Modify Payments to Medicare Advantage Plans for Health Risk

Background

Roughly a third of Medicare beneficiaries are enrolled in the Medicare Advantage program. Through that program, private health insurers receive a payment for each beneficiary they enroll and then take financial responsibility for covering that beneficiary's care. Almost all other Medicare beneficiaries receive care in the Medicare fee-for-service (FFS) program, which pays providers directly for each service or set of services covered by Part A (Hospital Insurance) or Part B (Medical Insurance). Payments to Medicare Advantage plans depend on three components: bids that plans submit to the Centers for Medicare & Medicaid Services (CMS), predetermined benchmarks that CMS sets on a county-level basis, and risk scores that reflect variation in beneficiaries' expected spending because of health conditions and other characteristics.

Plans' bids and Medicare's benchmarks together determine a base payment—or a per capita payment from CMS to the plan for an enrollee with average expected health costs. CMS determines base payments by comparing area-specific benchmarks to a plan's standardized bid—or a bid that reflects the plan's estimated cost for providing Medicare benefits in a given area to an enrollee in average health. If a plan's bid is above the benchmark, then CMS pays plans the benchmark. Plans must then charge enrollees a premium (which the enrollee pays in addition to the Part B premium) equal to the difference between the bid and the benchmark. If the plan's bid is less than the benchmark, then the base payment from CMS is the bid plus a rebate. That rebate is a percentage of the difference between the bid and the benchmark, which plans are required to devote primarily to reducing premiums for Part B or Part D (the prescription drug benefit), reducing cost sharing, or covering additional benefits that Medicare does not cover, such as vision or dental care. Both the benchmark and the rebate percentage are also modified to reflect a plan's average quality score. (Quality scores are discussed in detail in the option "Reduce Quality Bonus Payments to Medicare Advantage Plans.")

CMS further adjusts payments to plans to reduce insurers' incentives to selectively enroll beneficiaries on the basis of their expected spending. Specifically, CMS scales total payments to plans upward or downward by the risk scores of a plan's enrollees. Risk scores are constructed to reflect variation in enrollees' expected health care costs and are calculated for all Medicare beneficiaries on the basis of their diagnoses and other characteristics. Those scores are standardized so that a score of 1.0 reflects the health care spending of the average beneficiary in Medicare FFS—a type of calculation that is generally referred to as normalization. Higher risk scores indicate higher expected health care spending, and a plan is paid more for an enrollee with a higher risk score. Conversely, a plan is paid less for enrollees with lower expected health care spending.

More thorough documentation of beneficiaries' diagnoses increases their risk scores, and thus, plans have a financial incentive to record all diagnoses for their enrollees. In contrast, providers serving Medicare FFS patients have more limited financial incentives to code a beneficiary's diagnoses because their payments are not tied to risk scores. Recent research has, in fact, shown that Medicare Advantage enrollees have higher average risk scores than otherwise similar FFS beneficiaries and that the difference has increased over time. Therefore, that divergence in risk scores appears to reflect more thorough diagnostic coding by Medicare Advantage plans, rather than differences in enrollees' health (Hayford and Burns 2018; Medicare Payment Advisory Commission 2018).

To adjust for differences in coding, federal law currently requires CMS to apply an across-the-board reduction to Medicare Advantage plan payments that is intended to reflect the difference in coding intensity across the two populations. However, some research has found that the increase in payments that is attributable to coding intensity exceeds the current reduction being applied in the program (Medicare Payment Advisory Commission 2018; Kronick and Welch 2014). Additionally, evidence suggests that some plans code more intensively than others. For instance, health maintenance organizations (HMOs) are thought to be able to code diagnoses more completely than preferred provider organizations (PPOs) or private fee-for-service (PFFS) plans, which have broader provider networks and exercise less control over providers' practice patterns(Geruso and Layton 2018; Hayford and Burns 2018). Thus, an across-the-board reduction in payments to offset coding intensity penalizes plans that do not code as intensively and maintains incentives for plans to increase coding intensity.

Option

This option—which would affect risk-adjustment policy—consists of three alternatives, all of which would take effect in 2021. Under current law, CMS must reduce payments to all plans by a minimum of 5.9 percent to reflect differences in coding across populations. The first alternative would require CMS to reduce payments to all plans by at least 8 percent instead. Eight percent is the Medicare Payment Advisory Commission's most recent estimate of the average difference between Medicare Advantage and FFS risk scores for otherwise similar beneficiaries.

The second alternative would also require CMS to reduce average plan payments by a minimum of 8 percent, rather than 5.9 percent. However, it would further require CMS to scale that 8 percent reduction—that is, increase or decrease the reduction—on the basis of differences in coding intensity for each insurer in a given region. CMS would calculate that adjustment using the change in risk scores for beneficiaries who switched from Medicare FFS to an insurer's plan in a given region and then place plans into quartiles according to growth in those enrollees' average annual coding intensity since switching to Medicare Advantage. To simplify implementation, plans within the same quartile would have their risk scores adjusted by the same percentage so that the average reduction across all plans, weighted by enrollment, would be a minimum of 8 percent.

