Adherence and Persistence to Prescribed Medication Therapy Among Medicare Part D Beneficiaries on Dialysis: Comparisons of Benefit Type and Benefit Phase

BACKGROUND: The implementation of Medicare Part D provided insurance coverage for outpatient medications, but when persons reach the “gap,” they have very limited or no medication insurance coverage until they reach a second threshold for catastrophic coverage. In addition, some patients have a low-income subsidy (LIS), and their out-of-pocket costs do not reach the threshold for the gap. Little is known about how these Part D types (LIS versus non-LIS) and benefit phases (before the gap, during the gap, after the gap) affect medication adherence and persistence of dialysis patients. OBJECTIVES: To examine medication use, adherence, and persistence for Medicare-eligible dialysis patients by Part D benefit type and benefit phase. METHODS: A retrospective cohort study using data from the U.S. Renal Data System (USRDS) was conducted for Medicare-eligible dialysis patients. Outcomes included medication use, adherence, and persistence. Patients were categorized into 4 cohorts based on their Part D benefit phase that the beneficiaries reached at the end of the year and LIS receipt in 2007: Cohort 1 = non-LIS and did not reach the coverage gap; Cohort 2 = non-LIS and reached the coverage gap; Cohort 3 = non-LIS and reached catastrophic coverage after the gap; and Cohort 4 = received an LIS and none of the LIS patients reached the coverage gap. Outcomes were measured separately for 5 therapeutic classes of outpatient prescription drugs: antihyperglycemics, antihypertensives, antilipidemics, phosphate binders, and calcimimetics. RESULTS: A total of 11,732 patients met the study inclusion criteria. Patients were distributed among the cohorts as follows: 3,678 (31.3%) patients in Cohort 1 who did not reach the coverage gap; 4,349 (37.1%) patients in Cohort 2 who reached the coverage gap but not catastrophic coverage; 1,310 (11.2%) patients in Cohort 3 who reached catastrophic coverage; and 2,395 (20.4%) patients in Cohort 4 who had an LIS (none of whom reached the gap). Overall, the percentage of patients who were adherent to their medications (≥ 80% medication possession ratio) was low: 39% for antihyperglycemics, 59% for antihypertensives, 54% for antilipidemics, 22% for phosphate binders, and 35% for cinacalcet. There were wide ranges in adherence rates depending on the cohort. For patients on antihyperglycemics, antihypertensives, antilipidemics, phosphate binders, and cinacalcet, the odds ratios for adherence to therapy were 0.76 (95% C I =0.63-0.92), 1.06 (0.94-1.19), 0.80 (0.67-0.95), 0.65 (0.55-0.76), and 0.39 (0.30-0.49), respectively; the hazard ratios for discontinuation of therapy were 1.18 (95% CI 1.06-1.31), 1.01 (0.93-1.10), 1.25 (1.12-1.40), 1.13 (1.05-1.21), and 1.61 (1.75-1.82), respectively, for Cohort 2 patients who reached the coverage gap compared with those in Cohort 4 who received an LIS. In addition, when comparing adherence before and after the benefit gap, patients in Cohort 2 were significantly more likely to be nonadherent to medications for diabetes (relative risk (RR) = 1.71, 95% CI = 1.48-1.99), hypertension (RR = 1.69, 95% CI = 1.54-1.85), hyperlipidemia (RR = 2.01, 95% CI = 1.76-2.29), hyperphosphatemia (RR = 1.74, 95% CI = 1.55-1.95), and hyperparathyroidism (RR = 2.08, 95% CI = 1.66-2.60) after reaching the coverage gap. CONCLUSIONS: More than half of Medicare beneficiaries on dialysis reached the Part D coverage gap in 2007. Our findings suggest that the Part D coverage gap was significantly associated with decreases in adherence and persistence for medications frequently used in patients undergoing dialysis. Patients who reached the coverage gap (Cohort 2) often decreased use of or discontinued critical medications after reaching the coverage gap. Compared with patients who had an LIS (Cohort 4), patients in Cohort 2 had significantly lower medication adherence and persistence levels. The negative impact of the Part D coverage gap (high out-of-pocket cost sharing) on medication adherence and persistence for Medicare-eligible dialysis patients has implications for currently proposed Medicare end-stage renal disease bundled reimbursement payment and requires more research.


R E S E A R C H
• Poor adherence with prescribed medications is a widely recognized problem in dialysis patients due to the complexity of the regimen and lifelong duration of therapy. A recent systematic literature review found that more than half of the included studies reported nonadherence rates of ≥ 50% in the end-stage renal disease (ESRD) population. • The Part D coverage gap has been associated with reduced medication adherence in the general Medicare population; in other words, beneficiaries are significantly more likely to adopt costlowering strategies when entering the coverage gap. • ESRD patients have a higher risk of reaching the coverage gap compared with other Medicare beneficiaries with Part D benefits (48% vs. 23%).

