Adapting to CMS-HCC Model v28

For the first time in a decade, the Centers for Medicare and Medicaid Services (CMS) has overhauled its hierarchical condition categories (HCC), upgrading the underlying methodology to align with ICD-10-CM – which the rest of the healthcare system has been using since 2015. As a result, HCC version 28 requires greater specificity in documentation and code assignment to ensure accurate risk adjustment.

How v28 Differs

The HCC model is used to assign risk adjustment factor (RAF) scores and estimate future healthcare costs for Medicare Advantage patients. HCC v28 will be phased in over a three-year period. Thus, 2023 dates of service will be categorized with a blend of 33% from the v28 model and 67% from the v24 model, 67% will be v28 for 2024 dates of service, and it will be fully phased in at 100% for 2025 dates of service. 

The revisions made in v28 are designed to reflect more recent utilization, cost, and diagnostic patterns, and therefore focus primarily on those HCC codes with the highest variations between Medicare Advantage and fee-for-service. Newer data and clinically based adjustments have also been incorporated to ensure that conditions are stable predictors of costs for improved payment accuracy.

Among the most significant changes in v28 are:

• 115 HCC categories, up from 86 in v24.
• Fewer HCC codes (ICD-10-CM codes), down to 7770 from 9797.
• Elimination of approximately 2294 codes.
• Addition of 268 codes.
• Re-numbering and changing HCC categories.
• Changes to the HCC coefficient values (risk scores that map to each HCC category).

For example, HCC 35 has been added for pancreas transplant status while malnutrition has been excluded. Further:

• HCC 85 related to heart failure has been eliminated.
• Heart failure has been expanded to seven HCCs, including heart transplant status (HCC 221).
• The neoplasm disease hierarchy increases from five to seven levels. 
• The Skin disease group now groups to a hierarchy that differentiates pressure ulcers and chronic ulcers of the skin by severity or depth of the wound.

RAF scores, which determine the amount paid by CMS to a health plan per patient, have also been affected by HCC v28. Because the RAF scores and anticipated care costs of patients with multiple conditions and/or conditions with greater levels of severity are higher, Medicare Advantage plans are paid at a higher rate.

In 2023, Medicare Advantage plans were paid on HCC v24, which was established using ICD-9-CM claims coded data. Under v28, CMS finalized revisions for the Part C risk adjustment model that incorporate recalibration and clinical reclassification of HCCs, updating the data year to 2018 diagnoses and 2019 expenditures from v24’s 2014 and 2015, respectively. The denominator year used to calculate risk scores was also updated to 2020 in v28 from 2015 under v24.

Applying the Constraining Process

In reclassifying HCC mappings in v28, CMS used a process known as constraining, under which related HCCs are given the same coefficients. For example, diabetic disorders contribute the same to the RAF score whether the patient has uncomplicated diabetes or diabetes with complications. However:

• Type 2 diabetes mellitus without complications (E11.9) receives a slightly higher risk score in v28 than it currently does in v24 (from 0.105 to 0.166).
• A patient with diabetes with peripheral vascular disease in v24 has risk scores of 0.302 + 0.288 (0.590), but under v28 receives a risk score of only 0.166.

As a result, the RAF scores will be significantly reduced for patients with acute or chronic complications from diabetes.

CMS notes that v28 will result in a more appropriate relative weight, reflecting recent utilization, coding, and expenses. The CY 2024 impact on Medicare Advantage risk scores is also projected to be -3.12%, which translates into a net savings to the Medicare Trust Fund of $11 billion in 2024.

Documentation Specificity is Key

Specificity of documentation and diagnostic coding is foundational to accurate risk adjustment. Under HCC v28 even greater specificity is needed to ensure that the true level of the patients’ illness severity is captured, and that CMS is provided with coded data for future analysis in model recommendations.

This means it is important to continue documenting against the MEAT (Monitoring, Evaluation, Assessment, Treatment) criteria – the four factors that establish the presence of a diagnosis during a patient visit and ensure proper documentation. Additionally, because of the staged transition, providers will be reimbursed under two different models until v28 is fully implemented, based on both the date of service and the percentage of the new model that has been deployed. Documentation and coding will need to support both.

Calculating the RAF Score

Because the new model is being phased in over time, RAF score calculation during the transition phase requires using both v24 and v28 models. The first step is to calculate risk scores for both models, followed by calculating the risk score as the sum of 33% of the adjusted v28 CMS-HCC model risk score and 67% of the adjusted v24 risk score.

Example: Jane is a 93-year-old female who has diabetic amyotrophy, fatal familial insomnia, CKD stage 3A, and toxic liver diseases with hepatic necrosis and coma. The table below illustrates how the RAF score is computed for CY2024.

Diagnosis	v28 HCC Model	v24 HCC Model	RAF score (v28)	RAF score (v24)
Diabetic amyotrophy (E11.44)	HCC 37	HCC 18	0.166	0.302
Fatal familial insomnia (A81.83)	HCC 127	HCC 52	0.341	0.346
Paroxysmal atrial fibrillation (I48.0)	HCC 238	HCC 96	0.299	0.268
Chronic kidney disease, stage 3a (N18.31)	HCC 329	HCC 108	0.127	0.288
Toxic liver disease with hepatic necrosis, with coma (K71.11)	–	HCC 27	–	0.515
	4 payment HCC counts	5 payment HCC counts	0.050	0.077
93-year-old female (demographic factor)	–	–	0.737	0.783
	Total raw risk score		1.72	2.579
	Blending formula		33%*1.72	67%*2.579
	Blended risk score		0.568	1.728
	Final risk score		2.296

Key to a Successful Transition

Providers and health plans alike face several challenges when it comes to managing two HCC versions during the transition to v28. For example, conditions that are considered as HCC in one version may not be in the other. Further, even if a diagnosis is an HCC in both versions, the actual HCC and RAF scores may differ. 

To help overcome these challenges, health plans and providers should identify their patient population’s top HCCs to examine and understand the potential impact of the two versions. Doing so, along with investing in technologies allowing for more specific documentation and more accurate and efficient coding of large volumes of clinical documents, are important strategic steps that will enable health plans, providers, and other stakeholders to manage their risk adjustment program more effectively under HCC v28.