Abstract
BACKGROUND:
Direct oral anticoagulants (DOACs) are recommended for patients with atrial fibrillation (AF) given their improved safety profile. Suboptimal adherence to DOACs remains a significant concern among individuals with AF. However, the extent of adherence to DOACs following a cardiovascular or bleeding event has not been fully evaluated.
OBJECTIVE:
To evaluate the pattern of adherence trajectories of DOACs after a cardiovascular or bleeding event and to investigate the sociodemographic and clinical predictors associated with each adherence trajectory by using claims-based data.
METHODS:
This retrospective study was conducted among patients with AF prescribed with DOACs (dabigatran/apixaban/rivaroxaban) between July 2016 and December 2017 and who were continuously enrolled in the Texas-based Medicare Advantage Plan. Patients who experienced a cardiovascular or bleeding event while using the DOACs were further included in the analysis. The sample was limited to patients who experienced a clinical event such as a cardiovascular or bleeding event while using the DOACs. The clinical events considered in this study were cardiovascular (stroke, congestive heart failure, myocardial infarction, systemic embolism) and bleeding events. To assess adherence patterns, each patient with a DOAC prescription was followed up for a year after experiencing a clinical event. The monthly adherence to DOACs after these events was evaluated using the proportion of days covered (PDC). A group-based trajectory model incorporated the monthly PDC to classify groups of patients based on their distinct patterns of adherence. Predictors associated with each trajectory were assessed using a multinomial logistic regression model, with the adherent trajectory serving as the reference group in the outcome variable.
RESULTS:
Among the 694 patients with AF who experienced clinical events after the initiation of DOACs, 3 distinct adherence trajectories were identified: intermediate nonadherent (30.50%), adherent (37.7%), and low adherent (31.8%); the mean PDC was 0.47 for the intermediate nonadherent trajectory, 0.93 for the adherent trajectory, and 0.01 for low adherent trajectory. The low-income subsidy was significantly associated with lower adherence trajectories (odds ratio [OR] = 4.81; 95% CI = 3.07-7.51) and with intermediate nonadherent trajectories (OR = 1.57; 95% CI = 1.06-2.34). Also, nonsteroidal anti-inflammatory drug use was significantly associated with lower adherence trajectories (OR = 5.10; 95% CI = 1.95-13.36) and intermediate nonadherent trajectories (OR = 3.17; 95% CI = 1.26-7.93). Other predictors significantly associated with both nonadherent trajectories are type of DOACs (OR = 0.53; 95% CI = 0.35-0.79), presence of coronary artery disease (OR = 1.89; 95% CI = 1.01-3.55), and having 2 or more clinical events (OR = 1.65; 95% CI = 1.09-2.50).
CONCLUSIONS:
Predictors identified provide valuable insights into the suboptimal adherence of DOACs among Medicare Advantage Plan enrollees with AF, which can guide the development of targeted interventions to enhance adherence in this high-risk patient population.
Plain language summary
Patients with atrial fibrillation are at a higher risk of experiencing clinical events. Among such patients, long-term use of direct oral anticoagulants (DOACs) is recommended to prevent such events. We examined adherence patterns of DOACs among Medicare patients with atrial fibrillation after experiencing clinical events including stroke, cardiac events, systemic embolism, and bleeding. DOAC nonadherence was driven by low-income subsidy, nonsteroidal anti-inflammatory drug use, DOAC type, coronary artery disease, and having more than 1 clinical event.
Implications for managed care pharmacy
Our study is among the first to report adherence patterns of DOACs among patients with atrial fibrillation after a cardiovascular or bleeding event. This study emphasizes the need for enhancing patient education support and engagement to improve adherence to DOACs among these high-risk patients. Furthermore, the predictors associated with lower adherence trajectories can inform targeted interventions to improve adherence and highlight the need for medication management programs within managed care settings.
Atrial fibrillation (AF) is the most frequent arrhythmia treated in clinical practice and increases the risk of several negative outcomes such as stroke, congestive heart failure, myocardial infarction, systemic embolism, and death.1-3 The risk of stroke is increased by 5 times in those with AF.4 The relationship between AF and the risk of stroke becomes more pronounced with advancing age in the general population, from 0.12%-0.16% in those aged younger than 49 years to 3.7%-4.2% in those aged between 60 and 70 years.4 The occurrence of AF might reach 10%-17% beyond age 80.4 The introduction of long-term oral anticoagulants is recommended in the United States by the American College of Cardiology and American Heart Association for the reduction of risk of ischemic stroke and systemic embolism.5
The cornerstone of anticoagulation therapy for more than 50 years has been vitamin K antagonists (eg, warfarin), which effectively prevent stroke and other thromboembolic events in patients with AF. Since the initial approval of direct oral anticoagulants (DOACs) in 2010, they have become one of the most effective and practical therapeutic alternatives, giving physicians and patients more options for treating thromboembolic conditions.6,7 The DOACs currently available in the United States include apixaban, dabigatran, edoxaban, and rivaroxaban. Since their approval, DOACs have been shown to be superior or equivalent to warfarin, the previous standard of therapy.3 DOACs have unique advantages in effectiveness, safety, convenience, predictability of results, and reduced medication and dietary interactions.6 However, to maintain the desired antithrombotic effect, long-term treatment compliance and persistence are necessary because DOACs have short half-lives. Studies have reported that even short-term nonadherence to DOACs can quickly lead to subtherapeutic anticoagulant levels, resulting in poor clinical outcomes.8,9
