Medicare Part D Roulette: Potential Implications of Random Assignment and Plan Restrictions

Abstract

Background

Dual-eligible (Medicare/Medicaid) beneficiaries are randomly assigned to a benchmark plan, which provides prescription drug coverage under the Part D benefit without consideration of their prescription drug profile. To date, the potential for beneficiary assignment to a plan with poor formulary coverage has been minimally studied and the resultant financial impact to beneficiaries unknown.

Objective

We sought to determine cost variability and drug use restrictions under each available 2010 California benchmark plan.

Methods

Dual-eligible beneficiaries were provided Part D plan assistance during the 2010 annual election period. The Medicare Web site was used to determine benchmark plan costs and prescription utilization restrictions for each of the six California benchmark plans available for random assignment in 2010. A standardized survey was used to record all de-identified beneficiary demographic and plan specific data. For each low-income subsidy-recipient (n = 113), cost, rank, number of non-formulary medications, and prescription utilization restrictions were recorded for each available 2010 California benchmark plan. Formulary matching rates (percent of beneficiary's medications on plan formulary) were calculated for each benchmark plan.

Results

Auto-assigned beneficiaries had only a 34% chance of being assigned to the lowest cost plan; the remainder faced potentially significant avoidable out-of-pocket costs. Wide variations between benchmark plans were observed for plan cost, formulary coverage, formulary matching rates, and prescription utilization restrictions.

Conclusions

Beneficiaries had a 66% chance of being assigned to a sub-optimal plan; thereby, they faced significant avoidable out-of-pocket costs. Alternative methods of beneficiary assignment could decrease beneficiary and Medicare costs while also reducing medication non-compliance.

Background

The prescription drug benefit, Medicare Part D, available to Medicare beneficiaries through private insurance plans is entering its eighth year of operation. As of 2010, of the 46.5 million beneficiaries eligible for Medicare Part D nearly 4.67 million (10%) resided in California, the highest number of any state (CMS, 2010). Of all eligible beneficiaries, 68.9% (27.6 million) have enrolled in a Medicare Part D plan, with most others receiving benefits through their former employer or union group (CMS, 2010).

In 2009, the Centers for Medicare & Medicaid Services (CMS) estimated that almost 10 million (36%) beneficiaries enrolled in a Medicare Part D plan had lower costs through enrollment in the low-income subsidy (LIS; Kaiser Family Foundation, 2009). Individuals who qualify for the LIS have an annual income up to 150% of the Federal Poverty Level ($14,570 for a family of two in 2010) and limited assets (<$25,010 for a married couple living together in 2010). Of those eligible for LIS, 8.1 million automatically received the subsidy as full dual-eligible beneficiaries (Medicare and Medicaid recipients) or through receipt of either Supplemental Security Income or the Medicare Savings Programs (Kaiser Family Foundation, 2009). As of February 2010, 1.22 million LIS recipients resided in California; the highest number of LIS beneficiaries of any state (CMS, 2010).

The challenges faced by dual-eligible beneficiaries include, but are not limited to, having a higher reported prevalence of physical and cognitive impairments (Ponce, Ku, Cunningham, & Brown, 2006; Summer, Nemore, & Finberg, 2007; Donohue, 2006). Since the inception of Medicare Part D in 2006, full-benefit, dual-eligible beneficiaries have been automatically enrolled in a stand-alone prescription drug plan (PDP) in the hope of safeguarding this vulnerable population. This automatic enrollment into a regional low-income ‘benchmark’ plan was required by Congress and implemented by CMS (Nemore, 2005), a process that has since been expanded to include all LIS-receiving beneficiaries (Summer, Hoadley, Hargrave, Cubanski, & Neuman, 2008).

