Highlights
-
•
Medicare Part D included a coverage gap during which beneficiaries were responsible for substantially higher portions of medication costs, although the impact of this on oral anticoagulant prescription fills and subsequent stroke in patients with atrial fibrillation was previously unknown.
-
•
In this study, patients who entered the coverage gap had decreases in the proportion of oral anticoagulant days covered while in the gap, which continued to decrease for patients who left the gap, despite subsequently regaining coverage.
-
•
Patients had an increased stroke risk whilst in the coverage gap, and in the postgap period, that persisted for the rest of the calendar year.
-
•
Public policy decisions regarding Medicare can have unintended adverse population health consequences, highlighting the importance of assessing the impact of such policy changes.
ABSTRACT
Background
Prior to 2025, Medicare Part D included a coverage gap during which beneficiaries were responsible for substantially higher portions of medication costs. The impact of this on oral anticoagulant (OAC) prescription fills and subsequent stroke in patients with atrial fibrillation (AF) is not known.
Methods
Using Centers for Medicare and Medicaid Services claims data from 2016 to 2018, we evaluated OAC prescription fills in patients with AF by assessing their proportion of days covered on OAC before, during, and after their coverage gap. Hazard of stroke was assessed for patients who entered the coverage gap before entering, while in, and after exiting, the gap.
Results
Patients who entered the coverage gap had a 16% decrease in median proportion of days covered from pregap (0.92 [interquartile range 0.80, 0.97]) to in-gap (0.76 [0.41, 0.98]). Proportion of days covered continued to drop for patients who entered and then left the coverage gap, despite regaining coverage (pregap: 0.95 [0.85, 0.98]; in-gap: 0.88 [0.55, 0.97]; postgap: 0.70 [0.00, 1.00]). Patients who entered the gap were at significantly higher risk for stroke while in the gap (HR 2.21, 95% CI 1.91-2.54) and during the combined in-gap and postgap periods (HR 3.13, 95% CI 2.71-3.62).
Conclusions
OAC use decreased upon entering the coverage gap and was associated with an increased stroke risk, that persisted for the rest of the calendar year. Health policy decisions regarding Medicare can have unintended adverse public health consequences, highlighting the importance of assessing the impact of such policy changes.
Introduction
Medicare Part D beneficiaries previously entered a coverage gap after their yearly total medication costs exceeded a prespecified amount, known as the “donut hole”. Until its elimination on December 31st 2024, patients were responsible for larger portions of their medication costs while in this coverage gap, unless total medication costs exceeded another prespecified amount, at which point they regained financial coverage for their medications (“catastrophic coverage”). Thresholds to enter and exit the coverage gap were set annually, and in 2024, beneficiaries entered when their annual medication costs reached $5,030. Once total out-of-pocket costs for Part D medications exceeded $8,000, beneficiaries would exit the coverage gap.
The coverage gap has had significant implications for patients prescribed multiple expensive medications. For instance, a patient with atrial fibrillation (AF) and heart failure may take empagliflozin (2023 list price for 1-month supply: $573.00), sacubitril/valsartan ($628.00), and apixaban ($521.00). Though programs were available to mitigate these costs for Medicare users, patients filling a monthly prescription for even 1 of these medications would enter the coverage gap each year. Beneficiaries were therefore subject to high out-of-pocket payments, particularly those with chronic health conditions. While the coverage gap has been closed, complex coverage differences related to Advantage Plans and deductibles remain as barriers to medication affordability.
Surprisingly, the impact of increased costs related to the coverage gap on medication adherence and subsequent clinical outcomes has not been widely evaluated, especially in the case of anticoagulation for patients with atrial fibrillation. Importantly, 1 in 8 patients with cardiovascular disease report nonadherence to their medications related to cost, and increased out-of-pocket costs are associated with decreased adherence in patients with diabetes. For patients with AF on an oral anticoagulant, decreased use has been associated with higher inpatient and outpatient medical costs. , Nonadherence to oral anticoagulation has also been associated with substantially increased risks of stroke and death in those with AF. ,,,,, One prior study demonstrated that among commercially insured patients with AF initiating a DOAC for stroke prevention, higher copayments were associated with lower adherence and higher rates of discontinuation and switching to a different anticoagulant within the first year. However, this analysis was underpowered to detect an association between copayment status and thromboembolic events or major bleeding.
The impact of increased medication costs during the coverage gap on oral anticoagulant adherence and subsequent clinical outcomes is unknown. We therefore assessed gap-related changes in oral anticoagulant prescription fills in patients with AF who entered the coverage gap and evaluated the related risk of clinical events, including ischemic stroke and bleeding, by gap status.
Methods
We used administrative claims data from the Centers for Medicare and Medicaid Services from 2016 to 2018 to evaluate beneficiary-level proportion of days covered by an oral anticoagulant before, during, and after observed coverage gaps ( Figure 1 ). Medicare claims were ascertained from the 5% sample, which contains all final action claims for a 5% random sample of Medicare beneficiaries.