Previous literature evaluated DOAC adherence in patients with AF only, whereas the current study focused on patients with AF with a clinical event (cardiovascular [CV] or bleeding). For example, Maura et al examined adherence data for more than 22,000 oral anticoagulant–naive patients with nonvalvular AF, and 2 of 5 DOAC new users were identified as nonadherent to the treatment.10 Studies conducted by Charlton et al and Manzoor et al identified nonadherence to newly prescribed DOAC users and reported suboptimal adherence among a large segment of the patient population.11,12 Proportion of days covered (PDC) is used in most of the reported studies as the preferred method for measuring DOAC adherence using a claims database. However, PDC measures adherence as a single estimate and is not able to distinguish patients with varied DOAC adherence patterns.13 Group-based trajectory modeling (GBTM), which identifies longitudinal trends of drug use adherence over time, has also been used to depict adherence patterns and can capture the dynamic nature of adherence behavior that is not fully ascertained with PDC.14 GBTM identified a group of patients with similar medication adherence behavior and clustered them in one trajectory. Patients with a similar trajectory have similar characteristics. Distinct trajectories are driven by patient behavior, which is depicted through the analysis using GBTM.14 A study conducted by Mohan et al reported different trajectories of DOAC adherence based on patients’ use of DOACs by using GBTM, namely, gradual decline, adherent, and rapid decline.9,15 Although several studies have examined DOAC adherence in patients with AF for extended periods after initiating therapy, data on adherence among these patients after experiencing a cardiovascular or bleeding event have not been explored.16,17 Given the increased risk of experiencing future clinical events among this high-risk subpopulation (cardiovascular or bleeding) and the impact adverse clinical events may have on patient adherence behavior, the primary aim of this study was to evaluate DOAC adherence trajectories during the 1 year after experiencing a cardiovascular or bleeding event while using DOACs. Also, this study evaluated the association of patient sociodemographic and clinical characteristics with each adherence trajectory.
Methods
DATA SOURCE AND STUDY DESIGN
This retrospective cohort study was conducted using data from a Cigna Medicare Advantage Plan in Texas from January 2016 to December 2020 (Figure 1). Several electronic data files were used in the analysis, including a membership file, professional claims, institutional files, and pharmacy claims. The membership file was used to provide data about patient sociodemographic characteristics as well as Centers for Medicare & Medicaid Services (CMS) risk scores. The inpatient diagnostic information and outpatient encounters for AF, all comorbidities, and clinical events (stroke, congestive heart failure, systemic embolism, and bleeding) in the form of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10 CM) codes were identified from the institutional and professional claims file. Lastly, the DOAC-related information, fill dates, quantity dispensed, days supply, and dosing information were obtained from the pharmacy file.
FIGURE 1.
Study Design
The study protocol approval was obtained from the University of Houston research institutional review board on February 16, 2021 (IRB id: STUDY00002815).
STUDY POPULATION
The study included patients with AF who were aged 18 years and older with a DOAC prescription (dabigatran/apixaban/rivaroxaban) between July 2016 and December 2017. Patients using edoxaban were excluded, as there were only 9 patients who were prescribed edoxaban. The included patients were required to have a diagnosis of AF (ICD-10 code, I48.0) before the DOAC prescription date. Baseline characteristics were identified during the 6 months before the DOAC prescription. Patients who experienced a CV or bleeding event following the DOAC prescription and before January 1, 2020, were included for analysis to allow for a 365-day follow-up after the clinical event to evaluate adherence. Clinical events were identified using ICD-10 codes and included CV and bleeding events. The CV events comprised stroke, congestive heart failure, myocardial infarction, and systemic embolism (Supplementary Table 1 (277.1KB, pdf) , available in online article).
Patients were excluded if they had a diagnosis of preexisting systemic embolism due to distinct treatment guidelines,18 valvular disease, and valvular replacement conditions in which DOAC use is contraindicated and were concomitant users of warfarin. Our study did not exclude patients with a history of venous thromboembolism if they were receiving prophylactic DOAC therapy.
MEASUREMENT OF ADHERENCE AND TRAJECTORY MODELING
DOAC adherence was measured using PDC during the 1-year follow-up period following the index date. PDC, a frequently used adherence metric, accounted for oversupply and reported adherence as a binary outcome. The thresholds for full adherence and nonadherence were defined as greater than or equal to 0.80 and less than or equal to 0.80, respectively.19
During the 12-month follow-up period, PDC was calculated distinctly for each month. Then, these 12 binary indicators of DOAC adherence were modeled in a logistic GBTM as a longitudinal response and clustered patients together according to their distinct patterns of DOAC adherence. In a trajectory model, several regression models were estimated simultaneously, including a multinomial logistic model to calculate the probability of membership in each group along with logistic models to estimate the likelihood of being adherent over time, represented as a smooth function of time.
For the estimation of model parameters, maximum likelihood estimation was used and a final trajectory model was estimated using 2 to 5 adherence groups, by using the second order polynomial function of time.14,20 Each trajectory was assessed by comparison of Bayesian information criteria, clinical significance, and a sample size of 5% membership requirement.20 The Bayesian information criterion was used to select the best fit model. GBTM used the built-in function of the “Proc Traj” to depict adherence trajectories (add-on package to SAS; version 9.4; SAS Institute Inc.).21
STATISTICAL ANALYSIS
Descriptive statistics were performed to summarize patient demographic and clinical characteristics and patient characteristics were compared between different trajectory groups from the final model. Chi-square tests or analysis of variance were used to measure trajectory group–based variations for categorical and continuous variables, respectively. A multinomial logistic regression model was used to determine the predictors of adherence trajectories to DOACs 1 year after a clinical outcome.
