The study population consisted of 2,186 AF patients. Of them, 2,064 patients were continuously enrolled and had a prescription claim for any DOAC during July 2016-December 2017. Ninety-five patients were excluded as they had either valvular disease or pre-existing systemic embolism. The final cohort comprised of 1,969 patients with AF. Approximately, 55.81% were using rivaroxaban, 38.00% were using apixaban, 5.68% were using dabigatran, and 0.45% were on edoxaban. The mean (SD) age of the study cohort was 76 (7.52) years. Most of the patients were females (54.60%), had no subsidy (63.23%), were prevalent users (52.62%), with a CHA2DS2-VASc score 3 (54.34%), and HAS-BLED score ≥2 (63.33%). Most patients were prescribed with antihyperlipidemic agents (64.96%). Hypertension (19.40%) and CAD (12.09%) were the most common comorbidities.
When the 5-group adherence trajectory model was tested, the adherence trajectory was split into the “consistently adherent trajectory” and those who became adherent after the eighth month. The sample size in this additional adherence trajectory was small. The trajectory model with 5 groups was not selected as it did not give any additional valuable information compared to the model with 4 groups. Based on the BIC values and the clinical relevance, the model with four distinct trajectories of adherence were selected: Group 1, gaps in adherence (9.3%); Group 2, gradual decline in adherence (29.7%); Group 3, adherent (36.8%); and Group 4, rapid decline in adherence (24.2%) (Fig. 1). The mean PDC (standard deviation) during the one-year follow-up was 0.22 (0.12) for the rapid decline trajectory, 0.64 (0.13) for the gradual decline trajectory, 0.71 (0.11) for the gaps in adherence trajectory, and 0.93 (0.05) for the adherent trajectory.
The descriptive statistics of patients in the four adherence trajectories are presented in Table 1. There were a greater proportion of males in the rapid decline trajectory. Most of the patients in the lower adherence trajectories did not have any subsidy and had lower stroke and bleeding risk scores. The average CMS risk was highest in the adherent trajectory.
Table 1 Patient Demographics and Clinical Characteristics
Variables
|
Total Patients (N=1969)
|
Gaps in adherence (N=163)
|
Gradual Decline (N=567)
|
Adherent (N=757)
|
Rapid Decline (N=482)
|
P value
|
Age
|
|
<75 years
|
799 (40.58)
|
58 (35.58)
|
233 (41.09)
|
306 (40.42)
|
202 (41.91)
|
0.54
|
>=75 years
|
1170 (59.42)
|
105 (64.42)
|
334 (58.91)
|
451 (59.58)
|
280 (58.09)
|
Gender
|
|
Female
|
1075 (54.60)
|
94 (57.67)
|
329 (58.02)
|
426 (56.27)
|
226 (46.89)
|
0.001†
|
Male
|
894 (45.40)
|
69 (42.33)
|
238 (41.98)
|
