Will Medicare Advantage Benchmark Reforms Impact Plan Rebates and Enrollment?

Abstract

Objectives

To assess the relationship between Medicare Advantage plan rebates and enrollment and simulate the effects of Affordable Care Act payment reforms.

Study Design and Methods

First difference regressions of county-level Medicare Advantage payment and enrollment data from the Centers for Medicare and Medicaid Services from 2006 – 2010.

Results

A $10 decrease in the per member/per month rebate to Medicare Advantage plans was associated with a 0.20 percentage point decrease in MA penetration (p < 0.001) and a 7.1% decline in the average MA enrollee's risk score (p < 0.001). Changes were markedly larger in counties with low levels of Traditional Medicare spending, a $10 decrease in monthly rebates was associated with a 0.64 percentage point decline in MA penetration and a 10% decrease in risk score. Affordable Care Act reforms are predicted to reduce the level of rebates in low spending counties, leading to enrollment decreases of 2.7 percentage points in the lowest spending counties. The simulation predicts that the disenrollment would come from MA enrollees with higher risk scores.

Conclusions

Medicare Advantage enrollment responds to the generosity of supplemental benefits available. MA plans in low-cost counties may have difficulty offering the supplemental benefits that attract enrollees even when benchmarks are set at levels well above Traditional Medicare spending if plans do not find ways to deliver standard Medicare benefits at lower cost.

Private plans have been offered as an alternative to Traditional, fee-for-service Medicare (TM) for more than 30 years. The Medicare Advantage (MA) program has repeatedly changed rules governing payments and enrollment in effort to design a program that provides additional benefits to Medicare beneficiaries at no greater cost than if enrollees remained in TM. Despite optimism that managed care would reduce Medicare spending, policy changes starting with the 2003 Medicare Modernization Act have ensured higher payment rates relative to TM ¹.

In 2006, the Centers for Medicare and Medicaid Services (CMS) implemented a bid system in a new attempt to rein in spending on plans. Each year, CMS sets a payment benchmark, the maximum monthly amount they will reimburse for an enrollee of average risk, for each county. The benchmark updates prior years' payment rates by annual Medicare growth rates, and are not directly tied to cost care in the county in many cases.² MA plans in turn “bid” the amount it costs them to standard Medicare benefit package to an average beneficiary in a given market. When bids are below these benchmarks, plans receive a rebate covering part of the difference between the benchmark and their bid (risk-adjusted to reflect case-mix balance), which must be used to provide additional benefits or reduced premiums to enrollees.

The rebate amount was initially set at 75 percent of the difference between the benchmark and the bid, under the Affordable Care Act (ACA) this will be gradually reduced to 50, 65, or 70 percent of the difference, depending on a plan's quality rating.³ ACA provisions will also tie county benchmarks to Traditional Medicare spending (ranging from 95% of expected spending in the highest FFS spending regions to 115% in the lowest-cost cost regions), likely reducing the difference between benchmarks and bids in many regions.

Under the bid system, more generous benchmarks relative to plans' true costs of providing care imply larger rebate amounts. While previous research has found a positive relationship between payments to MA plans and enrollment, it is unknown whether changes in the MA payment formula will impact the size and composition of the MA population. Plans can use larger rebates to offer benefits such more comprehensive drug coverage that will attract sicker enrollees, or benefits that attract healthier enrollees such as fitness programs and preventive care.^4-6

While many studies have documented a positive relationship between overall payments to MA plans and program enrollment,^7,8 data assessing the proportion of payment available for supplemental benefits have only recently been released to researchers. To date, little is known about how the bid system has influenced MA market structure, payments to plans, and Medicare beneficiary enrollment decisions. To assess the potential for these payment changes to affect the size and composition of the Medicare Advantage population, I studied the relationship between MA rebates and enrollment between 2006 – 2010. These results were used to estimate the consequences of ACA changes to plan payments for MA enrollees in high and low spending regions of the country.

Methods

This study analyzed aggregated administrative data and was exempt from Institutional Review Board review.

Data

I analyzed publicly available data from the Centers for Medicare and Medicaid Services detailing county-level MA benchmarks from 2006 – 2010 linked to information about payments, enrollment and patient risk scores reported at the county by plan type level. These files were combined with information about average Fee-for-Service spending adjusted for labor costs from the Dartmouth Atlas of Health Care and characteristics of the commercial market from HealthLeaders Interstudy. I focused on MA enrollment in HMOs, local and regional PPOs and PFFS plans. Special needs and employer-sponsored plans, and other types of private plans only available to certain subgroups of beneficiaries were excluded.

