Risk Adjustment in ACA Marketplaces: The 2023 Final Rule

May 26, 2022 | Policy, Affordable Care Act, Risk Adjustment

Introduction

In a prior blog, we provided background on how the RA program works and highlighted relevant changes that were under consideration in the proposed 2023 Notice of Benefit and Payment Parameters. The U.S. Department of Health and Human Services (HHS) has since posted the final 2023 Notice of Benefit and Payment Parameters. This rule finalizes several changes to the way the risk adjustment (RA) program works for qualified health plans starting in 2023. In this blog, we review the changes that were finalized.

Risk Adjustment User Fee

As proposed, the risk adjustment user fee for 2023 will be $0.22 per member per month (PMPM), slightly lower than the $0.25 PMPM for 2022. However, HHS expects it will cost about the same—$60 million—to operate the RA program, and that increased enrollment will make up for the lower PMPM amount.

Recalibration

HHS finalized its proposal to recalibrate 2023 benefit year RA models using the three most recent consecutive years of enrollee-level External Data Gathering Environment (EDGE) data that are available to be included in the calibration at the time of the proposed rule. This means 2017, 2018, and 2019 enrollee-level EDGE data for 2023. Under this policy, 2020 enrollee-level EDGE data would begin to be used in the recalibration for the 2024 benefit year. There is concern about this given the substantial changes in healthcare utilization during the COVID-19 pandemic, particularly in the first year of the pandemic. Due to these concerns, several commenters on the proposed rule suggested that HHS compare RA model outputs with and without the 2020 EDGE data and solicit further comment. HHS stated it will consider the feedback as it contemplates options for including 2020 data in future recalibration efforts.

Risk Adjustment Model Updates

HHS is finalizing two of its three RA model specification changes for 2023: (1) removing the current severity illness factors from the adult models and adding interacted hierarchical condition category (HCC) count factors to both the child and adult models; and (2) replacing the current enrollment duration factors (EDFs) with HCC-contingent enrollment duration factors. As discussed in the prior blog, the new interacted factors help address the underprediction of plan liability for the costliest enrollees. The new EDFs are geared to improving the accuracy of the model for partial-year enrollees.

The federal government considered but did not adopt a proposal to add a two-stage weighted approach, which would have been the most significant change. The approach would have weighted healthier enrollees more heavily in the model, to help address systemic under-prediction of risk for healthier and lower-cost enrollees (i.e., those without HCCs). Comments by the economists at the non-profit think-tank Brookings appeared particularly influential. They raised concerns that this change could have anti-competitive effects. Their comments are predicated on the notion that the current RA model does not perfectly predict costs and results in transfer payments that are too low (i.e., fully mitigating risk would require transferring more money from insurers with low-risk enrollees to insurers with high-risk enrollees). This can result in strong incentives for plans to cater benefit designs to healthier enrollees (such as by using narrow networks). Implementing a change that would increase the risk scores for low-risk enrollees (or lower them for high-risk enrollees) could increase selection incentives (making healthier enrollees even more attractive and sicker ones less attractive) and reduce incentives for plans to offer more generous and flexible coverage. There is no plan by the federal government to pursue additional implementation strategies or analysis of the two-stage model at this time.

Hepatitis C Pricing Adjustments

A pricing adjustment for Hepatitis C drugs will continue to be applied for 2023 RA models. As discussed in the prior blog, this approach had been adopted since the 2020 models due to concern that the data used to estimate the RA models does not reflect the availability of less expensive generic versions of the Hepatitis C drugs. Most commenters supported this pricing adjustment. However, some did note that the professional ethical standards of providers should prevent them from prescribing drugs that they did not believe were medically appropriate, reducing the potential for issuers to game the model. These commenters were concerned about undercompensating issuers for enrollees with serious chronic conditions, which would incentivize issuers to avoid these enrollees. They encourage HHS to continually evaluate the models to ensure they fully capture the cost of the current standard of care for conditions in the model.

RXC Mapping for Recalibration

HHS is finalizing the proposal to use the fourth quarter prescription drug categories (RXC) mapping document for each benefit year of recalibration data to recalibrate the adult models. The exception will be for 2017, in which case the Q2 2018 RXC mapping document was the most recent available when the 2017 enrollee-level EDGE data was processed. This change aims to reduce the year-to-year volatility of some RXC factors and ensure better representation of the utilization and costs observed for the underlying drugs in use in that year for each RXC (see the prior blog for more background on RXCs and the role of the mapping document).

