How sick is this population?
Every chronic condition, documented face-to-face, every year, becomes an HCC and a coefficient.
The RAF score sets the money. Missed or unsupported codes move it in both directions.
Risk Adjustment and Value-Based Care
Fee-for-service pays for each thing you do. Value-based payment pays for the population you are responsible for, adjusted for how sick it is and for how well it does. Risk adjustment is the arithmetic that makes those two ideas meet: diagnosis codes become a risk score, the score scales a capitation payment or a benchmark, and audits check that every code was earned. This tutorial covers the CMS-HCC model as it stands in 2026, RAF math, documentation and audit rules, ACOs and shared savings, Star ratings and MIPS, the mandatory TEAM bundle, and the work revenue cycle teams do to keep all of it accurate.
Did we spend less and do better?
Benchmarks, minimum savings rates, quality scores and sharing rates decide the check.
Shared savings, Star bonuses and bundle reconciliations are all formulas you can compute ahead of time.
Can we prove it?
RADV, OIG and DOJ measure the chart against the code, not the code against the claim.
An unsupported HCC is an overpayment, and from 2026 every MA contract is audited every year.
The big picture
Risk adjustment predicts what a person will cost from who they are and what they have been diagnosed with, so that a plan or provider paid a fixed amount per person is paid more for sicker people and less for healthier ones. Value-based care (VBC) is the set of payment arrangements built on top of that prediction: capitation, shared savings, quality bonuses and episode payments that reward lower cost and better outcomes rather than more services.
For most of the revenue cycle, the unit of work is a claim. In risk-adjusted and value-based arrangements the unit of work is a person-year: every diagnosis captured, every quality gap closed and every dollar of spending attributed to that person over twelve months. The claim still has to go out, but it now also carries data that sets next year's payment. A diagnosis code that would have been irrelevant to a fee-for-service claim can be worth several thousand dollars a year in a Medicare Advantage capitation, and the same code without a supporting note is an overpayment the government can take back with interest.
That is why this topic belongs in a revenue cycle series. Coding accuracy, encounter data submission, attribution reconciliation and the audit trail behind every HCC are revenue cycle functions, even when the money arrives as a capitation check or a shared savings distribution instead of an 835.
Why it matters in numbers
Scale
35.2 million
People enrolled in Medicare Advantage in 2026, 55% of beneficiaries with Parts A and B. Every one of them is paid for through a CMS-HCC risk score, and 23% are in special needs plans where risk scores run highest.
Coding intensity
$22 billion
MedPAC's March 2026 estimate of 2026 MA payments attributable to coding intensity: MA risk scores about 10% above comparable fee-for-service before the statutory 5.9% adjustment, about 4% above after it. Favorable selection adds an estimated $57 billion more.
Shared savings
$2.4 billion net
What Medicare Shared Savings Program ACOs saved Medicare in performance year 2024, the program's record at the time (performance year 2025 results are published in late 2026). 476 ACOs took part, 75% earned performance payments, and $4.1 billion was paid out in shared savings.
Who is involved
- CMS
- Builds and calibrates the CMS-HCC, RxHCC and ESRD models, sets each year's normalization and coding adjustment in the Rate Announcement, runs RADV audits, and operates the Shared Savings Program, Star ratings and the Innovation Center models.
- Medicare Advantage organization (MAO)
- The plan paid a risk-adjusted capitation per member per month. It collects diagnoses from claims, encounter data and chart reviews, submits them to CMS, and is the entity audited and recouped under RADV.
- Risk-bearing provider group
- A medical group, IPA or health system that takes a percentage of the plan's premium (a "percent of premium" or capitation contract) or participates in an ACO. Its revenue rises and falls with the RAF scores and quality of the members attributed to it.
- ACO
- An accountable care organization: a legal entity of providers that accepts responsibility for the cost and quality of an attributed population in Traditional Medicare (MSSP, ACO REACH, from 2027 LEAD), Medicaid or a commercial plan.
- Treating clinician
- The physician, NP or PA whose face-to-face documentation is the only acceptable evidence that a condition exists. No note, no HCC.
- Coders, CDI and risk adjustment analysts
- Translate the note into ICD-10-CM codes, query when the note is ambiguous, run prospective and retrospective reviews, and reconcile submitted diagnoses against what CMS accepted.
- Revenue cycle and data teams
- Get the diagnoses onto the claim or encounter, fix 277/MAO-002 rejections, reconcile capitation and shared savings payments, and report the KPIs that show whether the population's risk and quality are being captured.
- Auditors and enforcers
- CMS's RADV contractors, the HHS Office of Inspector General, and the Department of Justice under the False Claims Act, which has produced settlements and judgments against plans and provider groups for unsupported diagnoses.
What goes wrong
Under-capture
A patient with heart failure, diabetes with CKD and depression is seen four times in a year but the notes say "follow-up, stable" and the claim carries one code. The plan is paid as if the patient were healthy. Next year's capitation is short by thousands of dollars, and nothing on an 835 shows it.
Unsupported codes
A diagnosis "recaptured" from a problem list, a prior year's claim, or a home assessment with no treatment plan. Under RADV the HCC is deleted, the payment for that member is recalculated, and the error rate may be extrapolated across the contract. Under the False Claims Act, knowing submission is fraud.
Data that never arrives
Encounters rejected by the plan's EDI edits, diagnoses truncated to the first four on a claim, supplemental data filed after the sweep deadline, or an attribution list nobody reconciled. The care happened, the documentation is perfect, and the score still does not move.
From fee-for-service to value
Payment models sit on a spectrum from "paid per service, no accountability for cost or quality" to "paid a fixed amount for everything, fully accountable." Most organizations live at several points on the spectrum at once, with different payers.
The Health Care Payment Learning and Action Network (HCP-LAN) framework is the standard map. Category 1 is pure fee-for-service. Category 2 is fee-for-service with a link to quality: pay-for-reporting, pay-for-performance, and penalties such as the hospital readmissions program. Category 3 is fee-for-service architecture with shared savings or shared risk layered on top: most ACOs and bundled payments. Category 4 is population-based payment: capitation, percent-of-premium contracts, and the Medicare Advantage payment itself. Risk adjustment matters everywhere from Category 2 upward, because any payment that is not tied to the service delivered has to be tied to the patient instead.
| Model | How the provider is paid | Who bears the risk | Where risk adjustment enters | Examples in 2026 |
|---|---|---|---|---|
| Fee-for-service | Per CPT/HCPCS or DRG, from a fee schedule or contract | Payer bears utilization risk; provider bears none | Only indirectly (DRG weights, MDM levels) | Traditional Medicare, most commercial PPO contracts |
| FFS plus quality adjustment | FFS with a bonus, withhold or penalty based on measures | Provider bears a small share of quality risk | Measures are risk-standardized (readmissions, mortality, cost measures) | MIPS (up to ±9%), Hospital VBP (2% withhold), HRRP (up to 3% penalty), HAC (1%) |
| Shared savings, one-sided | FFS continues; the ACO earns a share of spending below a benchmark | Upside only; payer still bears losses | Benchmark and performance-year spending are both risk-adjusted; risk score growth is capped | MSSP BASIC Levels A and B |
| Shared risk, two-sided | FFS continues; the ACO shares savings and repays a share of losses | Shared, with loss limits | Same as above, with higher sharing rates | MSSP BASIC C–E and ENHANCED, ACO REACH, LEAD (2027) |
| Bundled / episode payment | FFS during the episode, then reconciliation against a target price for the episode | Episode initiator bears cost risk for the window | Target prices adjusted for patient HCC burden, age, social risk and hospital bed size | TEAM (mandatory, 2026–2030), commercial maternity and joint bundles |
| Capitation / percent of premium | A fixed per-member-per-month amount for a defined set of services, or a share of the plan's premium revenue | Provider bears full cost risk for the covered scope | The PMPM is multiplied by each member's risk score; the plan's own revenue is risk-adjusted by CMS or the state | MA delegated groups, Medicaid managed care capitation, ACO REACH total care capitation |
The three questions every VBC contract answers
Attribution
Whose patients are these?
