What Is a RAF Score? How CMS-HCC Risk Scores Are Calculated in 2026 and 2027

By Jay Chowdappa, MD, co-founder and CEO, Zynix AI · Updated November 10, 2026

Key takeaways

A RAF (risk adjustment factor) score is a number that expresses how costly a Medicare beneficiary is expected to be next year compared with the average fee-for-service beneficiary, who scores 1.0. CMS builds it by adding demographic factors and the hierarchical condition categories (HCCs) supported by the prior year's diagnoses, then applies a normalization factor and, for Medicare Advantage, a coding adjustment.

• The score is prospective: diagnoses documented in one year drive the score used for the next year.
• Medicare Advantage scores use the 2024 CMS-HCC model (often called V28), with 115 payment HCCs, in 2026 and again in 2027.
• For 2027, CMS applies a normalization factor of 1.079 and the statutory minimum 5.90% MA coding pattern adjustment.
• ACO benchmarks in the Medicare Shared Savings Program are also adjusted with HCC risk scores, subject to a cap on growth.

What a RAF score means

A RAF score is a relative cost index. A beneficiary with a score of 1.0 is expected to cost about what the average fee-for-service Medicare beneficiary costs; a score of 1.5 means about 50% more, and 0.8 about 20% less. CMS uses these scores to adjust what it pays Medicare Advantage plans for each enrollee and to adjust spending benchmarks for ACOs, so that organizations caring for sicker patients are not penalized for it.

The score does not measure quality, and it is not a clinical severity rating a physician should use to make treatment decisions. It is a payment tool built from administrative data.

How a RAF score is built

CMS's Medicare Managed Care Manual (chapter 7) describes the CMS-HCC model's core design. Five features explain almost everything about how scores behave:

• Additive. A beneficiary's score is the sum of the coefficients for each demographic factor and each condition category that applies.
• Demographic factors. Age, sex, disabled status, original reason for entitlement, and Medicaid or low-income status.
• Condition categories. Diagnosis codes map to condition categories; only the categories included in the payment model (payment HCCs) add to the score.
• Hierarchies. When a patient has related conditions of different severity, only the most severe category in that hierarchy counts.
• Interactions. Certain combinations, such as two specific diseases or a disease plus disabled status, add extra factors.

The model is also prospective: it uses diagnoses from a base year to predict costs for the following year. There are separate coefficient sets (segments) for community residents, long-term institutional residents and new enrollees.

Which model applies when

The 2024 model increased the number of payment HCCs from 86 to 115. What changed between V24 and V28 is covered in depth in HCC V28 changes explained.

From raw score to payment score

The sum of the coefficients is the raw score. Two program-wide adjustments come next:

1. Normalization. Risk scores drift upward over time as the population and coding change. CMS calculates a normalization factor for each model to keep the average fee-for-service risk score at 1.0 in the payment year. For CY 2027, the normalization factor for the 2024 CMS-HCC model is 1.079.
2. MA coding pattern adjustment. Because Medicare Advantage plans tend to document more diagnoses than fee-for-service, the law requires CMS to reduce each MA risk score by a minimum percentage. For CY 2027, CMS applies the statutory minimum of 5.90%.

Illustrative arithmetic only (a hypothetical member, not a real coefficient set): a raw score of 1.200 adjusted for normalization becomes about 1.112 (1.200 ÷ 1.079), and the 5.90% coding adjustment brings it to about 1.047. The point is directional: the same documentation produces a lower payment score than the raw sum suggests.

Which diagnoses count

Risk-adjustment-eligible diagnoses come from hospital inpatient, hospital outpatient and physician settings, submitted through encounter data and fee-for-service claims. For CY 2027, CMS also finalized two exclusions for Medicare Advantage:

• diagnoses from audio-only services (modifiers 93 or FQ);
• diagnoses from unlinked chart review records, with an exception for members who switch from one MA organization to another.

The practical consequence: a diagnosis has to be assessed and documented in a qualifying encounter. Retrospective chart review that is not linked to an encounter no longer adds to 2027 MA scores.

Where RAF scores show up

For ACOs, the cap matters: an MSSP ACO cannot simply code its way to a higher benchmark. Accurate documentation keeps the benchmark honest; it does not replace the work of reducing avoidable spending. How that work produces savings is explained in what an ACO is and how MSSP, ACO REACH and LEAD differ.

What moves a RAF score (and what should not)

A RAF score should change because a patient's documented health status changed, not because of how a chart was worked. The legitimate drivers:

• Annual reassessment of chronic conditions. Because the model is prospective and year-based, a chronic condition documented last year does not carry into next year's score unless it is assessed and documented again in a qualifying encounter.
• Specific, supported documentation. Payment HCCs map from specific ICD-10-CM codes, and not every code maps to a payment HCC. The condition the clinician actually assessed, documented specifically, determines whether and where it counts.
• Demographic and status changes. Age, Medicaid status and institutional status change scores without any coding.

