How the Medicare Part D Drug Benefit Changed the Distribution of Out-of-Pocket Pharmacy Spending Among Older Beneficiaries

Abstract

Objective.

To examine the impact of the Medicare drug benefit (Part D) on the distribution of out-of-pocket pharmacy spending among older adults.

Methods.

We used a pre-post-with-comparison-group design to compare four groups of beneficiaries continuously enrolled in a Medicare Advantage plan between 2004 and 2007: three intervention groups with no or limited (quarterly caps of $150 or $350) prior coverage that obtained Part D benefits in 2006 and a comparison group with stable drug coverage from 2004 to 2007.

Results.

The comparison group’s out-of-pocket drug spending was stable throughout 2004–2007, whereas Part D reduced out-of-pocket spending 13.4% among those without prior coverage (95% confidence interval [CI] −17.1% to −9.1%) and 15.9% among those with $150 quarterly caps (95% CI −19.1% to −12.8%) relative to the comparison group. Individuals in the top decile of drug spending paid a greater share of their costs out-of-pocket than others in the top 5 deciles.

Discussion.

Although Part D reduced out-of-pocket expenditures for older adults, those with the highest drug spending still pay a substantial share of their drug costs out-of-pocket. Thus, the Part D benefit does not achieve a primary purpose of insurance—to offer the greatest financial protection to those at the highest risk.

A primary goal of the Medicare drug benefit (Part D), implemented in 2006, was to protect older adults from catastrophic drug spending. The standard Part D benefit includes a deductible ($250 in 2006), an insured period where beneficiaries pay 25% coinsurance before their total drug spending reaches a threshold ($2,250 in 2006), and a coverage gap where beneficiaries pay everything out-of-pocket until their out-of-pocket spending reaches a catastrophic limit ($3,600 in 2006). Most Part D plans have benefit designs that differ from but are actuarially equivalent to or better than the standard benefit, although beneficiaries pay higher premiums for richer benefits. Nonetheless, in 2006, only 13% of stand-alone prescription drug plans (PDPs) and 23% Medicare Advantage prescription drug (MA-PD) plans offered coverage of some generic drugs in the coverage gap, whereas only 2% of PDPs and 5% of MA-PDs covered any brand name drugs in the gap (Medicare Payment Advisory Commission, 2006), thus potentially exposing beneficiaries with high drug expenses to large out-of-pocket costs.

Prior to Part D, 48% of beneficiaries had relatively generous drug coverage through their former employer or Medicaid. One third had more limited coverage through a privately purchased Medigap plan or a Medicare Advantage plan, and 18% had no coverage whatsoever (Kaiser Family Foundation, 2005). After Part D’s implementation, the proportion of Medicare beneficiaries without drug coverage fell to 9%, and most of those with limited prior drug coverage gained improved coverage. On average, Part D increased drug use 6%–16% and reduced out-of-pocket spending 13%–23% (Ketcham & Simon, 2008; Lichtenberg & Sun, 2007; Yin et al., 2008). However, Part D’s effect on out-of-pocket costs of individual beneficiaries will vary depending on their drug spending and the richness of their pre-Part D coverage.

In an earlier study, we found that increases in total pharmacy spending associated with Part D were related to the extent of the prior coverage; beneficiaries’ transitioning from no coverage to Part D experienced larger increases in drug expenditures than those moving from limited coverage ($150 or $350 quarterly caps; Zhang, Donohue, Lave, O’Donnell, & Newhouse, 2009). Here, we examine changes in out-of-pocket drug spending following Part D’s implementation by level of prior coverage. We also assess whether Part D resulted in a shift in the “risk” of out-of pocket pharmacy spending consistent with the goals of insurance by determining whether the proportion of payments made out-of-pocket decreased as the level of drug spending increased after Part D’s implementation (Arrow, 1963).

METHODS

Setting and Study Design

We obtained enrollment, benefits, and pharmacy claims information on a 40% random sample of 36,858 individuals continuously enrolled in Medicare Advantage plans offered by a Pennsylvania insurer between January 1, 2004, and December 31, 2007. We excluded 1,756 members younger than 65 years eligible for Medicare because of disability and 926 who had low-income subsidies to cover medications, leaving a study cohort of 34,176 members.

To control for underlying trends affecting out-of-pocket payment (e.g., entry of generic drugs), we used a pre-post-with-comparison-group design. The pre–post periods were 2 years pre- and post-Part D’s implementation in January 2006. The comparison group had relatively generous employer-subsidized drug coverage (retiree health insurance) that did not change with Part D. They faced copayments of $10 (generic) to $20 (brand) per monthly prescription and had no deductible, coverage gap, or catastrophic limits. We divided the remaining enrollees into three “intervention” groups based on their drug coverage prior to Part D: those with no coverage and two groups who had capped quarterly expenditures of $150 or $350 and tiered copayments ($8 generic/$20 brand). The intervention group to which a beneficiary belonged depended on the county of residence.