Prospective (assigned at the start of the year from prior claims or enrollment) or retrospective (assigned at the end, from where the patient actually got primary care). Medicare Advantage uses enrollment; MSSP uses a plurality of primary care services by claims-based assignment, with voluntary alignment on top. Everything downstream (RAF, quality rates, spending) is computed only for attributed people.
Benchmark
What should these patients cost?
A historical baseline, trended forward, risk-adjusted, sometimes blended with regional spending. The benchmark is the number savings are measured against. Risk adjustment appears here twice: the baseline is risk-adjusted, and the performance year is re-adjusted for changes in the population's risk score.
Quality gate
Did care get worse to get cheaper?
A quality score (HEDIS, CAHPS, eCQMs, claims-based outcome measures) that scales or gates the financial result. Below the threshold, no savings are paid. In Medicare Advantage, quality is the Star rating that moves the plan's benchmark and rebate.
A fictional year in a risk contract
Riverside Medical Group (billing NPI 1234567893) holds a percent-of-premium contract with Northwind Health Plan (payer ID 99999) for 4,000 Medicare Advantage members. The example is fictional and simplified.
- 1
January: attribution and the starting score Month 1
Northwind sends a member roster with each member's current RAF. The initial 2026 scores came from diagnoses with dates of service July 2024 through June 2025. Riverside's revenue for January is base rate × each member's score × the contracted percentage.
- 2
Spring: annual wellness visits and recapture Months 2–6
Every member with a chronic condition on last year's list is scheduled. Each visit documents the status of each condition (MEAT) and the claim carries the specific ICD-10-CM codes. Conditions not seen by the end of the year drop off the score.
- 3
July: mid-year reconciliation Month 7
CMS's mid-year run (July 2026) uses calendar 2025 diagnoses. Northwind trues up capitation for members whose scores changed. Riverside reconciles the plan's accepted diagnoses against what it submitted and chases rejected encounters.
- 4
Fall: care gap closure and quality Months 8–12
HEDIS measures close on December 31. Mammograms, A1c tests, eye exams and medication adherence for the attributed population drive the plan's Star rating and the quality score that scales Riverside's bonus.
- 5
The following summer: final settlement Month 18
CMS's final 2026 run (about June 2027) uses all 2025 dates of service plus the late-submitted data. Northwind settles the year with Riverside: capitation adjustments, quality bonus, and any shared surplus or deficit on medical costs.
The one thing to remember
In fee-for-service the chart supports the claim. In risk-adjusted payment the chart supports the score, and the score pays for a year of care. Every chapter that follows is about keeping the chart, the code, the data and the money consistent with each other.
The CMS-HCC model and its cousins
A hierarchical condition category (HCC) model groups thousands of ICD-10-CM codes into a few hundred clinically related categories, keeps only the ones that predict next year's cost, ranks related categories so that only the most severe counts, and assigns each a dollar-calibrated coefficient. Add the coefficients to a demographic base and you have a risk score.
CMS's Part C model, the CMS-HCC model, pays Medicare Advantage and PACE. Its 2024 version, universally called V28, finished a three-year phase-in in 2026: 33% of the score in 2024, 67% in 2025 and 100% in 2026. The 2027 Rate Announcement (April 6, 2026) kept the same model for 2027 rather than adopting the update CMS had floated in the Advance Notice, so V28 is the model to learn.
How a diagnosis becomes a payment
- 1
Face-to-face encounter Date of service
A physician, NP, PA or other acceptable provider documents the condition during a face-to-face (or qualifying video telehealth) visit in a hospital inpatient, hospital outpatient or physician setting. Lab, imaging, DME and most ancillary claims are not acceptable sources.
- 2
ICD-10-CM code on the claim or encounter Days to weeks
The provider's claim (837) goes to the plan. The plan converts it to an encounter data record (EDR) and submits it to CMS through the Encounter Data System. Chart review records (CRRs) can add or delete diagnoses the claim did not carry; from 2027 CMS will exclude diagnoses from unlinked chart reviews and from audio-only encounters.
- 3
Filtering and mapping At CMS
CMS applies the encounter data filtering logic (acceptable provider type, CPT/HCPCS on the allowed list, facility type) and maps each surviving ICD-10-CM code to a condition category. Roughly 2,000 codes that mapped to payment in V24 no longer do in V28; the model has 115 payment HCCs compared with 86 in V24.
- 4
Hierarchies and interactions Model logic
Within a disease group only the most severe HCC counts (CKD stage 5 trumps stage 4, which trumps stage 3). Some pairs add an interaction coefficient (diabetes with heart failure, heart failure with CKD). A count-of-HCCs variable adds a little more for people with many conditions.
- 5
Normalization and coding adjustment Annual, in the Rate Announcement
The raw score is divided by a normalization factor (1.067 for the 2024 model in payment year 2026) so that the average fee-for-service beneficiary stays at 1.0, then reduced by the MA coding pattern adjustment (the statutory minimum of 5.90%, applied again for 2026 and 2027).
- 6
Payment Three runs per payment year
The plan's county base rate is multiplied by the member's final score. For payment year 2026: initial run January 2026 (dates of service July 2024–June 2025), mid-year run about July 2026 (calendar 2025), final run about June 2027 (calendar 2025 with late data). Delegated provider groups are paid the same way, one contract down.
What changed in V28
Structure
115 payment HCCs, renumbered
V28 reorganizes categories around ICD-10 rather than ICD-9 logic and recalibrates on 2018 diagnoses and 2019 spending. The HCC numbers changed (diabetes is now HCCs 36–38, heart failure 221–226, CKD 326–329), so crosswalks and reports built on V24 numbering are wrong.
Removed from payment
Low-specificity, high-variation codes
Conditions CMS found were coded very differently across plans lost payment status: angina pectoris, atherosclerosis of the extremities without complications, protein-calorie malnutrition, and many unspecified or mild codes. Documenting them still matters clinically, but they no longer move the score.
Constrained
Diabetes and heart failure flattened
All diabetes HCCs carry the same coefficient in V28, so "with complications" versus "without" no longer changes payment, although the complication (CKD, neuropathy) may map to its own HCC. Heart failure categories were similarly constrained. The specificity hunt moved from the diabetes code to the comorbidities.
Effect
Lower scores, less coding intensity
MedPAC estimates V28 reduced MA coding intensity by about 2.9 percentage points in each phase-in year. CMS's own estimate of the combined model and normalization effect for 2026 was −3.01% of payment. Plans that leaned on the removed codes felt it most.
The other models you will meet
| Model | Pays | Design | Diagnosis window | Notes for RCM |
|---|---|---|---|---|
| CMS-HCC (Part C, V28) | Medicare Advantage, PACE (blended), ACO benchmarks | Prospective: last year's diagnoses predict this year's cost. 115 payment HCCs, demographic base, dual and institutional segments | Prior calendar year (final run), with a mid-year and initial run on earlier windows | The model behind MA capitation and the risk-adjusted MSSP benchmark. PACE is 10% V28 in 2026 and 50% in 2027. |
| CMS-HCC ESRD | MA members on dialysis, with a functioning graft, or post-transplant | Separate 2023 ESRD model, 100% in 2026 for non-PACE | Same as Part C | Dialysis status is set by Medicare's ESRD data, not the plan's coding |
| RxHCC | Part D direct subsidy for MA-PD and PDP plans | Predicts drug spending; different HCC set and coefficients; separate MA-PD and PDP normalization (1.194 and 0.887 for 2026) | Same as Part C | Conditions that matter little in Part C (for example, some psychiatric and endocrine conditions) can matter a lot here |
| HHS-HCC (ACA) | Individual and small-group market issuers, zero-sum transfers within each state market | Concurrent: this year's diagnoses and drug utilization explain this year's cost. Adult, child and infant models by metal level; plan liability risk score includes enrollment duration and RXC prescription categories | Current benefit year; data collected in the EDGE server through April of the following year | Issuers chase diagnosis capture during the same year they are paid for; HHS-RADV audits a sample of each issuer's enrollees and adjusts transfers |
| Medicaid managed care models | State capitation to MCOs, then MCO payments to delegated providers | State's choice: CDPS+Rx, ACG, DxCG, HHS-HCC-derived or a state-built model, often with rate cells by eligibility group | Usually prior year, sometimes concurrent | Each state publishes its own model, run schedule and acceptable sources; many allow pharmacy data as a diagnosis proxy |
| Commercial VBC contracts | Provider groups under commercial ACO or capitation deals | Payer-chosen (Johns Hopkins ACG, DxCG/Cotiviti, Milliman MARA, or the payer's own) | Per contract | Ask for the model, the coefficient file and the run calendar before signing; the contract should say which one governs |
Prospective vs concurrent changes the workflow
Under the prospective CMS-HCC model, a condition documented in December 2025 pays throughout 2026. Under the concurrent HHS-HCC model, a condition documented in December 2026 affects 2026 transfers only. Medicare Advantage programs therefore run on an annual recapture rhythm; ACA issuers run on a continuous capture rhythm with a hard EDGE deadline in the spring.
