In this article
A RAF score, or Risk Adjustment Factor score, is a number that predicts a Medicare beneficiary's expected healthcare costs relative to the average beneficiary. A score of 1.0 means the member is expected to cost the average amount; 2.0 means twice the average. CMS uses these scores to adjust capitated payments to Medicare Advantage plans so plans treating sicker members receive higher reimbursement. This article explains what a RAF score is and walks through calculation examples using the CMS-HCC V28 model, with the figures CMS published for CY2026 and CY2027.
What is a RAF score?
A RAF score is a relative cost prediction. CMS uses it to adjust capitated payments to Medicare Advantage plans, paying more for sicker members and less for healthy ones. Each factor in the V28 model is a predicted dollar cost divided by a model denominator of $10,402.34, according to the CMS 2024 Rate Announcement. That is why the factors add up to a ratio rather than a dollar amount: a member who scores 1.394 is predicted to cost about 39% more than the average beneficiary.
The 1.0 baseline and what above or below means
CMS calibrates a normalization factor each payment year so the average fee-for-service beneficiary scores 1.0. A score above 1.0 predicts higher-than-average costs; a score below 1.0 predicts lower-than-average costs. A healthy 67-year-old woman with no payment HCCs scores 0.330, her demographic factor alone.
What feeds into a RAF score
Four kinds of factors build the raw score:
- Demographic factors: age, sex, whether the member was originally entitled to Medicare by disability, and the model segment, which reflects Medicaid status (non-dual, full-benefit dual or partial-benefit dual) and whether the member lives in the community or an institution.
- Clinical factors: Hierarchical Condition Categories (HCCs). Each HCC is mapped from ICD-10-CM diagnosis codes documented during the data collection year, and each payment HCC carries a coefficient.
- Disease interactions: extra weight when specific condition combinations exist, such as diabetes with heart failure (0.112 in the community non-dual aged segment).
- Payment HCC count: extra weight once a member has five or more payment HCCs.
Check a sample member's demographic segment and count how many chronic conditions on their problem list have ICD-10-CM codes at the highest specificity. Any condition coded to an unspecified code is a review candidate. Before pursuing capture, confirm that the clinical documentation supports greater specificity and that the condition meets applicable risk-adjustment requirements.
How does a RAF score get calculated?
A RAF score is calculated in three steps:
- Raw score = demographic factor + every surviving HCC coefficient + interaction terms + the payment HCC count factor.
- Normalized score = raw score ÷ the normalization factor (1.079 for CY2027; 1.067 for CY2026).
- Payment score = normalized score × (1 − the MA coding pattern adjustment of 5.90%).
For Beneficiary A in the worked example below, that is 1.394 raw, 1.292 normalized and 1.216 as the CY2027 payment score, using the factors in the CMS CY2027 Rate Announcement.
The four components of a RAF score
Most ICD-10-CM codes map to no payment HCC at all. Of the 266 HCCs in V28, only 115 are used for payment. A documented diagnosis that does map to a payment HCC adds that HCC's coefficient. The demographic factor provides the baseline. Interaction factors add weight for clinically significant condition pairs. The count variable adds weight once a member has five or more payment HCCs in the community segments, and ten or more in the institutional segment. CMS added count variables to comply with the 21st Century Cures Act requirement that the model take into account the total number of conditions a beneficiary has.
How hierarchy rules prevent double-counting
Within a clinical family, only the most severe HCC counts. If a member has CKD Stage 4 (HCC 327, Chronic Kidney Disease, Severe) and CKD Stage 3 (HCC 329, Chronic Kidney Disease, Moderate, Stage 3 Except 3B), only HCC 327 contributes to the score. This prevents double-counting related conditions that represent different severity levels of the same disease.
Why segment selection changes the score
The CMS-HCC model has eight segments: six community segments (non-dual, full-benefit dual and partial-benefit dual, each split into aged and disabled), one institutional segment and the new enrollee model. Each continuing-enrollee segment has its own coefficient set, so the same diagnoses produce different scores depending on which segment applies. Members with end-stage renal disease are scored with a separate ESRD model.