Study overview. OAC, oral anticoagulant.
Study population
Patients over age 65 with any diagnosis of AF or atrial flutter in the calendar years 2015, 2016, and 2017 were identified using diagnosis codes for any inpatient, outpatient, or carrier claim in that year (Table S1). Calendar year cohorts were created for 2016, 2017, and 2018 comprising patients with AF or atrial flutter in the year prior who were continuously enrolled in Medicare fee-for-service parts A and B for 12 months prior to the start of the calendar year through the end of the calendar year or until death. Patients were also required to have continuous Medicare Part D enrollment for 90 days prior to the start of the calendar year through the end of the calendar year or until death. Yearly cohorts were further limited to patients with at least 1 Part D prescription claim for an oral anticoagulant (warfarin, rivaroxaban, apixaban, dabigatran, or edoxaban) in the 90 days prior to the start of the calendar year and at least 1 claim for an oral anticoagulant in the first 90 days of the calendar year of interest.
Patient characteristics
Patient age at the start of the calendar year, race/ethnicity, sex, geographic region, and dual enrollment in Medicaid were drawn from Medicare enrollment information. Comorbidities were assessed in the 12 months prior to the start of the cohort calendar year using International Classification of Diseases diagnosis codes (Table S1). Use of other oral medications in the 90 days prior to the start of the calendar year was assessed using prescription claims.
Medicare part D coverage gap status
Entry into and/or exit from the Part D coverage gap in a given calendar year was derived from the benefit phase listed on the patient’s prescription fill claims for any medication in that year. Dates of entry into and exit from the coverage gap were sourced directly from prescription claim dates, and were used to calculate time to entrance into the coverage gap from a patient’s first oral anticoagulant prescription in the calendar year. Patients were categorized as having (1) never entered the gap during the year, (2) entered and remained in the gap until the end of the year, or (3) entered and left the gap before the end of the year.
Patterns of use of and adherence to oral anticoagulants
Using Part D prescription fill claims, we assessed the type of oral anticoagulant patients were taking at the start of the calendar year (warfarin or DOAC) and adherence to oral anticoagulants using proportion of days covered). Proportion of days covered is calculated as the percentage of days in a given time period for which a patient had an available dose of their medication in their possession based on prescription supply and fill dates. Each patient’s proportion of days covered was calculated for the entire calendar year, as well as for each coverage gap phase they entered (pregap, in-gap, and postgap).
Secondary outcomes
Secondary outcomes of interest included hospitalization for ischemic stroke/transient ischemic attack (TIA) and hospitalization for bleeding (Table S2). These were assessed using Medicare inpatient claims data for each patient from the date of their first oral anticoagulant prescription fill in the calendar year until December 31 of that calendar year.
Statistical analyses
Baseline patient characteristics were evaluated by coverage gap status at the end of the calendar year (never entered gap, entered and remained in gap, or entered and left gap), and were summarized using frequencies and percentages for categorical variables and means (SD) and medians (IQR) for continuous variables. Group differences were tested using Kruskal-Wallis tests for continuous variables and chi-square tests for categorical variables. Median time to reach and median time in the coverage gap were calculated by year, type of oral anticoagulant used at the start of the calendar year, and calendar year proportion of days covered. Group differences in distributions were tested for significance using the Wilcoxon rank-sum test and Mood’s median test. Mean proportion of days covered was modeled with adjustment for patient characteristics across coverage gap status (pregap, in-gap, postgap), by whether patients entered the gap by the end of the calendar year, and by type of oral anticoagulant using a generalized linear mixed model with binomial distribution. Models were adjusted for age at baseline, sex, race, geographic region, comorbidities including claims-based CHA 2 DS 2 -VASc score, medications at baseline, and hospitalizations in the prior calendar year. Among patients who entered the coverage gap at least once in the 3-year study period, a Cox proportional hazards regression model with a time-varying covariate for coverage gap status was used to estimate associations between gap status and the outcomes of any hospitalizations for bleeding and/or stroke/TIA during the study period. Patients entered the cohort on January 1 of the first calendar year that they were eligible for the study and contributed time to the analysis until end of available follow-up, death, or an outcome of interest. Coverage gap status was modeled both as a 3-stage variable (pregap, in-gap, and postgap) with the pregap status serving as the reference, and a 2-stage variable in separate models (pregap vs in-gap and postgap). Models were additionally adjusted for switching from warfarin to a DOAC or vice versa (time-varying) and entering the coverage gap in the first quarter of the calendar year (time-varying), in addition to the previously described adjustment variables.