RAF score math
A risk adjustment factor (RAF) is a sum: a demographic coefficient, plus one coefficient per payment HCC that survives the hierarchy, plus any interaction terms, plus a count term. CMS then normalizes it and applies the coding adjustment. One worked example makes every later conversation about "raising RAF" concrete.
The components
- Segment
- V28 has separate coefficient sets for community aged non-dual, community aged full dual, community aged partial dual, the same three for disabled (under 65), institutional, and new enrollees (who have no diagnosis history and get a demographics-only score). The segment is set by CMS from enrollment data, not by coding.
- Demographic coefficient
- Age band × sex, plus flags for originally disabled status. It rises with age and is higher for men at most ages. Typically 0.3 to 0.8 in the community aged segment.
- Disease coefficients
- One per payment HCC after hierarchies. Values range from under 0.1 (for example, moderate CKD) to several points (metastatic cancer, end-stage liver disease, HIV/AIDS with complications). Each is the model's estimate of the incremental annual cost of that condition relative to the average beneficiary's cost.
- Hierarchies
- If a member has HCCs in the same family, the lower ones are zeroed. Coding CKD stage 3 and stage 4 in the same year pays stage 4 only. Coding a condition that is trumped is not wrong, but it adds nothing.
- Interactions
- Added when specific pairs or triples are present (diabetes and heart failure; heart failure and CKD; heart failure and specified arrhythmias; disabled status with certain conditions). They capture the extra cost of comorbidity beyond the sum of the parts.
- Payment HCC count
- A small additional coefficient for members with many payment HCCs, stepping up from four or so conditions and capped at ten.
- Normalization factor
- Divides the raw score so that the average FFS beneficiary's score stays at 1.0 as coding and population trends drift. 1.067 for the 2024 model in 2026. It rises most years, which is why a member with unchanged diagnoses can be worth slightly less each year.
- MA coding pattern adjustment
- A statutory reduction (minimum 5.90% since 2019) for the difference in coding between MA and FFS. Applied as a multiplier of (1 − 0.059) to the normalized score for non-PACE MA plans.
RAF score calculatorTeaching model with illustrative coefficients, not the CMS V28 model file. The normalization factor (1.067) and coding adjustment (5.90%) are the actual 2026 values.
Documented conditions this year
Reading the result
Specificity pays through comorbidities now
In V28 the diabetes coefficient does not change with complications, but documenting the complication itself (CKD stage, diabetic neuropathy where it maps, heart failure) adds its own HCC and often an interaction. "Diabetes, stable" with no mention of the kidney disease the labs show leaves an HCC, and an interaction, on the table.
Scores are per member, per year, from scratch
No HCC carries over. A member with five conditions who is not seen face-to-face in the data year reverts to the demographic score. The entire difference between the two results in the calculator above is the recapture work described in the next chapter, and it is also the entire exposure in an audit if the recapture was not supported.
Where the score meets the money
| Arrangement | How RAF is used | What a 0.10 change is worth (illustrative) |
|---|---|---|
| MA capitation to the plan | County rate × member score, per month | About $1,100–1,400 per member per year at typical 2026 county rates |
| Percent-of-premium to a provider group | Plan revenue for the member × contracted percentage (often 80–90%) | The same amount × the percentage |
| MSSP benchmark | Benchmark PMPY scaled by the population's average score; positive growth capped at 3% over the agreement period | Up to 3% of benchmark over five years; beyond that, nothing |
| TEAM target price | Episode target adjusted for the patient's HCC burden (V28) and other factors | Shifts the target, not the payment; the hospital still gets DRG and the surgeon still gets FFS |
| ACA issuer | Plan liability risk score vs the state average sets transfers in or out | Zero-sum: one issuer's gain is another's payment |
Documentation and recapture
A diagnosis supports an HCC only if an acceptable provider documented, at a face-to-face encounter in the data year, that the condition exists and was addressed. Every rule in this chapter is a version of that sentence, and every RADV deletion is a failure of one of its parts.
The elements of a supportable diagnosis
- Acceptable provider
- Physicians (MD/DO), nurse practitioners, physician assistants, clinical nurse specialists and the other specialties on CMS's acceptable physician specialty list. Diagnoses from a dietitian, a pharmacist, a nurse's home visit or a lab report do not count on their own.
- Face-to-face encounter
- An in-person visit, or a video telehealth visit that meets CMS's criteria. From payment year 2027 CMS excludes diagnoses from audio-only encounters. Diagnoses from an acceptable inpatient, outpatient or professional setting; durable medical equipment, ambulance, lab and radiology claims are filtered out.
- Date of service in the data year
- For payment year 2026's final run, January 1 to December 31, 2025. A condition last documented in 2024 does not pay in 2026, however permanent it is.
- Signed and credentialed
- The note carries the provider's name, credentials and an authenticated signature (electronic is fine). Unsigned notes and notes signed by someone other than the treating provider fail RADV.
- The condition is assessed, not listed
- The note shows the provider thought about the condition at this visit. The common tests are MEAT (Monitor, Evaluate, Assess/Address, Treat) and TAMPER (Treat, Assess, Monitor/Medicate, Plan, Evaluate, Refer). One element is enough; a problem list entry with nothing else is not.
- Coded to the ICD-10-CM guidelines
- Code what the provider documented, to the highest specificity the note supports, following the Official Guidelines: no coding from "probable" or "rule out" in the outpatient setting, no coding of history-of as current, and status codes (amputation, transplant, dialysis) only when documented.
MEAT in practice
| Element | What it looks like in the note | Example that supports an HCC | Example that does not |
|---|---|---|---|
| Monitor | Signs, symptoms, disease progression or regression | "CHF: no orthopnea, weight stable at 182, mild ankle edema." | "CHF" on the problem list, nothing in the note |
| Evaluate | Test results, exam findings, response to treatment | "CKD stage 3b: eGFR 38 on 9/3, down from 44; BP 138/84." | "Labs reviewed" with no condition linked |
| Assess / Address | Status, severity, discussion, counseling | "Type 2 DM with diabetic CKD, suboptimally controlled, A1c 8.4, discussed diet and adherence." | "DM2, stable" copied forward from four prior visits |
| Treat | Medications, referrals, procedures, therapy, plan | "Continue furosemide 40 mg daily; cardiology follow-up in 3 months." | A medication list that includes furosemide with no condition named |
HCC documentation checkerDescribe the encounter and the note; the checker says whether the diagnosis would survive a validation review and what to fix.
The note shows
The annual recapture program
Because no HCC persists, risk-bearing organizations run a yearly cycle to see every member with chronic conditions and document each condition properly. The cycle below is typical of a provider group under an MA contract.
- 1
Build the suspect and recapture lists January
Recapture: HCCs paid last year not yet documented this year. Suspects: conditions implied by medications, labs, specialist claims or prior history that have never been coded, flagged for clinical confirmation, never auto-coded. Both lists are worked by the member's PCP.
- 2
Schedule the visit and prepare the chart Rolling
Annual wellness visits (G0438/G0439) and chronic care visits are the natural venue. Pre-visit planning puts the recapture list, the last labs and the care gaps in front of the clinician. The clinician decides what is present; the list is a prompt, not an instruction.
- 3
Document each condition with MEAT At the visit
One line per condition in the assessment and plan, with status and action. Clinicians who document well do not need to document more; they need to name the condition, say how it is doing, and say what they are doing about it.