What is already known about this subject
• Throughout 2007, non−low-income subsidy (LIS) Part D Medicare patients on dialysis who reached the coverage gap, but not the catastrophic phase, had mean out-of-pocket expenses of $1,854 (41.9% of medication costs), while patients who received an LIS had mean out-of-pocket expenses of only $192 (3.6% of medication costs).
The recent U.S. Renal Data System (USRDS) report estimated that 72% of hemodialysis patients received Medicare Part D coverage. 17 In 2008, 47% of dialysis patients reached the coverage gap, and 13% reached the catastrophic coverage phase, compared with 23% and 3%, respectively, in the general Medicare program. 18 Given their need for chronic medication therapy and multiple medications to treat comorbid conditions, patients with ESRD are at a particularly high risk of facing high out-of-pocket medication costs. 17 The high levels of clinical need for drug therapy combined with substantial out-of-pocket costs for patients with ESRD may make them more vulnerable to cost-related medication nonadherence.
Although inadequate adherence with prescribed medication is a widely recognized problem, very few studies have been conducted to understand cost-related medication nonadherence in the ESRD population. 19 The majority of the ESRD population is enrolled in Medicare, but little is known about how Part D benefit types and benefit phases affect medication adherence in this population.
The Affordable Care Act of 2010 is gradually closing the coverage gap by reducing the coinsurance in the donut hole until it reaches 25% in 2020. At the same time, the Centers for Medicare and Medicaid (CMS) implemented the new Medicare bundled ESRD prospective payment system (PPS) in 2011. Under this system, a single payment covers dialysis, supplies, laboratory tests, intravenous, and oral drugs. 20 The inclusion of dialysis-specific oral medications into the bundle is expected to be effective beginning in 2016, which is 2 years later than originally planned. 21 With these important Part D benefit changes in the near future, understanding how the coverage gap and low-income subsidy (LIS) affect medication use and adherence is critical to to designing adequate drug benefits for Medicare beneficiaries.
Therefore, the objectives of this study were the following: (a) to examine medication use among Medicare beneficiaries on dialysis stratified by the Part D benefit phase that the non-LIS beneficiaries reached at the end of the year (i.e., initial coverage, coverage gap, catastrophic coverage), as well as those with an LIS; (b) to compare medication adherence and persistence of dialysis patients by Part D benefit phase and LIS receipt; and (c) to determine whether medication adherence differs before and after the coverage gap was reached among patients reaching the coverage gap but not catastrophic coverage.

■■ Methods Data Source
A retrospective analysis using data from the USRDS was conducted for Medicare-eligible dialysis patients. The USRDS is a national registry of patients with ESRD based on Medicare claims submitted to the CMS by providers. This database has near universal inclusion of ESRD patients in the United States. 18 The USRDS used the billing information from the CMS A s a result of incomplete replacement of kidney function, chronic dialysis patients with end-stage renal disease (ESRD) face many health problems, including chronic inflammatory states, malnutrition, increased risk for cardiovascular morbidity and mortality, phosphate retention, secondary hyperparathyroidism, diabetes, and dyslipidemia. 1 Most dialysis patients, therefore, require many medications and have a high pill burden. 2,3 It has been reported that dialysis patients are prescribed, on average, 12-19 different medications. 2,3 Unfortunately, poor adherence with prescribed medications is a widely recognized problem in dialysis patients due to the complexity of the regimen and lifelong duration of therapy. 2,4,5 A recent systematic literature review found that more than half of the included studies reported nonadherence rates of ≥ 50% in the ESRD population. 5 High out-of-pocket costs borne by patients can be a deterrent to therapeutic adherence and, therefore, to the effectiveness of prescribed medications. 6 Even small increases in these costs can lead to potentially significant reductions in medication adherence, which, in turn, can have serious consequences for patients' health. 7 A national survey found that prescription coverage and out-of-pocket costs were determinants of underuse across medication types, although rates of underuse varied substantially across treatments. 8 To date, researchers have observed reduced drug utilization among Medicare Part D beneficiaries who reach the coverage gap and have no other financial assistance to pay for drugs. [9][10][11] Several other studies have demonstrated decreases in medication adherence and associated health outcomes for Part D beneficiaries who reached the coverage gap phase. [12][13][14][15][16] More specifically, Fung et al. (2010) reported that out-of-pocket expenditures were 189% higher, and adherence to 3 chronic medications-including oral medications used to treat diabetes, hypertension, and dyslipidemia-was significantly lower among beneficiaries with a coverage gap versus no gap. 14 Under the Medicare Part D standard plan cost-sharing structure in place in 2007, beneficiaries paid a deductible and 25% coinsurance up to $2,250 in total drug spending (which corresponds to $799 in out-of-pocket [OOP] expenses.) Once the $2,250 threshold was reached ($799 OOP), a coverage gap occurred and beneficiaries paid 100% of their medication costs OOP until a second (catastrophic coverage) threshold ($3,850 in OOP costs or $5,100 in total drug spending) was reached. After reaching the catastrophic level, patients paid the greater of 5% of medication costs or $2 copayment for generic drugs and $5 for branded drugs for the remainder of the year.
First, patients were categorized based on type of coverage-LIS versus non-LIS. LIS patients did not reach the coverage gap. The non-LIS group was then subcategorized into 3 cohorts based on their benefit phases at the end of the year. The resulting 4 cohorts included the following: Cohort 1 = non-LIS patients who did not reach the coverage gap (paid OOP costs < $799) by the end of the year; Cohort 2 = non-LIS patients who reached the coverage gap but not the catastrophic coverage phase (paid $799 ≤ OOP costs < $3,850) by the end of the year; Cohort 3 = non-LIS patients who reached catastrophic coverage (paid ≥ $3,850 OOP costs); and Cohort 4 = LIS patients, none reached the coverage gap.