The outcome variable was the trajectory groups with the “adherent” trajectory as reference. Other independent variables included age, sex, health plan (no subsidy vs low-income subsidy), stroke risk score (CHA2DS2-VASc score),5,22 bleeding risk scores (HAS-BLED score),5 primary care physician visits (yes vs no), number of clinical events, mean CMS risk score, concurrent medications, and comorbid conditions determined during the 6-month period before the index date. Comorbid conditions included diabetes mellitus (ICD-10 code, E11.0), hypertension (ICD-10 code, I-10, I11, I12), coronary artery disease (ICD-10 code, I24, I25), renal disease (ICD-10 code, N17-N19, R94.4), and anemia (ICD-10 code, D64) (Supplementary Table 1 (277.1KB, pdf) ). Concurrent medications included antiplatelet agents, antiarrhythmic agents, lipid-lowering agents, use of nonsteroidal anti-inflammatory drugs (NSAIDs), and type of DOACs (rivaroxaban, dabigatran, and apixaban).
All statistical analyses were conducted using SAS 9.4 (SAS Institute) with a significance level of 0.05.
Subgroup Sensitivity Analysis. We conducted 2 sensitivity analyses wherein patients were stratified based on whether they experienced the first event as a bleeding event or a CV event (stroke, congestive heart failure, myocardial infarction, and systemic embolism), and the trajectories among these subpopulations were identified and analyzed independently. The results of the final selected model of CV and bleeding events are provided in the Supplementary Appendix (277.1KB, pdf) . Additionally, a variable representing the type of first event (CV vs bleeding) was included as a predictor in the multinomial logistic regression model to assess its impact on the outcome.
Results
The final cohort included 694 continuously enrolled patients with AF who experienced a clinical event (N = 391; 56.3% encountered bleeding events, N = 303; 43.6% experienced CV events) following the DOAC prescription (Figure 2). The patients demographic results are presented in Table 1. The mean (SD) age of the study cohort was 77 (7.65) years. Most of the patients were female (56.48%), had no subsidy (61.24%), and had a CHA2DS2-VASc score greater than or equal to 3 (54.47%) and a HAS-BLED score greater than or equal to 2 (31.56%). Of these, 64.84% experienced 1 clinical event whereas 35.16% had more than 1 clinical event. In this study, 50.72% were taking rivaroxaban, 43.94% were using apixaban, and 5.33% were using dabigatran. Most patients were prescribed lipid-lowering agents (66.28%). Hypertension (16.43%) and coronary artery disease (CAD; 10.81%) were the most common comorbidities.
FIGURE 2.
Derivation of Study Sample
TABLE 1.
Patient Demographics and Clinical Characteristics
| Variables | Total patients (N = 694) | Adherent(N = 261) | Intermediate nonadherent (N = 213) | Low adherent (N = 220) | P value |
|---|---|---|---|---|---|
| Age, years | |||||
| 52-65 | 32 (4.61) | 14 (5.36) | 6 (2.82) | 12 (5.45) | 0.358 |
| 66-74 | 223 (32.13) | 91 (34.87) | 63 (29.58) | 69 (31.36) | |
| ≥75 | 439 (63.26) | 156 (59.77) | 144 (67.61) | 139 (63.18) | |
| Sex | |||||
| Female | 392 (56.48) | 149 (57.09) | 130 (61.03) | 113 (51.36) | 0.123 |
| Male | 302 (43.52) | 112 (42.91) | 83 (38.97) | 107 (48.64) | |
| Type of DOAC | |||||
| Dabigatran | 37 (5.33) | 14 (5.36) | 16 (7.51) | 7 (3.18) | 0.008a |
| Rivaroxaban | 352 (50.72) | 114 (43.68) | 119 (55.87) | 119 (54.09) | |
| Apixaban | 305 (43.95) | 133 (50.96) | 78 (36.62) | 94 (42.73) | |
| Clinical event | |||||
| 1 clinical event | 450 (64.84) | 186 (71.26) | 125 (58.69) | 139 (63.18) | 0.01a |
| >1 clinical event | 244 (35.16) | 75 (28.74) | 88 (41.31) | 81 (36.82) | |
| Health plan LIS | |||||
| No subsidy | 425 (61.24) | 124 (47.51) | 126 (59.15) | 175 (79.55) | 0.0001b |
| LIS | 269 (38.76) | 137 (52.49) | 87 (40.85) | 45 (20.45) | |
| CHA2DS2-VASc score | |||||
| Score < 3 | 316 (45.53) | 117 (44.83) | 91 (42.72) | 108 (49.09) | 0.395 |
| Score ≥ 3 | 378 (54.47) | 144 (55.17) | 122 (57.28) | 112 (50.91) | |
| HAS-BLED score | |||||
| Score < 2 | 475 (68.44) | 183 (70.11) | 139 (65.26) | 108 (49.09) | 0.481 |
| Score ≥ 2 | 219 (31.56) | 78 (29.89) | 74 (34.74) | 112 (50.91) | |
| PCP visits | |||||
| No | 522 (75.22) | 190 (72.80) | 161 (75.59) | 171 (77.73) | 0.454 |
| Yes | 172 (24.78) | 71 (27.20) | 52 (24.41) | 49 (22.27) | |
| Diabetes mellitus | |||||
| Yes | 71 (10.23) | 26 (9.96) | 21 (9.86) | 24 (10.91) | 0.921 |
| Hypertension | |||||
| Yes | 114 (16.43) | 41 (15.71) | 38 (17.84) | 35 (15.91) | 0.798 |
| Coronary artery disease | |||||
| Yes | 75 (10.81) | 28 (10.73) | 16 (7.51) | 31 (14.09) | 0.087 |
| Renal disease | |||||
| Yes | 45 (6.48) | 14 (5.36) | 17 (7.98) | 14 (6.36) | 0.513 |
| Anemia | |||||
| Yes | 49 (7.06) | 17 (6.51) | 11 (5.16) | 21 (9.55) | 0.186 |
| Antiplatelet agents | |||||
| No | 618 (89.05) | 237 (90.80) | 188 (88.26) | 193 (87.73) | 0.508 |
| Yes | 76 (10.95) | 24 (9.20) | 25 (11.74) | 27 (12.27) | |
| Antiarrhythmic agents | |||||
| No | 540 (77.81) | 204 (78.16) | 173 (81.22) | 163 (74.09) | 0.200 |
| Yes | 154 (22.19) | 57 (21.84) | 40 (18.78) | 57 (25.91) | |
| Antihyperlipidemic agents | |||||
| No | 234 (33.72) | 90 (34.48) | 72 (33.80) | 72 (32.73) | 0.920 |
| Yes | 460 (66.28) | 171 (65.52) | 141 (66.20) | 148 (67.27) | |
| NSAIDs | |||||
| No | 641 (92.36) | 252 (96.55) | 191 (89.67) | 198 (90.00) | 0.005a |
| Yes | 53 (7.64) | 9 (3.45) | 22 (10.33) | 22 (10.00) | |
| CMS risk score | |||||
| Mean ± SD | 2.29 (1.30) | 2.37 (1.36) | 2.39 (1.33) | 2.10 (1.18) | 0.0001b |
| Age, years | |||||
| Mean ± SD | 77.44 (7.65) | 77.29 (7.59) | 77.97 (7.70) | 77.12 (7.67) | |
Data are shown as n (%) unless otherwise noted.
