Five‐year impact of Medicare Part D coverage gap reform on drug expenditures and utilization

Abstract

Objective

To estimate the impact of the Medicare Part D coverage gap reform under the Affordable Care Act (ACA) on the utilization of and expenditures for prescription drugs within the first five years of the policy's implementation.

Data Sources

2008‐2015 Medicare Current Beneficiary Survey (MCBS).

Study Design

We used a difference‐in‐differences approach to estimate the year‐by‐year changes in prescription drug use and expenditures before (2006‐2010) and after (2011‐2015) the ACA’s Part D coverage gap reform between Part D beneficiaries not receiving the Low‐Income Subsidy (LIS) and those receiving the LIS.

Data Collection

The study sample included Part D beneficiaries (a) aged 65 years or older; (b) not disabled or having end‐stage renal disease; (c) continuously enrolled in a Part D plan (d) having at least one prescription fill in a given year. Survey‐reported and administrative Part D events data in the MCBS were used for the analyses.

Principal Findings

After the ACA reform, annual out‐of‐pocket drug spending significantly decreased by $88 (P < .01) among non‐LIS beneficiaries compared to LIS beneficiaries, with growing decreases over time (average decreases of $41 in 2011, $49 in 2012, $105 in 2013, and $135 in 2015, P < .01 or <.05). Changes in out‐of‐pocket costs were largely driven by significant decreases among brand‐name drugs (overall decrease of $106, P < .01). Despite significantly reduced out‐of‐pocket spending, there were no significant changes in the overall number of 30‐day drug fills and total drug spending; however, changes in the use of brand‐name and generic drugs were seen after the ACA (increase of 1.9 fills for brand‐name drugs and decrease of 2.3 fills for generic drug in 2015, P < .05).

Conclusions

The ACA coverage gap reform has helped to reduce the out‐of‐pocket drug cost burden for beneficiaries, although it had no noticeable impact on drug use or total drug spending.

What is already known on this topic

• Few studies have evaluated the effects of the ACA’s Part D coverage gap reform for Part D beneficiaries.

• Limited previous research found decreased out‐of‐pocket (OOP) drug costs and mixed effects on drug use after the ACA, but focused on specific subpopulations or used data with limited information on beneficiaries’ eligibility and drug claims.

What this study adds

• This study is the first to examine the year‐by‐year changes in drug expenditures and utilization among a nationally representative sample of the general Part D population through the first five years of the ACA reform.

• Significant and continued reductions in OOP costs were found over time, with particularly noticeable decreases among brand‐name drugs.

• No significant effects were found on total drug spending and drug use, but led to switching between brand‐name and generic drugs.

1. INTRODUCTION

Medicare beneficiaries have had access to prescription drug coverage through Medicare Part D since 2006, which brought several positive impacts for beneficiaries, such as decreased out‐of‐pocket cost burden, increased drug utilization, and improved health status. ¹ , ² , ³ , ⁴ Despite these benefits, the Part D benefit structure contains a coverage gap (or “doughnut hole”), during which beneficiaries are required to pay 100% of their drug costs. This gap has been criticized as a financial barrier to prescription drug access for beneficiaries. ² , ³

Furthermore, substantial evidence indicates that the coverage gap negatively impacts out‐of‐pocket costs, drug utilization, and medication adherence for beneficiaries. ² , ³ , ⁵ , ⁶ The entry of Part D beneficiaries into the coverage gap has been associated with increased out‐of‐pocket drug costs by as much as 89%, decreases in drug use, and cost‐related medication nonadherence such as medication cessation, skipping doses, or delaying or foregoing prescriptions. ² , ³ , ⁷ This medication nonadherence can negatively affect beneficiaries’ health outcomes and result in increased health care expenditures, which in turn increases costs to other parts of the Medicare program. ⁶ , ⁸ , ⁹

In response to these concerns, the Patient Protection and Affordable Care Act (ACA) of 2010 implemented provisions that initiated a ten‐year process to close the Part D coverage gap, by gradually phasing down the coinsurance rate in the gap from 100% in 2011 to 25% by 2020. ¹⁰ Recent changes made by the Bipartisan Budget Act of 2018 closed the coverage cap early for brand‐name drugs in 2019, with generic drugs on schedule for 2020. ¹¹ , ¹²