Changes in risk scores for beneficiaries who switch from FFS to Medicare Advantage capture differences in coding intensity because those beneficiaries' initial risk scores are based on coding patterns in Medicare FFS, whereas the change in risk scores reflects the increase in coding attributable to joining Medicare Advantage. Examining changes in risk scores for beneficiaries on an insurer-level basis allows CMS to determine how coding intensity varies across insurers, and applying adjustments that are specific to each insurer ensures that plans that code more intensively face larger payment reductions. Likewise, allowing those adjustments to vary across regions addresses the fact that plans in different parts of the country may have different relationships with providers or different coding practices. Under this second alternative, insurers that have operated in the market for fewer than three years would have the standard 8 percent reduction applied to their payments.

The third alternative would make two changes to risk-adjustment policy. First, CMS would be required to use two years of diagnostic data to calculate risk scores rather than one. Under the current system, risk scores are generated on the basis of a beneficiary's diagnoses from the previous calendar year. Empirically, using two years of diagnoses to generate risk scores rather than one would result in more diagnoses being captured among FFS beneficiaries—and would have minimal effects on the number of diagnoses captured among Medicare Advantage beneficiaries. Accounting for additional diagnoses among FFS beneficiaries therefore would reduce the gap between average Medicare Advantage risk scores and average FFS risk scores. (The 21st Century Cures Act gave CMS the authority to use two years of diagnostic data beginning in 2019; the agency did not use that authority in 2019 but may in future years.)

Second, risk scores would no longer reflect diagnoses captured from health risk assessments. Health risk assessments are visits by providers that can help determine a beneficiary's health needs and set a course for treatment. However, health risk assessments in Medicare Advantage are more likely than those in FFS to record a diagnosis for which a beneficiary receives no subsequent care. Excluding diagnoses recorded only during health risk assessments—rather than during other visits to providers—would therefore further reduce the disparity between FFS and Medicare Advantage risk scores.

Effects on the Budget

All three alternatives would reduce mandatory spending between 2021 and 2028, according to estimates by the Congressional Budget Office.

CBO estimates that changing the reduction in risk scores from the current 5.9 percent to 8 percent to better reflect coding differences—the first alternative—would lower mandatory spending by $47 billion between 2021 and 2028. Those savings would be the result of direct cuts to plan payments, but they include an offset that stems from the expectation that plans would adjust their bidding behavior in response to the payment reduction. (Because of shifts in the timing of payments between fiscal years, savings under all three alternatives would change minimally between 2022 and 2024 and increase in 2028.)

Under the second alternative—which would also change the reduction in risk scores from 5.9 percent to 8 percent but scale that reduction by insurer and region—CBO estimates that mandatory spending would be reduced by $47 billion, the same amount of savings resulting from the first alternative. Compared with the first alternative, plans could face larger or smaller reductions under the second alternative; however net savings would be equivalent to those resulting from the first alternative because reductions in risk scores would, on average, be the same. As in the first alternative, CBO anticipates that plans would adjust their bidding behavior to partially offset the effect of payment cuts. CBO also expects that changes in bids would, on average, be the same as in the first alternative because adjustments by plans facing larger cuts would be offset by adjustments from plans facing smaller cuts.

Under the third alternative—modifying how risk scores are constructed—mandatory spending would be reduced by $67 billion (including the timing shifts noted above), CBO estimates. That reduction would be driven by lower payments to plans resulting from a 3 percent reduction in average normalized risk scores. Those reductions would arise in two ways: First, excluding diagnoses that are solely recorded in health risk assessments generally would result in a greater reduction in risk scores for Medicare Advantage enrollees than for FFS beneficiaries. Second, basing risk scores on two years of diagnoses would result in a greater average increase in risk scores for FFS beneficiaries than in risk scores for Medicare Advantage enrollees. Risk scores are normalized around the average health of beneficiaries in FFS. Thus, if FFS risk scores increased without a corresponding increase in Medicare Advantage risk scores, average normalized risk scores for Medicare Advantage enrollees would be reduced. That reduction, in turn, would reduce payments. As with the first two alternatives, CBO anticipates that plans would adjust bids in response to those payment reductions. Those adjustments would be slightly larger than in the first and second alternatives because the average reduction in plan payments would be larger. However, on net, this alternative would result in larger reductions in mandatory spending than the previous two alternatives.

CBO anticipates that the amount of savings in the first two alternatives would increase or decrease proportionately with the reduction applied to risk scores. That is, if the reduction to risk scores was smaller than 8 percent, savings would be proportionately reduced, and if that reduction was greater, savings would increase—although there is likely a limit on how much risk scores could be reduced before plans would exit the program. In contrast, the third alternative represents a onetime change in the calculation of risk scores and therefore could not be increased or decreased without additional modifications to the risk-adjustment model.