Each Medicare region has at least one benchmark plan to which LIS-recipients can be automatically enrolled, provided that two criteria are met: (1) the monthly PDP premium is at or below the regional benchmark premium and (2) the PDP offers a standard benefit (Summer et al., 2008; Hoadley, Hargrave, Merrell, & Summer, 2008). According to federal law, the regional benchmark premium is based on an enrollment-weighted average of the monthly premiums between all PDPs and Medicare-Advantage prescription drug plans (MA-PDs) offered in the region (Summer et al., 2008). Moreover, CMS annually examines each Medicare Part D plan to determine if it is “actuarially equivalent,” or at least as good, as the standard benefit plan. Once a plan is designated as a benchmark plan, LIS beneficiaries are automatically and randomly assigned (“random assignment”) evenly between all available regional benchmark plans (U.S. GAO, 2007). Auto-assigned beneficiaries are also randomly reassigned to a new benchmark plan if their current plan no longer remains a benchmark plan in the upcoming year (Kaiser Family Foundation, 2009; Summer et al., 2008). Of the 47 PDPs available in California in 2010, 7 plans (15%) met the necessary criteria required for designation as a benchmark plan (Kaiser Family Foundation, 2009).

The advantages of random assignment include the assurance that each dual-eligible beneficiary is enrolled into a benchmark plan; thereby, eliminating the monthly plan premium and requiring only nominal co-pays for formulary covered medications. However, a beneficiary may be required to pay the full cost for a non-formulary medication unless a coverage determination is successfully made to the Part D plan, or a substitution to a formulary covered medication is made by the provider. This potential increase in out-of-pocket (OOP) costs for non-formulary medications is probable, based on observed formulary differences of California benchmark plans in 2008 (California HealthCare Foundation, 2008). A study by the Department of Health and Human Services reported that only 18% of LIS-receiving beneficiaries were assigned to benchmark plans covering the most commonly used drugs (Levinson, 2006). An analysis by Maine officials prior to the start of Medicare Part D found that random assignment resulted in one in five dual-eligible beneficiaries having formulary matching rates (the percentage of a beneficiary's medications included on the plan formulary) below 20% (U.S. GAO, 2007). Furthermore, only one-third of the beneficiaries in this study were automatically assigned to plans covering all of their medications, while approximately 30% were in plans that covered fewer than 60% of their drugs (U.S. GAO, 2007). Collectively, this research suggests that under random assignment, differences in plan formularies can lead to significant variability in both beneficiary and governmental costs and/or annual changes in medication utilization dictated by varying formulary coverage.

In recognition of the potential issues associated with random assignment, several state pharmacy assistance programs utilized beneficiary-centered assignment for certain Medicare beneficiaries during the first year of Medicare Part D plan availability (Hoadley, Summer, Thompson, Hargrave, & Merrell, 2007). Under beneficiary-centered assignment, a beneficiary will be reassigned to a different benchmark plan if the randomly assigned plan covers fewer than 85% of their regular prescription medications (formulary matching rates < 85%; Hoadley et al., 2007). Maine is the only state that continues to use beneficiary-centered assignment. A study examining the estimated cost differences between benchmark plans in five regions found that beneficiary-centered assignment can lower beneficiary OOP costs by approximately $450 from the median benchmark plan costs (Hoadley et al., 2008).

Although prior studies have determined the percentage of medications covered on benchmark plan formularies, to-date no study has looked at patient-level variability in beneficiaries' total OOP costs, nor the employment of utilization management tools for each available regional benchmark plan based on the exact array of prescription medications used by LIS-recipients. The present study seeks to calculate the estimated OOP cost for LIS-recipients, and the number of their medications affected by restriction processes under each California benchmark plan to which beneficiaries could be randomly assigned in 2010.

Methods

Study Design

This cross-sectional study examined the variation in beneficiary OOP costs between each of the 2010 benchmark plans offered in California. The increased cost of non-formulary medications and frequency of plan restrictions were also examined between benchmark plans.

Data Collection

In total, 11 outreach events were held in 7 different Central/Northern California cities during the 2010 annual election period (November 15 and December 31, 2009). Sites at which events were conducted included senior centers, housing complexes, community clinics, and institutions of worship. In total, 286 Medicare beneficiaries were assisted during the scheduled events, 40% (n = 113) of whom were full dual-eligible LIS-recipients without any additional creditable coverage (e.g., Veterans Affairs or employer health benefits).

