How Changes to Health Insurance Market Rules Would Affect Risk Adjustment

ACA risk adjustment was designed to be revenue neutral by state and market. Payment transfers are calculated separately for the individual and small group markets, except for states that merged the two markets. Within each state and market, total payments from insurers with a relatively healthier population are set equal to total payments to insurers with a relatively sicker population. No external funding is provided. Since risk adjustment only transfers funds among insurers, it can compensate for relative health status differences among insurers, but it does not ensure that overall market premiums are sufficient to cover the average claims within the state if the market as a whole has a worse-than-expected risk profile.
 
The ACA risk adjustment program includes two key components: (1) risk adjustment models to assess the relative risk for each individual; risk scores are combined for all of the plan’s enrollees to determine a weighted average risk score for each plan, and (2) a payment transfer formula, which compares the plan’s risk to the market average and adjusts for allowed rating variations and the statewide average premium. The risk adjustment models use demographic data such as age and gender, and diagnosis data from medical claims (and from prescription drug claims starting in 2018) to assign risk scores to each enrollee.
 
The ACA risk adjustment models and payment transfer formula support the single risk pool rating requirement that premiums for plans of different metal levels vary only due to benefit differences and not the risk of the population choosing plans in the metal level. Risk adjustment payment transfers occur between metal levels to compensate for the tendency of higher-cost enrollees to choose richer benefit plans. In order to accomplish this goal, the risk adjustment models provide risk factors by metal level and the payment transfer formula includes adjustments for metal level. Changes to the initial risk adjustment program are being made so that it better reflects differences in the underlying risk among participating insurers. These modifications include the incorporation of prescription drug data, the incorporation of preventive services, and better accounting for partial-year enrollees.¹
 
States have the flexibility to establish a state-run risk adjustment program or allow the Center for Medicare & Medicaid Services (CMS) to administer the risk adjustment program. The Commonwealth of Massachusetts was the only state to establish its own risk adjustment program, but it subsequently elected to have CMS administer the program. In 2017, CMS is administering risk adjustment for all states.
 

Implications for Risk Adjustment of Potential Changes to Current Laws and Regulations

Several types of changes to the health insurance market rules are being considered. These potential changes have various implications on the need for and the design and administration of risk adjustment.

Loosening Issue and Rating Rules
Prior to the implementation of the ACA, health plans in the individual market traditionally used individual medical underwriting to assess an applicant’s health status and charged premiums to reflect an individual’s underlying risk. The ACA eliminated the use of individual underwriting by requiring guaranteed issue; prohibiting pre-existing condition exclusions and premium variations by health status; and limiting premium variations by age, smoking status, and family size.
 
As long as guaranteed issue and rating restrictions are in effect, some form of risk adjustment is necessary to provide a level playing field among insurers. Otherwise, insurers would have incentives to structure benefits and marketing practices to avoid enrolling high-cost individuals.
 
If guaranteed issue and rating requirements are loosened or eliminated in the individual market, the need for risk adjustment decreases. However, if only rating restrictions are loosened but guaranteed issue remains, then risk adjustment would still be needed if there are limits on how high premiums could be increased to reflect health status. Otherwise, insurers would have incentives to avoid enrolling high-cost individuals.
 
Another scenario is to continue guaranteed issue and rating restrictions for individuals meeting continuous coverage requirements, but allow underwriting, pre-existing condition exclusions, or broader premium variations for those without continuous coverage. Risk adjustment would continue to be needed in this case. Without risk adjustment, insurers would be incented to structure benefits, provider networks, and marketing practices to avoid enrolling unhealthy people who were continuously covered who wanted to switch plans, especially if they desire to move to more generous coverage.
 