The largest source of uncertainty in the estimate of savings over the next 10 years under all three alternatives is CBO's estimate of how much plans would adjust their bids in response to reduced payments. CBO projects that plans would adjust their bids to partially offset that payment reduction. However, those adjustments could be larger or smaller than CBO anticipates. Additionally, enrollment in Medicare Advantage could be more responsive to changes in payments than the agency expects. CBO anticipates that plans would adapt to payment changes in ways that would preserve the benefits that enrollees value most; thus, in the agency's estimation, enrollment in Medicare Advantage would continue to grow as estimated under current law. Recent evidence suggests that, even when benchmarks have decreased, new and existing Medicare beneficiaries have continued to enroll in Medicare Advantage plans. However, if plans increased premiums or reduced the generosityof benefits in response to lower plan payments by more than CBO anticipates, then enrollment growth in Medicare Advantage could decrease over time. Whether changes in enrollment would increase or decrease savings depends on which beneficiaries disenrolled from or chose not to enroll in Medicare Advantage. If those beneficiaries, on average, cost more in Medicare FFS than they would in Medicare Advantage, then savings would be reduced. Conversely, if those beneficiaries cost more in Medicare Advantage than in Medicare FFS, then savings would increase.

There is an additional source of uncertainty associated with all three alternatives because spending reductions would be affected by the way in which risk scores changed under current law. If, under current law, plans increased the intensity with which they code diagnoses by more than anticipated, savings might grow over time. Conversely, other improvements in risk adjustment, such as changes in the data sources that CMS uses to calculate Medicare Advantage risk scores or improvements in coding accuracy for FFS beneficiaries, could decrease those savings over time by narrowing the gap between the risk scores of Medicare Advantage enrollees and otherwise similar FFS beneficiaries under current law. Estimates for the third alternative would be particularly affected by this source of uncertainty.

Other Effects

The main advantage of all three alternatives is that, in addition to reducing direct federal spending on plan payments, they would bring per capita payments for similar Medicare Advantage and FFS beneficiaries closer to parity. That is, reducing payments to Medicare Advantage plans would increase the likelihood that Medicare would make the same per capita payment for a beneficiary, regardless of whether that person was enrolled in Medicare FFS or Medicare Advantage. A disadvantage of all three alternatives is that insurers might reduce the generosity of the additional benefits that are funded by those additional payments, and some plans might either begin charging a premium or increase their premiums.

An advantage of the first alternative is that it would be easy to implement because it would reduce payments to all plans by the same amount. However, research has shown that coding intensity differs across plans: For instance, plans that have a more direct relationship with providers, such as HMOs, or plans that employ providers directly—that is, vertically integrated insurers—may exert more influence on diagnostic coding patterns. Other types of plans, such as PPOs and PFFS plans, have less influence over providers and therefore may have less influence on diagnostic coding patterns (Geruso and Layton 2018; Hayford and Burns 2018). Additionally, plans that conduct more health risk assessments, have better integrated electronic health records, or offer incentives to providers to code more diagnoses may all have higher risk scores than those that do not. Therefore, a uniform reduction to payments that reflects the average difference between Medicare Advantage and FFS beneficiaries' risk scores might exacerbate inequities in plan payments.

An advantage of the second alternative is that, unlike the first alternative, payment reductions would be scaled to reflect the degree to which plans in a given region coded more aggressively. Scaled reductions would have the benefit of applying lower payment reductions to plans that did not or could not code diagnoses as completely.

A disadvantage of the second alternative is that it would be more complicated for CMS to administer. Further, many of the activities that lead to more comprehensive coding of diagnoses could be desirable in other ways. For instance, diagnoses might be coded more comprehensively in plans that have better electronic health records and more integration with providers. Better integration with providers and more complete use of electronic health records might also improve patients' experiences and streamline the delivery of care. Thus, applying insurer-specific adjustments to risk scores might penalize plans that are engaged in behavior that otherwise would improve patient satisfaction or quality of care. Additionally, the alternative might give insurers incentives to change coding practices for beneficiaries who had recently switched from FFS to Medicare Advantage—that is, insurers might be inclined to delay documenting additional diagnoses until after the first three years of a beneficiary's enrollment.

An advantage of the third alternative is that it would work in part by improving the construction of risk scores rather than simply cutting payments. Using two years of diagnoses would result in conditions being coded more consistently for all Medicare beneficiaries, and thus should more accurately measure health risk among Medicare Advantage enrollees relative to FFS. Further, unlike the first two alternatives, this alternative would specifically discourage the use of health risk assessments primarily to uncover new diagnoses, rather than to define a plan of care for a beneficiary.

A disadvantage of the third alternative is that it would reduce plans' incentives to provide health risk assessments. If plans provided fewer health risk assessments, then they might also fail to detect conditions that might require additional care.