In January 2006, all members in the three intervention groups gained Part D coverage. Like those in most Part D plans, the beneficiaries we studied did not face a deductible. In the initial coverage period, they paid an $8/$20 (generic/brand name) copayment rather than 25% coinsurance in the standard benefit design. During the coverage gap, they had either no drug coverage or some coverage of generic drugs, depending on the plan they chose. Approximately 70% of our study population had some generic coverage, which was similar to MA-PD enrollees nationally (Centers for Medicare & Medicaid Services, 2008). The population we studied had the same catastrophic coverage limit as the standard benefit ($3,600 out-of-pocket).

Outcomes

We measured out-of-pocket pharmacy expenditures at the person-month and person-year levels and the proportion of total drug spending paid out-of-pocket at the person-year level. We adjusted expenditure data for inflation between 2004 and 2007 using the pharmaceutical component of the Producer Price Index. We also measured the absolute amount and the proportion of total drug spending that was paid out-of-pocket for each decile of total drug spending prior to Part D and the change in the distribution of out-of-pocket versus insurance payments from pre-Part D (2005) to post-Part D (2006).

Independent Variables

We created an indicator variable for each intervention group (no-coverage, $150-cap, and $350-cap). We also created a pre–post indicator and interacted this term with the intervention group variables to capture changes in outcomes before and after Part D in each of the intervention groups, relative to the changes in the comparison group.

Other covariates included age (65–74, 75–84, and >85 years), sex, and prospective risk scores. The risk score is a proprietary measure calculated by the insurance company to predict members’ future health care spending based on their prior year diagnosis and procedure codes reported on claims. The scores were calculated using Risk Grouper software from VeriskHealth (Waltham, MA) and are similar to the Centers for Medicare & Medicaid Services Hierarchical Condition Category weights used to adjust MA-PD payments. We used the risk score as a proxy for health status; a higher score indicates higher expected spending.

Statistical Analysis

We estimated regression models at the individual level, with the person-year as the unit of observation, and controlled for the covariates described earlier. To adjust for the correlation of repeated measures within individuals, we used a generalized estimating equation (GEE) model with a gamma distribution and a log-link function. We ran separate models comparing each intervention group with the comparison group.

To assess whether Part D shifted risk based on beneficiaries’ level of pharmacy spending, we examined changes in the proportion of total drug spending paid out-of-pocket for each decile of total pharmacy spending pre- and post-Part D. We used independent two-sample t tests to test for statistically significant differences in proportions across deciles.

RESULTS

Characteristics of Study Population

As is typical of those with retiree health insurance, members in the comparison group were younger and less likely to be female compared with those in the three intervention groups (p < .001). About half or more members in the three intervention groups and the comparison group had been diagnosed with hypertension and hyperlipidemia and 20% with diabetes (Table 1). Somewhat fewer members were diagnosed with the earlier three illnesses in the no-coverage group (p < .001). Importantly, the groups were reasonably comparable in their prospective risk scores.

Table 1.

Characteristics of Study Population

	Intervention groups
	No coverage	$150 Cap	$350 Cap	Comparison group
N (%)	3,703 (11)	2,536 (7)	18,528 (54)	9,409 (28)
Female, %	54.1	61.6	61.7	52.4*
Age (years), %
65–74	48.4	50.7	52.8	60.9*
75–84	44.0	40.0	39.4	33.9*
≥85	7.6	9.3	7.9	5.2*
Diagnosed chronic conditions, %
Hypertension	53.8*	62.7	62.5	61.2
Hyperlipidemia	48.4*	56.4	57.4	60.5
Diabetes	18.7*	22.2	22.1	23.2
Prospective risk scores, M (SE)
2004	0.80 (0.011)	0.83 (0.013)	0.85 (0.005)	0.84 (0.008)
2005	0.89 (0.012)	0.93 (0.016)	0.93 (0.006)	0.92 (0.009)

Notes: If asterisk is for the comparison group, it indicates statistically significant difference between each intervention group and the comparison group, otherwise only for that particular intervention group compared with the comparison group.

*p < .05.

Effects of Medicare Part D on Out-of-Pocket Spending

Figure 1 demonstrates the descriptive monthly trends of out-of-pocket pharmacy spending in each intervention group compared with the comparison group. Prior to Part D, monthly out-of-pocket spending in the $150-cap and the $350-cap groups spiked at the end of each quarter (p < .001), reflecting the quarterly caps. After Part D, the monthly out-of-pocket spending by the comparison group did not change, whereas that for the three groups whose coverage improved fell sharply immediately. Out-of-pocket spending then trended upward through the year as more beneficiaries entered the coverage gap; 25% of our sample reached the gap in 2006 (Zhang, Donohue, Newhouse, & Lave, 2009). Out-of-pocket spending peaked in October and decreased thereafter as some beneficiaries entered the catastrophic coverage region.