Although there are previous studies on the effects of the Part D coverage gap on drug utilization and expenditures, few studies have evaluated the effects of the ACA’s Part D coverage gap reform. Although these studies have indicated that out‐of‐pocket drug costs decreased after the reform, they have focused on limited populations, such as those with uncommon cancers ¹³ or diabetes, ¹⁴ or used survey data with limited drug claims and information on beneficiaries’ eligibility, which may have several limitations in selecting study samples. ¹⁵ Using nationally representative data from 2008 to 2015, this study examines the annual changes in drug utilization and expenditures after the ACA Part D coverage gap reform following its implementation in 2011. We hypothesized that the ACA reform would be associated with increased prescription drug use and decreased out‐of‐pocket costs. To our knowledge, this study is the first to examine the year‐by‐year changes in total and out‐of‐pocket drug spending and drug utilization among the general Part D population after the ACA.

2. METHODS

2.1. Study design and population

We adopted a difference‐in‐differences approach to estimate the effects of the ACA coverage gap reform on drug use and expenditures. The impacted group was Part D beneficiaries not receiving the Low‐Income Subsidy (LIS) (ie, non‐LIS beneficiaries) who may have been exposed to the coverage gap. Part D beneficiaries were selected as the comparison group if they received the LIS for at least one month in a given year (ie, LIS beneficiaries). ¹³ , ¹⁶ LIS beneficiaries were not affected by the ACA reform because they already had little or no cost‐sharing in the coverage gap before the ACA. ¹⁷ The following inclusion criteria were used to identify the study sample: (a) age 65 years or older; (b) not disabled or having end‐stage renal disease; (c) continuous enrollment in a Part D plan for a given year; and (d) having at least one prescription fill in a given year.

The study period was from 2008 to 2015, where the pre‐ACA period was defined as 2008 to 2010, and the post‐ACA period was defined as 2011 to 2015. Since the coverage gap has been phasing out gradually since 2011, we identified the policy effects separately for each year of the post‐ACA period compared to the pooled years of the pre‐ACA period, 2008‐2010 (ie, year‐by‐year changes difference‐in‐differences model). This approach has been used to trace out differential changes over time after the ACA policy changes. ¹⁸

2.2. Data source

We used data from the Medicare Current Beneficiary Survey (MCBS) for the years of 2008 to 2013 and 2015. MCBS data for 2014 were not released by the Centers for Medicare & Medicaid Services (CMS). ¹⁹ , ²⁰ The MCBS is a continuous, in‐person, longitudinal survey of a nationally representative sample of the Medicare population that is linked to administrative claims data, and provides comprehensive information on the health care utilization and expenditures of beneficiaries. ¹⁹ We used survey‐reported data to define the study cohorts and obtain sociodemographic data, and administrative Part D events data in the Prescribed Medicine Events (PME) file to estimate drug use and expenditures.

2.3. Outcome variables

We constructed two outcome measures: expenditures for and utilization of prescription drugs. Prescription drug expenditures were measured at two levels: mean annual total drug spending per person paid by all payment sources and mean annual out‐of‐pocket spending per person. Prescription drug use was measured as the mean annual number of 30‐day drug fills per person. Each record in the PME file is an individual outpatient prescribed medicine event, which is a single fill of a single drug in a single container. In order to account for the variability in the number of days supplied across fills, each drug fill was normalized to 30‐day fills (eg, 90‐day supply = 3 fills). ²¹ Additionally, all the outcomes were assessed by drug type for brand‐name and generic drugs. Due to limited information about drug type for each fill in the MCBS data, we utilized the First Databank drug names for each fill as a proxy. ²⁰

2.4. Statistical analysis

We used chi‐square tests to compare beneficiaries’ characteristics by year within each group. We then performed difference‐in‐differences regressions to compare the changes in drug expenditures and utilization among non‐LIS beneficiaries and LIS beneficiaries over the pre‐ACA (2008‐2010) and post‐ACA periods (2011‐2015). The parallel trends assumption was tested through visual analysis (Figure 1) and regression models (Appendix S4), ²² where the assumption was deemed valid.

FIGURE 1.