- 4
Code, query and submit Within days
Coders apply the guidelines, send compliant queries for ambiguous notes (AHIMA/ACDIS practice brief: non-leading, with clinical indicators, with the option of "unable to determine"), and release the claim with all supporting diagnoses. Professional claims carry up to 12 diagnoses; make sure the HCC-bearing ones are among them.
- 5
Confirm acceptance Monthly
Reconcile the plan's accepted-diagnosis report (or CMS's MAO-004 for plans) against what was sent. Rejections for provider type, place of service or code validity are fixed and resubmitted before the plan's supplemental data deadline.
- 6
Close the year November–December
Members still unseen get outreach, transportation help or a home visit by an acceptable provider that includes a genuine evaluation and plan. In-home assessments that only list diagnoses, without treatment or follow-up, are the audit target the OIG has written about most.
Practices that fail audits and attract enforcement
- Problem-list coding: pulling every historical diagnosis onto the claim without assessment.
- Coding from lab values alone: an eGFR of 45 is not a CKD diagnosis until a provider documents it.
- One-way chart reviews: retrospective reviews that add diagnoses but never delete unsupported ones. DOJ complaints against several plans turn on exactly this.
- Diagnosis-driven incentives: paying clinicians or coders per HCC added. Pay for completeness and accuracy, audited both ways.
- Cloned assessments: identical "stable, continue" text across years, which reviewers read as no evaluation.
- Acute codes for resolved events: coding a stroke or MI from years ago as current instead of as sequelae or history.
What good looks like
- Clinician education built on their own charts, not generic slides, with the HCC's clinical definition and the specificity the guidelines require.
- Prospective review that queries before the claim goes out, so the note and the code match from the start.
- Retrospective review with a documented delete process and a submission of deletes to the plan.
- An internal validation sample each quarter scored the way RADV scores: pass, fail, and why.
- A written policy on suspect lists, in-home assessments and chart review that a compliance officer has signed.
RADV audits and coding compliance
Risk Adjustment Data Validation (RADV) is CMS's audit of whether the diagnoses that produced a Medicare Advantage payment are supported by the medical record. For a decade it was small. In 2025 CMS announced it would audit every eligible contract every year, and in 2026 the rules on extrapolating the findings are being fought over in court.
How a RADV audit works
- 1
Contract and sample selection Audit start
CMS selects the contracts and, within each, a stratified sample of enrollees across low, medium and high risk-score strata. Before 2025 CMS audited about 60 plans a year with about 35 records each. Under the May 2025 expansion, newly initiated audits cover all eligible contracts, about 550 a year, with 35 to 200 records per plan depending on size.
- 2
Record request Weeks
The plan must produce, for each sampled enrollee and each HCC, the "one best medical record" that supports the diagnosis: a dated, signed note from an acceptable provider and face-to-face encounter in the data year. Plans request these from provider groups; the group's ability to find and send a clean note fast is the whole game.
- 3
Coding review Months
Certified coders check each HCC against the record. Outcomes: validated; not validated (no record, unacceptable provider or setting, condition not documented or not assessed, coded to the wrong specificity); or a different HCC than submitted. CMS raised its coder workforce from about 40 to about 2,000 to run this at scale.
- 4
Payment error and extrapolation Findings
For each sampled enrollee the audited score is recomputed without the unsupported HCCs and the overpayment is the difference. The 2023 RADV final rule (88 FR 6643) allowed CMS to extrapolate the sample's error rate to the whole contract from payment year 2018 forward, with no fee-for-service adjuster. See the status box below.
- 5
Appeals and recovery After findings
Plans can appeal medical record review determinations and the payment error calculation through CMS's RADV appeal process. Recoveries flow back down to risk-bearing provider groups under the terms of their contracts, which usually allow the plan to recoup unsupported-diagnosis payments.
Extrapolation status, as of October 2026
On September 25, 2025, the U.S. District Court for the Northern District of Texas vacated the 2023 RADV final rule in Humana Inc. v. Becerra, holding that CMS violated the Administrative Procedure Act when it dropped the fee-for-service adjuster without adequate notice. The government appealed to the Fifth Circuit in November 2025. Until the appeal or a new rulemaking resolves it, CMS can still audit and recover enrollee-level overpayments on sampled records, but contract-wide extrapolation of the audits it accelerated in 2025 is uncertain. CMS's original estimate was $4.7 billion in recoveries over ten years from extrapolation; its audits of payment years 2011–2013 found overpayment rates of 5% to 8%.
The enforcement landscape
CMS RADV
Contract-level, administrative
The May 21, 2025 strategy: audit all eligible MA contracts annually, finish the backlog of payment years 2018–2024 by early 2026, use technology to flag records, and expand coder review. Findings are classified rather than scored in the newer protocol.
HHS-OIG
Targeted audits and data studies
OIG audits specific plans for high-risk diagnosis groups (acute stroke, acute MI, embolism, major depression, vascular claudication, potentially mis-keyed codes) and publishes estimated overpayments. Its work on diagnoses supported only by in-home health risk assessments and chart reviews, with no other encounter in the year, drives the 2027 unlinked chart review exclusion.
DOJ, False Claims Act
Treble damages and per-claim penalties
Submitting or causing the submission of unsupported risk adjustment data, or failing to delete known unsupported codes, has produced settlements with plans and provider groups and ongoing litigation against large plans. The FCA's 60-day overpayment rule applies: an identified unsupported diagnosis is an overpayment that must be returned.
What RADV deletes most often
| Failure | Why it happens | Prevention |
|---|---|---|
| No record produced | The plan could not locate a face-to-face note for the enrollee and HCC in the data year; the diagnosis came from a claim, a chart review or a problem list | Link every submitted diagnosis to a retrievable encounter; keep a record-location index by member and date |
| Unacceptable provider or setting | Diagnosis documented by a nurse, dietitian or on a lab, DME or home health claim | Apply CMS's filtering logic before submission; do not submit what CMS would filter out |
| Condition not assessed | Named in a list but no MEAT; "history of" coded as active | Documentation standards and prospective queries |
| Wrong specificity or hierarchy | Record supports CKD stage 3, claim says stage 4; record supports depression unspecified, claim says major depression | Code from the note, not the list; coder validation audits |
| Unsigned or unauthenticated | Missing signature, wrong author, unsigned telehealth notes | EHR signature controls; monitor unsigned-note queues |
| Date of service outside the window | A note from the prior year submitted as current | Date validation at submission |
Coding intensity, the policy problem behind the audits
MedPAC's March 2026 report estimates Medicare will pay MA plans about 14% more than it would have spent on the same people in fee-for-service in 2026, roughly $76 billion. About $22 billion of that is coding intensity (MA risk scores about 10% higher than comparable FFS before the 5.9% adjustment, about 4% higher after it) and about $57 billion is favorable selection. CMS's May 2025 announcement cited federal estimates of about $17 billion a year in MA overbilling and MedPAC estimates as high as $43 billion. Whatever the true figure, it explains the direction of policy: a flatter model (V28), exclusion of chart-review-only and audio-only diagnoses from 2027, universal RADV, and a coding adjustment that stays at the statutory floor only because Congress set it there.
Compliance program elements for risk adjustment
ACOs and shared savings
An accountable care organization is paid fee-for-service all year and then settled against a benchmark. If the attributed population cost less than the benchmark by at least a minimum savings rate, and quality was good enough, the ACO keeps a share. In two-sided tracks it also repays a share of losses. The Medicare Shared Savings Program (MSSP) is the template; ACO REACH, its successor LEAD, Medicaid and commercial ACOs vary the parameters.
As of January 2025 the MSSP had 477 ACOs, more than 650,000 participating clinicians and 11.2 million assigned beneficiaries. In performance year 2024 the program's 476 ACOs saved Medicare $2.4 billion net ($241 per capita) after paying $4.1 billion in shared savings, and 75% of ACOs earned a payment. Those are the stakes for the arithmetic below.
