Not Your Grandmother’s Risk Adjustment

The pioneering realm of the ACA methodology Gregory G. Fann

Statements of fact and opinions expressed herein are those of the individual author(s) and are not necessarily those of the Society of Actuaries or the respective authors’ employers.


The largest casino in California, just a few miles from my residence, held its annual fireworks extravaganza on the Saturday after the Independence Day holiday. Having lost any desire we may have once possessed for getting mixed up in large crowds, my wife and I stepped out our front door into the darkness, walked up a hill to an empty field and quietly watched the flashes of light from a distance.

Returning down the hill in the calm aftermath, it occurred to me that the real fireworks of the holiday weekend—emanating from an organization responsible for optimizing risk models—had actually occurred the night before. The Wall Street Journal broke the story: The Centers for Medicare & Medicaid Services (CMS) was expected to suspend Affordable Care Act (ACA) risk-adjustment collections and payments because of a February federal court ruling in New Mexico. The article1 offered some belated clues to health actuaries who were inquisitive about the unexplained rationale of annual risk-adjustment results not being released as expected on June 30, 2018.

Follow-up stories on Saturday2 and Sunday3 largely focused on the negative implications for insurers due to the suspension of a program being administered in a budget-neutral fashion. It was briefly noted that the precipice for all of this activity was that several small insurers had filed lawsuits challenging the program methodology and that some insurers would also naturally benefit from such a suspension. Two industry groups—America’s Health Insurance Plans and the Blue Cross Blue Shield Association—immediately released statements4,5 critical of the legal maneuvering and suggested that significant 2019 rate increases were imminent without quick action and program reinstatement. The immediate public discourse that followed focused on contemplations of alternative legal strategies at CMS’ disposal and an imagined dispute over the efficacy of risk adjustment rather than the substantive discussion of improving the ACA risk-adjustment methodology, which was at the heart of the legal debate and the larger industry discussion over the past three years.

The lawsuit in New Mexico, and others across the country, did not question the general appropriateness of risk adjustment. The plaintiffs alleged that unpredictability and inequities resulting from the ACA methodology were harming some insurers and the market in general—topics that are still being debated today inside and outside of the courtroom. This article explores the history of risk adjustment in the ACA markets. As you will discover, the mechanics are quite unlike other risk-adjustment programs.

Risk-adjustment Basics

As it relates to health care, risk adjustment is a mechanism to normalize the relative health costs of different populations with different risk profiles. Risk adjustment is often used to adjust capitation payments to payers at risk for health care costs, to adjust payments to providers based on different risk levels of their patient panels, to develop appropriate benchmarks for risk-sharing arrangements and to normalize provider quality measurements. It is well-understood that individuals have unique risk characteristics, and appropriate compensation for provision of care or payment responsibility should vary at an individual level. There is not always consensus regarding the specific mechanics required to produce equity and a market that functions well.

In response to the need for risk adjustment, the private marketplace has responded with a multitude of proprietary models that seek to accurately predict costs based on known risk factors. Similarly, the federal government has developed its own models to be used in federally regulated markets.

While largely untested as a premium adjustment mechanism in the commercial insurance markets prior to the ACA, such risk-adjustment models have been well-established in Medicare Advantage and various state Medicaid programs. To the extent that premiums are not aligned with expected costs, these models attempt to adjust premiums to appropriately compensate health plans for the actuarial risk of their respective enrollees. A goal of any risk-adjustment program is to neutralize competition based on favorable selection and to encourage insurers to compete based on the efficiency and value of their plans. Specifically, the ACA aimed to “create market conditions in which insurers’ prices reflect the underlying value and efficiency of their products rather than the composition of their risk pools.”6 Well-constructed risk-adjustment programs foster market stability and predictable results.

While the technical details are rigorous, the general approach is straightforward. Individuals within a fixed population have different risk characteristics. If a government entity is going to pay various insurers to insure a portion of that population with quantifiable risk differences, it is equitable to vary those payments based on the risk of the population that each insurer enrolls.