Medication Use and Costs
Medication use was defined as 1 or more prescription fills in any of the therapeutic classes for the period of January 1, 2007, through December 31,2007. Pharmacy claims were identified using generic names from Part D prescription claims data (see Appendix A). Medication use was measured for 5 therapeutic classes of outpatient prescription drugs: antihyperglycemics, antihypertensives, antilipidemics, phosphate binders, and calcimimetics. Medication costs were defined as total drug costs and OOP costs for each therapeutic class and all medications. OOP costs for each prescription are equal to the amount paid directly by the patient. Total drug costs for each prescription are defined payments including the amount Medicare paid plus patient OOP costs.

Adherence
Before calculating medication adherence, medication treatment patterns were assessed (i.e., mono, dual, triple, or quadruple therapy). Monotherapy was defined as treatment with only 1 medication class within each therapeutic class (e.g., sulfonylureas, biguanides, or thiazolidinediones in antihyperglycemics). Dual (triple or quadruple) therapy refers to a coadministration of 2 (3 or 4) separate medication classes with at least 2 overlapping periods of 30 days or 1 overlapping period of 60 days (e.g., sulfonylureas and biguanides or sulfonylureas and thiazolidinediones, etc.).
Medication adherence was defined using the medication possession ratio (MPR), which is the sum of total days' supply for all fills divided by the number of days between first and last fills plus days supply of the last fill or the number of days between first fill and December 31, whichever ends later. 22 Patients who received at least 2 prescriptions for the same therapeutic class were included to estimate MPR. MPR for dual, triple, and quadruple therapies were determined by calculating the average of the MPRs of the individual medications that constituted the dual, triple, and quadruple therapies. Medication adherence was dichotomized, with adherence defined as MPR ≥ 80% and nonadherence defined as MPR < 80%. In addition, we conducted a separate analysis to determine whether medication adherence differed before and after the coverage gap was exceeded among patients in Cohort 2. For this analysis, patients were included if they received at least 2 prescriptions total and at least 1 prescription before the coverage gap date in each therapeutic class.

Persistence
Medication persistence was defined as the duration of therapy from the first fill date until discontinuation. Persistence was calculated by summing the number of days from the filling of the first medication to the end date of the last medication claim (fill date plus days supply before a 30-day gap and a 60-day gap). 22 25 Kaplan-Meier survival curves were used to depict the percentage of patients who remained persistent among the cohorts. Cox proportional hazards regression was used to measure the difference in persistence among the cohorts controlling for covariates.
All statistical analyses were two-tailed, with the significance level set a priori at α = 0.05. All analyses were conducted using SAS software (version 9.2; SAS Institute Inc., Cary, NC) and Stata (version 11.1; Stata Corp., College Station, TX).  For patients in Cohort 2 (coverage gap), the mean age was 73 years; 55% were male; 76% were white; and ESRD duration was 5 years. Patients in Cohort 4 were more likely to be younger (62 years) and African American (50%) than the other cohorts.