MSSP tracks (42 CFR 425.605 and 425.610)
| Track / level | Risk | Shared savings rate | Shared loss rate | Loss limit | Savings cap |
|---|---|---|---|---|---|
| BASIC Level A, B | One-sided | Up to 40% | None | None | 10% of benchmark |
| BASIC Level C | Two-sided | Up to 50% | 30% | 2% of participant revenue, capped at 1% of benchmark | 10% of benchmark |
| BASIC Level D | Two-sided | Up to 50% | 30% | 4% of revenue, capped at 2% of benchmark | 10% of benchmark |
| BASIC Level E (Advanced APM) | Two-sided | Up to 50% | 30% | 8% of revenue, capped at 4% of benchmark | 10% of benchmark |
| ENHANCED (Advanced APM) | Two-sided | Up to 75% | 40% | 15% of benchmark | 20% of benchmark |
Glide path tightens for agreements starting 2027
The CY 2026 Physician Fee Schedule final rule (October 31, 2025) caps one-sided BASIC participation at five performance years, all within the ACO's first agreement period, instead of seven years across two periods. Inexperienced ACOs must move to Level E or ENHANCED by their second agreement period. ACOs may start with fewer than 5,000 assigned beneficiaries in early benchmark years but only in BASIC, with capped savings and losses. The health equity adjustment to quality scores is removed from performance year 2026, and the "health equity benchmark adjustment" is renamed the population adjustment (share of dual or LIS beneficiaries).
The settlement, step by step
- 1
Assignment Prospective or retrospective, ACO's choice
Beneficiaries are assigned to the ACO whose participants delivered the plurality of their primary care services (allowed charges), in two steps: primary care physicians first, then specified specialists, NPs, PAs and CNSs. Voluntary alignment through MyMedicare.gov overrides claims. From 2026 the new behavioral health integration add-on services count toward assignment when furnished with advanced primary care management.
- 2
Benchmark Set at the start of the agreement, updated yearly
Three historical years of assigned-beneficiary spending, weighted equally, risk-adjusted with the CMS-HCC model, blended with regional FFS spending, trended by a mix of national and regional growth (the ACPT), adjusted by the prior savings and population adjustments, and capped: positive risk score growth is limited to 3% across the agreement period, so a sicker-coded population does not raise the benchmark indefinitely.
- 3
Performance-year expenditure After a three-month claims run-out
Part A and B spending for assigned beneficiaries, truncated at the 99th percentile per person, with IME, DSH and uncompensated care payments removed, risk-adjusted for the year's population.
- 4
Minimum savings or loss rate Test
One-sided ACOs face an MSR that depends on size: 3.9% at 5,000 beneficiaries down to 2.0% at 60,000 or more. Two-sided ACOs choose 0%, 0.5%, 1%, 1.5%, 2% or the variable rate; the MLR mirrors it. Savings below the MSR earn nothing; losses below the MLR cost nothing.
- 5
Quality Reported by March 31 of the following year
The APP Plus measure set (eCQMs, MIPS CQMs or Medicare CQMs, plus CAHPS and claims-based measures) produces a quality score. Meeting the quality performance standard earns the full sharing rate; otherwise savings are scaled or, for a failure of both the standard and the alternative standard, forfeited, and from 2026 repeated failure can mean termination.
- 6
Payment Late summer, following year
Shared savings = (benchmark − expenditure) × sharing rate × quality factor, capped. Shared losses = (expenditure − benchmark) × loss rate, limited. The ACO distributes savings to participants under its own plan, which must be filed with CMS, and must hold a repayment mechanism (escrow, line of credit or surety bond) for two-sided tracks.
Shared savings calculatorMSSP-style settlement math. Sharing rates, caps and MSR tables follow 42 CFR 425.604–425.610; the revenue-based loss limit is simplified to its benchmark cap.
The other Medicare ACO models
Ending Dec 31, 2026
ACO REACH
The Innovation Center's full- and professional-risk model (Global: 100% of savings and losses; Professional: 50%), with primary care or total care capitation options, a discount applied to the benchmark, and a health equity benchmark adjustment. 74 ACOs are in its final performance year, 2026.
Starts Jan 1, 2027
LEAD: Long-term Enhanced ACO Design
REACH's announced successor, a ten-year voluntary model running through December 31, 2036, with Global and Professional risk options, a long window without rebasing, and a separate CMS Administered Risk Arrangements track for episode-based deals with downstream providers. Applications ran through spring 2026.
Medicaid ACOs
State-designed
Several states run Medicaid ACO programs with their own attribution, benchmark and risk model. Risk adjustment usually rides on the state's managed care model; quality uses Core Set and HEDIS measures.
Commercial ACOs
Contract by contract
Shared savings and risk deals with commercial plans and self-funded employers. Everything (attribution, benchmark, MSR, risk model, quality measures, run-out, data rights) is negotiable and should be written down. The most common dispute is the attribution list.
Star ratings, MIPS and quality measures
Quality in value-based payment is not an abstraction. It is a set of numerators and denominators computed from claims, clinical data and surveys, rolled into a score that moves money. Three programs matter most to revenue cycle teams: Medicare Advantage Star ratings, the Quality Payment Program for clinicians, and the hospital quality programs.
Medicare Advantage Star ratings
What it is
A 1-to-5 rating per contract
Up to 43 measures for MA-PD contracts across HEDIS clinical quality, CAHPS patient experience, HOS outcomes, pharmacy measures (adherence, statin use), complaints, appeals timeliness and call center performance. Each measure is scored 1–5 against cut points, weighted (outcomes and intermediate outcomes 3, patient experience and access 2 since the 2026 ratings, process 1), and averaged, with reward factors for consistently high performers.
Why plans care
The quality bonus and the rebate
Contracts rated 4 stars or higher get a 5 percentage point increase in their county benchmarks (10 points in qualifying "double bonus" counties), and the share of the gap between bid and benchmark that the plan keeps as a rebate is 70% at 4.5+ stars, 65% at 3.5–4.0, and 50% below 3.5 (42 U.S.C. 1395w-23). MedPAC puts quality bonus spending at about $16 billion in 2026. Rebates fund the extra benefits and $0 premiums that win enrollment.
2026 ratings
64% of enrollees in 4+ star plans
The 2026 Star ratings (published October 2025, governing 2027 bonus payments; the 2027 ratings are due in October 2026) rated 516 MA-PD contracts with an enrollment-weighted average of 3.98 stars; 18 MA-PD contracts earned 5 stars, up from 7 the year before. The CY 2027 MA proposed rule would remove 12 administrative measures and add depression screening.
Where providers come in
Gap closure and the data behind it
Most clinical Star measures are computed from the plan's claims and supplemental data about the plan's members, which means from your claims, your lab results and your chart data. Plans pay providers for closed gaps, share gap lists, and increasingly tie capitation or bonuses to measure performance for attributed members.
Quality bonus illustratorHow a Star rating changes a plan's benchmark, rebate and revenue for one member at a 1.0 risk score. Statutory percentages; county figures are fictional.
The Quality Payment Program (MIPS and Advanced APMs)
| Element | 2026 performance year (2028 payment year) | Notes |
|---|---|---|
| MIPS performance threshold | 75 points | Held at 75 through the 2028 performance year. Below it, a negative adjustment up to −9%; above it, a positive adjustment scaled so the program stays budget neutral |
| Category weights | Quality 30%, Cost 30%, Promoting Interoperability 25%, Improvement Activities 15% | Statutory weights since 2022; reweighting applies for small practices and hardship exceptions |
| Reporting pathways | Traditional MIPS; 27 MIPS Value Pathways (6 new for 2026); APM Performance Pathway for ACO participants | MVPs remain optional in 2026; the APP Plus set is required for MSSP ACOs |
| Advanced APM qualifying participant | No MIPS; 0.75% conversion factor update instead of 0.25% | 2026 conversion factor $33.57 for QPs vs $33.40 for everyone else. The lump-sum APM incentive (5%, then 3.5%, then 1.88%) is paid for the last time in 2026 for 2024 performance |
| QP thresholds | 75% of Part B payments or 50% of patients through Advanced APMs | Statutory thresholds; determinations made at both the individual and APM-entity level from 2026 |
| Cost measures | Total per capita cost, MSPB-clinician and episode-based measures | Risk-adjusted with HCC-based models; specialty and MVP-specific episodes expanding |
Hospital and other quality programs
Hospital VBP
2% withhold, redistributed
Inpatient PPS payments are reduced by 2% and redistributed on a total performance score across clinical outcomes, safety, person and community engagement, and efficiency. Zero-sum across hospitals.