331 (43.73)
|
256 (53.11)
|
Health plan
|
|
No subsidy
|
1245 (63.23)
|
128 (78.53)
|
377 (66.49)
|
390 (51.52)
|
350 (72.61)
|
<0.0001†
|
Low-income subsidy
|
724 (36.77)
|
35 (21.47)
|
190 (33.51)
|
367 (48.48)
|
132 (27.39)
|
Prevalent users
|
|
No
|
933 (47.38)
|
53 (32.52)
|
259 (45.68)
|
314 (41.48)
|
307 (63.69)
|
<0.0001†
|
Yes
|
1036 (52.62)
|
110 (67.48)
|
308 (54.32)
|
443 (58.52)
|
175 (36.31)
|
CHA2DS2-VASc score
|
|
Score < 3
|
899 (45.66)
|
77 (47.24)
|
264 (46.56)
|
336 (44.39)
|
22 (46.06)
|
0.83
|
Score >=3
|
1070 (54.34)
|
86 (52.76)
|
303 (53.44)
|
421 (55.61)
|
260 (53.94)
|
HAS-BLED score
|
|
Score < 2
|
1247 (63.33)
|
113 (69.33)
|
374 (65.96)
|
495 (65.39)
|
265 (54.98)
|
0.0002†
|
Score >=2
|
722 (36.67)
|
50 (30.67)
|
193 (34.04)
|
262 (34.61)
|
317 (45.02)
|
PCP visits
|
|
No
|
1501 (76.23)
|
119 (73.01)
|
452 (79.72)
|
563 (74.37)
|
367 (76.14)
|
0.10
|
Yes
|
468 (23.77)
|
44 (26.99)
|
115 (20.28)
|
194 (25.63)
|
115 (23.86)
|
Comorbidities
|
Diabetes Mellitus
|
|
No
|
1749 (88.83)
|
146 (89.57)
|
503 (88.71)
|
671 (88.64)
|
429 (89.0)
|
|
Yes
|
220 (11.17)
|
17 (10.43)
|
64 (11.29)
|
86 (11.36)
|
53 (11.0)
|
0.98
|
Hypertension
|
|
No
|
1587 (80.60)
|
128 (78.53)
|
477 (84.13)
|
616 (81.37)
|
366 (75.93)
|
0.007†
|
Yes
|
382 (19.40)
|
35 (21.47)
|
90 (15.87)
|
141 (18.63)
|
116 (24.07)
|
Coronary Artery Disease
|
|
No
|
1731 (87.91)
|
143 (87.73)
|
500 (88.18)
|
676 (89.30)
|
412 (85.48)
|
0.25
|
Yes
|
238 (12.09)
|
20 (12.27)
|
67 (11.82)
|
81 (10.70)
|
70 (14.52)
|
Renal disease
|
|
No
|
1845 (93.70)
|
157 (96.32)
|
526 (92.77)
|
721 (95.24)
|
441 (91.49)
|
0.02*
|
Yes
|
124 (6.30)
|
6 (3.68)
|
41 (7.23)
|
36 (4.76)
|
41 (8.51)
|
Anemia
|
|
No
|
1828 (92.84)
|
156 (95.71)
|
527 (92.95)
|
707 (93.39)
|
438 (90.87)
|
0.15
|
Yes
|
44 (9.13)
|
7 (4.29)
|
40 (7.05)
|
50 (6.61)
|
44 (9.13)
|
Comedications
|
Antiplatelet agents
|
|
No
|
1798 (91.32)
|
150 (92.02)
|
529 (93.30)
|
685 (90.49)
|
434 (90.04)
|
0.20
|
Yes
|
171 (8.68)
|
13 (7.98)
|
38 (6.70)
|
72 (9.51)
|
48 (9.96)
|
Antiarrhythmic agents
|
|
No
|
1491 (75.72)
|
120 (73.62)
|
424 (74.78)
|
574 (75.83)
|
373 (77.39)
|
0.70
|
Yes
|
478 (24.28)
|
43 (26.38)
|
143 (25.22)
|
183 (24.17)
|
109 (22.61)
|
Antihyperlipidemic agents
|
|
No
|
690 (35.04)
|
70 (42.94)
|
208 (36.68)
|
240 (31.70)
|
172 (35.68)
|
0.03*
|
Yes
|
1279 (64.96)
|
93 (57.06)
|
359 (63.32)
|
517 (68.30)
|
310 (64.32)
|
NSAID
|
|
No
|
1817 (92.28)
|
156 (95.71)
|
517(91.18)
|
706 (93.26)
|
438 (90.87)
|
0.11
|
Yes
|
152 (7.72)
|
7 (4.29)
|
50 (8.82)
|
51 (6.74)
|
44 (9.13)
|
CMS Risk score
|
2.05 (1.20)
|
1.96 (1.33)
|
1.95 (1.09)
|
2.16 (1.26)
|
2.03 (1.19)
|
0.009†
|
* p < 0.05; † p < 0.001.