This exercise becomes more complicated in the commercial markets, where the enrolling population is largely unknown and each insurer independently develops different premium rates. The ACA model is further constrained by the budget-neutral methodology CMS implemented. As no funds are disbursed from the government, any risk-adjustment payments must come from other insurers. This necessitates the development of a complex formula to achieve funding balance.7

The varying levels of ACA net premiums in the individual market, constructed from high gross premiums minus a government subsidy derived from a market benchmark premium and personal income, result in enrollment incentives for some and disincentives for others.8 This creates greater unpredictability in assessing the average expected risk in the marketplace, something of major importance in a budget-neutral framework.

ACA Risk-adjustment Methodology: Rationale, Requirements and Challenges

The ACA disallows enrollee selection and limits rating variables that can be used in the individual and small-group markets. As insurers are not able to select enrollees or appropriately rate for the risks they accept, a risk-adjustment mechanism is intended to appropriately compensate insurers for the risk they enroll. The methodology CMS developed can be thought of as a two-step process. First, each enrollee in the marketplaces is assigned a risk score based on demographics, benefit plan and any identified high-cost health conditions. Second, to account for risk characteristics that cannot be differentiated by premium rates under the market rules, a “transfer payment” methodology is developed to transfer money from insurers that enroll lower-risk enrollees to insurers that enroll higher-risk enrollees.

As insurers are not able to select risks or set prices based on the risks received, they must rely on the CMS methodology for an appropriate and adequate financial accommodation. It is therefore imperative that the operational methodology is precise and impartial, as CMS has assumed accountability for equity among market participants.9 The risk-adjustment process should accurately assess risk based on health status and related predicted claim costs and not be influenced by other factors. A risk assessment model requires both appropriate data and appropriate methodology to function properly. Even with a perfect model of risk assessment, a biased transfer payment formula will have equity problems. The U.S. Department of Health and Human Services’ Notice of Benefit and Payment Parameters (NBPP) for 2019 notes the transfer payment formula intention: “The risk-adjustment transfer formula generally calculates the difference between the revenues required by a plan, based on the health risk of the plan’s enrollees and the revenues that a plan can generate for those enrollees.”10

Developing an equitable risk-adjustment model in the ACA realm is a challenging endeavor, as is projecting a relative risk score for a participating health plan. This is primarily due to the dynamic population, the multitude of factors in the transfer formula and the budget neutrality requirement that necessitates the average risk being determined prospectively by the enrolling population.

The risk-adjustment process should accurately assess risk based on health status and related predicted claim costs and not be influenced by other factors.

A risk-assessment model requires both appropriate data and appropriate methodology to properly function.

The individual market is more fragile due to the underlying incentives for prospective enrollees that are present in the net premium calculations. The market is sensitive to regulatory changes that impact enrollment dynamics. As a recent example, enrollment of individuals with income levels between 200 and 400 percent of the federal poverty level was expected to increase in 2018 due to additional premium subsidies.11

ACA Risk-adjustment Concerns

Many new and small insurers were caught by surprise when CMS informed them of their first annual risk-adjustment charges in 2015 (for 2014). A few were insolvent immediately. Others quickly tried to determine salvage plans. The Consumers for Health Options, Insurance Coverage in Exchanges in States (CHOICES) coalition was formed with “the primary objective of supporting health care consumers by advocating for improvements in the current risk-adjustment program.”12 Made up of mostly new consumer operated and oriented plans (co-ops) from different states, the group began discussing common challenges they were experiencing. The coalition engaged Richard Foster,13 who served as chief actuary of CMS from 1995 through 2012, as a consultant to understand some of the technical issues. CHOICES petitioned CMS to implement “immediate solutions to help alleviate the dramatic consequences these flaws are having on health insurers, including:

  • Exempting new and fast-growing plans from risk adjustment for the first three to five years.
  • Applying a credibility-based approach to participation in risk adjustment.
  • Placing an upper bound on the amount of a plan’s risk-adjustment transfer charge.”14

In response to this and solvency concerns, Maryland’s insurance commissioner endorsed capping risk-adjustment payments at 2 percent of premiums for certain plans and noted that the “[National Association of Insurance Commissioners (NAIC)] has urged CMS to review the formula and work with carriers and state regulators to make adjustments for 2015 and 2016 to ensure it is providing appropriate protection for all carriers, and not wait until 2017 or 2018 to enact reforms.”15