HRRP
Up to 3% penalty
Excess 30-day readmissions for six conditions and procedures, risk-adjusted and stratified by dual-eligible share. A penalty only; no bonus.
HAC Reduction
1% penalty, worst quartile
Hospitals in the worst-performing quartile on PSI 90 and CDC infection measures lose 1% of Medicare inpatient payment.
HEDIS
NCQA's measure set
The clinical measures behind Star ratings, Medicaid managed care scorecards and commercial accreditation. Administrative (claims-based) measures reward clean claims with specific codes; hybrid and ECDS measures let supplemental clinical data count, which is why plans want your lab and chart feeds.
CAHPS and HOS
Surveys
Patient experience (CAHPS) and health outcomes (HOS) surveys feed Stars, MSSP quality and MIPS. Weight reduced in Stars from 4 to 2 in the 2026 ratings, but still material, and the only measures that come from patients directly.
Medicaid Core Sets
Mandatory reporting
States report the Child Core Set and the behavioral health measures of the Adult Core Set to CMS; MCOs are scored on them and pass the pressure to providers through pay-for-performance.
Care gaps are a revenue cycle object
A care gap is a member in a measure's denominator who is not yet in its numerator: due for a mammogram, an A1c, a statin, a follow-up after discharge. Gap lists arrive from plans monthly. Closing a gap means delivering the service and, just as important, getting the right code (CPT II codes for results such as 3044F–3046F for A1c, 3074F–3080F for blood pressure) onto a claim or a supplemental data file before the measurement year closes. A service rendered and not coded is an open gap.
Bundled payments and TEAM
A bundled (episode) payment holds one party accountable for everything Medicare spends on a patient from a triggering event through a window after it: the surgery, the hospital stay, the surgeon, the anesthesiologist, the SNF, home health, readmissions and the outpatient visits. Claims are paid as usual; the accounting happens afterward against a target price. The Transforming Episode Accountability Model (TEAM) made this mandatory for selected hospitals on January 1, 2026.
TEAM at a glance
| Element | Rule | Notes |
|---|---|---|
| Dates | Five performance years, January 1, 2026 through December 31, 2030 | Finalized in the FY 2025 IPPS rule, updated in the FY 2026 and FY 2027 IPPS rules |
| Who | Acute care hospitals in selected core-based statistical areas; 716 participants listed by CMS | Mandatory for IPPS/OPPS hospitals in the selected CBSAs; new hospitals may get a one-year deferment; IHS and Tribal hospitals excluded in the FY 2026 rule |
| Episodes | Lower extremity joint replacement, surgical hip and femur fracture treatment, spinal fusion, coronary artery bypass graft, major bowel procedure | Inpatient and, for LEJR and spinal fusion, hospital outpatient initiations |
| Window | Anchor stay or procedure plus 30 days after discharge | Nearly all Part A and B spending in the window counts, with limited exclusions |
| Target price | Three-year baseline (2022–2024 for PY1) of regional episode spending, trended, risk-adjusted, minus a discount | Discount 1.5% for CABG and major bowel, 2% for LEJR, SHFFT and spinal fusion; national normalization from the FY 2026 rule |
| Risk adjustment | Patient age, HCC burden under V28, social risk (Community Deprivation Index, 180-day lookback), hospital bed size | Social risk and bed-size variables were refined in the FY 2026 rule |
| Tracks | Track 1: upside only, 10% stop-gain, PY1 (up to three years for safety-net hospitals). Track 2: 5% stop-gain and stop-loss, for safety-net, rural and certain other hospitals. Track 3: 20% stop-gain and stop-loss | Most hospitals move to Track 3 from PY2 |
| Quality | Composite quality score from hybrid hospital-wide readmission, THA/TKA PRO-PM and PSI 90 (PY1); PSI 90 replaced by falls with injury, postoperative respiratory failure and failure-to-rescue from PY2; information transfer PRO-PM for outpatient episodes from PY3 | CQS adjusts reconciliation by up to 10% on gains (and 15% on Track 2 losses); hospitals with insufficient data receive a neutral 50 |
| Waivers | SNF 3-day stay waiver (qualified SNFs rated 3+ stars; swing beds added), telehealth originating-site waiver, gainsharing with collaborators | Each waiver has its own billing requirements and documentation |
| Low volume | Hospitals with fewer than 31 episodes in the baseline are not held to financial risk | Finalized in the FY 2026 rule |
Episode math, worked example
Fictional PY2026 reconciliation for one LEJR episode at a Track 3 hospital. Real reconciliations aggregate all episodes in the year and apply the stop-loss and stop-gain to the total.
Target
Benchmark to target price
- Regional baseline LEJR episode spending, trended and risk-adjusted for this patient: $26,000
- Discount for LEJR: 2% × $26,000 = $520
- Target price = $26,000 − $520 = $25,480
Actual
Spending in the window
- DRG 470 payment $13,200 + surgeon and anesthesia $2,400 + SNF 9 days $5,100 + home health $2,000 + outpatient and DME $900 = $23,600
- Reconciliation amount = $25,480 − $23,600 = $1,880 in the hospital's favor
- Composite quality score adjustment (say 92nd percentile, 2% reduction of the 10% maximum): $1,880 × (1 − 0.02) = $1,842
Where the money moves
The hospital still received its DRG payment and every other provider still billed fee-for-service. The $1,842 is a separate reconciliation payment (or repayment, if spending had exceeded the target) settled after the performance year. The hospital's levers are the discharge destination (SNF versus home health), readmissions, implant cost under the DRG, and sharing arrangements with the surgeons and post-acute providers through gainsharing agreements that TEAM's fraud and abuse waivers permit.
Bundles beyond TEAM
History
BPCI, CJR, BPCI Advanced
Voluntary and mandatory Medicare bundles from 2013 to 2025 taught the lessons TEAM is built on: mandatory participation avoids selection, 30-day windows are long enough to capture post-acute spending, and regional pricing beats hospital-specific pricing over time. BPCI Advanced ended December 31, 2025.
Commercial
Maternity, joints, spine, oncology
Employer and payer bundles with prospective or retrospective payment. Prospective bundles pay one amount to a convener that pays the parties; retrospective bundles reconcile like TEAM. Each needs a contract that defines the trigger, the window, the exclusions and the data feed.
Coming
Specialty-focused models
The Ambulatory Specialty Model in the CY 2026 PFS rule holds specialists accountable for heart failure and low back pain cost and quality starting 2027, and LEAD's CMS Administered Risk Arrangements let ACOs pass episode risk to downstream providers. Expect more episodes, not fewer.
What the revenue cycle must get right in a bundle
How RCM teams support value-based care
Value-based contracts do not replace the revenue cycle; they add a second ledger to it. The claim still has to be clean and timely. On top of that, the organization needs to know who its attributed patients are, whether every diagnosis and quality result reached the payer, and whether the capitation, bonus and settlement checks match what the contract says they should be.
The four jobs
1 · Data capture and coding accuracy
Get the right codes on the claim
All supporting diagnoses on professional claims (up to 12), specific to the stage and type; CPT II result codes for quality; the correct rendering provider and place of service so the encounter passes CMS filtering. Coding audits scored both ways: missed and unsupported.
2 · Data submission and acceptance
Make sure the payer received it
Encounter submissions, supplemental data files for diagnoses and quality results, chart retrieval for the plan's reviews. Work the rejections (plan EDI edits, CMS encounter data rejections passed back, duplicate and provider-type errors). Track acceptance rates by payer and the deadlines for each run.
3 · Attribution and roster reconciliation
Know who you are paid for
Reconcile the plan's or ACO's attribution file against the practice's active patients every month. Unattributed patients you treat cost money without revenue; attributed patients you have never seen are open risk and quality gaps. Fix PCP assignments, deaths, moves and plan changes quickly.
