOACs were associated with a consistently lower risk of composite major bleeding (HR range across different coding schemas: 0.73-0.76, all P<0.05) and ICH (HR range: 0.43-0.63, all P<0.05) but not GIB (HR range: 0.87-0.90, all P>0.05), regardless of the coding schema applied. Restricting ICH definitions to primary diagnosis or spontaneous cases revealed a more pronounced reduction in ICH risk associated with DOACs.

Conclusions: While the choice of coding schemas has a negligible impact on overall bleeding risk comparisons between DOACs and warfarin, it significantly affects ICH risk assessment. This underscores the importance of careful coding schema selection in observational studies evaluating major bleeding risks.