The Impact of Medicare Part D on the Proportion of Out-of-Pocket Prescription Drug Costs Among Older Adults With Diabetes

Diabetes is a leading cause of excessive health care costs as well as increased morbidity and mortality among older adults in the U.S. (1). Approximately one in four (25.9%) adults aged 65 years and older in the U.S. has diabetes, a prevalence two to six times higher compared with other age-groups (1). Taking prescribed medications daily to maintain glycemic control is a component of diabetes self-management to prevent the complications associated with diabetes (2). Poor adherence with diabetes medications is significantly associated with higher risks for hospitalization and mortality (3).

Older adults with diabetes are especially vulnerable to prescription drug costs because of the need for multiple medications due to coexisting illnesses (4). In one sample of older adults with diabetes, subjects reported taking, on average, 8.2 (SD 4.0) medications each day (5). The cost of prescription drugs for adults aged 65 years and older with diabetes is two to seven times higher ($18,875 million annually) compared with those under age 65 years (6). High out-of-pocket medication cost contributes to poor medication adherence (4) due to lack of prescription drug coverage (7). Approximately 16% of older adults with diabetes report cost-related nonadherence to at least one medication (8).

To extend prescription drug coverage for Medicare beneficiaries, Medicare Part D was initiated in 2006 as part of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (9). However, the U.S. Department of Health and Human Services (10) estimates that more than one in four Part D beneficiaries discontinue taking their drugs as prescribed due to the burden of cost when they reach the Medicare coverage gap during which beneficiaries must assume full responsibility for their prescription drug costs.

This study evaluated the impact of Medicare Part D on reducing the burden for prescription drug costs among Medicare beneficiaries with diabetes by comparing a 6-year period prior to (2000–2005) and after (2006–2011) the implementation of Medicare Part D.

We performed a secondary data analysis using data from the Medical Expenditure Panel Survey (MEPS) provided by the Agency for Healthcare Research and Quality (AHRQ) for years 2000–2011, using the household component. The household variables used were demographics, health conditions and status, charges and payments for pharmacy expenditures, access to care, satisfaction with care, and health insurance coverage.

The sample for this study consisted of adults with a diagnosis of diabetes, either aged 65 years and older who were Medicare beneficiaries (Medicare group) or 50–60 years old who were not eligible for Medicare benefits (comparison group) who had a prescription record of diabetes medications. On the basis of their Part D benefit status, the Medicare group was classified as either Medicare Part D beneficiaries (Part D group) or nonbeneficiaries (non-Part D group). The younger comparison group was selected from adults 50–60 years old who did not receive the Medicare benefit. We excluded 50- to 60-year-old Medicare recipients (e.g., disability or end-stage renal disease, etc.) from the comparison group because they were affected by Part D and thus could not function as a proper control group. This age-group was selected for the comparison sample because of their proximity in age to the Medicare group and their lack of Medicare coverage. This group did not include those who were 61–64 years old because each MEPS panel was followed for 2 years, and therefore a minimum 3-year age gap between the Medicare and comparison groups was necessary to avoid any overlap.

Using the clinical classification software code variable in the MEPS, self-reported diabetes diagnosis (clinical classification software code 49 [diabetes without complications] or 50 [diabetes with complications]) was identified. Diabetes medications were identified using the Multum Lexicon therapeutic classification variable (TC1S1 99 [antidiabetes agents]) derived from Cerner Multum Inc. in the MEPS. Antidiabetes drugs consisted of the following 10 subtherapeutic classes: sulfonylureas, biguanide, insulin, α-glucosidase inhibitors, thiazolidinediones, meglitinides, antidiabetes combinations, dipeptidyl peptidase 4 inhibitors, amylin analogs, and incretin mimetics.