Enrollment

I assessed the impact of MA payments on two measures of enrollment; the proportion of Medicare beneficiaries in a county who were enrolled in a Medicare Advantage plan and the average risk score of these enrollees. Risk scores are a summary measure of enrollee health or expected utilization used to adjust payments to MA plans to account for enrollee health profiles that deviate from that of the “average” Medicare enrollee (with a risk score of 1). Because CMS reports information aggregated at the county by plan-type level, risk scores were calculated as the enrollment-weighted average of HMO, PPO, and PFFS average risk scores reported in each year's CMS Medicare Plan Payment Data file. Higher scores indicate higher levels of expected utilization. More generous supplemental benefits facilitated through larger rebates might attract larger shares of beneficiaries and those with higher risk scores to MA.

Rebates

Plan rebate amounts are weighted averages calculated in the same way as the risk scores and adjusted to 2010 dollars using the Consumer Price Index. CMS calculates each plan's actual rebate by case-mix adjusting the county-level benchmark to account for the plan's enrollee profile across all markets. Thus the average rebates analyzed capture variation in plans' bids, or cost of providing care in each market and each plan's risk profile across markets. Under the ACA, rebates are further adjusted based on plans' quality ratings. The ACA simulation assumes a 65% rebate, corresponding to a plan with a quality rating between 3.5 – 4.5 out of 5 stars. As of 2012, the majority of plans and enrollment were between 3 – 5 stars.⁹

Statistical Methods

To test the enrollment relationship to changes in plan rebates, I modeled the change in enrollment (risk score) from one year to the next as a function of changes in average rebates and control variables from the previous year. This was done by estimating first difference regressions of MA penetration and risk scores on the average rebate paid to plans to test the hypothesis that Medicare beneficiaries change their MA enrollment decisions in response to plan generosity.

I also controlled for the number of firms offering at least 1 MA plan, and the MA market Herfindahl index calculated from CMS enrollment data. The Herfindahl index ranges from 0 to 1 and measures market competition. Higher scores indicate lower levels of competition, with 1 indicating that market enrollment is concentrated in a single plan. Enrollment and patient acuity may increase in more competitive markets as beneficiaries have access to a more diverse choice set and may be more likely to find a plan appropriate for their needs. I controlled for the cost of providing care in a given region using the ratio of price adjusted average TM spending to unadjusted TM spending from the Dartmouth Atlas of Health Care to account for factors like local wage rates that make some markets more costly.¹⁰ Regressions included a constant term and indicator variables for years 2008 – 2010 to account for other factors influencing MA enrollment and plan rebates over time ¹¹.

Regressions were weighted by county Medicare enrollment (age 65 and up) and included all counties with a minimum 1% MA penetration rate. Regressions were estimated first pooling all counties, and then estimated separately for counties in the lower vs. higher halves of per capita TM spending because ACA county benchmarks will be tied to TM spending quartiles. Regression results from the first four years of the MA bidding regime were used to simulate the impacts of upcoming ACA changes to the MA benchmarks.

ACA Predictions

I used aggregate information about 2013 plan benchmarks, bids and payments and expected FFS costs reported in the 2013 Medicare Payment Advisory Commission (MedPAC) Report to the Congress, the most recent data available, to estimate the change in rebate amounts that would be expected under an immediate transition to ACA benchmarks slated to be fully implemented by 2017. The ACA benchmarks are 115% of expected TM spending in the quartile of counties with lowest TM spending, 107.5% in the next quartile, 100% and 95% in the third and fourth-highest quartiles respectively.

I used the midpoint of 2013 TM spending in each quartile and median bid (expressed as a proportion of TM spending) reported by MedPAC to calculate the rebate amount that a plan of average quality would receive under current policy and the ACA reforms. Recent quality ratings indicate that 63% of MA enrollment is concentrated in plans with fewer than 3.5 quality stars (out of 5), indicating that most plans are unlikely to qualify for the most generous rebate rate of 70%, so this analysis is based on the 65% rebate rate ⁹.

Estimated changes in rebates were used to predict changes in MA enrollment in counties with high and low FFS spending based on the parameters estimated in the 2006 – 2010 data. MA enrollees in the low-FFS counties are more likely to live in rural areas, where it may be difficult for MA plans to establish provider networks, leaving plans with less negotiating leverage to maintain rebate amounts by negotiating lower payments to providers ¹². MedPAC estimates that more than 75% of MA enrollment will be in counties in the highest two quartiles of FFS spending ⁹.