Note. OR: Odd Ratio; CI: Confidence Interval
CHA2DS2-VASc: Congestive heart failure, hypertension, age ≥75 (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74 and sex category (female); HAS-BLED score: Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol Concomitantly,
PCP: primary care physician; NSAID: Non-steroidal anti-inflammatory drugs; CMS: Centers for Medicare & Medicaid Services
Table 2 presents the odds ratio (OR) that measures the association between baseline characteristics and each of the adherence trajectories. Age, gender, health plan, prevalent users, CHA2DS2-VASc score, PCP visits, hypertension, renal disease, use of antihyperlipidemic agents, and NSAIDs use were significantly associated with lower adherence trajectory. Among the incident DOAC users’ gender, health plan, HAS-BLED score, and CAD were significantly associated with any trajectories (Table 3).
Table 2 Multinomial Logistic Regression to Estimate the Odds Ratio for the Association between patient characteristics and Each Trajectory Group (n=1969)
Variables
|
Gaps in adherence vs Adherent
|
Gradual decline vs Adherent
|
Rapid decline vs Adherent
|
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
Age
|
|
>=75 years vs <75 years
|
1.71 (1.06-2.74)
|
0.02*
|
1.07 (0.80-1.44)
|
0.61
|
0.88 (0.65-1.19)
|
0.41
|
Gender
|
|
Male vs. Female
|
0.70 (0.44-1.10)
|
0.12
|
0.86 (0.66-1.13)
|
0.30
|
1.36 (1.03-1.80)
|
0.03*
|
Health plan
|
|
Low-income subsidy vs. No subsidy
|
3.48 (2.29-5.27)
|
<0.0001†
|
1.77 (1.40-2.24)
|
<0.0001†
|
2.32 (1.79-3.00)
|
<0.0001†
|
Prevalent users
|
|
Yes vs. No
|
1.60 (1.08-2.36)
|
0.01*
|
0.80 (0.63-1.01)
|
0.06
|
0.42 (0.32-0.54)
|
<0.0001†
|
CHA2DS2-VASc score
|
|
Score ≥ 3 vs. Score < 3
|
0.51 (0.28-0.93)
|
0.02*
|
0.88 (0.62-1.25)
|
0.49
|
0.98 (0.68-1.42)
|
0.94
|
HAS-BLED score
|
|
Score ≥ 2 vs. Score < 2
|
0.66 (0.34-1.26)
|
0.21
|
0.85 (0.59-1.23)
|
0.40
|
1.21 (0.83-1.76)
|
0.30
|
PCP visits
|
|
Yes vs. No
|
1.31 (0.88-1.97)
|
0.18
|
0.75 (0.57-0.99)
|
0.04*
|
0.86 (0.65-1.14)
|
0.31
|
Comorbidities
|
Diabetes Mellitus
|
|
|
|
|
|
|
Yes vs. No
|
1.24 (0.67-2.29)
|
0.48
|
1.12 (0.76-1.64)
|
0.54
|
0.80 (0.53-1.20)
|
0.29
|
Hypertension
|
|
|
|
|
|
|
Yes vs. No
|
2.09 (1.05-4.16)
|
0.03*
|
0.83 (0.56-1.25)
|
0.39
|
0.94 (0.63-1.41)
|
0.79
|
Coronary Artery Disease
|
|
Yes vs. No
|
1.35 (0.78-2.34)
|
0.27
|
1.19 (0.83-1.70)
|
0.33
|
1.29 (0.90-1.87)
|
0.16
|
Renal disease
|
|
Yes vs. No
|
1.00 (0.38-2.64)
|
0.98
|
1.73 (1.03-2.91)
|
0.03*
|
1.34 (0.80-2.26)
|
0.25
|
Anemia
|
|
Yes vs. No
|
0.61 (0.26-1.40)
|
0.24
|
1.02 (0.65-1.60)
|
0.90
|
1.19 (0.76-1.86)
|
0.44
|
Comedications
|
Antiplatelet agents
|
|
Yes vs. No
|
0.94 (0.49-1.79)
|
0.85
|
0.70 (0.46-1.07)
|
0.10
|
0.82 (0.54-1.24)
|
0.36
|
Antiarrhythmic agents
|
|
Yes vs. No
|
1.03 (0.69-1.54)
|
0.87
|
1.10 (0.85-1.44)
|
0.44
|
0.97 (0.73-1.30)
|
0.87
|
Antihyperlipidemic agents
|
|
Yes vs. No
|
0.64 (0.45-0.91)
|
0.01*
|
0.83 (0.66-1.05)
|
0.13
|
0.80 (0.62-1.03)
|
0.09
|
NSAID Use
|
|
Yes vs. No
|
0.97 (0.39-2.39)
|
0.95
|
1.61 (1.01-2.60)
|
0.04*
|
1.23 (0.75-2.02)
|
0.39
|
CMS Risk Score
|
0.96 (0.82-1.12)
|
0.65
|
0.91 (0.82-1.00)
|
0.07
|
0.99 (0.89-1.10)
|
0.90
|
|
|
|
|
|
|
|
|
* p < 0.05; † p < 0.001.