Proportion of out-of-pocket pharmacy costs to total pharmacy expenditures was defined as annual out-of-pocket costs for all prescribed medications including both diabetes- and nondiabetes-related paid by self or family divided by annual total pharmacy expenditures. Out-of-pocket pharmacy costs as well as total pharmacy expenditures reported in MEPS were directly collected from pharmacists and measured as actual annual dollar values for each year. All dollar values during years 2000–2011 were adjusted to the first half of 2013 U.S. dollar values, the most recent year’s dollar values posted at the time of data analyses (11). These adjustments were in accordance with the inflation rate for each year using the Consumer Price Index (11).

All statistical analyses were performed using the SAS statistical software, version 9.4. As MEPS data were collected using a complex stratified multistage design, SAS procedures for survey sampling were used to calculate correct SEs and P values. To correct unequal probability sampling bias and nonresponse bias, the sampling weights were used in all analyses. Sampling weights were provided by the AHRQ to generate representative estimates. This study received approval from the Columbia University Medical Center Institutional Review Board prior to study initiation.

Descriptive statistics were used to describe the sample. Characteristics of the Medicare and comparison samples were compared. Within the Medicare group, the characteristics of Medicare Part D beneficiaries were compared with those of nonbeneficiaries. In addition, we examined the Medicare Part D enrollment rate among Medicare beneficiaries with diabetes as well as the proportion of Part D beneficiaries who reached the coverage gap and/or catastrophic coverage from 2006 to 2011 based on the Part D threshold limit for each year (12). Because the coverage gap threshold is based on actual dollar values and differs by year, dollar adjustments by discounting were not performed for this analysis. In addition to demographic characteristics (age, sex, race/ethnicity), variables for descriptive statistics included coexisting illness, health insurance type, and access to care (problems with access to prescribed medicines, delays in getting prescribed medicines). To examine the number and severity of coexisting illnesses, a Charlson comorbidity score was computed using the Charlson comorbidity index from D'Hoore et al. (13) because it uses the first three digits of ICD-9 codes consistent with the data provided in MEPS. Health insurance type was classified as any private insurance (this category includes supplemental insurance for those with Medicare), public insurance coverage during the year (this category includes those with Medicare only and dual eligibility for Medicare and Medicaid), and uninsured.

Using a segmented regression of interrupted time series analysis, we evaluated the longitudinal effect of implementation of Medicare Part D on out-of-pocket pharmacy costs by considering pre-Part D trends. This approach compared the trend of yearly out-of-pocket prescription drug cost burden between two time periods: before (2000–2005) and after (2006–2011) the implementation of Part D. The analysis was performed for the Medicare and comparison groups separately. Among the Medicare group during the post-Part D period (2006–2011), the level of yearly out-of-pocket pharmacy costs for Part D beneficiaries was compared with that of non-Part D beneficiaries.

For this analysis, the outcome variable, race/ethnicity-adjusted mean proportion of out-of-pocket pharmacy costs, was modeled as illustrated by Eq. 1. As sample sizes varied by year, weights were adjusted by the U.S. population race and ethnicity composition (non-Hispanic white 63.7%, non-Hispanic black 12.2%, Hispanic 16.3%, and other races 7.8%), using 2010 U.S. Census data (14) in order to treat each year’s data equally and make the weighted sample size statistically equivalent year to year. The model includes the variables of time (year), policy (Medicare Part D policy implementation status), timepost (year after Part D implementation), and Part D (Medicare Part D enrollment status):

To estimate causal effects of Medicare Part D on the mean proportion of out-of-pocket pharmacy costs, we compared average change during the two time periods (before [2000−2005] and after [2006−2011] Part D phases) using a difference-in-difference analysis for the Medicare group with that of the comparison sample. This analysis eliminates the effect of potential confounding factors by comparing pre- and post-Part D between the Medicare and comparison samples. The outcome variable, the race/ethnicity-adjusted mean proportion of out-of-pocket pharmacy expenditures, was modeled using the Eq. 2. For this analysis, the variables group (Medicare or comparison groups) and policy (Medicare Part D policy implementation status) were included in the following:

Table 1 presents characteristics of the sample before and after the implementation of Medicare Part D. The Medicare group consisted of 4,664 Medicare beneficiaries with diabetes age 65 years and older including 1,916 during the pre-Part D and 2,748 during post-Part D periods. Of the 2,748 individuals in the Medicare group during the post-Part D period, 56.7% were enrolled in Medicare Part D. The comparison group consisted of 2,938 adults with diabetes aged 50‒60 years who did not report receiving Medicare benefits. Both groups were similar in distribution of proportion of those reporting problems with access or delays in getting necessary prescribed medications and total pharmacy costs. However, compared with the Medicare group, the comparison group, on average, had a lower proportion of women, higher proportion of blacks or Hispanics, lower coexisting illnesses, lower out-of-pocket pharmacy spending before Medicare Part D, and lower proportion of out-of-pocket pharmacy costs during the years prior to implementation of Medicare Part D. Although there were no differences in total pharmacy costs between the Medicare and younger comparison groups, the average proportion of out-of-pocket pharmacy costs was higher for the Medicare group prior to Part D (51.4% versus 40.6%). After implementation of Medicare Part D, the mean proportion of out-of-pocket pharmacy costs was lower for the Medicare group compared with the younger comparison group (29.2% versus 37.7%).

Following implementation of Part D, both Part D and non-Part D groups were similar in distribution of age, coexisting illnesses, total pharmacy costs, and out-of-pocket pharmacy costs. However, the Part D group included more women, blacks or Hispanics, those reporting problems with access or delays in getting necessary prescribed medications, and lower proportion of out-of-pocket pharmacy costs to total pharmacy costs compared with the non-Part D group.

Figure 1 illustrates Part D enrollment and categorizes the level of pharmaceutical costs during the 2006–2011 for Medicare beneficiaries with diabetes. Part D enrollment gradually increased from 2006 (45.7%) to 2011 (52.4%). The proportion of Part D enrollees who remained within the Part D initial coverage limit increased from 39.9% to 59.1% during this period. In 2011, 40.9% of older adults with diabetes had pharmaceutical costs that placed them either in the Part D coverage gap, often referred to as the “donut hole,” or above the catastrophic threshold.

Figure 2 depicts the trend of race/ethnicity-adjusted proportion of out-of-pocket pharmacy costs during 2000–2011 for the Medicare and comparison groups and illustrates the sudden change in out-of-pocket spending that occurred in 2006 following implementation of Part D. For the Medicare group, a 1.3% yearly decreasing trend (SE 0.6, P = 0.03) was observed in the proportion of out-of-pocket pharmacy costs during the 6 years preceding Medicare Part D policy initiation. After the implementation of Medicare Part D, the yearly rate of decline of the proportion of out-of-pocket pharmacy costs decreased to 0.5% per year within the Medicare group; however, the trend change of 0.8% was not statistically significant (0.6, P = 0.5). Medicare Part D policy was associated with the reduction in the proportion of out-of-pocket costs for pharmacy expenditures by 13.5% (2.1, P < 0.0001) for both the Part D and non-Part D groups. However, when adjusted for race/ethnicity, the Part D group spent, on average, 5.3% less (0.8, P < 0.0001) on out-of-pocket expenses for pharmacy costs compared with the non-Part D group. For the non-Medicare comparison group, the baseline proportion of pharmacy out-of-pocket costs was 44.8%.

Figure 3 graphically presents the results of the difference-in-difference approach, showing the counterfactual out-of-pocket spending we estimated for the Medicare group using the non-Medicare group’s trend. We compared the difference between estimated and observed change in the mean proportion of out-of-pocket pharmacy costs for the Medicare group after Part D implementation to estimate the causal effect of Part D on the Medicare group. After Part D implementation, the mean proportion of out-of-pocket pharmacy costs decreased by 19.4% (SE 1.7, P < 0.0001) for the Medicare group compared with the younger comparison group.