Trends of drug expenditures and utilization among non‐LIS and LIS Part D beneficiaries, 2008‐2015. Trends were analyzed using the unadjusted estimates of out‐of‐pocket drug spending, total drug spending, and mean annual 30‐day drug fills. Dollar amounts were converted to inflation‐adjusted 2015 dollars. The time trends for out‐of‐pocket and total drug spending and the 30‐day drug fills did not differ significantly for the non‐LIS beneficiaries versus the LIS beneficiaries in the fully adjusted models (P = .701, .116, and .875, respectively, data are not shown). Vertical line indicates the implementation of the ACA reform. ACA, Affordable Care Act; LIS, Low‐Income Subsidy

First, year‐by‐year changes difference‐in‐differences models were estimated for each outcome using linear regression models with interaction terms between impacted group and year, which captured the changes attributable to the ACA reform in each year of the post‐ACA period, compared to the pooled pre‐ACA period. ¹⁸ Second, we estimated difference‐in‐differences models that pooled the pre‐ACA and post‐ACA periods to see the overall effects of the policy changes. Both regression equations are described in Appendix S2.

All regression models were adjusted for beneficiary demographics (age, sex, race/ethnicity), socioeconomic characteristics (attained education level, family income as percentage of poverty level, urban versus rural residence), health status measured by the number of chronic conditions, and enrollment in a Medicare Advantage Prescription Drug plan (MA‐PD). In order to obtain nationally representative estimates for the non‐LIS Medicare Part D population and to account for the complex sampling design of the MCBS, the Balanced Repeated Replication (ie, Fay's method) of variance estimation was used to adjust both serial and intra‐cluster correlation in the data, using replicate cross‐sectional weights for each year. ²³ All statistical analyses were conducted using Stata 15.1 (StataCorp). All estimates of drug spending were converted to inflation‐adjusted 2015 dollars using the all‐items Consumer Price Index. ²⁴

Sensitivity analyses were conducted using each year of the pre‐ACA period in place of the pooled pre‐ACA years to check if the policy effect would differ by the selection of baseline year in the pre‐ACA period. Our findings were robust to the selection of the year in the pre‐ACA period (Appendix S5). Additionally, we repeated our difference‐in‐differences analyses by rounding up to 1 for the fills <30‐day supply. Results were consistent with our main findings (Appendix S6). Lastly, we estimated difference‐in‐differences estimates excluding those enrolled in an MA‐PD. Our results were robust to this change (Appendix S7).

3. STUDY RESULTS

3.1. Study population

The study population consisted of 24 919 non‐LIS Part D beneficiaries and 9835 LIS beneficiaries. Table 1 presents the weighted characteristics of the two groups, which were systemically different from one another. Compared to the LIS beneficiaries, non‐LIS beneficiaries were more likely to be male, younger, non‐Hispanic white, more educated, higher income, living in an urban area, have fewer chronic conditions, and enrolled in an MA‐PD.

TABLE 1.

Characteristics of non‐LIS and LIS Part D beneficiaries by year, 2008‐2015

	Non‐LIS Part D beneficiaries					LIS Part D beneficiaries
	Pre‐ACA a	2011	2012	2013	2015	Pre‐ACA a	2011	2012	2013	2015
Sample (n)	9420	3317	3736	4126	4320	4150	1339	1393	1416	1537
Population (N)	39 124 342	14 563 427	16 304 256	19 376 540	24 751 318	14 844 481	5 283 088	5 504 294	5 761 143	7 453 439
Female (%)	60	60	58	58	58	72	71	69	70	68
Age (%)
65‐74	47	49	49	50	54	41	44	44	44	48
75‐84	37	36	36	31	35	39	36	35	34	33
85+	15	15	15	14	15	20	20	21	22	19
Race/Ethnicity (%)
White, non‐His	92	91	92	90	89	66	65	64	64	63
Black, non‐His	5	5	5	6	7	19	18	20	19	18
Hispanic	1	1	1	1	1	6	8	8	8	9
Other, non‐His	2	2	3	3	4	8	9	9	9	10
Education (%)
Less than high school	20	18	17	16	13	57	56	55	53	50
High school graduate	33	31	28	28	26	24	23	22	23	27
Some college or more	48	51	55	56	61	19	21	23	23	22
Family income (percent of poverty, %)
<125%	17	18	17	17	16	74	79	81	83	79
125%‐200%	25	25	25	24	22	18	17	15	14	16
200%‐400%	41	38	37	38	33	7	4	3	3	4
>400%	17	19	21	21	30	1	1	0	0	1
Residence in rural area (%)	14	14	15	14	13	16	16	17	16	15
No. of chronic conditions (%)
0‐2	40	38	37	37	28	27	25	24	22	17
3‐4	42	44	45	44	49	43	44	44	41	41
>5	18	18	18	19	23	30	32	33	37	41
MA‐PDs (%)	42	46	46	43	41	21	25	28	30	31

