Cut through the noise, find the signal: Intelligence for serious healthcare leaders
How Medicare Advantage, ACOs and ACA plans get paid for how sick their members are — and where the scrutiny is heading.
MedPAC estimates the federal government pays MA plans 20% more per person than it spends on similar people in traditional Medicare, totaling approximately $84 billion in excess payments in 2025.
Risk adjustment is the primary mechanism through which this overpayment occurs. CMS has multiple levers to address this: the coding intensity adjustment (currently at the 5.9% statutory minimum), RADV audits, model recalibration, and the proposed exclusion of unlinked chart reviews.
The question for MA plan executives is not whether CMS will continue tightening — it is how fast and how far.
An entire industry of retrospective chart review vendors has grown up around MA risk adjustment, generating billions of dollars in additional risk score value by mining medical records for diagnoses that were documented but never submitted.
The 2027 Advance Notice’s proposal to exclude unlinked chart reviews would effectively end this business model as a risk-score generator. The $7.2 billion impact CMS estimates represents the revenue currently flowing through this channel.
Organizations that have built their risk adjustment strategy around retrospective chart mining need to pivot rapidly toward prospective, encounter-based documentation improvement.
CMS is steadily aligning the risk adjustment approaches across its payment models. MSSP uses the same CMS-HCC model as MA for benchmark risk adjustment. ACO REACH uses CMS-HCC prospective for Standard ACOs and introduced a CMMI-HCC concurrent model for High Needs ACOs.
The incoming LEAD Model (replacing REACH in 2027) will further integrate specialist-focused risk arrangements. For health systems operating across multiple CMS programs, this convergence means a single CDI strategy can serve all programs simultaneously — but it also means that risk adjustment weaknesses are amplified across all payment streams.
Neither the CMS-HCC nor HHS-HCC models directly account for social determinants of health (SDOH). A beneficiary with diabetes who also faces food insecurity, housing instability, and transportation barriers will have higher actual costs than the model predicts, but these factors do not generate HCC codes.
MSSP has attempted to address this indirectly through the population and income adjustment. Some policy experts have proposed incorporating Z-codes (ICD-10 codes for social determinants) into future risk adjustment models. Organizations serving high-need, high-SDOH populations should track this policy development closely.
AI and NLP tools are rapidly becoming standard infrastructure for risk adjustment. AI-powered ambient listening tools can capture diagnoses during patient conversations. Computer-assisted coding platforms flag missed HCCs in real time. Machine learning algorithms predict which patients are likely to have undocumented conditions based on medication patterns and utilization data.
CMS itself is deploying AI to flag unsupported diagnoses in RADV audits. The organizations that adopt these tools early will have a dual advantage: more accurate risk scores AND better audit defensibility.
V28 did not affect all conditions equally. Some categories saw significant new weight (e.g., certain cancer categories, more granular diabetes classifications), while others lost value entirely (e.g., some forms of peripheral vascular disease, certain lower-severity conditions).
Organizations should analyze their population’s condition mix against V28 weightings to identify which conditions now carry disproportionate value and where documentation improvement will yield the highest ROI.
HCC codes must be re-documented every calendar year to count toward risk scores. A patient with well-managed diabetes in Year 1 must have that diabetes documented again in Year 2 for it to generate a risk score.
This creates a massive annual recapture challenge for MA plans, with significant revenue at stake if patients do not have encounters where chronic conditions are documented. The proposed exclusion of unlinked chart reviews in 2027 will make this problem even more acute.
Organizations need to shift to prospective, point-of-care recapture strategies driven by their EHR systems and care management workflows.