Figure 1.

Monthly out-of-pocket drug spending.

Notes. Monthly out-of-pocket spending prior to Part D in the $150-cap and $350-cap groups was cyclical, reflecting the quarterly caps prior to Part D. For example, in the $150-cap group, out-of-pocket costs on average in Month 3 of the quarter were $16 higher than costs in Months 1 and 2 of the quarter in the pre-period, at statistical significant level with p < .001 (t value = 11.48). In the $350-cap group, out-of-pocket costs in Month 3 of the quarter were $10 higher than costs in Months 1 and 2 of the quarter in the pre-period (t value = 22.64, p < .001).

Table 2 shows the numerical results from GEE regression models. Part D was associated with a reduction in out-of-pocket spending in two of the intervention groups. Relative to the comparison group, the absolute level of out-of-pocket spending declined 13.4% (95% confidence interval [CI] 9.7–17.1) in the group with no prior coverage and 15.9% (95% CI 12.8–19.1) among those in the $150 group. Absolute out-of-pocket spending did not change among those in the $350 group or the no-cap (comparison) group.

Table 2.

Out-of-Pocket Pharmacy Spending Changes Post-Part D

		Pre	Post	Percent changea	(95% CI)
A. Mean of annual out-of-pocket drug spending	Comparison	474	454	n/a
	No coverage	537	458	−13.4*	−17.1 to −9.7
	$150 Cap	682	558	−15.9*	−19.1 to −12.8
	$350 Cap	587	579	1	−0.6 to 2.5
B. Proportion of drug spending paid by out-of-pocket	Comparison	0.32	0.30	n/a
	No coverage	1.00	0.51	−44.9*	−45.9 to −43.9
	$150 Cap	0.57	0.40	−25.9*	−27.3 to −24.5
	$350 Cap	0.43	0.37	−9.5*	−10.6 to −8.5

Notes: ^aPre–post changes in an intervention group, adjusted for pre–post secular changes in the comparison group. CI = confidence interval.

*p < .05.

For individuals with drug spending below the median, there was regression to the mean as absolute out-of-pocket costs increased 67% in the no-coverage group, 9% in the $150-cap group, and decreased 1% in the $350-cap group (p < .01; raw data of out-of-pocket costs in Figure 2 and legend). Nonetheless, absolute out-of-pocket costs were low for these five deciles below the median. For individuals above the median, there was also regression to the mean; absolute out-of-pocket costs decreased by 14% in the no-coverage group, 30% in the $150-cap group, and 8% in the $350 cap group (p < .01).

Figure 2.

Out-of-pocket and insurance payment by deciles of total drug spending in 2005 and 2006.

Notes. Deciles are based on total drug spending. Percentages on top of the bars are ratios of out-of-pocket spending divided by total drug spending. Numbers in the bars are pharmacy spending paid out-of-pocket (top) and insurance plan payment (bottom). In the no-coverage group, numbers in the bars for first few three deciles are too small to be shown: in the first decile, out-of-pocket (OOP) costs were both zero; in the second decile, OOP costs increased from $7 in 2005 to $13 in 2006 (an 83% increase); in the third decile, OOP costs increased from $29 to $52, or an 82% increase; in the fourth decile, OOP costs increased from $68 to $109, or a 59% increase; and in the fifth decile, OOP costs increased from $128 to $181, or a 42% increase. On average, it was a 67% increase among those with total drug spending below the median.

Effects of Medicare Part D on Distribution of Plan Versus Out-of-Pocket Payments

The share of total drug spending paid out-of-pocket was affected even more than absolute out-of-pocket spending; it decreased 45% (95% CI 43.9–45.9) in the group previously lacking coverage, adjusting for trends in the comparison group (Table 2). The proportion paid out-of-pocket declined 25.9% (95% CI 24.5–27.3) in $150-cap group and by 9.5% (95% CI 8.5–10.6) in the $350 group.

In the intervention groups, for almost all deciles of baseline drug spending, the share paid out-of-pocket was lower in 2006 than in 2005 (Figure 2). The effect of Part D on the distribution of out-of-pocket versus plan payments, however, varied by prior drug coverage and level of drug spending. For those with drug spending below the median, the proportion paid out-of-pocket went from 100% to 66% in the no-coverage group, 56% to 46% in the $150-cap group, and 47% to 41% in the $350-cap group. Beneficiaries with drug costs above the median experienced an even greater percentage reduction in the proportion of drug costs paid out-of-pocket, from 100% to 42% for the no-coverage group, 60% to 33% for the $150-cap group, and 41% to 32% for the $350-cap group, on average, in the top five deciles.