Foster later stated that “the current HHS-HCC risk-adjustment model established by CMS is known to understate risk scores for relatively healthy individuals and to overstate them for those with significant health conditions,”16 and highlighted that “there is a fairly easy way to address this bias in the [risk-adjustment] model that could be used on a practical basis.”17 CMS elected not to address this concern immediately, but acknowledged it in the 2018 NBPP:

“We will not implement any of these approaches for 2018, but will consider changes in future years … we are still evaluating the tradeoffs that would need to be made in model predictive power among subgroups of enrollees. We continue to focus on encouraging plans to attract high-risk enrollees through the risk-adjustment model, but agree with commenters that we should further evaluate solutions prior to making any adjustments to the model.”18

While the equity concerns were largely voiced by small, growing plans, some established health plans also noted the inequities. Anthem’s chief financial officer stated that the risk-adjustment methodology “overcharges for healthy (members) and over-reimburses for certain moderately unhealthy disease states.”19 Compounding the risk assessment inequities, transfer payments are often magnified as the nature of the statewide average premium formula “severely exaggerates risk transfers for efficient insurers by mandating an inflated transfer amount relative to their cost structure.”20

The expressed concerns and solvency challenges ultimately led three of the health plans affiliated with CHOICES to file separate lawsuits in U.S. district courts challenging the ACA risk-adjustment methodology. They allege the formula includes several items not associated with actuarial risk and resulted in inequitable advantages for established and more expensive plans. A key argument was that the transfer formula is based on the state’s average premium, effectively calculating higher-than-appropriate payments from lower-cost, lower-premium plans.

The case in New Mexico is the only one of the three that has received a verdict partially favorable to the plaintiff. The 2018 ruling effectively suspended the risk-adjustment collections and payments for effective years 2017 and 2018.21 While the case remains outstanding (with CMS asking that the ruling be reconsidered), additional federal regulations already have been issued. A final rule was issued for 2017 (which allowed payments to be reinstated), and a proposed rule was issued for 2018. Both rules seek to clarify the language regarding the rationale of CMS in selecting the statewide average premium methodology, which was found to be “arbitrary and capricious.”22 The 2017 rule was challenged for not allowing public comments.23 Comments on the 2018 proposed rule have been submitted, and a final rule has not been issued as of November 2018.

The New Mexico plaintiff’s attorney views the original ruling as an opportunity for productive discussion. He said: “This gives a real opening to CMS and the industry to really come up with a better solution and a better path going forward that will help consumers and the ACA. That’s our end game. We want to make the ACA work better and want a risk-adjustment formula that works right.”24

CMS Response

The concerns expressed with the risk-adjustment methodology did not go unnoticed. In March 2016, CMS released a discussion paper25 and facilitated an industry conference to discuss the ongoing concerns. Many of the specific items were addressed in the 2018 NBPP,26 resulting in improvements in the risk-adjustment methodology.27 Other concerns were not immediately addressed and continue to be discussed.

The 2019 NBPP, the first authored by the Trump administration, largely continued implementation of the risk-adjustment enhancements in the 2018 rule. A new notable allowance for states is an opportunity to request up to a 50 percent reduction in risk-adjustment transfer payments.

Industry Perspectives

In April 2016, the American Academy of Actuaries’ (the Academy’s) Risk Sharing Subcommittee reviewed the first year (2014) performance of the risk-adjustment methodology in the individual market.28 The summary finding was that the risk-adjustment methodology operated in a directionally correct way as it “compressed the loss ratio differences among insurers.”29 An unavoidable data deficiency of this loss ratio analysis is the reliance on premiums in the loss ratio equation. Premium levels were sporadic and generally inadequate in the ACA marketplace’s initial year. Industry studies have demonstrated that premium inadequacy continued in 2015 and 2016.30 A similar analysis using 2017 experience will have the same inherent limitation, but it should reflect more accurate pricing and not have the complication of the presence of federal reinsurance.