The Medicare Plan Finder tool on the Medicare Web site (www.medicare.gov) was used to confirm the subsidy status, and calculate the estimated annual cost (EAC) of each benchmark plan for each LIS-recipient outreach attendee. The EAC is the total estimated OOP cost under a Part D plan that a beneficiary could be expected to incur during a given calendar year and includes all premium, deductible, and co-payment/co-insurance amounts at each coverage level. The EAC also includes the expected OOP costs for all non-formulary medications until they reach catastrophic coverage. The only OOP costs incurred by full dual-eligible beneficiaries enrolled in a benchmark plan are the nominal co-pay amounts associated with formulary covered medications. Beneficiaries receiving a partial LIS subsidy, or those who enroll in a non-benchmark plan, will also incur a monthly premium that is included in the EAC. Ultimately, the EAC of each Part D plan for each beneficiary is based on their medication profile (medication name, strength, and refill quantity and frequency), subsidy status, and pharmacy preference, along with the plan's formulary and cost-sharing structure.

The EAC and rank (based on EAC) of each of the six California benchmark plans in 2010 to which beneficiaries could be randomly assigned were recorded for all 113 confirmed full dual-eligible LIS-recipients. Due to extenuating circumstances, CMS did not auto-assign any beneficiaries to a 7^th California benchmark plan in 2010 (WellCare Classic offered by WellCare). In addition, the number of prescription medications that were not on formulary, required a prior authorization, had quantity limits, or had step therapy restrictions were recorded for the same six plans based on each individual beneficiary's prescription medication profile.

The formulary matching rates were calculated as the percent of the beneficiary's medications on each benchmark plan formulary. For example, a beneficiary taking eight medications with only seven on the plan formulary would have a formulary matching rate of 87.5%. The percent increase in EAC was calculated for participants with at least one non-formulary drug using the benchmark plan with the lowest EAC as the reference. This cost increase was further subdivided to include only those who had a formulary matching rate of less than 80%. For example, if a benchmark plan with a formulary matching rate of 75% was estimated to cost $1000, and the lowest EAC benchmark plan was $50, the percent increase would be 2000%. Lastly, the average EAC increase of suboptimal plan selection (a benchmark plan with a lower EAC was available) was calculated for each beneficiary and each benchmark plan.

Beneficiary demographic and plan cost data were collected and recorded via a standardized survey. Both the outreach encounter and data collection were performed by trained student pharmacists. All activities were performed under the oversight of licensed pharmacists employed as faculty members at the University of the Pacific, Thomas J. Long School of Pharmacy and Health Sciences. Approval to conduct this research was obtained from the Institutional Review Board at the University of the Pacific.

Statistical Analysis

Descriptive statistics were used to summarize the demographic characteristics of study participants. Cost data comparing the EAC of the six benchmark plans were analyzed using the Repeat-measures ANOVA test. Tukey's HSD post-hoc test was used to determine differences between individual plans. All inferential statistics were performed at an alpha level of 0.05. Data analyses were performed using SPSS, version 17.0 (Chicago, IL).

Results

The demographic characteristics of the study population are summarized in Exhibit 1. On average, participants used approximately six prescription medications per month, with over 30% of participants taking at least seven medications on a monthly basis. Approximately 62% of medications taken by LIS-recipients were generic.

Exhibit 1. Dual-eligible (Medicare/Medicaid) beneficiary demographics (n = 113).

Characteristics	Count	%
Age, y
Mean ± SD	73.2 ± 9.1
< 65	11	(10)
65-74	51	(46)
75-84	39	(36)
≥ 85	9	(8)
Health Status, No. (%)
Excellent	2	(2)
Good	32	(28)
Fair	38	(34)
Poor	21	(19)
Refused to answer	1	(1)
Prescription Medications Per Month
Mean ± SD	5.8 ± 3.3
1–2	13	(12)
3–6	64	(57)
7–10	26	(23)
11+	10	(9)
Generic, %		62.3

SOURCE: Information collected via beneficiary survey during the personalized intervention.

The average EAC between benchmark plans varied from a low of $302 to a high of $1223 (Exhibit 2). First Health Part D Premier and Health Net Orange Option 1 had significantly lower estimated annual costs than the other four benchmark plans (P < 0.01; Exhibit 2). The frequency of being the lowest EAC PDP ranged from 18.6% to 46.9%, a 2.5-fold difference across benchmark plans. Three of the six benchmark plans had at least one instance in which they were the highest cost PDP in the entire region. The largest range observed between the lowest and highest EAC benchmark plans was over $12,000 for one beneficiary.

