NBER Working Paper 24038; Service-level Selection: Strategic Risk Selection in Medicare Advantage in Response to Risk Adjustment

By Sungchul Park, Anirban Basu, Norma Coe, and Fahad Khalil
National Bureau of Economic Research, November 2017

Abstract

The Centers for Medicare and Medicaid Services (CMS) has phased in the Hierarchical Condition Categories (HCC) risk adjustment model during 2004-2006 to more accurately estimate capitated payments to Medicare Advantage (MA) plans to reflect each beneficiary’s health status. However, it is debatable whether the CMS-HCC model has led to strategic evolutions of risk selection. We examine the competing claims and analyze the risk selection behavior of MA plans in response to the CMS-HCC model. We find that the CMS-HCC model reduced the phenomenon that MA plans avoid high-cost beneficiaries in traditional Medicare plans, whereas it led to increased disenrollment of high-cost beneficiaries, conditional on illness severity, from MA plans. We explain this phenomenon in relation to service-level selection. First, we show that MA plans have incentives to effectuate risk selection via service-level selection, by lowering coverage levels for services that are more likely to be used by beneficiaries who could be unprofitable under the CMS-HCC model. Then, we empirically test our theoretical prediction that compared to the pre-implementation period (2001-2003), MA plans have raised copayments disproportionately more for services needed by unprofitable beneficiaries than for other services in the post-implementation period (2007-2009). This induced unprofitable beneficiaries to voluntarily dis-enroll from their MA plans. Further evidence supporting this selection mechanism is that those dissatisfied with out-of-pocket costs were more likely to dis-enroll from MA plans. We estimate that such strategic behavior led MA plans to save $5.2 billion by transferring the costs to the federal government.

From the Introduction

Recent health care reforms have facilitated the transition from volume- to value-based payment models in hopes of achieving cost control and enhancing the quality of care. One such example is that the Centers for Medicare and Medicaid Services (CMS) reimburses Medicare Advantage (MA) plans with a capitated amount per beneficiary to encourage coordinated care in managed care settings. It has been shown that MA plans save money without sacrificing quality. However, as an unintended consequence, MA plans have selectively enrolled healthier people to receive overpayments, known as favorable selection. It remains inconclusive whether cost savings are attributable to cost-effective care management or to risk selection. To reduce risk selection, CMS has adjusted payments to MA plans to reflect the health status of their enrollees, a process known as risk adjustment. To more accurately estimate capitated payments, in 2004, CMS introduced a new risk adjustment model—the CMS-Hierarchical Condition Categories (HCC) model—which uses extensive inpatient and outpatient diagnostic information from the prior year to generate risk scores.

It is debatable whether the CMS-HCC model has been effective in reducing risk selection or whether it has led to strategic evolutions of risk selection. On one hand, it has been shown that the CMS-HCC model considerably reduced the phenomenon of avoiding sicker beneficiaries (i.e., those with high-risk scores) in traditional Medicare (TM) plans. As risk adjustment leads to neutral payments for beneficiaries with conditions included in the risk adjustment formula, MA plans no longer have incentive to avoid those with high-risk scores if their conditions are included in the CMS-HCC model.

On the other hand, there is suggestive evidence showing that MA plans could strategically respond to the CMS-HCC model. Brown et al. (2014) argue that the HCC model merely shifted the profitable population from healthy people (i.e., those with low-risk scores) to sick ones who are over-compensated, within their risk-score. This can be achieved because, first, there is considerable variability in actual expenditures of beneficiaries around their risk-adjusted payments. For all beneficiaries with a given health condition, the CMS-HCC model is designed to adjust payments to MA plans by the same rate. However, the severity of the condition and thus the cost of treating it can vary within a given condition (Medicare Payment Advisory Commission). Second, the variability of the within-risk-score expenditures is larger for those with higher risk scores. Because the CMS-HCC model only accounts for about 100 major conditions, this generates underpayments for those whose conditions are not accurately measured by the model, who tend to have multiple chronic conditions.

We examine these competing claims and analyze the risk selection behavior of MA plans in response to the CMS-HCC model. Specifically, we hypothesize that MA plans engage in service-level selection, in which they provide relatively lower coverage levels for some services to discourage enrollment of certain beneficiaries. While the CMS-HCC model encouraged MA plans to accept TM beneficiaries with high-risk scores, we claim that MA plans strategically behave to avoid beneficiaries who could be unprofitable under the CMS-HCC model (i.e., those with higher expenditures than their risk-adjusted payments). Although there is a large literature on investigating service-level selection as a risk selection strategy, to the best of our knowledge, there is no research explaining mechanisms through which MA plans could engage in service-level selection as a strategic risk selection behavior in response to risk adjustment.

We find that the CMS-HCC model achieved the goal of reducing favorable selection based on health; however, we also find that it led to increased disenrollment of unprofitable beneficiaries from MA plans, leading MA plans to save costs of $5.2 billion in 2007-2009. We explain this phenomenon via service-level selection. Building upon Ellis and McGuire (2007), we theoretically show that MA plans have incentives to effectuate risk selection through service-level selection, as unprofitable beneficiaries are more likely to use services that are expensive and are thus more vulnerable to under provision by MA plans. Specifically, we show that those with higher expenditures than their risk-adjusted payments are more likely to use services that health plans would ration more tightly (i.e., services with higher service-level selection index). This phenomenon is more likely to be pronounced for those with higher risk scores. Then, we find evidence supporting our theoretical prediction that MA plans have actually raised copayments disproportionately more for services with higher service-level selection index (i.e., ambulance, home health service, partial hospitalization, and inpatient hospital service) than services with lower service-level selection index (i.e., outpatient substance abuse services, outpatient X-rays, and outpatient hospital services). Such disproportionate increases in copayment induced unprofitable beneficiaries to voluntarily disenroll from MA plans. In additional analyses, we find evidence supporting this selection mechanism that those with dissatisfaction with out-of-pocket costs were more likely to disenroll from MA plans. Consequently, the variation of total Medicare expenditures for MA enrollees with high-risk scores reduced over time. These findings indicate that service-level selection allowed MA plans to avoid the risk of enrolling unprofitable beneficiaries.

