Medicare Advantage Advance Notice: An Overview of the High-Impact Proposed Changes

Projected Growth Trends

MA rates are set based on what is called the United States Per Capita Cost (USPCC) which sets an average year over year per beneficiary per month amount for their total cost of care. This rate can be compared to prior years to establish predicted growth rates. For 2024, CMS has lowered the expected growth rate, which means the USPCC went down. The prior estimated USPCC for 2024 was $1,195.40 in the 2023 Rate Letter. That has been revised downward to $1,158.53, a change of -3.17%. It is important to note that this is a national average and MA rates are set regionally, where the projected per beneficiary costs can vary significantly, however overall CMS has projected lower growth than anticipated.

Changes to the Risk Scoring Methodology

The second impactful proposed change relates to the Hierarchical Condition Coding (HCC) methodology used to calculate risk scores. Risk scores are used to compare beneficiaries “illness burden” and to predict the amount of money a beneficiary will require for their care during a performance year. Risk scores are based on diagnoses from the prior year, so 2024 risk scores will be based on diagnoses captured for that beneficiary during 2023. A beneficiary with a risk score greater than one is expected to cost more than the average beneficiary to treat, whereas a beneficiary with a risk score less than one is expected to cost less.

The HCC model created by CMS is an algorithm that has multiple inputs:

1. A mapping of ICD codes to HCC codes (model classification);
2. Demographic factors (age, dual status, disability, gender);
3. Claims data for a base year and a measurement year; and,
4. A denominator year

The output of this algorithm is “relative factors” – both demographic relative factors and disease relative factors. A beneficiary’s raw risk score is the sum of their demographic relative factors and disease relative factors. Each factor represents the incremental cost to treat that beneficiary because of that condition.

The proposed new version 28 of the HCC model will have changes to inputs (1), (3), and (4) above. These changes have caused concern amongst our customers and other stakeholders.

HCC Model Reclassification

By far the (potentially) most impactful proposed changes come with the HCC Model Reclassification. CMS is proposing moving to v.28 of the HCC Model, which is based on ICD-10 diagnosis codes instead of ICD-9 codes. Importantly three categories of changes are proposed: (1) some HCCs are being deleted entirely; (2) some (~2700) ICD-10 codes are no longer being associated with an HCC; and (3) the relative factors associated with some HCCs are changing as a result of the model.

Overall in 2024, CMS is proposing 115 HCCs, which is more than the current 86 HCCs. However, a few HCCs are being discarded entirely: Protein Calorie Malnutrition, Angina Pectoris, and Atherosclerosis of Arteries of the Extremities, with Intermittent Claudication. The reasoning given for the deletion of these HCCs is Principle 10 from a 2000 report on HCC Models:

Principle 10 – Discretionary diagnostic categories should be excluded from payment models. Diagnoses that are particularly subject to intentional or unintentional discretionary coding variation or inappropriate coding by health plans/providers, or that are not clinically or empirically credible as cost predictors, should not increase cost predictions. Excluding these diagnoses reduces the sensitivity of the model to coding variation and coding proliferation.

CMS proffers their analysis shows that these three HCCs were disproportionately used by Medicare Advantage over FFS and additionally are not correlated with increase spend in the performance year.

In addition to removing these three HCCs, the new model is based on ICD-10 diagnosis codes instead of ICD-9. The HCC model works by associating diagnosis codes (ICD codes) with conditions, so multiple diagnosis codes will map up to one HCC code. HCC codes are “hierarchical” (thus the name), such that if a person has “Diabetes with Glycemic, Unspecified, or No Complications” (HCC38) and they subsequently develop complications such that they have “Diabetes with Chronic Complications” (HCC37), HCC37 subsumes HCC38 and only the HCC37 relative factor is added to the raw risk score. In the new methodology, the ICD-10 codes and how they map to HCCs has changed such that about 2700 ICD codes no longer map to an HCC code. For example, in the current version of the HCC model the diagnosis code E210 for Primary hyperparathyroidism maps to HCC code 23, however in 2024 this diagnosis code will not map to any HCC codes.