a P < 0.05.
b P < 0.001.
CHA 2 DS 2 -VASc = congestive heart failure, hypertension, age ≥ 75 (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74, and sex category (female); CMS = Centers for Medicare & Medicaid Services; DOAC = direct oral anticoagulant; HAS-BLED score = hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly; LIS = low-income subsidy; NSAID = nonsteroidal anti-inflammatory drug; PCP = primary care physician.
ADHERENCE TRAJECTORIES
For DOAC users who experienced clinical events, the trajectory model was selected based on Bayesian information criterion value and clinical relevance. This model was composed of 3 adherence trajectories: adherent (37.7%), intermediate nonadherent (30.50%), and low adherent (31.8%) (Figure 3). The mean (SD) PDC during the 1-year follow-up period was 0.01 (0.04) for the low adherent trajectory, 0.47 (0.10) for the intermediate nonadherent, and 0.93 (0.11) for the adherent trajectory.
FIGURE 3.
Adherence Trajectories of Patients Taking Direct Oral Anticoagulants for Atrial Fibrillation After a Clinical Event
Among the 3 adherence trajectories, descriptive statistics of patients are presented in Table 1. There was a greater proportion of female patients (61.03%) in the intermediate nonadherent trajectory and male patients in the low adherent trajectories (48.64). The majority of the patients in the lower adherence trajectory did not have any subsidy, had lowest PCP visits, lower HAS-BLED score, and more than one clinical event. The average CMS risk score was highest in the intermediate nonadherent trajectory.
In the first sensitivity analysis, the results of the selected 3 group trajectories for CV (N = 303) and bleeding (N = 391) were not significantly different from the previous trajectories of adherence for the combined population (Appendix 1).
MULTIVARIABLE LOGISTIC REGRESSION ANALYSIS
The findings of multinomial regression are presented in Table 2. Patients with low-income subsidy were more likely to be associated with the low adherence trajectory and intermediate nonadherence trajectory than the adherent trajectories (low adherence, OR = 4.81;95% CI = 3.07-7.51; intermediate nonadherence, OR = 1.57; 95% CI = 1.06-2.34).
TABLE 2.
Multinomial Logistic Regression to Assess Predictors Associated With Each Trajectory (n = 694)
| Variables | Low adherent vs adherent | Intermediate nonadherent vs adherent | ||
|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | |
| Age, years | ||||
| 66-74 vs ≤65 | 1.24 (0.67-2.30) | 0.48 | 1.15 (0.64-2.71) | 0.63 |
| ≥75 vs ≤65 | 1.35 (0.67-2.71) | 0.39 | 1.16 (0.60-2.26) | 0.64 |
| Sex | ||||
| Male vs female | 0.94 (0.57-1.53) | 0.80 | 0.66 (0.41-1.07) | 0.0927 |
| Health plan | ||||
| Low-income subsidy vs no subsidy | 4.81 (3.07-7.51) | 0.0001a | 1.57 (1.06-2.34) | 0.02b |
| CHA2DS2-VASc score | ||||
| Score ≥ 3 vs score < 3 | 0.67 (0.36-1.25) | 0.21 | 0.64 (0.35-1.18) | 0.16 |
| HAS-BLED score | ||||
| Score ≥2 vs score < 2 | 0.67 (0.34-1.34) | 0.26 | 0.91 (0.48-1.71) | 0.76 |
| PCP visits | ||||
| Yes vs no | 0.89 (0.55-1.43) | 0.64 | 0.72 (0.46-1.13) | 0.16 |
| Diabetes mellitus | ||||
| Yes vs no | 1.18 (0.60-2.31) | 0.62 | 1.01 (0.51-1.99) | 0.96 |
| Hypertension | ||||
| Yes vs no | 0.88 (0.40-1.92) | 0.76 | 1.21 (0.59-2.47) | 0.59 |
| Coronary artery disease | ||||
| Yes vs no | 1.89 (1.01-3.55) | 0.04b | 0.68 (0.34-1.37) | 0.28 |
| Renal disease | ||||
| Yes vs no | 1.32 (0.53-3.26) | 0.54 | 1.63 (0.69-3.81) | 0.26 |
| Anemia | ||||
| Yes vs no | 1.25 (0.59-2.61) | 0.55 | 0.73 (0.32-1.66) | 0.45 |
| Antiplatelet agents | ||||
| Yes vs no | 1.29 (0.68-2.44) | 0.44 | 1.34 (0.71-2.49) | 0.36 |
| Antiarrhythmic agents | ||||
| Yes vs no | 1.43 (0.91-2.26) | 0.12 | 0.90 (0.56-1.45) | 0.67 |
| Antihyperlipidemic agents | ||||
| Yes vs no | 1.27 (0.84-1.93) | 0.25 | 1.05 (0.70-1.57) | 0.80 |
| NSAID use | ||||
| Yes vs no | 5.10 (1.95-13.36) | 0.0009a | 3.17 (1.26-7.93) | 0.01b |
| Type of DOAC | ||||
| Dabigatran vs rivaroxaban | 0.37 (0.14-1.02) | 0.05 | 1.07 (0.48-2.35) | 0.87 |
| Apixaban vs rivaroxaban | 0.67 (0.44-1.01) | 0.05 | 0.53 (0.35-0.79) | 0.002b |
| CMS risk score | 0.93 (0.78-1.09) | 0.39 | 1.04 (0.89-1.20) | 0.61 |
| Clinical event | ||||
| 1 clinical event vs >1 clinical event | 1.30 (0.85-2.00) | 0.21 | 1.65 (1.09-2.50) | 0.01b |
a P < 0.001.