Note. OR: Odd Ratio; CI: Confidence Interval
Table 3 Results of Multinomial Logistic Regression Among the Incident Users (N==993)
Variables
|
Gradual decline vs Adherent
|
Rapid decline vs Adherent
|
Age
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
≥75 years vs <75 years
|
0.93 (0.61-1.41)
|
0.74
|
1.07 (0.70-1.63)
|
0.73
|
Gender
|
|
Male vs. Female
|
2.09 (1.42-3.07)
|
0.0002†
|
1.13 (0.77-1.66)
|
0.52
|
Health plan
|
|
Low-income subsidy vs. No subsidy
|
2.56 (1.78-3.65)
|
<0.0001†
|
2.50 (1.76-3.54)
|
<0.0001†
|
CHA2DS2-VASc score
|
|
Score ≥ 3 vs. Score < 3
|
0.79 (0.48-1.31)
|
0.36
|
0.68 (0.41-1.13)
|
0.14
|
HAS-BLED score
|
|
Score ≥ 2 vs. Score < 2
|
1.69 (1.01-2.81)
|
0.04*
|
0.92 (0.55-1.54)
|
0.75
|
PCP visits
|
|
Yes vs. No
|
0.99 (0.68-1.44)
|
0.98
|
0.98 (0.69 -1.41)
|
0.94
|
Comorbidities
|
Diabetes Mellitus
|
|
Yes vs. No
|
1.07 (0.64-1.77)
|
0.78
|
1.51 (0.93-2.45)
|
0.09
|
Hypertension
|
|
Yes vs. No
|
0.73 (0.43-1.22)
|
0.24
|
0.84 (0.49-1.43)
|
0.52
|
Coronary Artery Disease
|
|
Yes vs. No
|
1.16 (1.03-2.71)
|
0.03*
|
1.13 (0.69-1.85)
|
0.62
|
Renal disease
|
|
Yes vs. No
|
0.99 (0.52 -1.90)
|
0.99
|
1.22 (0.63-2.36)
|
0.54
|
Anemia
|
|
Yes vs. No
|
1.18 (0.67-2.07)
|
0.56
|
0.68 (0.37-1.24)
|
0.21
|
Comedications
|
Antiplatelet agents
|
|
Yes vs. No
|
0.64 (0.38-1.08)
|
0.09
|
0.73 (0.44-1.21)
|
0.22
|
Antiarrhythmic agents
|
|
Yes vs. No
|
0.83 (0.53-1.26)
|
0.37
|
0.88 (0.58-1.34)
|
0.57
|
Antihyperlipidemic agents
|
|
Yes vs. No
|
0.86 (0.60-1.23)
|
0.42
|
1.09 (0.77-1.56)
|
0.60
|
NSAID use
|
|
Yes vs. No
|
1.09 (0.58-2.03)
|
0.78
|
1.05 (0.56-1.98)
|
0.86
|
CMS Risk Score
|
1.08 (0.93-1.25)
|
0.27
|
1.05 (0.91-1.21)
|
0.49
|
* p < 0.05; † p < 0.001.
