Assessing methods to ascertain persistence and adherence of oral anticoagulants in patients with atrial fibrillation.

Background: Persistence and adherence to oral anticoagulants (OACs) is crucial for its effectiveness in stroke prevention in atrial fibrillation (AF). We aimed to assess the impact of different ascertainment methods on estimated persistence rates.

Methods: We conducted a retrospective cohort study based on the Medicare claims data (01/01/2013-12/31/2019). We built an incident user cohort of OAC (apixaban, dabigatran, edoxaban, rivaroxaban, and warfarin) prescription filling. We measured OAC medication persistence and adherence using the following approaches: (1) treatment-anniversary based persistence: if there is an active prescription overlapping the 180th and 365th day with vs. without a 15-day buffer period (i.e., overlapping with 165th-195th and 350th-380th day); (2) dispensing-gap-based persistence: if there is OAC discontinuation defined as having gap between prescriptions more than a threshold (e.g., 5-60 days) and secondarily, (3) proportion of days covered (PDC) adherence: proportion of days in which patient had filled medication available over the 365-day interval.

Results: We identified 1,398,692 patients who initiated OACs during the study interval. With the treatment-anniversary based approach, only 51.2% to 65.4% of the patients persisted with the medication for either warfarin or DOACs at 180 days, and the number dropped to 43.4% to 60.7% at 1 year. Adding a 15-day buffer period increased the treatment-anniversary based persistence rates by 6.5% to 10.5%. When the allowable gap increased from 5 to 60 days, the persistence rates increased by 36.3% to 42.4% for all OACs. Apixaban users had the highest PDC (74%-75%) over the 365 days, compared to other OACs (60%-69%).

Conclusions: We found that the estimated persistence rates are sensitive to the choice of ascertainment methods. When reporting and comparing persistence findings using the claims database, definitions of OAC discontinuation must be clearly delineated.