Updated Data and Denominator Years

The new methodology will update both the data years and the “denominator year.” Version 28 will look at diagnoses captured in 2018 to calculate risk scores in 2019. It will use 2020 as the “denominator year.” Risk scores are calculated by adding up “condition codes” each of which have an associated value called a “relative factor.” For example, “Severe Persistent Asthma” has a relative factor for a non-dual, non-disabled, aged beneficiary of 0.818. These factors are created using the HCC algorithm and represent the relative increase in cost over the average for a beneficiary with that condition. The denominator year is an input to the model and represents the average cost for a Fee for Service beneficiary in that year. For the denominator year 2020, that value is $10,402.34.

Normalization Factor

Once raw risk scores are calculated by summing up all of the relative factors (demographic and disease relative factors), they must be normalized, since risk scores are really a way of comparing beneficiaries to each other. CMS publishes a normalization factor each year and raw risk scores are divided by this factor to create normalized risk scores. Larger normalization factors lead to larger decreases in the conversion of raw risk scores to normalized risk scores.

Normalization factors are calculated by looking at the average risk scores over a five-year period and projecting this growth trend out to the sixth year. CMS then uses a formula¹ to calculate the normalization factor required to make the projected average risk score for the performance year equal to that of the denominator year, which is necessary for the model to accurately calculate forecasted costs for a beneficiary. The risk scores for the denominator year were all normalized to 1, so the effect of the normalization factor is to essentially bring the MA risk scores for the performance year down to the point they would have been relative to fee for service scores in 2020. It is important to note that as with PY2023, CMS is excluding 2021 risk scores from this 5-year trend due to pandemic-related substantially decreased utilization in 2020, which is the year that diagnoses for 2021 risk scores were captured.

This normalization factor will change more dramatically when the denominator year changes. In this case, since 2020 is more recent than 2015, and risk scores have steadily increased year over year, we would expect the normalization factor to be smaller. Indeed CMS has said that if the proposed new methodology is utilized, the normalization factor will be 1.015, vs. 1.146 if the current methodology is utilized.

What are the Expected Impacts?

CMS is predicting that risk scores will be 3.12% lower under the new methodology, which equates to a savings of $11B to the Medicare Trust Fund in 2024. They further assert that the impact of the different normalization factor is zero since that change originates from shifting the denominator year to 2020 from 2015 and is therefore built into the calculation of dollar coefficients and therefore the relative factors.

What Concerns are We Hearing?

At CareJourney we have been in many meetings with members and industry experts over the last two weeks, each of whom have unique concerns and arguments both in favor of and against these proposed modifications. The simplest to explain concern is that decreasing the funds for Medicare Advantage plans will result in decreased beneficiary enhancements such as no premium plans, Part D cost sharing, dental benefits, and other benefits MA plans typically offer their beneficiaries.

Another common concern that we have heard is that many of the eliminated ICD codes, and the three removed HCCs, are routinely used in the provision of preventative care. Even if they are not associated with higher costs in the performance year immediately after those diagnoses are captured, they overall save the Trust Fund money by delaying or stopping the progression of disease. The programs that support beneficiaries with these now removed conditions are not free to administer and therefore they do add an incremental cost for that beneficiary. Removal of these codes will result in increased costs for beneficiaries or decreased services.

Finally, our conversations have centered around concerns over whether these changes will disproportionately affect racial minorities, dual-eligible beneficiaries, or specific geographic regions.

What are We Doing?

At CareJourney, we are utilizing our access to the full Medicare Advantage encounters data to analyze the impacts to MA plans and to MA enrolled beneficiaries. We are working to understand the impact of shifting to the proposed risk scoring methodology at a beneficiary level and at larger demographic groupings. In the coming days and weeks, we hope to provide data to our members that will help them craft data-driven comment letters to submit by the March 3, 2023 deadline. We will also publish a follow-up to this post with broad stroke analysis of our findings.