b P < 0.05.
CHA2DS2-VASc = congestive heart failure, hypertension, age ≥75 (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74, and sex category (female); CMS = Centers for Medicare & Medicaid Services; HAS-BLED score = hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly, drugs/alcohol concomitantly; NSAID = nonsteroidal anti-inflammatory drug; OR = odds ratio; PCP = primary care physician.
Patients with CAD were more likely to be associated with the low adherent trajectory (OR = 1.89; 95% CI = 1.01-3.55) than the adherent trajectory. Patients using NSAIDs were at higher risk of being associated with the low adherent trajectory (OR = 5.10; 95% CI = 1.95-13.36) and intermediate nonadherent trajectory (OR = 3.17; 95% CI = 1.26-7.93) than the adherent trajectory. Patients who were prescribed rivaroxaban vs apixaban were more likely to be associated with the intermediate nonadherent trajectory (OR = 0.53; 95% CI = 0.35-0.79). Finally, patients who had 1 clinical event were more likely to be associated with the intermediate nonadherent trajectory (OR = 1.65; 95% CI = 1.09-2.50) than the adherent trajectory. However, age; sex; stroke risk score (CHA2DS2-VASc score); bleeding risk score (HAS-BLED score); primary care physician visits; comorbid conditions such as diabetes mellitus, hypertension, and renal disease; anemia status during the baseline period; concurrent use of antiplatelet agents, antiarrhythmic agents, and lipid-lowering agents; and mean CMS risk score were not significantly associated with the various adherence trajectories.
Sensitivity Analysis. For the sensitivity analysis (Supplementary Table 2 (277.1KB, pdf) ), patients with low-income subsidy were more likely to be associated with the low adherence trajectory and intermediate nonadherent than the adherent trajectories (low adherence, OR = 4.70; 95% CI = 3.01-7.35; intermediate nonadherence, OR = 1.55; 95% CI = 1.04-2.31). Patients using NSAIDs were at higher risk of being associated with the low adherent trajectory (OR = 4.82; 95% CI = 1.84-12.66) and intermediate nonadherent trajectory (OR = 2.85; 95% CI = 1.13-7.14) than the adherent trajectory. Patients who were prescribed rivaroxaban were more likely to be associated with the intermediate nonadherent trajectory (OR = 0.56; 95% CI = 0.37-0.85) compared with apixaban. Further, patients who had 1 clinical event were more likely to be associated with the intermediate nonadherent trajectory (OR = 1.66; 95% CI = 1.10-2.51) than the adherent trajectory. However, age; sex; stroke risk score (CHA2DS2-VASc score); bleeding risk score (HAS-BLED score); primary care physician visits; comorbid conditions such as diabetes mellitus, hypertension, renal disease, and CAD; anemia status during the baseline period; concurrent use of antiplatelet agents, antiarrhythmic agents, and lipid-lowering agents; and mean CMS risk score were not significantly associated with the various adherence trajectories.
Discussion
This retrospective study identified 3 distinct adherence trajectories by using GBTM and predictors associated with lower adherence trajectories among DOAC users who experienced CV or bleeding events. Four trajectory models ranging between 2 and 5 trajectory patterns were identified. Among these, the 3-group trajectory model (adherent, intermediate nonadherent, and nonadherent) best summarized the DOAC adherence trajectories and was ultimately selected to identify predictors associated with each adherence trajectory. Findings from this study showed that among patients with AF who are enrolled in Texas Medicare Advantage plan, DOAC therapy adherence is suboptimal within a year of experiencing a clinical event.
Prior research used the 3-group trajectory model to describe adherence patterns of DOACs among patients with AF, whereas the current study focused on patients with AF with a clinical event (CV or bleeding). To the best of our knowledge, this is the first study to provide information regarding longitudinal adherence trajectories to DOAC therapy after developing a clinical event. This study indicates that even after developing a clinical event, nearly two-thirds of the patients with AF are still nonadherent to DOAC therapy. DOAC adherence was higher among patients with AF with more than 1 clinical event compared with a single clinical event, an association that has not been previously reported. A potential explanation for this particular finding could be that patients with multiple clinical events require frequent monitoring and contact with health care providers, which might help reinforce the importance of medication adherence among such high-risk patients.