No Medicare Current Beneficiary Survey data released in 2014. All estimates are rounded off to the nearest whole number. Chi‐square tests were used for equality in frequencies across categories within each study group by year. All comparisons were statistically significant (P < .05), except for rural residence for the non‐LIS beneficiaries and sex, race/ethnicity, and rural residence for the LIS beneficiaries.

Abbreviations: ACA, Affordable Care Act; LIS, Low‐Income Subsidy; MA‐PDs, Medicare Advantage Prescription Drug plans.

^a The numbers for the pre‐ACA reflect pooled estimates for the years 2008‐2010.

3.2. Effects on out‐of‐pocket drug spending

After the ACA, out‐of‐pocket spending significantly decreased among non‐LIS beneficiaries relative to LIS beneficiaries. Unadjusted mean out‐of‐pocket spending decreased after the ACA in both groups, from $768 to $664 in the non‐LIS group and $135 to $108 in the LIS group, respectively (Table 2). The fully adjusted pooled difference‐in‐differences analyses showed that the out‐of‐pocket costs significantly decreased by $88 in the non‐LIS group relative to the LIS group (Table 3). The average marginal effect estimated from the year‐by‐year changes difference‐in‐differences model were significant in all years (P < .05), with decreases of $41 in 2011, $49 in 2012, $105 in 2013, and $135 in 2015. This corresponds to decreases of 5%, 6%, 14%, and 18% from the baseline of $768 in the pooled pre‐ACA period.

TABLE 2.

Unadjusted estimates of drug expenditures and utilization before and after the ACA Part D coverage gap reform

	Pre‐ACA				Post‐ACA
	Pooled (2008‐2010)	2008	2009	2010	Pooled (2011‐2015)	2011	2012	2013	2015
Non‐LIS beneficiaries
OOP costs ($)
Overall	768	789	777	742	664	723	693	647	625
Brand‐name	603	631	613	568	476	547	500	458	431
Generic	165	158	164	174	189	176	193	189	194
Total costs ($)
Overall	2257	2226	2296	2248	2370	2210	2275	2371	2524
Brand‐name	1773	1746	1809	1764	1828	1728	1739	1831	1942
Generic	484	480	487	484	542	482	535	541	582
30‐d drug fills
Overall	44.8	43.7	44.9	45.7	47.8	46.0	47.9	49.1	47.8
Brand‐name	16.9	17.8	16.9	16.1	13.4	15.0	14.0	13.5	12.0
Generic	27.9	26.0	28.0	29.6	34.4	31.0	33.9	35.6	35.8
LIS beneficiaries
OOP costs ($)
Overall	135	146	126	132	108	131	97	110	98
Brand‐name	93	107	88	85	62	82	56	64	51
Generic	41	39	38	48	46	49	41	46	47
Total costs ($)
Overall	4045	4035	3989	4110	4480	4144	4000	4310	5204
Brand‐name	3290	3268	3248	3354	3545	3342	3093	3412	4126
Generic	755	768	741	756	935	802	907	898	1078
30‐d drug fills
Overall	61.8	61.1	61.1	63.3	66.6	64.4	66.3	68.6	66.9
Brand‐name	25.3	26.4	25.4	24.2	20.8	23.6	21.6	21.0	18.2
Generic	36.5	34.7	35.7	39.1	45.8	40.8	44.7	47.6	48.7

The sample size and characteristics are described in Table 1. Dollar amounts were converted to inflation‐adjusted 2015 dollars. All estimates indicate the mean annual estimates per beneficiary.

Abbreviations: ACA, Affordable Care Act; LIS, Low‐Income Subsidy; OOP, out‐of‐pocket.

TABLE 3.