Whether there would be separate risk pools for enrollees who are underwritten and those who are not would have implications for this scenario. Risk adjustment is designed to address differences in risk among insurers within a risk pool. Risk adjustment does not address risk differences between risk pools. For instance, currently, grandfathered and transitional (grandmothered) policies are not included in the risk adjustment pool and these blocks of business have premium rates set based on their claims experience. If healthy individuals can get preferred rates by undergoing underwriting, even if they were continuously covered, or if non-ACA compliant plans are opened up to new enrollees, healthier individuals would choose to undergo underwriting or enroll in non-ACA compliant plans with fewer benefits or looser rating requirements. Less-healthy individuals would choose community rated ACA-compliant plans. As a result, less-healthy individuals would face higher premiums, even if they were continuously covered, potentially destabilizing the community rated ACA-compliant market without risk adjustment transfers between the markets to account for such selection differences.
 
The small group market differs from the individual market in a few areas related to the need for risk adjustment. For instance, the small group market was already guaranteed issue prior to the introduction of the ACA. And compared to the individual market, the pre-ACA small group market had less underwriting and more limited rating flexibility, but still more than is allowed under current law. Risk adjustment was not used in the small group market prior to the ACA and the markets were competitive in most states. Risk adjustment is less needed in the small group market because there is typically less variation in health status across groups of individuals than there is within the individual market, and active employees are typically healthier than those without access to employer coverage. In addition, the employer generally chooses the plan for the employees and provides premium subsidies to encourage enrollment, leading to less adverse selection and a more balanced risk pool. Since there is less variation in health status in the small group market, risk adjustment transfers as a percent of premium are lower in the small group market than in the individual market. For the 2015 risk adjustment results, the absolute value of risk adjustment transfers was 6 percent of premium for the small group market and 10 percent of premium for the individual market.²
 
If the rating rules in the small group market are changed to permit more variation in premium to reflect underlying risk, including health status, the need for risk adjustment would be lessened or eliminated. However, if premium variation restrictions remain, risk adjustment could still be helpful to reduce incentives for insurers to avoid high-cost groups. If small group rating rules are changed to be different than individual market rating rules, the risk adjustment methodology for the small group market would need to differ from that in the individual market. In addition, if states are given flexibility to change small group rating rules, the risk adjustment changes may need to vary by state.
 
Incorporating High-Risk Pooling
High-risk pools are being considered as a potential mechanism to provide coverage to individuals with pre-existing conditions and to lower premiums and stabilize the individual health insurance market. There are different ways to structure high-risk pools. One way is a traditional high-risk pool approach. Prior to the ACA, many states used traditional high-risk pools to provide coverage in a separately run insurance pool to individuals who were not able to get insurance due to pre-existing health conditions. In addition, to create a bridge to guaranteed issue coverage in 2014, the ACA established the Pre-Existing Condition Insurance Plan (PCIP) program, under which local or federally run high-risk pools would be created in every state. Another high-risk pool approach is to use “invisible” risk pools, where enrollees remain in the individual market, but all or a portion of their claims are reimbursed by the high-risk pool. Invisible high-risk pools are typically characterized as determining eligibility based on conditions. Alaska’s program, for instance, provides payments to insurers for individual enrollees who have one or more of 33 identified high-risk conditions. Reinsurance is a third approach, which is similar to invisible high-risk pools in that enrollees would remain in the individual market. But rather than being condition based, payments to plans for high-risk enrollees would be based on claims exceeding a specific dollar threshold. The ACA’s transitional reinsurance program followed this approach. Funding high-cost claims in the individual market through external sources would result in lower premiums.
 
Even with high-risk pool programs, risk adjustment would still be necessary to address differences in relative enrollee risks among insurers.³ If a high-risk pool program is in place, it should be coordinated with the risk adjustment program, otherwise insurers would be compensated twice for the same risk. For instance, if a portion of claims is covered by an invisible high-risk pool or reinsurance program, then the risk adjustment program only needs to consider the portion of claims not covered. For an invisible high-risk pool program based on conditions, the medical conditions used to identify individuals eligible for the high-risk pool may or may not be the same ones used to identify the risk score for the risk adjustment program, which could introduce complications in coordinating the two programs. In addition, identifying individuals eligible for an invisible high-risk pool or reinsurance program and risk adjusting the portion of claims not covered by such a program would be complex if the essential health benefits (EHB) requirements are loosened and benefits aren’t consistent among plans.
 