Note. OR: Odd Ratio; CI: Confidence Interval
CHA2DS2-VASc: Congestive heart failure, hypertension, age ≥75 (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74 and sex category (female); HAS-BLED score: Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol Concomitantly,
PCP: primary care physician; NSAID: Non-steroidal anti-inflammatory drugs; CMS: Centers for Medicare & Medicaid Services
Only 9 patients were prescribed with edoxaban and hence they were excluded from the analysis. Dabigatran users were less likely to be associated with the rapid decline trajectory as compared to the rivaroxaban users (OR, 0.49:95% CI 0.26-0.93) (Table not given).
Sub-group analysis among incident DOAC users
A total of 933 incident DOAC users were identified. Based on the GBTM, patients were classified into three adherence trajectories: Group 1, rapid decline in adherence (32.5%); Group 2, gradual decline in adherence (35.6%); and Group 3, adherent trajectory (31.8%) (Fig. 2). The mean PDC during the one-year follow-up was highest in the adherent trajectory 0.94 (0.05), followed by gradual decline trajectory 0.68 (0.12), and rapid decline trajectory 0.20 (0.12).
Separate multinomial regression analysis was conducted for incident users to estimate the odds ratio for the association between baseline characteristics and trajectory group. The results are presented in Table 3. Males were at higher odds to be in gradual decline trajectory (OR, 2.09; 95% CI 1.42-3.07). Incident users with low-income subsidies had higher odds of being associated with the gradual decline trajectory (OR, 2.56; 95% CI, 1.80-3.65) and rapid decline (OR, 2.50; 95% CI, 1.76-3.54). Patients with a HAS-BLED score ≥2 had higher odds of being in the rapid decline trajectory than the adherence trajectory (OR, 1.69; 95% CI, 1.01-2.81). Finally, patients with CAD were more likely to be associated with the rapid decline trajectory than the adherence trajectory (OR, 1.16; 95% CI, 1.03-2.71).
Sub-group analysis among prevalent DOAC users
A total of 1036 prevalent DOAC users were identified. Based on the BIC values and clinical relevance, the GBTM with 5 adherence trajectories were selected. They were Group 1, perfect adherence (5.4%); Group 2, rapid decline in adherence (15.5%); Group 3, gradual decline in adherence (29.6%); Group 4, adherent trajectory (37.6%); Group 5, gaps in adherence (11.8%) (Fig. 3). The mean PDC was highest in the perfect adherence trajectory 1 (0.005) followed by adherent trajectory 0.91 (0.05), gaps in adherence trajectory 0.75 (0.09), gradual decline adherence 0.59 (0.12), and rapid decline 0.24 (0.10).
Males, higher stroke risk score, use of antihyperlipidemic agents were significantly less likely associated with lower adherence patterns. Finally, health plans with no low-income subsidy, high bleeding risk score, and CMS risk score were more likely associated with lower adherence trajectories (Table 4).
Table 4 Results of the Multinomial Logistic Regression Among the Prevalent Users (N =1036)
Variables