Previous studies reported significant predictors that are known to be associated with DOAC nonadherence were sex, race and ethnicity, region of residence, higher number of comorbidities, higher risk of thromboembolism, HAS-BLED score, and use of antiarrhythmic medications.4,12,17 In this study, nonadherence was significantly associated with low-income subsidy health plans, NSAID use, type of DOAC therapy, having 2 or more clinical events, and presence of CAD.
Most of the studies previously published evaluated long-term DOAC adherence among patients with AF. A study performed by An et al assessed long-term (3.5 years) adherence trajectories to DOAC therapy using monthly PDC and GBTM and reported 85.2% of those who initiated DOAC treatment continuously adhered to therapy in the first year after AF diagnosis. In this study, patients who had received a minimum of 2 prescriptions of DOACs were considered adherent, with a 45-day gap permitted in the enrollment period. However, these criteria may have led to an overestimation of DOAC adherence.17 Furthermore, a recent meta-analysis of 48 real-world studies involving more than 570,000 patients with AF revealed suboptimal adherence to DOAC, with patients failing to take their DOAC once every 4 days. Also, nonadherence among patients with AF was associated with an increased risk of stroke.8
Patients with low-income subsidy health plans were at greater risk of being associated with low adherent and intermediate nonadherent trajectories. These findings of our study reaffirmed the results from Hernandez et al, who reported that eligibility for low-income subsidy was significantly associated with higher odds of not using oral anticoagulants.15 It is important to consider the characteristics of low-income subsidy enrollees as they are more likely to be non-white and less educated.23,24 Moreover, low-income subsidy enrollees have limited income; hence, it is crucial for health care professionals to ensure that medication cost should not impact the adherence rate.25
High-risk patients with AF who had CAD were at higher risk of being associated with the low adherent trajectories. To our understanding, this association has been demonstrated recently in only one study, which highlighted the association of poor oral anticoagulant therapy adherence with comorbid conditions such as diabetes, CAD, and history of major bleeding.26 This finding may be due to the complexity of managing multiple medications, fear of bleeding events, and potential for drug interactions.27
Patients prescribed NSAIDs were also more likely to be associated with low adherent and intermediate nonadherent trajectories, which could be related to the fear of elevated risk of bleeding. A nested case-control study using data from Korean National Health Insurance services found that DOAC users with AF who received NSAIDs had higher risk of bleeding events.28
Interestingly, patients with AF after a clinical event who were using apixaban were less likely to be associated with intermediate nonadherent trajectories compared with those on rivaroxaban. There are several potential reasons for low adherence rates of rivaroxaban, such as its increased likelihood of causing bleeding problems among older adults.28-30 However, our findings suggest that dosing regimen may be a minor contributing factor because we observed better adherence rates with twice-daily apixaban compared with once-daily rivaroxaban. These results differ from previous literature on adherence that typically find once-daily regimens have higher adherence compared with twice-daily regimens.31,32 Other factors that may contribute to higher apixaban adherence may include stronger safety and efficacy evidence for apixaban in patients with chronic kidney disease, its cost-effectiveness, and greater amounts of direct-to-consumer advertising expenditure for apixaban compared with other agents.33-35
Identifying the intermediate nonadherent trajectory as a crucial subgroup is significant in our study, given the increased risk of thromboembolism. Patient demographic and clinical predictors associated with the intermediate nonadherent trajectory identified from this study include lower-income subsidy plan, NSAID use, type of DOAC, and having at least 1 clinical event. These findings would be instrumental in identifying patients who may benefit from designed adherence intervention. Our study is a valuable addition to the existing literature because it identifies longitudinal patterns of DOAC adherence among patients with AF following clinical events, which is a high-risk group, and demonstrates that adherence in such patients remains suboptimal. Optimal medication adherence among these high-risk patients is crucial to prevent future life-threatening complications.
The sensitivity analysis of the selected 3-group trajectories for CV and bleeding events were not significantly different from the trajectories of adherence for the combined population. This suggests that although there may be subtle differences in patient behavior following specific events, these distinctions did not substantially alter the overall adherence patterns observed in our study population. Moreover, in the second sensitivity analysis, we found that the distinction between bleeding events and CV events did not significantly predict adherence behavior in the model. This means that, within our study population, the occurrence of a bleeding event compared with a cardiovascular event did not significantly influence adherence patterns. Also, adding the variable (CV vs bleeding) did not have a major impact on other predictors in the model; the significance remained the same.
From a clinical and economic standpoint, this study reveals the need for various health care professionals and clinicians to educate patients regarding the need for being adherent even after they develop any safety or efficacy outcomes. Based on the results of this study, we recommend that clinicians should monitor their patients on a regular basis regarding their medication-taking behaviors. These clinical events substantially contribute to an increased economic burden on Medicare beneficiaries and payers. This study highlights the importance of minimizing clinical events and improving long-term adherence to DOACs.
LIMITATIONS
Our study has several limitations. First, we were unable to ascertain the reason for discontinuation of therapy with the data. The study also did not account for medication switching to other oral anticoagulants. Second, the study only measured longitudinal patterns of adherence over a 1-year period and did not assess the link between lower adherence trajectories and clinical outcomes. Future studies should investigate adherence patterns among a larger sample size for a more extended period. Third, drug exposure was tracked using pharmacy claims; therefore, we could not confirm whether patients took the medications. The database did not include any information regarding direct measures of adherence. The claims can capture only claims that are paid via insurance; however, the database did not include over-the-counter NSAID use. Fourth, as the study only included beneficiaries of a Texas-based Medicare Advantage Plan, it may not be generalizable to other populations. Finally, DOACs have shown promising results in phase 3 clinical trials for postoperative venous thromboembolism prophylaxis. Our study did not exclude patients with a history of venous thromboembolism if they were receiving prophylactic DOAC therapy.