4 · Financial reconciliation
Check the checks
Capitation = members × risk score × rate, by month, with retroactive adjustments. Settlements = the formulas in this tutorial, computed independently before the payer's statement arrives. Quality bonuses against the measure results you tracked. Dispute within the contract's window.
The calendar
| When | Medicare Advantage risk | MSSP / ACO | Quality |
|---|---|---|---|
| January | Initial payment year run (July–June DOS); build recapture and suspect lists; new rosters | Prospective assignment list for the year; quarterly reports begin | New measurement year; prior-year gap lists close for supplemental data |
| January–March | Late submissions for the prior year's final run | Quality reporting for the prior year due March 31 | Medical record review season for hybrid HEDIS measures |
| April | Rate Announcement for the next payment year (normalization, coding adjustment, model changes) | Benchmark and assignment updates | ACA EDGE server data for the prior benefit year closes |
| July | Mid-year run (prior calendar year DOS); mid-year capitation true-ups | Claims run-out ends; performance year results prepared | Stars preview; plan gap lists for the second half |
| August–October | Final deadline for the prior year's risk adjustment data (plan deadlines earlier) | Shared savings and losses settled; distributions to participants | Star ratings published in October for the next year's bonuses |
| November–December | Last chance to see unseen members with chronic conditions; home visits | Participant list changes for next year certified | Last chance to close care gaps for the measurement year |
| The following June | Final run for the payment year; final settlements with delegated groups | Final reconciliation of prior performance year | Final HEDIS submissions |
KPIs for a value-based revenue cycle
| Measure | Formula | Why it matters |
|---|---|---|
| Recapture rate | Prior-year HCCs re-documented this year ÷ prior-year HCCs | The share of last year's score you have already secured; aim for most of it by mid-year |
| Annual visit completion | Attributed members with a qualifying face-to-face visit ÷ attributed members | No visit, no HCC, no gap closure |
| Encounter acceptance rate | Encounters accepted by the plan (and CMS) ÷ encounters submitted | Rejected encounters are invisible diagnoses |
| Coding accuracy, two-sided | Validated HCCs ÷ audited HCCs; and missed HCCs ÷ audited charts | Audit exposure and lost revenue, measured together |
| Average RAF, attributed population | Mean normalized score, trended by month and by site | Watch for drops (unseen members) and implausible rises (audit risk) |
| Care gap closure rate | Gaps closed ÷ gaps open at start of year, by measure | Drives Star ratings, ACO quality score and bonus payments |
| Attribution match rate | Attributed members in the practice's active panel ÷ attributed members | Unmatched members are unmanaged risk |
| Settlement variance | Payer's settlement − internally computed settlement | Every variance is either a data problem or a dispute |
| Medical loss ratio (capitated) | Medical cost for attributed members ÷ capitation revenue | The P&L of the risk contract; needs IBNR reserves |
Supplemental data and chart retrieval
- Supplemental diagnosis data
- Diagnoses the plan submits to CMS from sources other than the original claim, including chart reviews, which must be linked to a face-to-face encounter. From 2027 unlinked chart review diagnoses no longer count for payment, so a chart review must point to a real encounter.
- Supplemental quality data
- Lab results, vitals, screening results and immunizations sent to the plan in a standard file so HEDIS hybrid and ECDS measures can use them without a chart pull. The fastest way to close administrative gaps you have already met clinically.
- Chart retrieval
- Plans request records for RADV, HEDIS medical record review and their own retrospective reviews. Agree the method (EHR access, HIE, secure upload), the turnaround and who pays. Count the requests; they peak January to May.
- Interoperability
- CMS-0057-F's Payer-to-Payer and Provider Access APIs (January 1, 2027) and bulk FHIR exchange are gradually replacing file drops and faxes for both diagnosis and quality data. Ask what your plans and your EHR support.
The line the revenue cycle must not cross
RCM teams improve capture by making sure what the clinician documented gets coded, submitted and accepted. They do not decide what the patient has. Suspect lists go to clinicians as questions; queries follow the compliant query rules; retrospective reviews delete as readily as they add; and nobody is paid per HCC. A program that treats RAF as a revenue target rather than an accuracy target is building the evidence for its own False Claims Act case.
Rules and regulations
Risk adjustment and value-based payment are governed by the Social Security Act, CMS's annual notices, a set of program regulations in 42 CFR, and the fraud and abuse laws that apply to any federal payment. These are the ones to know.
Statute · Medicare Advantage
Social Security Act §1853 (42 U.S.C. 1395w-23)
Requires risk adjustment of MA payments for health status (§1853(a)(1)(C)), sets the MA coding pattern adjustment with a statutory minimum of 5.9% from 2019 (§1853(a)(1)(C)(ii)), establishes county benchmarks at 95% to 115% of FFS spending, the quality bonus for 4+ star plans and the 50/65/70% rebate percentages (§1853(n) and (o)). Together with §1857 it gives CMS the authority to audit MA organizations' risk adjustment data and recover overpayments, implemented through RADV at 42 CFR 422.311.
Regulation · Risk adjustment data
42 CFR 422.310 and 422.311
MA organizations must submit risk adjustment data that conform to CMS's requirements, certify its accuracy, and submit encounter data; CMS may audit and recover. 422.311 lays out RADV: medical record review, payment error calculation, the appeal process, and (as amended by the 2023 rule, now vacated and on appeal) extrapolation from payment year 2018 without a fee-for-service adjuster. Related: 42 CFR 422.504(l) (certification of data) and the annual Advance Notice and Rate Announcement issued under 42 CFR 422.312.
Rate Announcements · 2026 and 2027
CY 2026 (April 7, 2025) and CY 2027 (April 6, 2026)
2026: 100% of the 2024 CMS-HCC model, normalization factor 1.067, coding adjustment 5.90%, PACE at 10% new model, expected revenue change +5.06% before coding trend. 2027: same model retained, PACE moves to 50/50, diagnoses from unlinked chart review records and audio-only encounters excluded (an estimated −1.53% impact), coding adjustment 5.90%, expected revenue change +2.48%.
Regulation · Shared Savings Program
42 CFR Part 425
Eligibility and governance (Subpart B), assignment (Subpart E, §425.400–425.404), benchmarks and the MSR/MLR tables (§425.601–425.605 and 425.610), quality (§425.500–425.512), the BASIC and ENHANCED track terms, repayment mechanisms (§425.204), and the program integrity provisions. Updated annually in the Physician Fee Schedule rule; the CY 2026 rule (CMS-1832-F) made the 2026 and 2027 changes described in Chapter 07.
Regulation · Quality Payment Program
42 CFR Part 414, Subpart O (MIPS and APMs)
Implements MACRA's Quality Payment Program: MIPS scoring, performance threshold (75 points for 2026), payment adjustments (up to −9%), MVPs, the APM Performance Pathway, qualifying participant thresholds, and the differential conversion factor update (0.75% for QPs, 0.25% for others) that replaced the lump-sum APM incentive.
Regulation · Innovation Center models
42 CFR Part 512, Subpart B (TEAM); SSA §1115A
TEAM's rules (participants, episodes, target pricing, tracks, quality, waivers, beneficiary protections) are codified at 42 CFR 512.500 et seq. from the FY 2025 IPPS rule as amended. Section 1115A authorizes the Innovation Center to test models like ACO REACH, LEAD and the Ambulatory Specialty Model and to waive Medicare rules (SNF 3-day stay, telehealth, gainsharing) for participants.
Regulation · ACA risk adjustment
45 CFR Part 153 (Subparts D, G, H)
The permanent HHS-operated risk adjustment program for the individual and small group markets: the HHS-HCC model, the EDGE server data requirements, the user fee, the transfer formula, and HHS-RADV. Parameters are set each year in the Notice of Benefit and Payment Parameters.
Regulation · Medicaid managed care
42 CFR 438.4–438.7 (actuarial soundness and rate setting)
State capitation rates to MCOs must be actuarially sound and may be risk-adjusted using a methodology described in the rate certification. Encounter data requirements (§438.818) make MCO and provider data submission a condition of the federal match.