Exhibit 2. Benchmark plan estimated annual costs (EAC), non-formulary medications (NF Rx), formulary matching rates, and potential cost ramifications for study beneficiaries (n = 113).

Metric	Benchmark Plan Name1
	AS	BC	BR	FH	FV	HN
EAC in $, Mean ± SD	860 ± 1600	1223 ± 1552	1174 ± 1772	302 ± 5902	1157 ± 1847	369 ± 9782
Lowest EAC Frequency, No. (%)	37 (32.7)	21 (18.6)	31 (27.4)	53 (46.9)	53 (46.9)	51 (45.1)
Lowest EAC, Mean $	80	60	74	90	64	94
Suboptimal plan EAC increase in $, Mean ± SD	1100 ± 1637	1359 ± 1440	1442 ± 1736	339 ± 731	1884 ± 1958	432 ± 1035
Beneficiaries w/ ≥ 1 NF Rx, No. (%)	47 (41.6)	65 (57.5)	56 (49.6)	17 (15)	56 (49.6)	16 (14.2)
No. NF Rx per beneficiary, Mean	1.43	1.55	1.71	1.18	1.38	1.25
Mean EAC inc. w/ ≥ 1 NF Rx, %	1556	1837	1830	1167	1855	1059
Formulary matching rates < 80%, No. (%)	19 (16.8)	35 (31)	35 (31)	6 (5.3)	22 (19.5)	4 (3.5)
Mean EAC inc. w/ formulary matching rates < 80%, %	1960	2426	2255	2122	1828	1614

¹AS, Advantage Star Plan by RxAmerica; BC, BlueCross; BR, Bravo Rx; FH, First Health Part D Premier; FV, Fox Value Plan; HN, Health Net Orange Option 1

²P < 0.01 (Tukey's HSD test) for comparison of FH or HN to other plans

SOURCE: Medicare Plan Finder Tool

On average, only two of the six benchmark plans were the lowest EAC PDP in 2010 for each beneficiary, indicating that study participants had a two in three chance of being randomly assigned to a sub-optimal benchmark plan. When ranked as the lowest EAC benchmark plan in the region, the mean EAC varied between $60 and $94. A non-benchmark PDP was associated with the lowest EAC in 2010 for 5 (4.4%) of the study's 113 participants.

Formulary coverage varied significantly between the six benchmark plans with 16 (14.2%) participants taking at least one medication not covered on the Health Net plan formulary compared to 65 (57.5%) beneficiaries taking one or more medications absent on the Blue Cross formulary. Compared to the lowest cost benchmark plan, EAC increased by over 1500% if one or more of the participant's medications was not covered by the plan formulary. The EAC increase due to non-formulary medications ranged from a low of 1,059% (Health Net) to a high of 1,855% with Fox Value (Exhibit 2). Enrollment in a suboptimal plan was associated with a $339 to $1,884 increase in EAC (Exhibit 2).

Differences in formulary matching rates were also found. Only 3.5% of study participants had a formulary matching rate less than 80% with Health Net, as compared to 31% with either Blue Cross or BravoRx (Exhibit 2). Approximately 18% of study participants had a formulary matching rate less than 80% across all benchmark plans. The mean cost increase for benchmark plans with a formulary matching rate less than 80% was 2034% and ranged from a low of 1614% to a high of 2426%.

Quantity limits were the most common restriction observed and had the largest variation across benchmark plans (Exhibit 3). This restriction is especially pertinent to LIS-receiving beneficiaries who are able to purchase a three-month medication supply at the cost of a single month's co-pay. For those participants with one or more restricted medications, prior authorizations were observed less frequently than step-therapy requirements under all benchmark plans except Blue Cross (Exhibit 3). Both prior authorizations and step therapy restrictions have the potential to decrease prescription drug plan costs, increase provider involvement (and therefore provider costs) and potentially delay treatment of both acute and chronic conditions (Raper et al., 2010).

Exhibit 3. Benchmark plan restrictions for study beneficiaries (n = 113).