Medicare Part D increased access to prescribed drugs for older adults with diabetes. Compared with the non-Medicare group, Medicare beneficiaries had, on average, 10.8% higher spending on prescription drugs before implementation of Medicare Part D. After Part D, although all in the Medicare group benefited from lower out-of-pocket spending regardless of their Part D status, Part D group had 5.3% lower out-of-pocket spending for prescription drugs compared with those without Part D. Overall, the effect of Medicare Part D was a 19.4% decrease in prescription drug spending during 2006–2011. This is particularly noteworthy as it reflects a period of time where overall diabetes expenditures and medication costs have increased (15), with the highest costs attributed to older adults with diabetes (6).

During 2000–2005, the proportion of out-of-pocket pharmacy costs for the Medicare group was declining. Although the reason for this decrease is not known, there is a potential explanation. Prior to Medicare Part D, approximately one in four older adults with diabetes lacked prescription drug coverage; of these, one in two reported cost-related nonadherence (16). Also, there was a 2-year gap between enactment and initiation of Medicare Part D (17). It is possible that in anticipation of the new Part D insurance benefit and lower out-of-pocket costs, some Medicare beneficiaries opted not to purchase their medications (18) while they awaited prescription drug cost relief through the new program.

Collectively, our findings are consistent with other research (19–24). Safran et al. (21) and Millett et al. (22) reported decreased out-of-pocket prescription drug costs particularly for those Medicare beneficiaries who previously had limited prescription drug coverage. Briesacher et al. (19) also found that beneficiaries with Part D had decreased out-of-pocket pharmacy costs, with the exception of those in poor health and low-income beneficiaries without Medicaid. Similarly, Li et al. (24) used an interrupted time series analysis using the 1996–2008 MEPS participants with diabetes and found that the Medicare group with diabetes had a greater reduction ($530) in out-of-pocket pharmacy costs than the overall Medicare population ($143–$230) reported in the previous studies, probably because of higher spending on medications in patients with diabetes than those without diabetes (19).

Our study adds to this growing body of work. The findings of this study suggest that women, racial/ethnic minorities, and beneficiaries without private supplemental coverage had the greatest uptake of the Part D prescription drug benefit. Although out-of-pocket spending for prescription drugs in Medicare beneficiaries with diabetes decreased after Part D was implemented, we found almost two of every five Medicare beneficiaries with diabetes experienced the coverage gap in 2011. Although earlier studies have found that having a diagnosis of diabetes is significantly correlated with coverage gap entry (25), our finding is significantly higher than previously reported (26). This is of concern because one in four Part D beneficiaries discontinues their prescription drugs when they enter the coverage gap (10).

The high cost of recently approved diabetes medications as well as the need for multiple medications to treat coexisting illnesses (4) may contribute to the high costs faced by the elderly with diabetes to maintain their health. In 2013, out-of-pocket health care spending was 2.5 times higher ($1,922), on average, for those with diabetes compared with those without diabetes ($738) (27) and expenditures for prescription drugs in patients with diabetes have increased (28). This may be due to the increased use of newer, more expensive diabetes drugs, such as dipeptidyl peptidase 4 inhibitors, with a concomitant decline in use of less expensive generic diabetes medications, such as sulfonylureas and metformin (29).

In the descriptive statistics, we also found that the Part D group reported greater problems or delays in accessing prescribed drugs compared with the non-Part D group. Part D may threaten patients’ access to prescription drug because of unintended consequences, such as delays in reimbursement for pharmacists. According to Goyal et al. (30), the delayed processing time for reimbursement is a major problem that is discouraging pharmacists. As such, there is a possibility of pharmacies declining Medicare Part D beneficiaries, which would eventually lead to reduced access to their prescription drugs.