Conclusions
The adherence to DOACs among patients with AF after experiencing a clinical outcome is suboptimal. Predictors associated with suboptimal adherence such as having low-income subsidy, NSAIDs use, type of DOACs, presence of CAD, and having 2 or more clinical events should be considered when designing future tailored interventions for such high-risk patients to overcome barriers to adherence and enhance health outcomes.
REFERENCES
- 1.Vizzardi E, Curnis A, Latini MG, et al. Risk factors for atrial fibrillation recurrence: A literature review. J Cardiovasc Med (Hagerstown). 2014;15(3):235-53. doi:10.2459/JCM.0b013e328358554b [DOI] [PubMed] [Google Scholar]
- 2.Bassand J-P, Virdone S, Goldhaber SZ, et al. Early risks of death, stroke/systemic embolism, and major bleeding in patients with newly diagnosed atrial fibrillation: Results from the GARFIELD-AF registry. Circulation. 2019;139(6):787-98. doi:10.1161/CIRCULATIONAHA.118.035012 [DOI] [PubMed] [Google Scholar]
- 3.Chen A, Stecker E, Warden BA. Direct oral anticoagulant use: A practical guide to common clinical challenges. J Am Heart Assoc. 2020;9(13):e017559. doi:10.1161/JAHA.120.017559 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Chen N, Brooks MM, Hernandez I. Latent classes of adherence to oral anticoagulation therapy among patients with a new diagnosis of atrial fibrillation. JAMA Netw Open. 2020;3(2):e1921357. doi:10.1001/jamanetworkopen.2019.21357 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.January CT, Wann LS, Calkins H, et al. 2019. AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS Guideline for the Management of Patients With Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society in collaboration with the Society of Thoracic Surgeons. Circulation. 2019;140(2):e125-e151. doi:10.1161/CIR.0000000000000665 [DOI] [PubMed] [Google Scholar]
- 6.Zirlik A, Bode C. Vitamin K antagonists: Relative strengths and weaknesses vs. direct oral anticoagulants for stroke prevention in patients with atrial fibrillation. J Thromb Thrombolysis. 2017;43(3):365-79. doi:10.1007/s11239-016-1446-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Oertel LB, Fogerty AE. Use of direct oral anticoagulants for stroke prevention in elderly patients with nonvalvular atrial fibrillation. J Am Assoc Nurse Pract. 2017;29(9):551-61. doi:10.1002/2327-6924.12494 [DOI] [PubMed] [Google Scholar]
- 8.Ozaki AF, Choi AS, Le QT, et al. Real-world adherence and persistence to direct oral anticoagulants in patients with atrial fibrillation: A systematic review and meta-analysis. Circ Cardiovasc Qual Outcomes. 2020;13(3):e005969. doi:10.1161/CIRCOUTCOMES.119.005969 [DOI] [PubMed] [Google Scholar]
- 9.Mohan A, Majd Z, Trinh T, Paranjpe R, Abughosh SM. Group based trajectory modeling to assess adherence to oral anticoagulants among atrial fibrillation patients with comorbidities: A retrospective study. Int J Clin Pharm. 2022;44(4):966-74. doi:10.1007/s11096-022-01417-4 [DOI] [PubMed] [Google Scholar]
- 10.Maura G, Pariente A, Alla F, Billionnet C. Adherence with direct oral anticoagulants in nonvalvular atrial fibrillation new users and associated factors: A French nationwide cohort study. Pharmacoepidemiol Drug Saf. 2017;26(11):1367-77. doi:10.1002/pds.4268 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Charlton A, Vidal X, Sabaté M, Bailarín E, Martínez LML, Ibáñez L. Factors associated with primary nonadherence to newly initiated direct oral anticoagulants in patients with nonvalvular atrial fibrillation. J Manag Care Spec Pharm. 2021;27(9):1210-20. doi:10.18553/jmcp.2021.27.9.1210 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Manzoor BS, Lee TA, Sharp LK, Walton SM, Galanter WL, Nutescu EA. Real-world adherence and persistence with direct oral anticoagulants in adults with atrial fibrillation. Pharmacotherapy. 2017;37(10):1221-30. doi:10.1002/phar.1989 [DOI] [PubMed] [Google Scholar]
- 13.Nau DP. Proportion of days covered (PDC) as a preferred method of measuring medication adherence. Pharmacy Quality Alliance. 2012;6:25. [Google Scholar]
- 14.Franklin JM, Shrank WH, Pakes J, et al. Group-based trajectory models: A new approach to classifying and predicting long-term medication adherence. Med Care. 2013;51(9):789-96. doi:10.1097/MLR.0b013e3182984c1f [DOI] [PubMed] [Google Scholar]
- 15.Hernandez I, He M, Chen N, Brooks MM, Saba S, Gellad WF. Trajectories of oral anticoagulation adherence among Medicare beneficiaries newly diagnosed with atrial fibrillation. J Am Heart Assoc. 2019;8(12):e011427. doi:10.1161/JAHA.118.011427 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hori K, Okumura Y, Koichi N, et al. ; SAKURA AF Registry Investigators. Association of patient satisfaction with direct oral anticoagulants and the clinical outcomes: Findings from the SAKURA AF registry. J Cardiol. 2020;76(1):80-6. doi:10.1016/j.jjcc.2020.01.007 [DOI] [PubMed] [Google Scholar]