Results
We analyzed data from 57,029 to 63,824 patients per year and 182,192 patient-years overall between 2016 and 2018. Median available follow-up time was 1.96 (IQR 0.98, 2.92) years. About half (51.2%) of the patients were on warfarin and the other half (48.8%) were prescribed a DOAC. In 2016, 28.2% of patients entered the coverage gap and remained there for the rest of the year, while 15.4% entered and subsequently exited the coverage gap. The remaining 56.4% of patients did not enter the gap during the calendar year. These trends remained stable in 2017 and 2018 (Table S3). In patients who reached the coverage gap, median time to reach the gap was 174 days, with no significant difference between warfarin and DOAC users. Median time in the coverage gap was 174 days overall, and was on average slightly shorter for warfarin users compared with DOAC users (119 [IQR 71, 166] vs 134 [97, 172] days; P <.001).
Baseline characteristics
There were significant differences in baseline characteristics between patients who never entered the coverage gap, who entered and stayed in the gap, and who entered and left the gap later that calendar year ( Table 1 , Tables S4 & S5). Compared with those who never entered the coverage gap, patients who either stayed in or left the gap were more likely to be older, female, Black, and have medical comorbidities. They also had, on average, higher CHA 2 DS 2 -VASc scores, were prescribed more medications, and were more likely to be prescribed a DOAC than warfarin.
Table 1
Baseline characteristics of patients by maximum coverage gap status in 2016
|
No time in gap
(N = 32,186) |
Entered and remained in gap
(N = 16,082) |
Entered and left gap
(N = 8761) |
P -value | |
|---|---|---|---|---|
| Age, mean (SD), yrs | 74.9 (8.8) | 78 (8.9) | 75.3 (10.3) | <.001 |
| Race, % | ||||
| White | 29,306 (91.1%) | 14,523 (90.3%) | 7,284 (83.1%) | <.001 |
| Black | 1,710 (5.3%) | 779 (4.8%) | 763 (8.7%) | |
| Other | 1,170 (3.6%) | 780 (4.9%) | 714 (8.1%) | |
| Male | 16,138 (50.1%) | 7,530 (46.8%) | 3,923 (44.8%) | |
| Medicaid dual-enrolled, % | 3874 (12%) | 3,396 (21.1%) | 4299 (50.2%) | <.001 |
| Medical history, % | ||||
| Hypertension | 28,398 (88.2%) | 14,524 (90.3%) | 8,203 (93.6%) | <.001 |
| Hyperlipidemia | 24,223 (75.3%) | 12,529 (77.9%) | 7,129 (81.9%) | <.001 |
| CV disease | 8,472 (26.3%) | 4,385 (27.3%) | 2,737 (31.2%) | <.001 |
| Prior MI | 3,533 (11%) | 1,847 (11.5%) | 1,376 (15.7%) | <.001 |
| Diabetes | 11,121 (34.6%) | 6,150 (38.2%) | 5,418 (61.8%) | <.001 |
| Heart failure | 13,106 (40.7%) | 6,661 (41.4%) | 4,696 (53.6%) | <.001 |
| Prior stroke/TIA | 817 (2.5%) | 489 (3%) | 244 (2.8%) | .005 |
| Prior bleeding | 2201 (6.8%) | 1014 (6.3%) | 784 (8.9%) | <.001 |
| Any renal disease | 7455 (23.2%) | 3,777 (23.5%) | 3,053 (34.8%) | <.001 |
| CKD stage 1-3 | 5,190 (16.1%) | 2,634 (16.4%) | 1,984 (22.6%) | <.001 |
| CKD stage 4 | 1,204 (3.7%) | 584 (3.6%) | 598 (6.8%) | <.001 |
| ESRD on dialysis | 538 (1.7%) | 291 (1.8%) | 481 (1.4%) | <.001 |
| CHA 2 DS 2 -VASc score, mean (SD) | 4.1 (1.3) | 4.1 (1.3) | 4.4 (1.4) | .03 |
| Medications in previous 90 days | ||||
| Antihypertensives, % | 27,606 (85.8%) | 13,901 (86.4%) | 7,826 (89.3%) | <.001 |
| Lipid-lowering, % | 17,646 (54.8%) | 9,233 (57.4%) | 5,827 (66.5%) | <.001 |
| P2Y12 antagonist, % | 1,090 (3.4%) | 586 (3.6%) | 562 (6.4%) | <.001 |
| ACE/ARB/ARNI, % | 8,684 (27.0%) | 4,453 (27.7%) | 2,780 (31.7%) | <.001 |
| No. of prescriptions, mean (SD) | 3.2 (1.3) | 3.3 (1.3) | 3.7 (1.5) | <.001 |
| OACs in previous 90 days, % | ||||
| Warfarin | 24,908 (77.45%) | 5,944 (37%) | 3,646 (41.6%) | <.001 |
| Rivaroxaban | 3,041 (9.4%) | 4,243 (26.4%) | 2,181 (24.9%) | <.001 |
| Dabigatran | 1,473 (4.6%) | 1,908 (11.9%) | 1,002 (11.4%) | <.001 |
| Apixaban | 3,020 (9.4%) | 4,206 (26.2%) | 2,095 (23.9%) | <.001 |
| Edoxaban | <11 (<0.01%) | <11 (<0.01%) | <11 (<0.01%) | .97 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