As noted above, the individual market risk adjustment program is revenue neutral; money is transferred among participating issuers. In contrast, high-risk pool programs typically require external funding. If external funding is provided, the level of risk adjustment transfers among issuers would decrease since a portion of the high-cost claims would be covered by the external funding. Nevertheless, the risk adjustment program itself would remain revenue neutral.
 
Increased Flexibility in Cost-Sharing Levels
ACA-compliant plans must meet actuarial value (AV) requirements conforming to different metal level tiers as measured by the CMS AV calculator (e.g., 60%, 70%, 80%, 90%); risk adjustment risk scores and transfer payments reflect how spending differs by metal level. Loosening the AV requirements to give insurers more flexibility or eliminating the AV requirements altogether would necessitate adjustments to the risk adjustment model. As long as plans continue to cover EHBs, differences in plan levels could still be estimated using the AV calculator and incorporated into the risk adjustment methodology. The model would likely need to be re-calibrated, however, and plan levels could be incorporated into the risk adjustment methodology in one of two ways. One approach to assigning plan liability risk scores would be to develop risk models by various ranges of the AVs, consistent with how it is done today. Another approach would be for CMS to develop a single model that uses AV as an input variable and bases the plan liability risk scores on actual AV as well as demographics, enrollment duration, and medical and pharmacy diagnoses.
 
Lower-risk individuals tend to select leaner plans and higher risk people tend to select richer plans. Because risk adjustment transfers have to be incorporated into premiums, leaner plans need to set premiums higher than what’s needed to cover their claims and richer plans set premiums lower than what would be needed to cover their claims. If AV requirements are loosened to allow insurers to sell plans with AVs lower than 60 percent, risk adjustment could still work to mitigate the selection effects between plan levels. However, the larger differences in relative plan generosity plans would lead to more selection risk, which would increase the risk adjustment transfers across AV levels.

Loosening or Eiminating Essential Health Benefit Requirements
Insurers are required to cover a specific set of federally defined essential health benefits. Those requirements could potentially be loosened or eliminated at the federal level, thus leaving it to states to set benefit requirements. If insurers are still required to cover the major EHB categories, but are allowed more flexibility regarding the scope of covered services within each category, the risk adjustment methodology may not need major changes. However, if health plans are allowed to offer major categories of benefits, such as prescription drugs, mental health/substance abuse benefits, or maternity benefits, as an optional benefit or have the ability to place internal limits on them, enrollees needing the benefits will enroll in plans covering those benefits without limitations. Depending on the type of optional benefit, these enrollees may be higher risk overall. Under this kind of benefit design flexibility, insurers would face increased selection risk and it would be difficult to develop a risk adjustment methodology that would adequately compensate for the resulting risk differences among plans.
 
In particular, varying benefits among plans would complicate the task of developing risk model factors. The current ACA risk adjustment models predict the expected cost of conditions assuming uniform EHB coverage. The risk adjustment methodology would need to be integrated with the rating rules, which could treat optional benefits in one of two ways—(1) the optional benefits are paid entirely by the enrollees electing the benefits (i.e., the full impact of selection is borne by those opting for the benefit), or (2) the benefits are rated according to the premium differential that would be required if the entire risk pool selected the benefit (i.e., the impact of selection is spread across all enrollees). The first case would lead to higher premium differentials for the optional benefit and the second case would have lower premium differentials but require larger risk adjustment transfers.
 