- 17.An J, Bider Z, Luong TQ, et al. Long-term medication adherence trajectories to direct oral anticoagulants and clinical outcomes in patients with atrial fibrillation. J Am Heart Assoc. 2021;10(21):e021601. doi:10.1161/JAHA.121.021601 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wann LS, Curtis AB, Ellenbogen KA, et al. ; Heart Rhythm Society. Management of patients with atrial fibrillation (compilation of 2006 ACCF/AHA/ESC and 2011 ACCF/AHA/HRS recommendations): A report of the American College of Cardiology/American Heart Association Task Force on practice guidelines. Circulation. 2013;127(18):1916-26. doi:10.1161/CIR.0b013e318290826d [DOI] [PubMed] [Google Scholar]
- 19.Mohan A, Majd Z, Trinh T, Paranjpe R, Abughosh S. Group-based trajectory modeling to assess adherence to oral anticoagulants among atrial fibrillation patients with comorbidities: A retrospective study. Int J Clin Pharm. 2022;44(4):966-74. doi:10.1007/s11096-022-01417-4 [DOI] [PubMed] [Google Scholar]
- 20.Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6(1):109-38. doi:10.1146/annurev.clinpsy.121208.131413 [DOI] [PubMed] [Google Scholar]
- 21.Arrandale V, Koehoorn M, MacNab Y, Kennedy SM. How to use SAS® Proc Traj and SAS® Proc Glimmix in respiratory epidemiology. UBC; 2006. [Google Scholar]
- 22.Kaplan RM, Koehler J, Ziegler PD, Sarkar S, Zweibel S, Passman RS. Stroke risk as a function of atrial fibrillation duration and CHA2DS2-VASc score. Circulation. 2019;140(20):1639-46. doi:10.1161/CIRCULATIONAHA.119.041303 [DOI] [PubMed] [Google Scholar]
- 23.Stuart B, Hendrick FB, Xu J, Dougherty JS. How low-income subsidy recipients respond to Medicare Part D cost sharing. Health Serv Res. 2017;52(3):1185-206. doi:10.1111/1475-6773.12520 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Shoemaker JS, Davidoff AJ, Stuart B, Zuckerman IH, Onukwugha E, Powers C. Eligibility and take-up of the Medicare Part D low-income subsidy. Inquiry. 2012;49(3):214-30. doi:10.5034/inquiryjrnl_49.03.04 [DOI] [PubMed] [Google Scholar]
- 25.Qiao Y, Steve Tsang CC, Hohmeier KC, et al. Association between medication adherence and healthcare costs among patients receiving the low-income subsidy. Value Health. 2020;23(9):1210-7. doi:10.1016/j.jval.2020.06.005 [DOI] [PubMed] [Google Scholar]
- 26.Akao M, Tsuji H, Kusano K, et al. Clinical characteristics and outcomes of Japanese atrial fibrillation patients with poor medication adherence: A sub-analysis of the GENERAL study. J Cardiol. 2023;81(2):209-14. doi:10.1016/j.jjcc.2022.07.022 [DOI] [PubMed] [Google Scholar]
- 27.Pundi KN, Perino AC, Fan J, et al. Direct oral anticoagulant adherence of patients with atrial fibrillation transitioned from warfarin. J Am Heart Assoc. 2021;10(23):e020904. doi:10.1161/JAHA.121.020904 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.PRADAXA. Prescribing information. Boehringer Ingelheim Pharmaceuticals; 2018. [Google Scholar]
- 29.Ng KH, Hart RG, Eikelboom JW. Anticoagulation in patients aged ≥75 years with atrial fibrillation: Role of novel oral anticoagulants. Cardiol Ther. 2013;2(2):135-49. doi:10.1007/s40119-013-0019-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Mentias A, Heller E, Vaughan Sarrazin M. Comparative effectiveness of rivaroxaban, apixaban, and warfarin in atrial fibrillation patients with polypharmacy. Stroke. 2020;51(7):2076-86. doi:10.1161/STROKEAHA.120.029541 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Claxton AJ, Cramer J, Pierce C. A. systematic review of the associations between dose regimens and medication compliance. Clin Ther. 2001;23(8):1296-310. doi:10.1016/S0149-2918(01)80109-0 [DOI] [PubMed] [Google Scholar]
- 32.Coleman CI, Roberts MS, Sobieraj DM, Lee S, Alam T, Kaur R. Effect of dosing frequency on chronic cardiovascular disease medication adherence. Curr Med Res Opin. 2012;28(5):669-80. doi:10.1185/03007995.2012.677419 [DOI] [PubMed] [Google Scholar]
- 33.Shah A, Shewale A, Hayes CJ, Martin BC. Cost-effectiveness of oral anticoagulants for ischemic stroke prophylaxis among nonvalvular atrial fibrillation patients. Stroke. 2016;47(6):1555-61. doi:10.1161/STROKEAHA.115.012325 [DOI] [PubMed] [Google Scholar]
- 34.Siontis KC, Zhang X, Eckard A, et al. Outcomes associated with apixaban use in patients with end-stage kidney disease and atrial fibrillation in the United States. Circulation. 2018;138(15):1519-29. doi:10.1161/CIRCULATIONAHA.118.035418 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Lowenstern A, Al-Khatib SM, Sharan L, et al. Interventions for preventing thromboembolic events in patients with atrial fibrillation: A systematic review. Ann Intern Med. 2018;169(11):774-87. doi:10.7326/M18-1523 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Lee M-T, Park K-Y, Kim M-S, You S-H, Kang Y-J, Jung S-Y. Concomitant use of NSAIDs or SSRIs with NOACs requires monitoring for bleeding. Yonsei Med J. 2020;61(9):741-9. doi:10.3349/ymj.2020.61.9.741 [DOI] [PMC free article] [PubMed] [Google Scholar]