Study Results

Sample Characteristics

Table 1 and –Figure 1 describe sample characteristics overall and by levels of TM spending. The ACA quartiles are designed to have an equal number of counties, rather than an equal number of Medicare beneficiaries in each quartile; nearly 45% of Medicare beneficiaries lived in the highest-spending. During the study period, MA penetration averaged 22.9% (increasing from 19% in 2006 to 26% in 2010). MA benchmarks, payments, and rebates were highest in counties with high TM spending. Average monthly, per-member rebate payments were $87 in the highest-spending counties, compared to $52-$57 in the first through third spending quartiles. MA enrollment is more common in high-TM counties and includes enrollees with higher risk scores; the mean risk score was 0.87 in the lowest-spending counties and 1.01 in the highest. Figure 1 shows patterns in the key enrollment and payment measures from 2006 – 2010; rebates and risk scores fell over time though MA penetration and county benchmarks (measured in constant dollars) increased.

Table 1. County Characteristics by Traditional Medicare Spending Quartile, 2006 - 2010.

		Quartile of TM Spending
Variable	Overall	Q1	Q2	Q3	Q4
MA Enrollment (%)	22.9	21.6	19.6	21.6	25.3
Share of Medicare Beneficiaries	100	12.8	19.4	23.1	44.7
Average MA risk score	0.95	0.87	0.90	0.94	1.01
MA Benchmark	838	747	763	786	923
Average Plan Rebate (PM/PM)	69	52	52	57	87
Average Plan Payments (PM/PM)	737	696	709	721	768
Fee-for-Service Spending (PM/PM)	742	561	631	699	863
MA market Herfindahl Index	0.42	0.41	0.42	0.44	0.41
Under 65 Market Herfindahl	0.25	0.28	0.27	0.26	0.23
MA Firms in County	10.1	7.6	7.8	8.5	12.7
Under 65 Firms in County	18.7	16.0	16.5	17.7	20.9
Number of Counties	3,180	785	785	785	785

Notes: Summary statistics are weighted by county population and are nationally representative. Under Affordable Care Act reforms, counties will be divided into quartiles with equal numbers of counties in each group. Medicare Advantage benchmarks will be set as a proportion of county-level Traditional Medicare spending ranging from 115% in the lowest TM spending counties to 95% in the highest spending counties.

Figure 1. Medicare Advantage Plan Enrollment and Payment Characteristics By Level of Traditional Medicare Spending, 2006 – 2010.

Notes: County-level observations weighted by Medicare enrollment. High and Low TM refer to the 1,570 counties in each year with the highest (lowest) Traditional Medicare spending per capita. Rebates and benchmarks adjusted for inflation and reported in 2010 dollars. Higher scores indicate higher expected utilization, with 1 representing a beneficiary of average risk.

Plan Rebates and Medicare Advantage Enrollment

Table 2 shows the estimated relationships between changes in average plan rebates and changes in county-level MA enrollment and risk scores. After adjusting for MA and commercial market characteristics, a $10 decrease in the per member/per month rebate amount was associated with a 0.20 percentage point decrease in MA penetration (p < 0.001). However, there was a four-fold difference in the impact of a rebate change in high versus low TM spending regions. A $10 decrease in rebates was associated with a 0.16 percentage point decrease in enrollment in high-cost regions (p < 0.001), and a 0.64 percentage point decrease in low-cost regions (p < 0.001). Decreases in rebates were also associated with declines in the average MA risk score (-0.07 (7.1%), p < 0.001). The change in average enrollee risk was modestly larger in magnitude in low-cost regions (-0.09 (10%), p < 0.001 versus -0.07 (7.1%), p < 0.001).

Table 2. Relationship between Medicare Advantage Rebates and Enrollment, 2006 - 2010.

	Coefficient	P-Value	Impact of $10 PM/PM Decrease
MA Penetration
Overall	0.020**	< 0.001	- 0.20pp
Low TM Spending	0.064**	< 0.001	-0.64pp
High TM Spending	0.016**	< 0.001	-0.16pp
MA Risk Score
Overall	0.007**	< 0.001	- 0.07 (7.3%)
Low TM Spending	0.009**	< 0.001	- 0.09 (10%)
High TM Spending	0.007**	< 0.001	- 0.07 (7.1%)

Notes: Results from first-difference regressions of county-level MA penetration and risk score on average MA plan rebates per member per month. Higher scores indicate higher expected utilization, with 1 representing a beneficiary of average risk. High and Low TM refer to the 1,570 counties in each year with the highest (lowest) Traditional Medicare spending per capita. Regressions are estimated for all counties and separately for counties in the lowest two and highest two quartiles of TM spending with at least 1% MA penetration using annual data from 2006 - 2010. Standard errors are clustered at the county level.