What should not move scores: diagnoses without clinical support in the note, conditions copied forward without reassessment, and audio-only or unlinked chart-review diagnoses that CMS now excludes in MA.

A year-round checklist for physician groups and ACOs

1. Before the visit: list the conditions documented last year that have not been reassessed this year, with the evidence (labs, medications, specialist notes). See pre-visit planning for value-based care.
2. At the visit: the clinician decides what is clinically present and documents assessment and plan for each condition addressed.
3. After the visit: check that the claim carries the documented diagnoses.
4. For patients with no visit booked: invite them in early in the year, not in December.
5. Quarterly: review open recapture and suspect lists by clinician and site, and retire suspects the clinician has ruled out.

How Zynix fits

Zynix AI does not assign codes or decide diagnoses; clinicians do. What it does is organize the work around documentation so open conditions are reviewed at a real visit:

• Population Intelligence combines claims, EHR, lab and pharmacy data to show, per patient, which previously documented conditions have not been reassessed this year and what evidence exists.
• HCC gap workflows rank open gaps, prepare clinicians before scheduled visits and invite patients with no visit booked; see HCC gap closure for ACOs and risk adjustment outreach for Medicare Advantage plans.
• AI Patient Engagement voice and text agents handle the scheduling outreach within the rules and hours your team sets.

If you are comparing tools for this work, read HCC gap closure software for ACOs.

Frequently asked questions

What is a good RAF score?

There is no good or bad RAF score. A score near 1.0 means a beneficiary is expected to cost about the fee-for-service average; higher scores mean higher expected cost. What matters is accuracy: a score that reflects every condition a clinician actually assessed and documented that year, and nothing that is not supported in the record.

How is a RAF score calculated?

CMS adds the coefficients for a beneficiary's demographic factors and for each payment HCC supported by diagnoses from the prior year, applying hierarchies and interaction terms. That raw score is then adjusted with the normalization factor, and Medicare Advantage scores are reduced by the coding pattern adjustment, which is 5.90% for 2027.

What is the difference between HCC and RAF?

An HCC (hierarchical condition category) is a group of related diagnosis codes with a cost coefficient in the CMS-HCC model. The RAF score is the patient-level total: demographic factors plus the coefficients for every HCC and interaction that applies. HCCs are the building blocks; the RAF score is the result.

Does a RAF score reset every year?

Yes, in effect. The CMS-HCC model is prospective and uses one year's diagnoses to predict the next year's costs, so chronic conditions must be assessed and documented again each year to count in the following year's score. Demographic factors are recalculated too. This is why recapture work is a year-round task.

Do RAF scores matter for ACOs, or only Medicare Advantage plans?

Both. Medicare Advantage payments are adjusted per enrollee. In the Medicare Shared Savings Program, CMS uses prospective HCC risk scores to adjust each ACO's benchmark, with a cap on positive growth. ACO REACH pays risk-adjusted capitation, and CMS says LEAD's benchmarks will use more accurate risk adjustment for high-needs patients.

Which CMS-HCC model is used in 2027?

For calendar year 2027, CMS continues to use the 2024 CMS-HCC model (V28) for Medicare Advantage, with a normalization factor of 1.079. CMS proposed an updated model in the CY 2027 Advance Notice but did not adopt it in the April 6, 2026 Rate Announcement.

Related reading

• HCC V28 Changes Explained: V24 vs V28 and What Applies in 2026 and 2027
• HCC Gap Closure Software for ACOs: What It Should Do and How to Evaluate It
• Transforming HCC Risk Adjustment: From Q4 Scramble to Year-Round Success

Source notes

CMS: Medicare Managed Care Manual, Chapter 7 – Risk Adjustment. https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/downloads/mc86c07.pdf

CMS (April 6, 2026): Announcement of Calendar Year 2027 Medicare Advantage Capitation Rates and Part C and Part D Payment Policies. https://www.cms.gov/files/document/2027-announcement.pdf

CMS (April 6, 2026): 2027 Medicare Advantage and Part D Rate Announcement fact sheet. https://www.cms.gov/newsroom/fact-sheets/2027-medicare-advantage-part-d-rate-announcement

CMS (April 7, 2025): 2026 Medicare Advantage and Part D Rate Announcement fact sheet. https://www.cms.gov/newsroom/fact-sheets/2026-medicare-advantage-part-d-rate-announcement

CMS (March 31, 2023): Announcement of Calendar Year 2024 Medicare Advantage Capitation Rates and Part C and Part D Payment Policies. https://www.cms.gov/files/document/2024-announcement-pdf.pdf

eCFR: 42 CFR 425.605 (Shared Savings Program BASIC track benchmark risk adjustment). https://www.ecfr.gov/current/title-42/section-425.605

CMS Innovation Center: LEAD Model. https://www.cms.gov/priorities/innovation/innovation-models/lead

About the author

Jay Chowdappa, MD is co-founder and CEO of Zynix AI. He is a physician, and ACOs led by him generated $300M+ in shared savings before he started Zynix AI to help care teams follow through on what their data already shows.