Individuals in the top decile of total drug spending, the threshold for which differs across the intervention groups, however, had a markedly different pattern. They paid a greater share of costs out-of-pocket than others in the top five deciles. Such individuals in the no-coverage group paid 51% of drug expenditures out-of-pocket post-Part D, statistically significantly higher compared with the other four deciles above the median: 38% (t value = 14.57, p < .001) in the ninth decile, 36% (t value = 17.64, p < .001) in the eighth decile, 40% (t value = 12.10, p < .001) in the seventh decile, and 45% (t value = 6.65, p < .001) in the sixth decile (Figure 2). The analogous percentages in the $150 group were 43% in the top decile compared with 34%, 27%, 28%, and 31% from ninth to sixth deciles; again, all statistically significant (p < .001). In the $350 group, the proportion of total drug spending paid out-of-pocket was 45% in the top decile versus 29% on average among other four deciles above the median (p < .001). All beneficiaries in the top decile spent some portion of the year in the coverage cap.

DISCUSSION

Medicare Part D was associated with an increase in overall drug use. Beneficiaries who had no prior drug coverage, or who had their pharmacy benefits capped at $150 or $350 per quarter, increased their total drug spending 74%, 27%, and 11%, respectively (Zhang, Donohue, Lave, et al., 2009). Because of the 74% increase in total drug spending among those who previously lacked drug coverage, absolute out-of-pocket spending level decreased relatively little (13%) in this group. Out-of-pocket spending declined 16% in the $150-cap group and remained the same in the $350-cap group. Although the decline in out-of-pocket spending among those who previously lacked coverage may appear modest, Part D financed half the drug spending for that group. Our estimates of the reduction in out-of-pocket costs were smaller than those estimated before Part D went into effect because those simulation analyses assumed no change in drug utilization from the expanded coverage (Gellad, Huskamp, Phillips, & Haas, 2006). However, our estimates of the change in out-of-pocket costs were similar to others based on data from a single large pharmacy chain (Lichtenberg & Sun, 2007; Yin et al., 2008).

Effects at the group level, however, mask differences among individuals with varying expenditure levels. Good insurance should offer more protection to those with the highest expenditures. On this criterion, Medicare Part D does not fare well. For groups with total drug spending above the median, the proportion of drug spending Part D covers is lowest among the highest spending decile. This is due to the unusual benefit design of Part D, most notably its coverage gap. Like our sample, recent national studies estimate that approximately one quarter of Medicare beneficiaries spent some time in the coverage gap in 2007, half of whom spent at least 4 months without coverage (Kaiser Family Foundation, 2008). Furthermore, beneficiaries exposed to the coverage gap reduce their prescriptions filled by 14%, whereas those enrolled in plans with coverage of generic drugs reduce medication use by only 3% (Zhang, Donohue, Newhouse, et al., 2009, Health Affairs). Requiring plans to cover generic drugs in the coverage gap even if they must impose moderately higher cost-sharing up front to finance it would result in better financial protection for those with the highest drug costs. Only 4% of beneficiaries overall (15% of those who reached the coverage gap) had out-of-pocket drug spending high enough ($3,850 in 2007) to put them in the catastrophic coverage region where they paid only 5% of drug costs (Kaiser Family Foundation, 2008).

Our study is subject to some limitations. Our data were from a single Medicare Advantage plan. Thus, our findings may not generalize to other managed care plans or to stand-alone PDPs. We would overestimate the impact of Part D on out-of-pocket costs if some beneficiaries who lacked or had limited drug coverage filled prescriptions in pharmacies outside the plan’s network prior to Part D. We believe this bias is small because beneficiaries faced a strong financial incentive to fill prescriptions in network (they received a 15% discount off-prices charged cash payers on average) and because the plan’s pharmacy network was extensive (58,000 nationwide). Nevertheless, some beneficiaries may have filled prescriptions for which claims were not submitted to the plan. Finally, there may be some measurement error in the proportion of total drug costs paid out-of-pocket versus the plan for members who have supplemental coverage through the Veteran’s Administration or some other source.

In sum, our findings show that Part D reduced beneficiaries’ financial burdens associated with prescription drugs, in many cases substantially. But Part D has not yet accomplished a principal goal of insurance—covering a higher proportion of costs for higher risk beneficiaries.

FUNDING

The National Center for Research Resources, a component of the National Institutes of Health (NIH), NIH Roadmap for Medical Research (KL2-RR024154-01 to J.M.D.); the University of Pittsburgh's Graduate School of Public Health Computational and Systems Models in Public Health Pilot Program (Y.Z.); and the Alfred P. Sloan Foundation (J.P.N.).