Risk Adjustment Has to Move Upstream

by | Oct 18, 2026 | White Papers | 0 comments

Executive point of view · Medicare Advantage

MA plans are unlikely to find the next margin gain by coding harder after the fact. The better opportunity comes earlier: use clinical data during the encounter, document what is actually managed and carry that evidence into quality, payment integrity and audit work.

By Health Sigma Academy

Evidence current through October 16, 2026 · Written for MA plan CEOs, CFOs and enterprise risk leaders

114%
MedPAC’s 2026 estimate of MA spending versus expected FFS spending; $76B above FFS.
72%
of MA plans operated at negative EBITDA in 2024, according to McKinsey.
100%
V28 weighting for non-PACE organizations in payment year 2026; the three-year phase-in is complete.
Jan 2027
deadline for four FHIR APIs under CMS-0057-F.

01 · The thesis: risk adjustment belongs in the clinical workflow

Medical-cost pressure, the V28 transition, tighter documentation rules and volatile quality payments are now hitting at the same time. A retrospective revenue-recovery program is poorly suited to that environment.

“Risk adjustment should begin with the clinical encounter, not with a chart-retrieval campaign months later.”

The operating model needs timely data, clear clinical ownership and an evidence trail that can be reused for care, quality, payment and audit review.

For CEOs

Risk, quality, utilization management and payment integrity draw on much of the same clinical information. Managing them as a connected portfolio reduces duplicate work and competing requests to providers.

For CFOs

Coding yield is only part of the return. The business case also has to account for documentation quality, audit exposure and the cost of retrieving and reviewing the same record more than once.

For clinical leaders

Technology can surface evidence and uncertainty, but the clinician remains responsible for deciding whether a condition is present, evaluated and managed.

For technology leaders

A normalized clinical event should be captured once and made available to prior authorization, care management, risk adjustment, quality and audit teams.

02 · Evidence base: what the current record shows

Payment policy, litigation and Star Ratings all point in the same broad direction: cost pressure will persist, documentation will face more scrutiny and the rules will continue to move. Plans need an operating model that can hold up through those changes.

Topic Current evidence Executive implication
MA versus FFS MedPAC estimates that MA spending in 2026 will equal 114% of expected FFS spending, or $76 billion above FFS. Its methodology changed between reports, limiting simple trend comparisons. Affordability will remain central to payment policy. Plans will need to show that their risk scores are accurate and clinically supported.
Medicare spending CMS expenditure data show Medicare spending rising from $309 billion in 2004 to $1.118 trillion in 2024; 2025 spending was projected at $1.204 trillion. At Medicare’s scale, small gains in clinical accuracy, administrative efficiency and payment integrity can have meaningful financial value.
Rate environment CMS projected its final policies would increase MA payments by 5.06% in CY2026 and by 2.48% in CY2027; the latter rises to 4.98% when estimated risk-score trend is included. A positive rate update does not solve medical-cost trend, benefit pressure or weak documentation. Those issues still have to be managed together.
RADV and audits A federal court vacated the RADV extrapolation rule in September 2025, and CMS appealed. Separately, CMS announced plans to audit all eligible MA contracts and materially expand its coder workforce. Whatever happens on appeal, plans need encounter-level provenance and records that are ready for review.
Star Ratings equity policy CMS did not implement the Health Equity Index/EHO4All in the CY2027 final rule; the historical reward factor continues. Investment should follow finalized measures without making the operating model too rigid to accommodate later policy changes.

The practical lesson: policy will keep moving. CMS can tighten acceptable documentation while pausing a model recalibration, and a court can interrupt one recoupment method even as audit volume grows. Plans should build controls that remain useful across more than one version of the rules.

03 · Converging forces: these pressures are likely to last

Four changes are reinforcing one another: tighter margins, a new risk model, closer scrutiny of documentation sources and federal interoperability deadlines. Each one raises the cost of waiting until after the encounter to assemble the record.

1

Margin compression

McKinsey reports that MA EBITDA approached breakeven overall in 2024, with about 72% of plans below zero. Its outlook calls for a gradual recovery of 100 to 150 basis points beginning in 2026 and margins of roughly 2% by 2029.

2

Model modernization

V28 reached 100% in 2026 after 33% and 67% blends in 2024 and 2025. It expands payment HCCs, removes or reclassifies thousands of diagnosis codes and changes coefficients. Historical crosswalks are no longer a sound operating foundation.