Difference‐in‐differences estimates of the effects of the ACA Part D coverage gap reform on drug expenditures and utilization for non‐LIS beneficiaries

	Pre vs Post a	Year‐by‐year changes b
	Post‐ACA	2011	2012	2013	2015
OOP spending ($)
Overall	−88.1**	−41.4*	−48.7*	−104.8**	−135.2**
Brand‐name	−106.0**	−44.2*	−75.4**	−124.0**	−155.9**
Generic	18.0*	2.8	26.8**	19.1**	20.7**
Total spending ($)
Overall	−333.5	−113.9	54.7	−117.6	−943.8**
Brand‐name	−210.3	−67.8	157.5	−40.6	−714.8*
Generic	−123.2**	−46.1	−102.8**	−77.1*	−228.2**
30‐d drug fills
Overall	−1.3	−0.9	−1.1	−1.1	−1.8
Brand‐name	1.0	−0.02	0.8	1.2	1.9*
Generic	−2.3**	−0.9	−1.9	−2.3*	−3.7**

The sample size and characteristics are described in Table 1. Results show difference‐in‐differences estimates for the non‐LIS beneficiaries versus the LIS beneficiaries, by year and by subgroup.

Abbreviations: ACA, Affordable Care Act; OOP, out‐of‐pocket; LIS, Low‐Income Subsidy.

^a Pooled difference‐in‐differences estimates for the years 2011‐2015 (post‐ACA period), compared with the pooled estimates for the years 2008‐2010 (pre‐ACA).

^b Year‐by‐year changes difference‐in‐differences estimates, compared with the pooled estimates for the years 2008‐2010 (pre‐ACA period). All analyses were adjusted for sex, age, race/ethnicity, education, family income as a percentage of poverty, urban versus rural residence, number of chronic conditions, and Medicare Advantage Prescription Drug plan status.

^* P < .05.

^** P < .01.

Our analyses by drug type shows that the significant decreases in overall out‐of‐pocket spending were largely driven by significant decreases in out‐of‐pocket costs for brand‐name drugs of $106 (Table 3). In contrast, out‐of‐pocket spending on generic drugs significantly increased by $18 in the non‐LIS group relative to the LIS group. Similar growing trends over time were seen in the year‐by‐year changes estimates by drug type.

3.3. Effects on total drug spending

Overall, there was no significant effect of the ACA coverage gap reform on total drug spending. Unadjusted mean annual total drug spending increased after the ACA, with larger increases seen among LIS beneficiaries (Table 2). The fully adjusted difference‐in‐differences analyses show a decrease of $334 among non‐LIS beneficiaries relative to LIS beneficiaries, but it was not statistically significant (Table 3). However, the year‐by‐year changes difference‐in‐differences analysis showed a significant decrease of $944 for non‐LIS beneficiaries in 2015 relative to LIS beneficiaries, which seems to be mainly due to the considerable increases in total spending for LIS beneficiaries in 2015. When explored by drug type, total drug spending increased for both brand‐name and generic drugs in both the non‐LIS and LIS groups, with much larger increases in the LIS group, especially for brand‐name drugs (Table 2 and Appendix S3).

3.4. Effects on drug utilization

Overall, there were no significant changes in drug utilization for non‐LIS beneficiaries after the ACA. Although the unadjusted number of 30‐day drug fills slightly increased in both groups in the post‐ACA period (Table 2), the fully adjusted difference‐in‐differences estimates from the both models (pooled and year‐by‐year changes) showed non‐significant decreases in the use of prescription drugs among non‐LIS beneficiaries relative to LIS beneficiaries after the ACA (Table 3). In our analyses by drug type, brand drug utilization decreased in both groups, with a statistically significant larger decrease seen in 2015 among LIS beneficiaries. On the other hand, the use of generic drug increased in both groups over time, with a significantly larger increases were seen among LIS beneficiaries, resulting in negative effects on generic drug utilization for LIS beneficiaries.

4. DISCUSSION

This study estimated the effects of the ACA Part D coverage gap reform on drug expenditures and utilization among non‐LIS beneficiaries within the first five‐year of implementation. We found that the ACA reform was significantly associated with decreased out‐of‐pocket spending for prescription drugs among non‐LIS beneficiaries, with no major effects on the use of prescription drugs and total drug spending without statistical significance. The reductions in out‐of‐pocket spending grew over time after implementation of the reform. The significant decreases in out‐of‐pocket spending indicates that the ACA reform may have helped to reduce the financial burden of prescription drugs for beneficiaries, supporting the intent of the policy. This finding is consistent with the limited previous literature that has focused on specific subpopulations. ¹³ , ¹⁴ , ¹⁵