|
Rapid decline vs Perfect Adherence
|
Gradual Decline vs Perfect Adherence
|
Gaps in Adherence vs Perfect Adherence
|
Adherent vs
Perfect Adherence
|
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
OR (95% CI)
|
P-Value
|
Age
|
|
≥75 years vs <75 years
|
0.61 (0.28-1.36)
|
0.23
|
1.06 (0.50-2.23)
|
0.87
|
1.22 (0.52-2.87)
|
0.64
|
0.74 (0.36-1.53)
|
0.42
|
Gender
|
|
Male vs. Female
|
0.47(0.22-1.02)
|
0.06
|
0.32(0.15-0.67)
|
0.002†
|
0.40(0.17-0.94)
|
0.04*
|
0.48 (0.24-0.97)
|
0.04*
|
Health plan
|
|
Low-income subsidy vs. No subsidy
|
2.82(1.52- 5.25)
|
0.001†
|
1.65 (0.94-2.91)
|
0.08
|
4.81 (2.32-9.96)
|
<0.0001†
|
1.02 (0.59- 1.76)
|
0.94
|
CHA2DS2-VASc score
|
|
Score ≥ 3 vs. Score < 3
|
0.43 (0.16- 1.14)
|
0.09
|
0.30 (0.12- 0.75)
|
0.01
|
0.22 (0.07- 0.65)
|
0.01*
|
0.38 (0.16- 0.92)
|
0.03*
|
HAS-BLED score
|
|
Score ≥ 2 vs. Score < 2
|
2.43 (0.81- 7.27)
|
0.11
|
2.72 (0.96- 7.74)
|
0.06
|
1.34 (0.38- 4.62)
|
0.64
|
3.28(1.18-9.12)
|
0.02*
|
PCP visits
|
|
Yes vs. No
|
1.65 (0.71- 3.82)
|
0.24
|
1.16 (0.51- 2.60)
|
0.71
|
2.32 (0.96- 5.63)
|
0.06
|
1.96 (0.91- 4.21)
|
0.08
|
Diabetes Mellitus
|
|
Yes vs. No
|
1.23 (0.32- 4.69)
|
0.76
|
2.68 (0.79- 9.02)
|
0.11
|
1.56 (0.35-6.91)
|
0.55
|
2.04 (0.62- 6.73)
|
0.24
|
Hypertension
|
|
Yes vs. No
|
0.89 (0.26- 3.08)
|
0.86
|
0.69 (0.21- 2.24)
|
0.54
|
1.80 (0.45- 7.22)
|
0.40
|
0.59 (0.18- 1.88)
|
0.38
|
Coronary Artery Disease
|
|
Yes vs. No
|
0.66 (0.23- 1.83)
|
0.43
|
0.95 (0.38- 2.37)
|
0.92
|
1.243 0.436 3.546
|
0.68
|
0.88 (0.36- 2.13)
|
0.78
|
Anemia
|
|
Yes vs. No
|
0.77 (0.21- 2.72)
|
0.69
|
0.92 (0.29- 2.91)
|
0.89
|
0.23 (0.04- 1.36)
|
0.11
|
0.60 (0.18- 1.93)
|
0.40
|
|
|
|
Antiplatelet agents
|
|
Yes vs. No
|
2.88 (0.60-13.64)
|
0.18
|
1.99 (0.43- 9.21)
|
0.38
|
0.71 (0.09- 5.38)
|
0.75
|
2.30 (0.52-10.25)
|
0.27
|
Antiarrhythmic agents
|
|
Yes vs. No
|
1.38 (0.68- 2.78)
|
0.37
|
1.40 (0.72- 2.72)
|
0.31
|
1.39 (0.66- 2.95)
|
0.38
|
1.53 (0.80- 2.92)
|
0.20
|
Antihyperlipidemic agents
|
|
Yes vs. No
|
0.51 (0.25- 1.01)
|
0.06
|
0.37 (0.19- 0.71)
|
0.003†
|
0.38 (0.18- 0.79)
|
0.01*
|
0.50 (0.26- 0.94)
|
0.03*
|
NSAID use
|
|
Yes vs. No
|
0.78 (0.17- 3.54)
|
0.75
|
1.10 (0.26- 4.55)
|
0.88
|
0.85 (0.15- 4.89)
|
0.86
|
0.34 (0.07- 1.46)
|
0.15
|
CMS Risk Score
|
0.95 (0.74- 1.21)
|
0.70
|
0.78 (0.62- 0.98)
|
0.04*
|
0.78 (0.58- 1.04)
|
0.09
|
1.00 (0.80- 1.24)
|
0.99
|
* p < 0.05; † p < 0.001.
Note. OR: Odd Ratio; CI: Confidence Interval
CHA2DS2-VASc: Congestive heart failure, hypertension, age ≥75 (doubled), diabetes, stroke (doubled), vascular disease, age 65 to 74 and sex category (female); HAS-BLED score: Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol Concomitantly,
PCP: primary care physician; NSAID: Non-steroidal anti-inflammatory drugs; CMS: Centers for Medicare & Medicaid Services
Apixaban users were more likely to be associated with gradual decline (OR, 1.86;95% CL 1.03-3.36) and gaps in adherence trajectory (OR, 2.35;95% CI 1.19-4.67) as compared to the rivaroxaban users (Table not given).