MedPAC vs. CMS Coding Intensity Estimates: Why 4% and 2% Are Both Right
A 2026 *Health Affairs* Forefront piece by Michael Chernew, Andy Johnson, Paul Masi, and Karen Stockley tackles an apparent contradiction: MedPAC estimates MA coding intensity at roughly 4% for 2026, while a CMS staff analysis put uncorrected coding at just 1.5–2.0%. Both are credible. They measure different things. (Source: Health Affairs Forefront, "Aligning MedPAC And CMS Estimates Of Coding Intensity")
The two methods
MedPAC's approach: measure uncorrected coding intensity under the risk model *actually in effect* for the current payment year. For 2026 — the first year V28 is fully phased in — that's about 4%, down from roughly 10% in 2022 under V24.
CMS staff approach: ask a counterfactual — what *would* coding intensity have been in 2022 had V28 already been in effect? Answer: 1.5–2.0%, versus ~10% under the V24 model that was actually in place.
The CMS number applies a new model to an old year. The MedPAC number applies the current model to the current year. Neither is wrong; they're answering different questions.
The trend is the reconciliation
Coding intensity grows over time — plans get better at capture every year. So a 2022 counterfactual (1.5–2.0%) and a 2026 current-year estimate (4%) are consistent once you account for four years of coding-intensity growth layered on top of the model change.
Two forces, both real:
- The model change cut coding intensity dramatically — V28 removed or constrained the diagnoses that drove much of the MA–FFS gap.
- Coding intensity regrows — plans and providers adapt to whatever model is in force.
Why anyone outside actuarial should care
Coding intensity isn't an abstraction. Because FFS claims calibrate the risk model *and* set MA benchmarks, excess MA coding raises program-wide spending — and therefore beneficiaries' Part B premiums, which cover roughly 25% of Part B costs. That's why Congress mandates the coding-intensity adjustment (currently 5.9%) and why MedPAC keeps arguing it doesn't fully offset the gap.
The Part D parallel
The same dynamic applies to RxHCC: CMS updates the model, coding patterns adapt, the gap reopens, CMS updates again. The CY2027 RxHCC refresh — new segments, re-derived coefficients, audio-only exclusions — is the Part D turn of this cycle. Plans should assume today's coding advantage has a half-life, and build documentation quality rather than coding aggression as the durable asset.
*Read next: MedPAC's V28 first-year findings — the 18%-to-9% gap and the $10B payment effect.*