Pick one de-identified member encounter from your queue and identify the member's segment. Then list every ICD-10-CM diagnosis code to see which map to a payment HCC and which are suppressed by hierarchy.
How does V28 change RAF scores compared to V24?
For CY2026, CMS calculates 100% of risk scores using the 2024 CMS-HCC model (V28), according to the CMS 2026 Rate Announcement. This completed a three-year phase-in that began in 2024, when 33% of the score came from V28, rising to 67% in 2025. CMS kept the same model for CY2027 without recalibrating it, so the V28 coefficients in this article apply to both years.
What V28 changed from V24
V28 reclassified conditions using the ICD-10-CM classification system instead of the ICD-9 system that V24's categories came from, and it updated the underlying fee-for-service data from 2014 diagnoses and 2015 expenditures to 2018 diagnoses and 2019 expenditures. It classifies approximately 74,000 ICD-10-CM codes into 266 HCCs, 115 of which are payment HCCs, according to CMS. V28 also constrained related HCCs to share one coefficient: diabetes HCCs 36, 37 and 38 all carry 0.166 in the community non-dual aged segment, and CKD HCCs 328 and 329 both carry 0.127.
| Feature | V24 | V28 |
|---|---|---|
| Code mapping basis | ICD-9-era categories, calibrated on 2014 diagnoses and 2015 costs | Reclassified on ICD-10-CM, calibrated on 2018 diagnoses and 2019 costs |
| Total HCCs | 204 | 266 |
| Payment HCCs | 86 | 115 |
| Interaction terms | 22 (16 disease, 6 disabled) | 11 (6 disease, 5 disabled) |
| CY2026 and CY2027 weight | 0% | 100% |
Source: CMS 2020 and 2024 Rate Announcements; CMS Report to Congress, December 2024.
Reading a score above or below 1.0
Plans use these scores for risk stratification and financial planning. According to the CMS 2026 Rate Announcement, the CY2026 normalization factor is 1.067 and the MA coding pattern adjustment is 5.90%. For CY2027, the normalization factor is 1.079 and the coding pattern adjustment stays at 5.90%.
CMS applies both after the raw score to derive the payment score. Compare one member's HCC list under V24 versus V28 mappings using CMS's published mapping files. Note which categories changed, disappeared, or gained a coefficient under V28. Angina pectoris is one example: it moved to HCC 230, a non-payment HCC, while unstable angina stayed in the payment model.
How do you calculate a RAF score from real diagnoses?
Two worked examples from the CMS Innovation Center's ACO REACH PY2026 risk adjustment paper show how the formula applies in practice. Both members are community-residing, non-dual and aged.
A simple V28 calculation
Beneficiary A is a 67-year-old female with Crohn's disease (HCC 80), CKD Stage 4 (HCC 327) and CKD Stage 3 (HCC 329). Hierarchy suppresses HCC 329 because HCC 327 is more severe in the same family. The raw RAF score is 1.394, according to CMS.
| Component | V28 coefficient |
|---|---|
| Female, age 65-69 | 0.330 |
| HCC 80, Crohn's disease | 0.550 |
| HCC 327, CKD Stage 4 | 0.514 |
| HCC 329, CKD Stage 3 (suppressed by HCC 327) | 0.000 |
| Raw score | 1.394 |
Source: CMS ACO REACH PY2026 Risk Adjustment paper; CMS 2024 Rate Announcement, Table VIII-1.
A complex case with multiple HCCs
Beneficiary B is an 88-year-old male with five payment HCCs: diabetes with chronic complications (HCC 37), chronic liver failure or end-stage liver disorders (HCC 63), rheumatoid arthritis (HCC 93), moderate dementia (HCC 126), and unstable angina or other acute ischemic heart disease (HCC 229). Five payment HCCs trigger the count variable, which adds 0.050. No interaction applies because he has no heart failure. His raw score is 3.040.
| Component | V28 coefficient |
|---|---|
| Male, age 85-89 | 0.664 |
| HCC 37, Diabetes with chronic complications | 0.166 |
| HCC 63, Chronic liver failure/end-stage liver disorders | 0.962 |
| HCC 93, Rheumatoid arthritis | 0.617 |
| HCC 126, Dementia, moderate | 0.341 |
| HCC 229, Unstable angina and other acute ischemic heart disease | 0.240 |
| Payment HCC count, 5 | 0.050 |
| Raw score | 3.040 |
Source: CMS ACO REACH PY2026 Risk Adjustment paper; CMS 2024 Rate Announcement, Table VIII-1.