Fraud and abuse
False Claims Act (31 U.S.C. 3729–3733) and the 60-day rule (42 U.S.C. 1320a-7k(d))
Knowingly submitting unsupported risk adjustment data, or knowingly retaining an overpayment that results from it, is a false claim: treble damages plus per-claim penalties (inflation-adjusted annually), and qui tam exposure. The 60-day rule (42 CFR 401.305, revised in the CY 2025 PFS rule) requires return within 60 days of identifying an overpayment, with up to 180 days to investigate related overpayments. The Anti-Kickback Statute and Stark law govern gainsharing and incentive arrangements inside VBC; the 2020 value-based safe harbors and exceptions (42 CFR 1001.952(ee)–(gg), 42 CFR 411.357(aa)) protect qualifying arrangements.
Coding standards
ICD-10-CM Official Guidelines for Coding and Reporting, FY 2027
Adopted under HIPAA (45 CFR 162.1002), the guidelines govern which diagnoses may be reported: no uncertain diagnoses in the outpatient setting, code to the highest specificity documented, report chronic conditions only when they affect care, history-of versus current. The AHIMA/ACDIS compliant query practice brief is the industry standard for queries, and CMS's Risk Adjustment Data Validation medical record reviewer guidance explains how RADV coders apply the guidelines.
Key numbers 2026–27
| Item | Figure | Context |
|---|---|---|
| Medicare Advantage risk adjustment | ||
| 2024 CMS-HCC model (V28) share of the 2026 score | 100% | 33% in 2024, 67% in 2025; retained unchanged for 2027 |
| Payment HCCs in V28 (vs V24) | 115 (vs 86) | About 2,000 ICD-10-CM codes lost payment mapping |
| Part C normalization factor, 2026 | 1.067 | 2024 CMS-HCC model; 2017 model (PACE) 1.187 |
| MA coding pattern adjustment | 5.90% | Statutory minimum, applied in 2026 and 2027 |
| RxHCC normalization, 2026 | 1.194 / 0.887 | MA-PD / PDP |
| Expected MA revenue change, 2026 | +5.06% | Over $25 billion; coding trend of +2.10% excluded |
| Expected MA revenue change, 2027 | +2.48% | Over $13 billion; unlinked chart review exclusion −1.53% |
| PACE blend | 10% / 50% | Share of the new model in 2026 / 2027 |
| PY2026 data windows | Jul 2024–Jun 2025; CY2025; CY2025 | Initial (Jan 2026), mid-year (Jul 2026), final (Jun 2027) runs |
| Scale and spending | ||
| MA enrollment, 2026 | 35.2 million | 55% of beneficiaries with Parts A and B; 23% in SNPs (KFF) |
| MA payments above FFS, 2026 (MedPAC) | 14% · $76 billion | Coding intensity $22 billion; favorable selection $57 billion |
| MA risk scores vs FFS (MedPAC) | ~10% before, ~4% after | Before and after the 5.9% coding adjustment |
| Quality bonus spending, 2026 (MedPAC) | ~$16 billion | 64% of enrollees in bonus plans |
| Average rebate per MA enrollee | $2,660 per year | $222 per month; average capitation $16,242 per year |
| RADV and enforcement | ||
| Contracts audited per year (new audits) | ~550 (from ~60) | 35–200 records each (from 35); CMS May 21, 2025 |
| CMS coder workforce | ~2,000 (from 40) | Target September 1, 2025 |
| Backlog target | PY2018–2024 by early 2026 | Extrapolation rule vacated Sept 25, 2025; appeal pending |
| Overpayment rate, PY2011–2013 audits | 5–8% | Projected $4.7 billion over 10 years from extrapolation (2023 rule) |
| ACOs | ||
| MSSP net savings, PY2024 | $2.4 billion | 476 ACOs; 75% earned payments; $4.1 billion paid out |
| MSSP size, Jan 2025 | 477 ACOs · 11.2 million | More than 650,000 clinicians |
| Sharing rates | 40% / 50% / 75% | BASIC A–B / BASIC C–E / ENHANCED |
| One-sided MSR range | 2.0%–3.9% | By assigned beneficiaries; two-sided choose 0–2% |
| One-sided BASIC limit, agreements from 2027 | 5 years | Down from 7 years across two agreement periods |
| ACO REACH ends / LEAD starts | Dec 31, 2026 / Jan 1, 2027 | LEAD runs 10 years to Dec 31, 2036; 74 REACH ACOs in 2026 |
| Quality programs | ||
| MIPS performance threshold, 2026 | 75 points | Through the 2028 performance year; max penalty −9% |
| MIPS Value Pathways, 2026 | 27 | 6 new; optional |
| Conversion factor, 2026 | $33.57 / $33.40 | QPs / non-QPs (+0.75% / +0.25% statutory updates) |
| Star rating bonus and rebates | +5 pts · 50/65/70% | Benchmark bonus at 4+ stars; rebate by rating tier |
| 2026 Star ratings | 64% · 3.98 · 18 | Enrollees in 4+ star MA-PDs · average · 5-star MA-PD contracts (516 rated) |
| TEAM | ||
| Dates | Jan 1, 2026–Dec 31, 2030 | 716 hospital participants; five episode categories; 30-day window |
| Discounts | 1.5% / 2% | CABG and major bowel / LEJR, SHFFT, spinal fusion |
| Stop-gain / stop-loss | 10% / 5% / 20% | Track 1 (gain only) / Track 2 / Track 3 |
| Low-volume threshold | 31 episodes | In the baseline period |
What is changing
In effect 2026
V28 complete, TEAM live, universal RADV
Scores are 100% on the 2024 model; mandatory surgical bundles started January 1; every MA contract is in the audit cycle. Documentation and record retrieval are now continuous obligations.
Coming 2027
Chart-review and audio-only exclusions, LEAD, tighter MSSP glide path
Unlinked chart review and audio-only diagnoses stop paying; LEAD replaces REACH; new MSSP agreements get five years one-sided at most; CMS-0057-F APIs go live; Ambulatory Specialty Model begins.
Unsettled
RADV extrapolation
The 2023 rule is vacated and on appeal at the Fifth Circuit. Enrollee-level recoveries continue; contract-level extrapolation back to PY2018 depends on the appeal or a new rule. Plan for the stricter outcome.
Sources
- CMS fact sheet: 2026 Medicare Advantage and Part D Rate Announcement (April 7, 2025)
- CMS: Announcement of CY 2026 MA capitation rates and Part C and D payment policies (normalization factors, coding adjustment)
- CMS: CY 2026 risk adjustment implementation memo (model blends, run dates)
- CMS fact sheet: 2027 Medicare Advantage and Part D Rate Announcement (April 6, 2026)
- CMS press release: aggressive strategy to enhance and accelerate Medicare Advantage audits (May 21, 2025)
- WilmerHale: decision vacating the CMS RADV rule (Humana v. Becerra, N.D. Tex., Sept. 25, 2025)
- Healthcare Dive: administration appeals decision vacating the RADV rule (Nov. 2025)
- MedPAC March 2026 Report to Congress, Chapter 12: the Medicare Advantage program
- KFF: Medicare Advantage in 2026, enrollment update and key trends
- CMS fact sheet: Medicare Shared Savings Program changes in the CY 2026 PFS final rule
- Fierce Healthcare: MSSP ACOs saved $2.4 billion in 2024
- CMS Innovation Center: LEAD (Long-term Enhanced ACO Design) model
- CMS Innovation Center: Transforming Episode Accountability Model (TEAM)
- CMS: TEAM overview (discounts, tracks, quality measures)
- CMS fact sheet: CY 2026 Physician Fee Schedule final rule (conversion factors)
- CMS: 2026 Quality Payment Program final rule fact sheet (MIPS threshold, MVPs)
- CMS: 2026 Medicare Advantage and Part D Star Ratings fact sheet
Glossary
Knowledge check
Eight questions drawn from the chapters above. Pick an answer to see the explanation.
Score: 0 / 8
For training and reference. This page summarizes general rules and does not replace payer contracts, plan documents, CMS manuals, the CMS-HCC model software, state law or legal advice. Risk coefficients in the calculator are illustrative, not CMS values. Example names, IDs, rates and payers are fictional.
Figures checked October 15, 2026.