Medication Costs and Adherence Before and After Coverage Gap
Medication costs and adherence before and after the coverage gap among patients in Cohort 2 are shown in Table 6. Patients' OOP costs increased significantly after the patient reached the gap in Cohort 2 (P < 0.05), although total medication costs decreased significantly (P < 0.05). The GEE model showed that

Medication Use and Costs by 5 Therapeutic Classes of Outpatient Prescription Medications Among Study Cohorts
Unlike previous studies, we also calculated adherence and persistence measures for 4 cohorts of Medicare Part D beneficiaries categorized by type of coverage (LIS vs. non-LIS) and benefit phase at the end of the year. In the present study, we found that a significantly higher proportion of the Part D non-LIS enrollees (Cohorts 1, 2, and 3) reached the coverage gap than in dialysis patients, which found antihypertensive adherence to be 58%, 26 and cinacalcet adherence to be 29% (with a refill rate of 37%). 27 The adherence to phosphate binders was lower in the current study (22%) than in a previous study (38%), which was conducted in 3 different dialysis centers, 2 but the difference could be explained by differences in study populations.

Adherence and Persistence to Prescribed Medication Therapy Among Medicare Part D Beneficiaries on Dialysis: Comparisons of Benefit Type and Benefit Phase
proportions reported in the 2012 USRDS annual report (61% vs. 47%). 18 Our findings suggest that patients who reached the coverage gap but not catastrophic coverage (Cohort 2) had significantly lower adherence and persistence levels to therapy for diabetes, lipid-lowering medications, phosphate binders, and cinacalcet compared with those who received an LIS (Cohort 4). Interestingly, after adjustment for the covariates measured in our study, patients in Cohort 1 (non-LIS who did not reach the coverage gap) had the lowest adherence and persistence, while those in Cohort 3 (reached catastrophic coverage) had the highest adherence and persistence among the 4 cohorts. This result was somewhat surprising because patients who did not reach the coverage gap would be expected to be relatively healthier and to not require many medications. However, only 17%-47% of patients were adherent to their medications among those who did not reach the coverage gap. Dialysis patients may be especially sensitive to the coverage gap issue even prior to the coverage gap, given the large number of medications needed to treat their comorbid conditions. A recent study found that Part D enrollees' adherence decreased prior to the gap because patients anticipated that continued spending would make them more likely to reach the gap. 28 Patients use different strategies to face the coverage gap and mitigate the effects of the coverage gap. Patients may fall into the initial coverage phase (Cohort 1), evidenced by OOP spending < $799, due to lack of medication adherence and persistence. These patients might have avoided or delayed reaching the coverage gap by reducing and/or discontinuing their medications. In addition, patients may fall into the catastrophic coverage phase (Cohort 3), evidenced by OOP spending ≥ $3,850, because they were adherent and persistent with filling prescriptions. Some sicker patients, who entered the coverage gap early in the year, might increase their spending, trying to come out of the coverage gap as soon as possible to reach the catastrophic coverage phase. 10 Other patients may adopt cost-lowering strategies (i.e., using less of medications, discontinuing medications, or not filling prescriptions) during the coverage gap, 29,30 making them more likely to reach the coverage gap but not the catastrophic coverage phase (Cohort 2). Further analysis of medication adherence before and after reaching the coverage gap (Cohort 2) provides information regarding the association between the coverage gap and adherence. After adjustment for the covariates, patients were up to 2 times more likely to be nonadherent to prescription medications after reaching the coverage gap. These findings