Plan Restrictions	Benchmark Plan Name1
	AS		BC		BR		FH		FV		HN
	#	%	#	%	#	%	#	%	#	%	#	%
Frequency of Restriction
Prior Authorization	10	8.8	9	8.0	8	7.1	7	6.2	9	8.0	6	5.3
Quantity Limit	16	14.2	92	81.4	95	84.1	85	75.2	34	30.1	106	93.8
Step Therapy	11	9.7	0	0.0	20	17.7	41	36.3	25	22.1	22	19.5
% Medications Restricted in Beneficiaries w/ ≥1
Restricted Medication
Prior Authorization		15.4		16.2		11.4		11.8		10.7		13.3
Quantity Limit		14.8		36.6		34.0		34.4		15.9		47.6
Step Therapy		17.7		N/A		16.9		20.6		15.7		14.3

¹AS, Advantage Star Plan by RxAmerica; BC, BlueCross; BR, Bravo Rx; FH, First Health Part D Premier; FV, Fox Value Plan; HN, Health Net Orange Option 1

SOURCE: Medicare Plan Finder Tool

Discussion

The present study revealed large variation in costs and plan restrictions under each available benchmark plan in California in 2010. Based on the participants in this study, only two of the six plans were associated with the lowest OOP cost if all were randomly auto-enrolled in 2010. Also, the presence of a single non-formulary medication resulted in substantial OOP cost increases.

The primary study objective focused on the variability in benchmark plan EAC and drug utilization restrictions at the individual beneficiary level. Exhibit 2 highlights potential financial consequences of random assignment in study participants in 2010, as revealed through the fourfold inter-plan variation in EAC and formulary coverage differences between benchmark plans. It was also noted that on average only two of the six benchmark plans were the lowest EAC benchmark plan for each participant; range 18.6% to 46.9% (Exhibit 2). If extrapolated to all California beneficiaries, this finding indicates that, on average, 66% of auto-enrolled California beneficiaries in 2010 could have been assigned to a benchmark plan that was not associated with the lowest annual cost. It has been reported that 20% of LIS-recipients were enrolled in benchmark plans in 2009 that did not qualify as benchmark plans in 2010. As such, it is estimated that CMS randomly reassigned 1.2 million LIS enrollees to a new benchmark plan in 2010 (Summer, Hoadley, & Hargrave, 2010).

Through its own analysis, CMS has found that nearly 32% of randomly-assigned beneficiaries took at least one non-formulary prescription medication (Levinson, 2006). In the present study, an average 38% of study participants had at least one non-formulary medication, with a fourfold difference between benchmark plans (14.2% versus 57.5%, Exhibit 2). Additionally, assignment to a plan with one or more non-formulary medications was shown to, on average, increase EAC over 15-fold.

Previous data from Maine indicated that approximately 46% of Medicare beneficiaries would be auto-enrolled into a benchmark plan covering fewer than 80% of their current medications (U.S. GAO, 2007). Participants in this study had, on average, a 17.8% chance (range of 3.5% to 31% between all 6 benchmark plans) of being assigned to a benchmark plan with a formulary matching rate less than 80%. The smaller proportion of participants with formulary matching rates < 80%, as compared to previous studies, must be weighed against the observed 20-fold potential increase in OOP cost. Clearly, the benefit design requiring Medicare beneficiaries, including those with LIS, to pay full price for non-formulary medications resulted in a large increase in OOP costs for study participants and offset copayment and coinsurance reductions afforded by their dual-eligible subsidy status.

In addition to supporting potential cost implications of random assignment seen in other studies, the present study also found that 4% of participants would pay lower total OOP costs by enrolling into an enhanced plan rather than an available benchmark plan. For these participants, the increased expense of paying a monthly premium differential above the benchmark amount was less costly than paying the full cost for non-formulary medications. Despite the small sample size of this study, this finding lends weight to the potential implication of allowing beneficiary assignment to enhanced plans (Hoadley et al., 2008).

Though the Medicare Part D benefit includes significant financial provisions for LIS-recipients, the benefit can only be maximized if beneficiaries become fully aware of all available afforded resources. The necessity of beneficiary-centered assignment has been downplayed by CMS due to avenues available to LIS-receiving beneficiaries, including the prerogative to switch PDPs at any time during the course of the year, or by taking a more proactive approach and switching to medications that are on their existing plan formulary (U.S. GAO, 2007). Despite these provisions, our data revealed that only 9% of the study sample switched plans outside of the open enrollment period. Although there are potential channels available to LIS beneficiaries whose medications are not on the plan formulary, including requesting an exception, obtaining a new prescription, taking medication less frequently than prescribed, paying for the full cost of the medication, or changing to a plan which covers their medication, all of the these avenues can be complex, burdensome, and/or put the patient at undue risk (Hoadley et al., 2007; U.S. GAO 2007; Nemore & Gottlich, 2006).