Gradual and statistically significant decreases in out‐of‐pocket spending occurred after the ACA reform, with the largest reduction seen in 2015. This may be due in part to the unique characteristics of the gradual phase‐in schedule of the reform. ²⁵ The decreases in out‐of‐pocket spending were mainly driven by the decreases in spending on brand‐name drugs, which may be because of the uneven nature of how the coverage gap was closed, including brand‐name drug manufacturer discounts and more rapid decreases in coinsurance rates for brand‐name than generic drugs within the first five‐year of the ACA. ²⁵ Our findings show that although no significant changes in the use of brand‐name drugs were seen among non‐LIS beneficiaries after the ACA, out‐of‐pocket spending on brand‐name drug decreased more among non‐LIS beneficiaries relative to LIS beneficiaries, which may be due to the mandated 50% manufacturer discount on the price of brand‐name drugs. ²⁵ Although the 50% discount for brand‐name drugs was large enough to see an immediate impact of the policy, it might have affected a relatively small proportion of beneficiaries as the majority of drugs used by Part D beneficiaries are generic drugs, with an average generic dispensing rate of 87% in 2015. ²¹ The significant reductions in out‐of‐pocket spending in the later years of the post‐ACA period may reflect the delayed effects of the policy. ²⁵

In addition, several top‐selling brand‐name drugs patent expirations during this period could be another possible contributing factor to the decreases in out‐of‐pocket spending observed in this study. ²⁶ For example, Lipitor^®, Caduet^®, Combivir^®, and Solodyn^®, which accounted for more than $7 billion in sales, lost patent protection in 2011, and Cymbalta^® went generic in 2013. ²⁶ These patent expirations in widely used branded drugs leads to lower drug prices due to generic substitutions and competition. The price of atorvastatin (Lipitor^®) has fallen by more than 95% since 2011, ²⁶ and Cymbalta^® was listed among the top 10 Medicare traditional drugs in 2013, but it was substituted by duloxetine in 2014. ²⁷ , ²⁸ In this regard, our findings of decreased out‐of‐pocket spending for brand‐name drugs but increases for generic drugs could be partially explained by the entry of generic versions of popular brand‐name drugs during this study period.

Although the ACA’s Part D coverage gap reform helped beneficiaries reduce their out‐of‐pocket spending for prescription drugs, it did not lead to a significant increase in the overall use of prescription drugs; this finding is consistent with previous studies that found no significant changes after the ACA. ¹³ , ¹⁵ This seems to be largely driven by the opposite changes in the use of brand‐name and generic drugs after the ACA. After the ACA, non‐LIS beneficiaries have significantly decreased generic drug use but increased brand‐name drug use relative to LIS beneficiaries. This may be associated with larger and more rapid decreases in beneficiaries’ cost‐sharing for brand‐name drugs than generics under the ACA reform. Another possible factor would be Part D plans’ utilization management. Part D plans applied tightening formulary management responding to the high and rising costs of prescription drugs, as well as increased availability of generic alternatives. ²⁹ , ³⁰ This may lead to increased use of generic drugs for both non‐LIS and LIS beneficiaries; although their cost‐sharing is relatively very low, LIS beneficiaries still have a larger copayment or coinsurance rate for brand‐name drugs than generics. ³¹ Although both groups increased generic drug use after the ACA, non‐LIS beneficiaries increased generic drug use to a lesser extent than LIS beneficiaries, which might be due to the lower costs of brand‐name drugs after the ACA.

Our findings in drug utilization by drug type contrasts with a previous study using MEPS data. ¹⁵ This study found that, although the trends in unadjusted utilization of brand‐name and generic drugs in the non‐LIS group were similar, the non‐LIS group decreased brand‐name drug use but increased generic drug use relative to the LIS group after the ACA. This discrepancy might be due to differences in the measures of drug utilization, characteristics of the LIS group, and use of a different data source. Since our normalized drug utilization measure (30‐day fills) accounted for the variability in number of days dispensed across fills, we believe it captured more detailed differences in drug use across different therapeutic classes. Additionally, as used in previous studies on the Medicare Part D coverage gap, ¹³ , ³² LIS beneficiaries are more similar to non‐LIS beneficiaries than those aged 55 years or older with private health insurance that were used in the previous study. ¹⁵