3

Documentation source scrutiny

The CY2027 final policy excludes unlinked chart-review diagnoses and audio-only encounter diagnoses from risk scores, with an exception for unlinked chart-review records when a beneficiary switches MA organizations. Separately, an OIG analysis associated HRA-only or HRA-linked chart-review diagnoses with an estimated $7.5B in 2023 payments.

4

Interoperability deadlines

CMS-0057-F shortened non-drug prior-authorization decisions for MA plans to 72 hours for expedited requests and seven calendar days for standard requests in 2026. Four FHIR APIs are required in 2027.

Signals worth watching

CMS kept the 2024 model for CY2027.The agency declined to proceed with the proposed Part C model recalibration. The question is when, and on what data, it returns.
More contracts will be audited.CMS announced plans to audit all eligible MA contracts, about 550 each year, and increase its coder workforce from about 40 to about 2,000.
Clinical outcomes carry more weight in Stars.Patient experience, complaints and access measures moved from weight 4 to 2 for the 2026 ratings. An AJMC contributor analysis estimates that these measures now account for 22% of the Star Rating.
The MA population is getting older.Nearly half of the MA-eligible population is projected to be 75 or older by 2030, up from roughly 40% when McKinsey published its 2023 outlook.

04 · Operating model: start with the clinical encounter

A retrospective program asks which codes can be recovered after the visit. A stronger model asks whether the condition was supported, addressed at the right time and recorded in a form that other teams can use.

Retrospective recovery

Batch claims and delayed chart retrieval

Broad suspect lists and repeated vendor handoffs

Post-visit addenda and disconnected reviews

Incremental RAF as the main measure of success

Evidence assembled only after an audit begins

 →

Prospective evidence

Clinical context available before and during the visit

Condition-level evidence with uncertainty shown

Clinician evaluation, management and attestation

Revenue, quality and exposure measured together

Traceability created with the clinical event
Normalized clinical event  →  Shared evidence and governance layer
PRIOR AUTHDecision context, reasons and turnaround evidence CARE MANAGEMENTTransitions, acuity and intervention signals RISK ADJUSTMENTEncounter-linked diagnosis support and gaps QUALITY & STARSCare-gap closure and clinical outcomes PAYMENT INTEGRITYPre-submission validation and audit traceability
A necessary limit: no platform can make a diagnosis “audit-proof.” AI can retrieve evidence, prioritize gaps and test consistency. It cannot replace clinical judgment or supply support that is missing from the record. Human review, source provenance and measured false-positive rates should be built into the workflow from the start.

05 · Six decisions for the next 18 months

The immediate work reaches well beyond coding. Leadership has to decide who owns the result, which data can be reused, where to invest and how performance will be proved.

01

Use a risk-adjusted contribution view.

Connect premium, medical cost, Stars, risk-score durability, audit reserve and administrative expense at the contract and cohort levels.

03

Invest in the evidence layer first.

Normalize identity, encounters, provenance, terminology and timestamps before adding more prompts at the point of care. A faster interface cannot repair weak data lineage.

05

Give clinicians authority over AI-supported work.

Measure precision, confidence, source traceability, clinician feedback and drift at the condition level. Automation should reduce false positives and provider abrasion as well as produce useful leads.

02

Track capture and substantiation separately.

Report suspected, clinically evaluated, submitted, accepted and later validated conditions as distinct stages. A single “opportunity” number hides important differences among them.

04

Use CMS-0057-F to shape the architecture.

Build the 2027 API work so that care, quality and risk teams can use it. Treating the rule as a stand-alone compliance project would create another silo.

06

Remove duplicate work.

McKinsey cites administrative cost below $80–$100 PMPM as a benchmark, not a guarantee. Set a plan-specific baseline, then retire duplicate retrieval, abstraction and workflow costs only when service levels hold.

Where the next investment should go.

Retrospective coding work can still produce a near-term return. But the more durable investment is an operating model that improves what is known and documented while care is being delivered. That capability is better suited to V28, broader audits, tighter source rules and FHIR-based data exchange.

The standard should be clinical accuracy and a record that can be traced from source to submission. Plans that create that evidence early will have more options than those trying to reconstruct it later.

06 · How Health Sigma can help

Health Sigma brings near-real-time FHIR-based clinical data retrieval, acuity analytics and risk-adjustment operations into one platform. The aim is to replace repeated handoffs with a traceable process from source information to CMS submission.

One clinical information backbone, used across the organization.

Health Sigma links FHIR-based clinical data retrieval with Risk Disconnect Analytics and an integrated operating workflow. Charts and discrete resources can support risk adjustment and other teams without being retrieved and normalized again for each use case.