Service-level selection

By Law, MA plans are not allowed to deny coverage based on beneficiaries’ health status. However, MA plans might practice risk selection in subtle ways so that unprofitable beneficiaries voluntarily disenroll from MA plans. To achieve such risk selection, MA plans could risk-select through collecting additional data or advertising. However, because such selection mechanisms would lead to substantial screening costs, MA plans are likely to seek screening approaches with lowest costs. One such approach that generates relatively low screening costs is service-level selection because MA plans do not need to predict each beneficiary’s expenditures but rather only need to predict services more likely used by beneficiaries who could be unprofitable under the CMS-HCC model.

Service-level selection is one type of risk selection, which is based on the phenomenon that unprofitable individuals are more likely to use services that are expensive to health plans subject to capitated payments and are thus more vulnerable to under-provision by health plans. As with risk selection, service-level selection occurs due to asymmetric information between two parties, in which health plans do not know individuals’ private information about health status and preferences for health care. The health plan only knows the probability of using the service at the population level, while the individual knows her need, or probability of need, for each health care service and chooses the best health plan that can satisfy her need. Since rational individuals respond to health plan design when selecting plans, reducing coverage levels for services related to financial losses (i.e., services more likely used by unprofitable individuals) would induce unprofitable individuals to voluntarily disenroll from the plan. In this way, service-level selection would allow health plans to reduce the scope of enrolling those who could be costly to them. Although unprofitable individuals enroll in the plan, service-level selection would also enable health plans to reduce their financial loss as they shift the costs to the individual.

From the Discussion and Conclusion

The goal of this paper is to shed light on the competing claims on the effectiveness of the CMS-HCC model and to comprehensively understand strategic risk selection behaviors of MA plans. We find that the CMS-HCC model reduced the phenomenon that MA plans avoid beneficiaries with high-risk scores in TM plans, whereas it led to increased disenrollment of high-cost beneficiaries, conditional on risk score, in MA plans. We explain this phenomenon through service-level selection. Through theoretical and empirical analysis, we show that after the full phase-in period of the CMS-HCC model, MA plans have the incentive to and did increase copayments disproportionately more for services that appeal to beneficiaries who could be unprofitable under the CMS-HCC model than other services. The disproportionate changes in copayments led to voluntary disenrollment of beneficiaries with need for these services, who tend to incur higher expenditures than their risk-adjusted payments. We also find evidence supporting our hypothesis that those who were less satisfied with out-of-pocket costs were more likely to disenroll from MA plans. Such strategic behavior led to MA plans to save $5.2 billion in 2007-2009 by simply transferring the costs to the federal government, thereby placing significant financial burdens on the federal government.

Findings from this study indicate that the CMS-HCC model reduced the MA plans’ risk selection of avoiding TM beneficiaries with high-risk scores, whereas it induced MA plans to strategically behave in response to the CMS-HCC model via service-level selection. MA plans have raised copayments disproportionately more for services needed by high-need beneficiaries than for other services, thereby inducing unprofitable beneficiaries to voluntarily disenroll from their MA plans, mainly due to increased out-of-pocket costs. This allows MA plans to avoid the risk of enrolling unprofitable beneficiaries.

http://www.nber.org…

Since private Medicare Advantage plans initially were paid a flat capitation rate for each enrollee, they selectively marketed their plans to healthier Medicare beneficiaries and that was successful in increasing their profits since their patients were healthier than those enrolled in the traditional Medicare program. To counter this, the government established the Hierarchical Condition Categories (HCC) risk adjustment model so that capitation payments were were made based on the health status of each beneficiary.

To no surprise, the plans then upcoded their diagnoses so that they received capitation rates that were higher than the actual level of care required. It was suspected that the plans were also gaming the system through service-level selection, and this study confirms that they were. Specifically they determined which services would be used by high-cost beneficiaries, and then they disproportionately increased copayments for those services. The dissatisfaction caused by the high out-of-pocket costs for individuals using these services caused them to voluntarily disenroll from the Medicare Advantage plans.

Although the insurers had already been paid at higher capitation rates for this population, they were able to avoid paying for those individuals with costs even higher than the capitation payments, by this devious scheme of shifting more costs to these patients, thus causing them to disenroll. The authors calculated that this shifted $5.2 billion in costs from the insurers to the federal government, that is, to us the taxpayers.

Ironically the system of capitation payments supposedly represents a transition from volume- to value-based payments, yet the insurers have figured out ways to decrease value instead by selectively marketing to the healthy, by upcoding to qualify for higher risk-adjusted payments, and now by chasing away patients who need more care by penalizing them for accessing services normally used by patients with more serious disorders (service-level selection).

How much longer do we put up with this? Throwing these crooks out and replacing their plans with our own single payer national health program would fix this problem right now. What’s holding us back?

Stay informed! Visit www.pnhp.org/qotd to sign up for daily email updates.