From raw score to payment score
The raw score is not what CMS pays on. Dividing by the normalization factor and applying the coding pattern adjustment gives the payment score:
| Beneficiary | Raw score | CY2026 payment score | CY2027 payment score |
|---|---|---|---|
| A | 1.394 | 1.229 | 1.216 |
| B | 3.040 | 2.681 | 2.651 |
Source: calculated from the CMS CY2026 (normalization 1.067) and CY2027 (normalization 1.079) Rate Announcements, with the 5.90% MA coding pattern adjustment.
List one member's demographic coefficient and every surviving HCC coefficient from the V28 table. Add interaction terms and the count factor, divide by the normalization factor, and apply the coding pattern adjustment. Compare your result to what your risk adjustment software reports.
What should you do next to improve RAF capture?
Review your organization's documentation to ensure all chronic conditions are coded to the highest specificity the record supports. Documentation gaps cost both HCC coefficients and the interactions they trigger. The model is prospective and starts over each year: a condition not captured from an eligible encounter in the current data collection year does not carry forward into next year's score. The most commonly missed HCC codes are a good place to start an audit.
- Audit recent encounters for conditions documented to unspecified ICD-10-CM codes and query the provider for specificity.
- Verify that every chronic condition on the problem list was documented and addressed at an eligible encounter during the data collection year. Many organizations use MEAT (Monitor, Evaluate, Assess, Treat) as an internal documentation standard; it is an industry convention, not a CMS rule.
- Confirm each member's segment flags (age, sex, Medicaid eligibility, institutional status, original reason for entitlement) are correct in your enrollment data.
- Ask Claire about a de-identified encounter and compare the ICD-10-CM codes and V28 HCCs it returns with what was actually submitted. Any gap is a documentation opportunity. Claire maps codes to HCCs with a deterministic CMS crosswalk rather than a guess. See Claire for risk adjustment coders.
- Track your before-and-after RAF scores monthly to measure whether documentation improvements are translating into captured HCCs.
Frequently asked questions
Are HCC and RAF scores the same thing?
An HCC is one clinical category in the model, each carrying a coefficient. The RAF score is the member's total after combining the demographic factor, all surviving HCCs, interaction terms and the count factor, then applying normalization and the coding pattern adjustment. HCCs are the building blocks; the RAF score is the finished structure.
What is the RADV audit?
Risk Adjustment Data Validation audits verify that diagnoses submitted by Medicare Advantage organizations are supported by medical record documentation. CMS recovers overpayments when diagnoses cannot be substantiated, making documentation accuracy financially significant. A valid record comes from an acceptable inpatient, outpatient or physician encounter during the data collection period and is authenticated with the provider's signature and credentials. Telehealth counts when it uses audio and video together; audio-only services do not.
What is the MA coding pattern adjustment for 2026?
For CY2026, the MA coding pattern adjustment is 5.90%, the statutory minimum, per CMS. It offsets higher coding intensity in Medicare Advantage compared to Original Medicare and is applied after the normalization factor of 1.067, lowering the payment score. CY2027 keeps the 5.90% adjustment, with a normalization factor of 1.079.
Written by
Valentina GallegosBA, CPC, CRC
Co-founder of Claire and a medical coding professional and Subject Matter Expert in Risk Adjustment and HCC coding, recognized as a healthcare leader focused on igniting transformative change and bridging opportunity gaps. Known for her emphasis on accuracy, compliance, and documentation quality, her work connects clinical detail with better coding outcomes, supporting both healthcare standards and patient-centred care — and shaping how Claire reasons about risk adjustment.