No significant changes in total drug spending were seen after the ACA. Although mean annual total drug spending remained relatively unchanged after the ACA among non‐LIS beneficiaries, it increased considerably among LIS beneficiaries, resulting in a significant negative effect of the ACA on total drug spending for non‐LIS beneficiaries in 2015 (a decrease of $944). The increase in total spending among LIS beneficiaries may be due in part to the higher proportion of people reaching the catastrophic coverage threshold relative to non‐LIS beneficiaries, which reflects the fact that the majority of high‐cost Part D enrollees are likely to receive the LIS. ²¹ Additionally, as LIS beneficiaries are six times more likely to have hepatitis C than non‐LIS beneficiaries, new hepatitis C medicines could be another potential driver of increases in total drug spending. ²¹ New oral therapies for hepatitis C such as Sovaldi, Olysio, or Harvoni began to come out since 2013 at very high costs (eg, $84 000 per treatment regime or $1000 per pill for Sovaldi), which resulted in significant increases in Part D spending in 2014. The growth in Part D spending, mainly due to the increased spending for high‐cost beneficiaries, has been a growing concern for the Medicare program. ²¹ , ³³ , ³⁴ Policymakers should consider changes to the Part D program to promote more cost‐effective and high quality medication use by Part D beneficiaries, especially by high‐cost beneficiaries.

We note key contributions to the limited existing literature on the effects of the ACA Medicare Part D coverage gap reform. We provide the first estimates of the year‐by‐year changes of the policy change on drug expenditures and use through the policy's first five years of implementation. As the reform had a phased‐in schedule by year, our estimates provide information on how the policy's impact has changed as the coverage gap has gradually closed. In addition, using a leading source of information on the Medicare population based on linked survey and claims data, we document the effects of the ACA coverage gap reform in the general Medicare Part D population, by providing a nationally representative estimates for Part D beneficiaries not receiving the LIS.

This study has several limitations to note. First, although LIS beneficiaries have been used as comparison group in previous studies on Part D coverage gap reform, ¹³ , ³⁵ , ³⁶ , ³⁷ , ³⁸ we acknowledge that LIS beneficiaries differ from non‐LIS beneficiaries in terms of socioeconomic status or having more generous drug coverage, which could change over time. These time‐varying unobserved differences between the two groups could also affect the outcomes of this study. However, since we found parallel trends in both groups, we believe our difference‐in‐differences estimates are unbiased. ²² Second, we used a proxy to identify brand‐name and generic drugs for each fill due to limited information in the MCBS data. Therefore, our estimates by drug type may not be as accurate as estimates using actual data on the brand/generic status of each fill. Third, given the phase‐in schedule until 2020, the policy effect is expected to increase over time. Since additional reductions in cost‐sharing were scheduled following our study period, the later years of the policy may better reflect the full impact of the policy, particularly on generic drugs. Lastly, our study sample included beneficiaries enrolled in MA‐PDs, which could have had a differential impact on beneficiaries’ drug use and expenditures, as they are likely to face lower out‐of‐pocket costs than those with a stand‐alone prescription drug plan. ¹⁵ A sensitivity analysis excluding these individuals found results similar to the main analysis but with larger effects (Appendix S7).

5. CONCLUSION

Over the first five years after implementation of the ACA reforms to close the Part D coverage gap for non‐LIS beneficiaries, significant reductions were seen in out‐of‐pocket spending for prescription drugs that continued to decrease over time. However, despite seemingly large reductions in cost‐sharing to Part D beneficiaries while in the coverage gap, the ACA had no significant effect on the use of prescription drugs or on total drug spending. The findings from this study suggest that the ACA reform has helped to reduce out‐of‐pocket drug cost burden for Part D beneficiaries, which is expected to increase in the later years of the policy as more generous coinsurance rates are phased in. However, since such reductions in out‐of‐pocket costs did not lead to an increase in overall prescription drug use but instead led to switching between brand‐name and generic drugs, policymakers should consider additional initiatives to ensure beneficiaries’ access to needed prescription drugs.

CONFLICT OF INTEREST

The authors have no conflicts of interest.

Supporting information

Author matrix

Appendix S1‐S9

Park J, Look KA. Five‐year impact of Medicare Part D coverage gap reform on drug expenditures and utilization. Health Serv Res. 2021;57:56–65. 10.1111/1475-6773.13660