Under the first case, risk adjustment could be structured to apply only to the required set of EHBs, with the risk factors recalibrated accordingly. However, the risk models may not be precise enough to fully reflect the selection costs among plans with and without the optional benefits, especially if the optional benefits are prescription drug or mental health benefits. Enrollees selecting the plans with these optional benefits may have higher severity for the same conditions. In other words, enrollees opting for prescription drug benefits could have not only higher prescription drug spending than enrollees passing up drug coverage, but also higher medical spending even given the same health condition.
 
Under the second case, plans with and without the optional benefits would be risk adjusted together, but there would be complications in developing the risk scores. For example, if one health plan covers prescription drugs and another does not, the expected cost for an enrollee with a condition requiring expensive prescription drugs will be different for the health plans and would therefore require different risk scores.
 
Eliminating significant benefit categories, such as mental health, from essential health benefits may create additional problems in risk adjustment methodologies. When benefits like mental health are not covered, certain diagnoses will not be submitted and risk scores will not fully reflect major conditions that people have. Plans could decide to employ strategies to cover some of these service categories when the additional reimbursement from risk scores is expected to outweigh the costs and adverse selection of covering the conditions, especially if internal limits are imposed. These types of strategies could disadvantage certain plans without necessarily improving the quality of care delivered to enrollees. Therefore, the risk adjustment methodology should be calibrated to the essential health benefits by considering the impact of materially different essential health benefits requirements not only on risk weights, but also on the mapping of condition categories.
 
In summary, increased flexibility in benefit designs could increase the need of risk adjustment but simultaneously make the implementation of risk adjustment more challenging.
 
Allowing Insurance Sales Across State Lines
Allowing insurance sales across state lines has been proposed as a way to increase insurer competition and consumer choice. Such a policy change would also have implications for risk adjustment. As discussed above, risk adjustment occurs at the state and market level and is coordinated with the market’s rating rules. The risk adjustment methodology adjusts for expected claims costs that cannot be included in the rating structure. For example, if the state has a 3:1 limit on age rating, the payment transfer calculation assumes that younger enrollees are paying lower premiums than older enrollees based on the 3:1 age rating curve, whereas if the state is fully community rated, the payment transfer formula assumes younger enrollees pay the same premium as older enrollees. Under full community rating, risk adjustment transfers are larger to reflect larger differences in health costs between young and older enrollees that are not accounted for in premium differences.
 
For risk adjustment to work as intended, the methodology must be consistent with the applicable market rules and how the risk pool is defined. In other words, if people are allowed to buy coverage from an insurer licensed in another state and that coverage abides by the licensing state’s market rules, the out-of-state enrollees need to be part of the licensing state’s risk pool, as opposed to the risk pool in the enrollees’ home state. For example, if an insurance company is licensed in Nebraska and enrolls a resident of New York under Nebraska’s rating rules, the enrollee would be need to be part of the Nebraska risk pool for risk adjustment.
 
Under this scenario, insurers would rate the out-of-state enrollees as part of their licensing state’s single risk pool, using their licensing state’s rating rules and average morbidity, adjusted for the cost of care in the residence state. The state’s average morbidity would reflect the morbidity of both in-state and out-of-state enrollees; out-of-state enrollees would also affect other state average factors that are used in the state’s risk adjustment payment transfer calculation. As a result, all insurers in the state would be affected, not just those with out-of-state enrollees. Premium rates need to reflect expected risk adjustment transfers, but developing these premium rates would be more difficult in both the enrollee’s residence state and the insurance company’s licensing state, since each state’s risk pool would no longer include all and only enrollees who reside in that state. The risk pool in each state could change considerably over time and could create extreme volatility in risk adjustment transfers and financial results.
 
Another complicating factor relates to geographic rating factors. The risk adjustment payment transfer formula currently calculates transfers by geographic rating area within the state and includes geographic cost factors to adjust for how the costs of care vary by rating area. It is unclear how additional geographic areas outside of the state could be handled in risk adjustment for out-of-state enrollees.