Based on the congruency of our findings with larger studies, we too espouse the position that CMS should move away from random assignment to one that considers the medication regimen of each LIS-recipient when placing them in a benchmark plan (Hoadley et al., 2007; U.S. GAO 2007; Hoadley et al., 2008). We argue that the primary objective of beneficiary-centered assignment should be to lower beneficiary, and not federal government, costs associated with Medicare Part D. By preventing OOP cost increases, medication compliance is more likely maintained, and unnecessary medical, hospital and emergency room costs might be averted. To help curb the cost of implementing a program which unilaterally adopts beneficiary-centered assignment, we suggest that those beneficiaries who are at high risk of non-compliance related adverse events (e.g., based on prescription utilization or previous Medicare costs) should be prioritized to receive such a service. In terms of the feasibility of implementing a system based on beneficiary-centered assignment, a report conducted for the Medicare Payment Advisory Commission concluded that such a system is possible for LIS-receiving beneficiaries, with minimal disruptive effect or economic costs of providing such service (Hoadley et al., 2007).

To date, the idea of beneficiary-centered assignment has been met with some resistance from CMS. With regards to assigning beneficiaries to a benchmark plan based on their prescription drug utilization pattern, CMS has previously taken the inflexible position that “we do not accept the premise that exact matches are necessary or desirable” (U.S. GAO, 2007). Additionally, CMS has retorted that it is statutorily mandated to randomly enroll LIS-receiving beneficiaries to a benchmark plan (U.S. GAO, 2007). As such, CMS argues that it does not have the discretion to enroll beneficiaries into a plan based on their specific prescription drug usage patterns (U.S. GAO, 2007). An alternative approach that could be employed by CMS is to require broader formulary coverage for benchmark plans. This would alleviate some of the variation in plan formulary coverage and decrease the potential for suboptimal plan assignment.

In response to concerns raised by CMS, the U.S. Government Accountability Office recommended that states be given the authority to use alternative enrollment methods, such as beneficiary-centered assignment, when assigning LIS-recipients to benchmark plans (U.S. GAO, 2007). Furthermore, as of 2012, beneficiary-centered assignment is permissible instead of random assignment by Section 1205 of the Health Care reform proposal (HR 3200; America's Affordable Health Choices Act, 2009). Though we recognize the potentially onerous implications of designing a system for individual-level reexamination of benchmark plan assignment, the potential economic savings to both the LIS-recipient and the government seem to justify the implementation of such a system.

The congruency of our findings with other reports indicates that expansion into other regions and larger sample sizes may be justified. An extension of this study that would likely be of interest to policy-makers would examine the cost savings at a societal level, including both beneficiary and governmental savings of a beneficiary-centered system of assignment. Additionally, a prospective study which examines actual beneficiary behavior when faced with potentially high OOP medication costs, or barriers to prescription drug access through the implementation of restriction processes, is recommended for future consideration and may shed light on the potential impact of drug costs on non-drug related Medicare costs.

A key limitation of this study was the small sample size (n = 113) of LIS-receiving Medicare beneficiaries who self-selected to participate in this study. Participants may have been more likely than the general LIS-eligible Medicare population to need assistance and we recognize that this was not a randomized sample of Medicare beneficiaries residing in California. However, the results from this study tend to coincide with previous research and analyses; therefore, they lend weight to the validity of our findings. This is the first study to report individual-level cost and drug-restriction use for regional benchmark plans in any region. As such, the novelty of these findings must be weighed against the potential for generalization.

Conclusion

The present study highlights the potential variability in OOP costs for LIS-recipients under each of the available benchmark plans in California. Much of this variability was explained by the number of beneficiary medications that were not on the benchmark plan formulary. The similar results of this study, as compared to other larger reports, indicate that renewed attention should be placed on the viability of beneficiary-centered assignment for all LIS-recipients. Additionally, we found that targeted assistance to a vulnerable population of beneficiaries may help minimize both their OOP costs and potential challenges they face with regard to their Medicare Part D benefit.