^**denotes statistical significance at p < 0.01. Column 3 indicates the change in enrollment (risk score) associated with a $10 per member per month decrease in plan rebate payments.

Predicting Effects of ACA Benchmark Reductions

Figure 2 shows median, 25^th and 75^th percentile 2013 MA bids, representing plans' costs of providing standard MA benefits relative to the ACA benchmarks for counties in each quartile of TM spending. Once ACA reforms are fully implemented, benchmarks in counties in the lower half of TM would continue to exceed TM spending. Current plan costs in more than 25% of plans in TM quartiles exceed the county's ACA benchmarks. These plans would have to charge premiums to MA enrollees, find ways to deliver the standard benefits at lower cost, or leave the market in response to ACA reforms.

Figure 2. Medicare Advantage Plan Bids Relative to Affordable Care Benchmarks, 2013.

This figure plots median 2013 Medicare Advantage plan bids (cost of providing standard Medicare benefits) with 25^th and 75^th percentile ranges by quartile of average Traditional Medicare spending reported in MedPAC (2013) and county benchmarks once Affordable Care Act provisions are fully implemented. MedPAC reports medians for terciles of average TM spending within the highest-TM spending counties because of the wide range of spending levels and bids in these counties. Average TM spending ranges from $560 per member/per month - $750 PM/PM across quartiles 1 – 3, and from $751 - $1,350 in the highest quartile. Bids are typically a smaller share of FFS spending in the highest cost counties. 86% represents the median bids in the middle ($825 - $900 PM/PM) range of high-TM spending, to give a sense of the variation in bids within this quartile, endpoints represent the 75^th percentile from the lower tertile and the 25^th percentile from the highest tercile.

Under the ACA benchmarks, predicted median rebates would fall in the lowest 3 quartiles of FFS spending and increase in the highest-cost quartile. Figure 3, Panel A presents estimated rebates that plans would receive if their 2013 bids had been subject to the ACA county benchmarks and the 65% quality-adjusted rebate and actual rebates received in 2010, the most recent year of available data. Holding other market characteristics constant, rebates are predicted to fall by $20.40 (56%), $42.70 (117%), and $48.76 (124%) per member/per month in the lowest three quartiles of TM spending, and increase by $4.10 (6.7%) in the highest quartile.

Figure 3. Predicted Medicare Advantage Rebates, Enrollment and Average Risk Score under ACA Benchmarks By Quartile of Traditional Medicare Spending Relative to 2010 Median Values.

Predicted median values under Affordable Care Act payment changes assumed that all MA plans submitting bids in 2013 remain in the market without changes to their cost of providing the standard part A and part B benefits. Calculations based on regression coefficients reported in Table 2 and 2013 plan bids and expected Traditional Medicare costs reported by MedPAC and assume plan quality of 3.5 – 4.5 stars (out of 5), corresponding to rebates of 65% of the difference between the plan's bid and the county benchmark.

Figure 3, Panels B and C describe the expected changes in enrollment share and risk score, respectively, that would result from implementing the ACA rebate changes. Relative to 2010 levels, enrollment would decrease by 2.7 percentage points to 18.9% in the lowest TM counties, 1.3 percentage points to 20.3% in the next lowest, 0.8 percentage points to 20.8% in the third quartile, and increase by 0.07 percentage points in the counties in the highest quartile of TM spending. The simulation predicts that the disenrollment would come from MA enrollees with higher risk scores. ACA changes would lead to stark differences in the average risk score of MA beneficiaries in low versus high-cost TM counties. In 2010, average risk scores ranged from 0.87 in the lowest spending counties to 1.01 in the highest. Under ACA implementation these would range from 0.52 to 1.04. These projections assume that plans maintain the same cost structure and market conditions are held constant under the ACA, though these factors may change in response to lower benchmarks. Only 25% of counties in the lower TM spending counties had average risk scores below 0.84 in 2010, plans may find ways to improve efficiency in plans that cater to enrollees with greater health needs. As of 2013, 90% of Medicare beneficiaries have access to at least plan that currently bids below their county's ACA benchmark, enrollees may face fewer plan choices but will retain the MA option.⁹