01 FHIR-based clinical data retrieval.

A standards-based extraction layer retrieves and normalizes charts alongside discrete FHIR resources, including diagnostic reports, medications, upcoming appointments and care plans, on a near-real-time basis. Analyzed in context, that information becomes actionable across prior authorization, care and disease management, risk adjustment, Stars and HEDIS — creating a common clinical information backbone rather than a separate feed for each program.

02 Risk Disconnect Analytics.

Longitudinal clinical evidence, claims and other relevant signals are brought together to estimate a “true risk score” that more closely reflects the member’s health acuity. Teams can focus on meaningful gaps and identify conditions that may not have enough support.

03 An integrated path to CMS submission.

The platform manages clinical data retrieval, evidence normalization, analytics, clinical review, coding validation and CMS submission in one workflow. Source provenance and status remain visible as the work moves from one team to the next.

04 Clinical governance.

Human review, condition-level precision, source traceability, clinician feedback and false-positive monitoring are part of the process. Performance can then be assessed through substantiation, cycle time, provider abrasion and audit findings.

07 · What remains unsettled

Several important variables are still open. Plans should model the plausible outcomes rather than build a forecast around a single policy assumption.

RADV extrapolation

The appeal of the September 2025 decision remains unresolved. Plans should model multiple legal outcomes while maintaining documentation controls.

Next model recalibration

CMS shelved its proposed CY2027 recalibration. The timing, data years and coefficient impact of the next proposal remain open.

Overpayment methodology

MedPAC’s estimates are contested, and its methodology changed materially between reports. Treat the 114% figure as a current estimate, not an uncontested measure of plan efficiency.

AI economics

Claims of broad AI savings are not substitutes for a plan-specific business case. Validate against coding accuracy, clinical abrasion, cycle time, audit findings and displaced operating cost.

Sources

  1. Medicare Payment Advisory Commission, “Chapter 12: The Medicare Advantage Program: Status Report,” in Report to the Congress: Medicare Payment Policy, March 2026.
  2. McKinsey & Company, “Healthcare trends for 2026 and beyond,” January 12, 2026.
  3. Centers for Medicare & Medicaid Services, “Healthcare Spending Growth Rate Continues to Decline in 2004,” press release, January 10, 2006.
  4. Centers for Medicare & Medicaid Services, “NHE Fact Sheet,” Historical NHE, 2024.
  5. Centers for Medicare & Medicaid Services, “2026 Medicare Advantage and Part D Rate Announcement,” April 7, 2025.
  6. McKinsey & Company, “The future of Medicare Advantage,” published July 11, 2023; updated December 12, 2024.
  7. Centers for Medicare & Medicaid Services, “2026 Medicare Advantage and Part D Rate Announcement,” April 7, 2025.
  8. Centers for Medicare & Medicaid Services, “2027 Medicare Advantage and Part D Rate Announcement,” April 6, 2026.
  9. Crowell & Moring LLP, “Court Vacates CMS’s 2023 Final Rule on RADV Audits,” September 26, 2025, and “CMS Appeals Humana v. Becerra,” November 26, 2025.
  10. Centers for Medicare & Medicaid Services, “CMS Rolls Out Aggressive Strategy to Enhance and Accelerate Medicare Advantage Audits,” press release, May 21, 2025.
  11. Centers for Medicare & Medicaid Services, “Contract Year 2027 Medicare Advantage and Part D Final Rule,” April 2, 2026.
  12. Centers for Medicare & Medicaid Services, “CMS Interoperability and Prior Authorization Final Rule CMS-0057-F,” January 17, 2024.
  13. Centers for Medicare & Medicaid Services, “CMS Interoperability and Prior Authorization Final Rule CMS-0057-F,” January 17, 2024.
  14. HHS Office of Inspector General, “Medicare Advantage: Questionable Use of Health Risk Assessments Continues To Drive Up Payments to Plans by Billions,” October 2024.
  15. Centers for Medicare & Medicaid Services, “2026 Medicare Advantage and Part D Star Ratings,” November 18, 2025; AJMC, “Contributor: The Stars Have Realigned (Again)—What Medicare Advantage Plans Need to Know,” March 6, 2024.
  16. U.S. Department of Justice, “The Villages Health System LLC Agrees to $541.5M Settlement to Resolve False Claims Act Allegations,” August 26, 2026.
  17. U.S. Department of Justice, “Aetna Agrees to Pay $117.7 Million to Resolve False Claims Act Allegations,” March 11, 2026.
  18. CMS Office of the Actuary / Health Affairs, “National Health Expenditure Projections, 2025–34: Strong Utilization Growth Initially, Legislative Impacts Later,” January 2026, Exhibit 3.