Medicare Advantage Risk Adjustment Model Continues to Phase In

Update 1/13/20: This article has been updated to reflect that the American Academy of Actuaries’ paper and the Tenth Circuit Court decision were in regards to the Affordable Care Act’s risk adjustment model. A previous version stated that these entities were reflecting on the Medicare Advantage risk adjustment model.

CMS has released the first of a two-part advance notice, which states that it plans to phase in the Medicare Advantage risk adjustment model and to introduce encounter data into calculating Medicare Advantage risk adjustment models.

“Medicare Advantage has been successful in providing Medicare beneficiaries with options so that they can choose the healthcare that best fits their individual health needs,” the advance notice begins. “A key element in the success of Medicare Advantage is ensuring that payments to plans reflect the relative risk of the people who enroll. A critical tool that CMS uses to accomplish that goal is the risk adjustment models that adjust payments based on the characteristics and health conditions of each plan’s enrollees.”

The advance notice goes on to explain the proposals that are currently up for public review.

CMS uses the CMS Hierarchical Condition Categories (CMS-HCC) risk adjustment model in Medicare Advantage and Programs of All-Inclusive Care for the Elderly (PACE) entities to identify costly diagnoses. The risk score estimates a beneficiary’s treatment costs expected for the next year, relative to similar patients in one of eight segments. The segments include categories such as new enrollees and dual eligible beneficiaries.

For payment year (PY) 2020, risk scores will be the sum of 50 percent of the risk score calculated with the 2020 CMS-HCC model and 50 percent of the score calculated with the 2017 CMS-HCC model.

This is a progression from last year, in which 25 percent of the score was based on the 2019 CMS-HCC model and 75 percent was based on the 2017 CMS-HCC model.

The Social Security Act requires that any changes to the risk-adjusted payments be phased in over the course of three years. Thus, the changes initiated in PY 2019 will be fully incorporated into the risk adjustment model by PY 2022.

Next year, the proposed risk score calculation would be based on 75 percent of the 2020 CMS-HCC model and 25 percent of the 2017 CMS-HCC model.

In the past, CMS drew primarily from CMS Risk Adjustment Processing System (RAPS) data to calculate the risk score. RAPS data consists of diagnoses that MA plans submitted to CMS.

However, CMS prefers using encounter data, which consists not only of diagnoses but also includes records of providers’ products and services used to treat the patient.

Similar to CMS-HCC models, switching to encounter data from RAPS data requires a phase in. Thus, in 2020, CMS will calculate risk scores with 50 percent encounter data and 50 percent RAPS data.

This is a step forward from last year when CMS arrived at the risk score by adding 25 percent of data as encounter data and 75 percent as RAPS data.

Already, CMS is proposing that the 2021 risk score be calculated by adding together 75 percent of the encounter data and 25 percent of the RAPS data.

Determining risk adjustment models is a challenging process. Less than a week before CMS published this advance notice, the risk adjustment for the Affordable Care Act’s risk adjustment—which uses a different model—cleared the Tenth Circuit Court of Appeals.

Some argued that by using statewide average premiums, the ACA model favors larger payers. When calculating how much of a risk adjustment payment a payer receives in return, the plan’s average risk score is measured against the average statewide premium.

Larger payers are more likely to be close to the statewide premium average as opposed to payers with a smaller market share, the American Academy of Actuaries explained in a 2016 public policy issue paper regarding the ACA risk adjustment model.

However, the Tenth Circuit Court of Appeals could only narrowly evaluate the ACA risk adjustment model, assessing whether it was “arbitrary and capricious.” The Tenth Circuit Court of Appeals said the ACA risk adjustment model would be “arbitrary and capricious” if HHS skipped assessing an element of the issue or failed in one of three other criteria.

The Tenth Circuit Court found that HHS had considered other possibilities and contemplated the necessary factors in establishing the ACA's risk adjustment model.

While the MA risk adjustment model has not been brought to court, it has faced other challenges including MA plans reporting services in the encounter data that never occurred. Some may do this in order to attain a high risk score and obtain greater funding.

To avoid unintentionally inaccurate reporting, some payers are utilizing advanced technologies. In January 2019, Change Healthcare and Health Fidelity declared their partnership to use artificial intelligence to improve reporting processes for both their MA and their ACA risk adjustment models.