Findings of this study must be interpreted within the context of the following limitations. First, this study used different cohorts each year for the 12-year period rather than follow one cohort over time; therefore, the outcomes each year are likely partly influenced by the different characteristics of each cohort. It remains unclear whether Medicare Part D is effective in lowering out-of-pocket spending or if patients with high out-of-pocket spending choose other prescription drug insurance options rather than Medicare Part D. Further, only one modeling approach, linear regression, was used and not verified using a logistic model or other statistical modeling options. An additional limitation of the analysis is that we removed those who received the Medicare benefit because of disability or end-stage renal disease from the 50- to 60-year-old comparison group, as they were likely also affected by Medicare Part D. This restriction may have biased the results because patients with these conditions, likely among those with the highest drug costs, were excluded from the comparison group but not from the Medicare group.

Despite these limitations, this study contributes significant new evidence to the field. Using interrupted time series analyses, this study assessed the trends in maturation of Medicare Part D over time compared with years before Part D began to avoid being misled by the results. Use of an interrupted time series analysis enabled statistical control of the trend prior to Part D implementation. In addition, the causal effects of Medicare Part D on the Medicare adults with diabetes aged ≥65 years were estimated using a non-Medicare group of adults with diabetes aged 50–60 years using a difference-in-difference analysis to eliminate the effect of potential confounding factors by comparing pre- and post-Part D between the Medicare and non-Medicare groups. Finally, the findings are generalizable to all noninstitutionalized U.S. older adults with diabetes because data analysis was performed using the MEPS national-level data and appropriately weighted based on complex stratified multistage sampling that enables to the generation of national estimates.

Although this study presents several meaningful findings, the evidence regarding whether the closure of Part D coverage gap is effective is preliminary as closure of the Part D coverage gap will not be fully implemented until 2020 (31). Limited evidence is available regarding the impact of recent efforts to close the Medicare Part D coverage gap on reducing financial burden and improving medication adherence in Part D beneficiaries. Future research is needed to evaluate the full impact of the Medicare Part D coverage gap closure on out-of-pocket spending through implementation of the Patient Protection and Affordable Care Act of 2010.

This study aimed to evaluate the effect of Medicare Part D in reducing the financial burden of prescription drugs in Medicare beneficiaries with diabetes. Findings demonstrated that implementation of the Medicare Part D policy was effective in lowering the proportion of out-of-pocket costs to total pharmacy expenses for Part D beneficiaries compared with nonbeneficiaries and provided access to prescription drugs, especially for medically underserved populations including women and racial/ethnic minorities. Part D enrollment rates gradually increased among Medicare beneficiaries with diabetes during 2006–2011. Further, Part D has impacted the prescription drug insurance market regardless of Part D enrollment status because both beneficiaries and nonbeneficiaries spent a lower proportion of out-of-pocket pharmacy costs after the implementation of Part D compared with before.

Current efforts through the Affordable Care Act to reduce the coverage gap in Part D must be continued to decrease the proportion of those affected by the coverage gap. To reduce the risks of cost-related medication nonadherence, health care providers should enhance their understanding of Medicare Part D design and its coverage gap to offer the most affordable prescription drug options for their patients.

Funding. This study was supported by the Columbia University School of Nursing Center for Health Policy and Sigma Theta Tau International Honor Society of Nursing Alpha Zeta Chapter.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.J.C. contributed to the study design, methods, and discussion; performed the statistical analyses; interpreted the data; and drafted and revised the manuscript. H.J. and T.G. contributed to methods, interpreted the data, and reviewed and edited the manuscript. K.W. and P.W.S. contributed to discussion, interpreted the data, and reviewed and edited the manuscript. A.M.S. contributed to the study design, methods, and discussion; interpreted the data; and reviewed and edited the manuscript. Y.J.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. The preliminary findings of this study were presented at the 75th Scientific Sessions of the American Diabetes Association, Boston, MA, 7 June 2015, and at the AcademyHealth Annual Research Meeting, Minneapolis, MN, 15 June 2015.