If out-of-state enrollees are required to be in the risk pool of the enrollee’s state of residence, then the rating rules and benefit requirements of the state of residence should also apply for risk adjustment to work as intended. Under this scenario, out-of-state insurers would be required to develop premium rates for each state they intend to sell in based on each state’s benefit and rating requirements and average morbidity. This option wouldn’t lead to shifts in risk profiles among states because insurers would be selling under the same rules in each geographic market. However, premiums will be more similar among insurers selling in a geographic market because out-of-state insurers will not be able to offer different benefit packages or premiums using different rating rules.

Impact of State Variation
Beyond the issue of selling across state lines, increased state variation in benefit or rating rules would affect the risk adjustment program. The ACA made the regulatory environment more consistent among states. Prior to the ACA, there was much more variation in underwriting, rating, and benefit rules. The current risk adjustment methodology uses factors that do not vary by state or market, although the risk adjustment transfers are calculated within each state and market. The use of a national risk adjustment model is possible because of the similar market rules across states.

If states have the ability to tailor their market rules and risk sharing programs in the future, they may need to also modify the risk adjustment model and/or methodology. For example, if states have the ability to design their own high-risk pooling mechanism, they may need to modify the risk adjustment models and transfer payment formula to coordinate with the parameters of the high-risk pool program so that the programs do not both compensate for the same risk.

If similar market rules continue to exist across states, a national risk adjustment methodology may continue to be appropriate, but it may also be possible to incorporate some state flexibility to reflect local utilization and spending patterns. Additionally, it may be desirable to develop risk adjustment models based on regional or state experience, to better fit a state’s experience.

Operationally, it would be relatively straightforward for CMS to allow states to vary certain elements of the risk adjustment models and transfer payment formula, and still have CMS administer the risk adjustment program. This approach would involve identifying the elements of the risk adjustment methodology that could be easily varied and allowing states to use actuarially supported inputs that differ from the national model. It would be more difficult for CMS to allow states to modify the fundamental structure of the methodology (for example, changing the set of diagnoses used to develop risk scores) and still have CMS administer the program.

A state could choose to implement non-zero sum risk adjustment by dedicating external funding to the risk adjustment mechanism. This could limit or decrease the amounts paid into risk adjustment by insurers with low-risk enrollees, while still compensating insurers of high-risk enrollees at full value based on the risk adjustment model. This option would involve modifying the current risk transfer formula without affecting the fundamental risk scoring methodology.

State variation of market rules or of risk adjustment methodology would make it more administratively complicated and less cost effective for multistate insurers. Similar to issues regarding the CMS administration of the risk adjustment program, different risk weights and parameters within the risk adjustment methodology would have less of an effect on multistate issuers’ operations than fundamental differences in the structure of the methodology across states, such as using different diagnoses to develop risk scores.

Administrative Considerations
Risk adjustment administration is very complicated. As noted above, the risk adjustment program is administered by CMS in all states. CMS defines the risk adjustment methodology, updates the models, and incorporates ongoing improvements. In order to administer risk adjustment, CMS developed the data gathering methodology, set up the infrastructure, performs the transfer calculations, and performs data validation. If states were required to administer risk adjustment, it would require significant resources to build the models and the data collection and payment transfer capabilities. Flexibility would be gained, but economies of scale would be lost.
 

Conclusion

Potential changes to the rules applying to the health insurance market have various implications for risk adjustment. Short of returning to the pre-ACA environment of underwriting and risk rating, risk adjustment would still be necessary to reduce incentives for insurers to avoid high-cost enrollees. Some changes, such as incorporating high-risk pooling and increasing flexibility in cost-sharing requirements, could require only adjustments to the risk adjustment design. Other changes, such as loosening or eliminating the EHB requirements and allowing sales across state lines could greatly complicate the design and effectiveness of a risk adjustment mechanism. If states have flexibility in setting benefit and rating rules, the risk adjustment models and payment transfer factors may need to vary by state. The administration of risk adjustment is complicated. Moving administration of a risk adjustment program from CMS to the states would require a significant investment of state resources.