To better understand why changes in the benchmarks would have a more pronounced impact on rebates and enrollment in low-cost counties, I regressed average rebates on county benchmarks and prior year market characteristics (Appendix 1). In high-cost counties, a $1 increase in the benchmark per member per month translates to a $0.36 increase in the rebate, while a $1 increase in low-cost counties translates to a $0.89 increase in the rebate, suggesting that plans in low TM spending markets may respond more competitively.²

Discussion and Implications

Using administrative plan data from 2006 – 2010, I found that decreases in rebates to Medicare Advantage plans used to offer supplemental benefits are associated with decreases in both the number of MA enrollees and their average expected utilization (risk score). Efforts to reduce payments to MA plans under the Affordable Care Act will likely have the biggest impact on beneficiaries in counties with low average Traditional Medicare spending, though plans in these regions will continue to bid against benchmarks that exceed TM spending.

These findings highlight a long-time challenge in the Medicare Advantage program that has important implications not only for short-term changes to the program itself, but for more comprehensive reforms such as premium support, which would provide Medicare beneficiaries with a defined contribution to purchase health insurance.¹³ Providing a single amount nationally will produce enormous inequity in the level of coverage Medicare beneficiaries are able to purchase. ^14,15 However, diverse market conditions across counties challenge rates set as a fixed proportion of local Traditional Medicare spending. ^16,17 While plans in low-spending regions will struggle to provide any supplemental benefits at payment rates of up to 115% of TM spending, those in relatively high spending counties will be able to offer generous benefit packages at costs below average TM spending.

This analysis is based on aggregate data, precluding study of variation in plan costs and offerings within counties. It is unknown whether the increased enrollment within a county actually moves into the plans with the lowest bids and highest rebates. Plans that submit the lowest bid within a county may have low costs because they are good at attracting enrollees who will use less care than their risk score suggests (positive selection), or have restrictive provider networks that would discourage sicker beneficiaries from enrolling.

The ACA simulation assumed a 65% rebate, the level that would be offered to the majority of plans based on current quality metrics. In theory, implementation of the ACA payment changes could motivate improved MA quality of care as plans strive to reach the 4-star designation, conducting quality improvement efforts or exiting the market. Similarly, relatively inefficient or high-cost plans may be driven out of the market. There is a long history of plan withdrawal from unprofitable markets in response to a variety of payment changes to the Medicare Advantage program, though the extent of changes in response to the ACA will not be fully known until benchmarks are fully implemented over the coming years.^15,18,19

At first glance, exits by lower-quality plans and improvements to quality scores should help MA enrollees. However, this may leave Medicare beneficiaries in some areas with limited plan choice. Improving performance on many of the elements of the star-rating quality scores may increase plan administrative costs through compliance with data collection and increased patient monitoring to achieve process and intermediate outcome endpoints. Plan investments to improve quality ratings may have the unintended consequences of crowding out spending on supplemental benefits beneficiaries may value more than those targeted by the CMS rankings.

Finally, it is important to note that some MA enrollees already choose plans that charge additional monthly premiums. It will be important to track enrollee expenditures under the ACA reforms to determine whether plans lower costs in response to lower benchmarks or shift costs to enrollees.

Appendix 1

Relationship between MA Benchmarks, Market Characteristics, and Plan Payments and Rebates, 2006 - 2010.

	Total Plan Payments				Rebates
	All	Low FFS	High FFS	All	Low FFS	High FFS
MA Benchmark	0.24**	-0.33	0.23**	0.38**	0.89**	0.36**
	(0.04)	(0.40)	(0.05)	(0.07)	(0.22)	(0.08)
Lagged MA Firms in County	-2.0**	-2.20**	-1.83*	1.04**	1.21**	0.96*
	(0.60)	(0.52)	(0.77)	(0.34)	(0.32)	(0.43)
Lagged MA Herfindahl Index	1.25	6.8*	-0.37	-2.62	-7.91**	0.39
	(3.22)	(3.16)	(5.02)	(1.78)	(1.31)	(2.97)
Lagged Commercial Firms	-8.75	-2.63	-11.12	-1.34	3.85	-6.93
	(5.88)	(5.21)	(10.87)	(3.37)	(2.34)	(6.53)
Lag Commercial Herfindahl	-0.53**	-0.50**	-0.40#	0.20#	0.29**	0.21
	(0.15)	(0.18)	(0.23)	(0.11)	(0.08)	(0.16)

Notes: County-level regressions of total payments to MA plans and plan rebate amounts used to finance extra benefits on MA benchmarks set by the Centers for Medicare and Medicaid services. Cluster-robust standard errors in parentheses.

^#statistically significant at 10%.

^*significant at 5%,

^**significant at 1%.