HHS is also continuing to make targeted changes to the RXC mapping, regardless of which version is used. This could occur when significant changes occur in the market between the data year and the applicable benefit year of risk adjustment, such as substantial off-label prescribing, increases or decreases in drug costs relative to others in the same RXC, or changes in clinical indications or prescribing patterns. For 2023, this will involve the removal of hydroxychloroquine sulfate to Immune Suppressants and Immunomodulators (RXC 09) in the 2018 and 2019 benefit year enrollee-level EDGE data used for RA recalibration for the 2023 benefit year (which was as discussed in the proposed rule).

Requests for State-Specific Adjustments

Beginning with the 2024 benefit year, states will no longer be able to request a reduction in RA state transfers. The one exception will be Alabama, which has been the only state to make such a request. Alabama’s request for 50% reductions in 2023 transfer payments was approved, but at lower rates (25% for transfers in the individual market and 10% for the small group market). While Alabama asserted that any premium increase from their request in the individual market would not exceed a de minimis threshold (1%), public comments were provided estimating a larger increase for at least one insurer. This concern contributed to the approval at a lower rate. According to the final rule, any future requests would also need to meet the de minimis threshold standard. However, approvals are not guaranteed and HHS plans to phase them out completely by 2025 through future rulemaking. States also have the option to run their own RA program.

New Data Collection Requirements for Risk Adjustment

HHS finalized a series of new data collection and reporting requirements. The federal government will extract three new data elements (plan ID and rating area beginning for the 2021 benefit year, and subscriber indicator for 2022) that issuers already provide. Beginning with the 2023 benefit year, issuers will be required to collect and make available five data elements: zip code, race, ethnicity, enrollment in an individual coverage health reimbursement arrangement (ICHRA), and a subsidy indicator.

HHS will adopt a transitional period for the 2023 and 2024 benefit years, during which issuers can populate race and ethnicity fields using only enrollee data that is accessible or already collected (optional data on race and ethnicity is already collected through exchange applications at the time of enrollment). This data will be required for the 2025 benefit year, which will help ensure that the information is collected in the same format for on- and off-exchange enrollments, as well as for all exchange types for the individual, small group, and merged markets.

Many commenters supported the additional data collection to better help track information related to social determinants of health. However, there were also concerns about administrative burdens and privacy and security. In recognition, the federal government increased estimates of the burden associated with data collection to nearly $1.9 million in one-time fees for an estimated 650 insurers. Additionally, it is estimated that there would be $314,000 in cumulative annual costs.

Separately, HHS asked for comments on how to encourage consistent use of z codes, which are a subset of ICD-10 codes that document social determinants of health. Most commenters supported more consistent use of z codes by providers, which could help further assess risk in the individual, small group, and merged market risk pools. Others urged increased education for and outreach to providers on the value and use of z codes. One commenter stated that using z codes for the RA program without substantial preparation could widen existing gaps in recognized coding standards. HHS will keep considering potential uses of z codes and soliciting comments.

RADV Error Rate Calculation

HHS finalized refinements to the HHS Risk Adjustment Data Validation (HHS-RADV) to better align the calculation and application of error rates with the enrollee RA calculations. Specifically, CMS will extend application of super HCCs more broadly throughout the HHS-RADV error rate calculation process. Currently, super HCCs are applied only in the sorting step that assigns HCCs to failure rate groups. (More details on the HHS-RADV process are provided in the prior blog).

HHS also finalized proposals to define super HCCs separately according to the enrollee’s age group model. The super HCCs were defined based on the adult model, due to concerns of small sample sizes in the infant and child models. Some public comments on this proposal reiterated this concern. However, an analysis found that using each age group model’s factor definitions separately to define super HCCs could in many cases provide more stability than using only the adult models.

In addition, the final rule constrains any failure rate group outlier with a negative failure rate to zero. As discussed in the prior blog, this refinement is consistent with HHS’s intent to reduce potential incentives for issuers to use RADV to identify more HCCs than were reported to their EDGE servers for an applicable benefit year.

Recoupment of High-Cost Risk Pool (HCRP) Funds

HHS finalized the provision to use recouped funds to reduce the HCRP charges for all insurers during the next benefit year, if HCRP payments were already calculated for the current benefit year. If HCRP payments were not already calculated, the recouped funds would be disbursed in the next benefit year in the form of reduced charges for all insurers. The changes do not change the amount of HCRP payments or charges made. Commenters expressed general support for these changes.