Medication Persistence Until 30-Day Gap by 5 Therapeutic Classes of Outpatient Prescription Medications Among Study Cohorts
were consistent with previous studies in general Medicare populations. Several studies assessing the effects of the Part D coverage gap used the number of prescriptions filled as the outcome variable, 10,11,18 while few studies compared the adherence or discontinuation before and after reaching the coverage gap. 13,31 Gu et al. (2010) found that compared with Part D beneficiaries with full coverage, beneficiaries with no coverage were 62% less likely to be adherent to diabetic medication after reaching the coverage gap. 31 Polinski et al. (2011) also found that gap-exposed patients were twice as likely to discontinue their medications for cardiovascular conditions, diabetes, depression, dementia, or rheumatoid arthritis. 13 In addition, this study found that for patients who reached the coverage gap (Cohort 2), adherence and persistence to dialysis-specific medications (i.e., phosphate binders and cinacalcet) were lower than to nondialysis-specific medications (i.e., antihyperglycemics, antihypertensives, and antilipidemics). A possible explanation for these differences could be differences in medication costs by therapeutic classes. Based on the findings from pharmacy costs, total and OOP costs were significantly higher for dialysis medications (e.g., phosphate binders and cinacalcet) compared with medications for diabetes mellitus, hypertension, and dyslipidemia.  Beginning in 2016, payment for dialysis-specific medications, currently covered under Medicare Part D, will be integrated into the Medicare Part B ESRD bundled PPS, whereas payment for other oral medications (e.g., antihyperglycemics, antihypertensives, and antihyperlipidemics) will continue to be covered by Part D. This study adds a rationale for filling the coverage gap in the Part D phase design because the coverage gap was significantly associated with cost-related nonadherence in dialysis patients. Furthermore, the implementation of the bundled PPS is expected to lead to pronounced adjustments in the treatment of dialysis patients. In the present study, dialysis-specific medications accounted for 70% of total medication costs. However, integrating Part D drugs into Part B payments is believed to lack any policy precedent, and there is concern regarding the potential for inadequate funding. 21 Our study provides a timely, needed assessment of medication use, costs, and adherence under Part D and can help in setting an adequate bundle rate for dialysis-specific medications. Under Part B, patients are responsible for 20% of drug costs and may have supplemental insurance (Medigap) that covers their coinsurance responsibility for Part B drugs, whereas many of those with an LIS under Part D will have no subsidy under Part B. 21 It is uncertain whether the new ESRD bundled PPS will increase or decrease patients' OOP spending and how the bundled rate will affect patients' access to their medications. Our findings provide important policy implications for the beneficiaries who have high drug use and high OOP spending under Part B. Policy decision makers are encouraged to include adequate reimbursement for medications when bundle rates are being set to ensure that patients have reasonable access to critical medications.

Limitations
Studies using administrative claims have a number of limitations. While the standard benefit under Medicare Part D was used to classify subjects, the exact plan in which an individual was enrolled is unknown. Many prescription drug plans have no deductible or use drug copayments instead of coinsurance, and some include coverage during the coverage gap. In 2010, 60% of hemodialysis subjects had no deductible, and 15% had gap coverage (typically for generic medications). 18 The structure of the standard Part D plans used for this study may differ from some subjects' nonstandard plans. However, we used actual OOP drug spending to categorize subjects, which ascertained that they had actuarially equivalent OOP spending before reaching the coverage gap regardless of plan type. In this study, most of the OOP drug spending was observed for phosphate binders and cinacalcet-for which generics are not available. Thus, generic gap coverage would have a limited impact on OOP spending, and relatively few individuals have insurance for branded medications during the coverage gap.
Because this was an observational study, we could not conclude that a causal effect was present between high OOP expenses and poor adherence/persistence. It is a critical question whether our findings indicate direct consequences from Part D OOP expense levels or inherent differences between cohorts. However, we found consistent associations for patients who reached the coverage gap (Cohort 2)-they had significantly poorer adherence and persistence to most outpatient prescription medications compared with patients who received an LIS (Cohort 4), after adjusting for demographic and clinical factors. Furthermore, for Cohort 2 (those reaching the coverage gap), adherence decreased significantly for all 5 therapeutic classes of prescription medications after the patient reached the gap (again after adjusting for covariates). Nevertheless, unmeasured confounding variables might influence the results. In addition, as a general limitation with the use of a claims database, MPR was used as a proxy measure of adherence because it was not possible to determine if the patients actually used the medications as prescribed, but merely that they had received their medications. There is the possibility that MPR calculations might have under-or overestimated adherence. Patients were classified as using polytherapy (i.e., dual, triple, or quadruple therapy) only during the period when the medications overlapped and using monotherapy for the remaining part of the study period. This might have underestimated adherence measures. Patients classified as using monotherapy may have received 2 different medications in the same class of medication (e.g., 2 beta blockers), which might have overestimated adherence measures. In addition, patients may have filled prescriptions outside the Part D benefit. The extent to which dialysis patients fill prescriptions outside of their Part D plans is unknown. 32 However, the inclusion/exclusion criteria in this study probably were conservative, since patients were excluded if they were dual-eligible, received a retiree drug subsidy, or were on an employer-sponsored health benefit plan.

■■ Conclusions
More than half of Medicare beneficiaries on dialysis reached the Part D coverage gap in 2007. Our results indicate that patients often decreased the use of or discontinued critical medications after reaching the coverage gap. In addition, compared with patients who had an LIS, patients who reached the coverage gap had significantly lower medication adherence and persistence levels. This finding may reflect the large economic burden of medication costs faced by Medicare dialysis patients without any subsidy and raises concerns that the lack of drug coverage could lead to adverse health consequences for financially vulnerable patients. To improve medication adherence for necessary drug use for dialysis patients, a Medicare drug plan may consider offering more generously subsidized coverage than would be provided in the current proposal for the new bundled payment.