OBJECTIVES: This study aimed to investigate the impact of utilizing different diagnosis coding schemas on evaluating the major bleeding risk associated with non-vitamin K antagonist oral anticoagulants (NOACs) and warfarin.

METHODS: A retrospective cohort study was conducted using Taiwan's National Health Insurance Research Database to identify atrial fibrillation patients newly prescribed NOACs or warfarin between June 2012 and June 2019. The primary outcome was major bleeding events, including gastrointestinal bleeding (GIB), intracranial hemorrhage (ICH), and other major bleeding (OMB). Different coding schemas proposed by Cunningham et al., the FDA, and Yao et al. for defining major bleeding were compared. The impact of including or excluding patients with a bleeding history was also explored. Propensity score matching was performed to ensure covariate balance. Incidence rates and hazard ratios (HR) were estimated to evaluate major bleeding risks between NOACs and warfarin.

RESULTS: After matching, both NOAC and warfarin groups comprised 20,704 patients. Across coding schemas, NOACs consistently showed significantly lower rates of all major bleeding (HR range: 0.73~0.76) and ICH (HR range: 0.43~0.63) compared to warfarin, while GIB rates were similar between the treatment groups (HR range: 0.87 to 0.90; p-value all >0.05). The number of events varied based on the coding stringency, with Cunningham et al. algorithm yielding the highest, followed by the FDA and YAO et al. For ICH, focusing on the primary diagnosis or considering only spontaneous cases (excluding traumatic ICH) resulted in a greater risk reduction associated with NOAC. Risk estimates remained consistent when excluding patients with a bleeding history.

CONCLUSIONS: In general, different coding schemas had minimal impact on comparing major bleeding between NOACs and warfarin, except for ICH. Relying solely on the primary diagnosis and considering only spontaneous ICH yielded a greater risk difference in ICH favoring NOACs over warfarin.