In 2015, the Health Section Council of the Society of Actuaries (SOA) launched a new committee to identify areas worthy of focused research to supplement the education of health actuaries. One of the first two initiatives was a focused study on the ACA markets. The group of volunteers completed a series of papers late in 2016. Interestingly, a variety of current complications with the ACA were documented31 in actuarial literature at that time, but the volunteer group elected to focus its series of papers exclusively on ACA risk adjustment and not report the other concerns, highlighting the overwhelming impact and associated prominence that risk adjustment has on market results.32

The authors of this series were intentionally selected to represent different types of organizations and provide unique experiences with the ACA: co-ops, Blue Cross Blue Shield organizations, consulting firms, smaller health plans and large national payers. Notably, their themes were very similar and mirrored actuarial comments related to the 2018 proposed NBPP regulation,33 focusing on the statewide average premium methodology, HCC scoring inequities, volatility concerns and disadvantages for low-cost insurers that effectively manage care.

Mitigation

State regulators also struggled with comprehension of risk-adjustment results as solvency concerns arose with little warning. In 2016, the state of New York released emergency regulation to “stabilize” the impact of the ACA risk-adjustment methodology in the small-group marketplace. The superintendent of insurance “expressed concern that the CMS risk-adjustment program has created inappropriately disparate impacts and unintended consequences among health insurers in New York.”34 The regulation allowed the superintendent to reduce a percentage of the payments in the risk-adjustment formula. Along the same lines, CMS proposed allowing states to request to reduce risk adjustments by 50 percent in the 2019 NBPP. These “dilution” methods remove some of the shock of risk-adjustment payments, but they are regarded as crude adjustments by some actuaries who believe “it is unlikely that the application of a flat percentage reduction to the transfer amounts would produce equitable outcomes for all the issuers in a state.”35

Other states (e.g., Maryland) have explored mechanisms of capping risk adjustment at a fixed percentage of an insurer’s premium. This adds some predictability to the “risk-adjustment charge” in the pricing process, but it requires a balancing adjustment if the intent is to preserve a budget-neutral methodology; detractors argue insurers may be incented to avoid high-risk enrollees if they believe their risk-adjustment liability is limited. This capping mechanism can be structured in many ways, and actuarial models have been developed to analyze the opportunities and risks of this approach.

Some model enhancements have been suggested to align the risk-adjustment methodology with the program intent. As one author notes in the SOA’s series of papers: “The risk-adjustment transfer formula does not account for plans that have more advanced approaches to care management or more progressive value-based contracting methods that help drive premiums lower in a market. The transfer formula penalizes plans with lower premiums and rewards those with higher premiums in relation to the statewide average premium.”36 A proposed solution is a formula enhancement that “will remove the health plan’s perverse incentive and will motivate more issuers to improve their [care management effectiveness] performance and reduce the related cost of health care.”37

As more ACA market experience becomes available, actuaries will utilize this data to validate the effectiveness of these and other mitigation efforts. More important, robust data can be used to inform continued improvements to the ACA risk-adjustment methodology.

Conclusion

The nature of the ACA market is challenging in many respects. Risk adjustment is no exception. CMS developed an intricate model to maintain budget neutrality and sought to transfer payments from insurers that enrolled low-risk individuals to insurers that enrolled high-risk individuals in an actuarially fair manner.

While initial analyses indicate some positive directional changes, stakeholders have raised concerns about the equity and predictability of the results disseminated from the current model. Some improvements have been implemented, but some of the most troubling elements of the methodology (as expressed by some market segments) remain intact. Arguably, this is a barrier to new entrants in ACA markets, many of which suffer from limited competition.

Successful risk-adjustment models foster predictability and eliminate incentives for enrollee selection based on specific health conditions. They equitably adjust premium levels to reflect the health status or actuarial risk of an enrolled population. They provide impartial treatment for all health plans and do not offer advantages based on size, growth patterns, breadth of network, efficiencies, medical management or cost structure. Many stakeholders believe the current risk-adjustment model results in imbalanced assessments that penalize the types of insurers and enrollees the ACA seeks to attract. Improvements are being made, and actuarial analysis of emerging experience data likely will facilitate further improvements.

In today’s high-cost environment, it is imperative that programs such as risk adjustment foster an environment where insurers that offer efficient, quality coverage can participate without unnecessary volatility or the risk of being disadvantaged. Admittedly, that’s not easy to do. If we can make it happen, it will be a story worth telling our grandchildren.

Gregory G. Fann, FSA, FCA, MAAA, is a senior consulting actuary at Axene Health Partners LLC in Temecula, California.

Copyright © 2018 by the Society of Actuaries, Schaumburg, Illinois.