A chart-pull letter used to be a rare event. Now it arrives on a schedule. Payment Year 2026 runs entirely on the CMS-HCC V28 model, and CMS has told Medicare Advantage plans to expect audits every quarter. For the physician practice sitting behind that plan’s provider network, the paperwork just changed.

What Is Driving the RADV Audit Surge Physician Practices Are Feeling in 2026?

CMS confirmed in a January 2026 memo that Payment Year 2020 audits began that February, with new audits launching roughly every three months. Sample sizes now run 35 to 200 enrollees per contract, and CMS intends to reach every eligible plan. Extrapolated recoveries are paused pending appeal, but per-sample repayment demands continue without interruption.

That pause matters less than it sounds. A federal court vacated the extrapolation method in September 2025, so a plan cannot yet project a 10 percent sample error rate across 50,000 members. But CMS still recovers dollars on every unsupported diagnosis it finds in the sample itself, and the appeal could restore extrapolation at any point. Plans are not waiting to find out. They are pushing chart requests down to the practices that generated the original diagnosis codes, because the medical record, not the claim, is what validates an HCC.

Physicians rarely see a RADV notice with CMS letterhead. They see a records request from a payer’s risk adjustment vendor, often bundled with dozens of other patient names, due back within a tight window. That request is the practical face of RADV for most practices, and it is the piece most competing content skips entirely.

The restored five-month record submission window gives plans, and by extension practices, slightly more breathing room than the accelerated 2025 timeline threatened. But a wider window does not fix a practice that cannot locate a signed note, cannot confirm which provider saw the patient, or stores paper charts in a format a vendor cannot process quickly. The clock still runs, and a missed deadline is treated the same as a missing diagnosis.

Why Does the Shift From V24 to V28 Raise Documentation Risk for Practices?

V28 removed more than 2,000 previously valid diagnosis codes and recalibrated how remaining codes map to Hierarchical Condition Categories. Coefficients shifted across diabetes, cardiovascular, and behavioral health categories. A note that supported a valid HCC under V24 can map to nothing, or to a lower-weighted category, under V28’s tighter rules.

The industry shorthand for this shift is coding intensity versus coding integrity. V24 tolerated some documentation looseness because volume alone often carried a diagnosis through validation. V28 closes that margin. Every code now needs a face-to-face encounter, a definitive assessment, and language specific enough to survive a reviewer’s checklist, not just a plausible mention buried in a problem list.

Practices that never touched their documentation habits between 2022 and 2026 are the ones most exposed right now. The model changed under them mid-cycle, and CY2026 is the first year the full weight of V28 lands without a V24 buffer softening the transition.

How Do RADV Audits Actually Reach a Physician Practice?

RADV audits are contract-level reviews of Medicare Advantage organizations, not direct federal audits of physician practices. Exposure arrives secondhand: an MA plan pulls charts from its network providers to defend the diagnosis codes it submitted to CMS, making the practice’s documentation the evidence a plan’s audit outcome depends on.

CMS allows a maximum of two medical records per audited HCC, and only one valid record is needed to sustain payment. That sounds forgiving until a practice realizes it may not control which two records the plan selects, or whether its own retrieval system can produce a complete, signed note inside the payer’s deadline. Plans that cannot retrieve records in time automatically fail those HCCs, and that failure traces straight back to whichever practice held the chart.

Delegated groups and independent practices affiliated with multiple MA plans face compounding exposure. A single high-risk diagnosis category, say chronic kidney disease staging or a cardiovascular hierarchy code, can surface in several plans’ samples in the same audit cycle, multiplying the number of times one ambiguous note gets reviewed.

Which Documentation Failures Most Often Invalidate a V28-Era Diagnosis?

OIG’s March 2026 compliance audit of high-risk diagnosis codes found 247 of 271 sampled enrollee-years unsupported, a 91 percent error rate concentrated in nine high-risk categories. Across the broader OIG audit series, comparable categories have shown error rates between 80 and 95 percent, pointing to a systemic documentation gap rather than isolated plan failures.

The recurring failure points are mundane, not exotic. A missing provider signature invalidates an otherwise accurate note. A diagnosis carried forward year over year without a current face-to-face encounter fails outright, because V28’s risk score calculation is additive and resets on new evidence, not on habit. Specificity gaps, missing laterality, stage, or type where V28 now requires it, quietly downgrade or void a code that used to pass under V24.

Diagnoses sourced only from a health risk assessment or an unlinked chart review carry the highest strike rate of all. OIG has repeatedly flagged HRA-only diagnoses as disconnected from any acceptable clinical encounter, and CMS’s own rules exclude a diagnosis that never touched a qualifying provider visit.

What Financial and Legal Exposure Follows When Documentation Fails?

Unsupported diagnoses expose a practice to downstream recoupment pressure from MA plans and, in serious patterns, to False Claims Act scrutiny tied to how diagnosis data reached the payer. CMS estimates a 2025 Part C improper payment rate above 6 percent, most of it traced to documentation that did not substantiate the submitted code, not to fabricated care.

The legal stakes escalated sharply in January 2026, when a major health system’s affiliates agreed to a nine-figure False Claims Act settlement over diagnosis-mining practices that pressured physicians toward addenda and internal coding targets. That case involved a large integrated system, not an independent practice, but it recalibrated how aggressively MA plans now scrutinize every upstream diagnosis source, including small practices, before submission.

For a practice, the exposure is rarely a direct federal penalty. It is a plan clawing back reimbursement tied to a code the practice originally submitted, a corrective action plan imposed by the network, or exclusion from a value-based contract after repeated documentation failures. None of that requires a courtroom to hurt.

MedPAC has estimated Medicare Advantage payments approaching $615 billion for 2026, with coding intensity alone contributing tens of billions of that figure above what equivalent fee-for-service spending would cost. That scale is exactly why CMS keeps pushing audit volume upward instead of trimming it back, and why plans keep pushing chart requests further down the delivery chain toward the practices that hold the original documentation.

How Should a Practice Build Documentation That Survives the Next Chart Pull?

A practice reduces RADV exposure by treating every chronic condition note as if a reviewer will read it cold, months later, with no clinical context beyond the page. That means naming the condition, stating its current status, and recording the specific evaluation, assessment, or treatment tied to that visit, not a restatement of last year’s problem list.

Two-way review deserves equal weight to upcoding prevention. A condition still coded as active when the record shows it resolved is exactly the pattern a reviewer flags, and undercoding a genuinely active condition creates its own compliance and revenue exposure. Practices that run periodic internal chart reviews against V28’s current mapping rules, rather than legacy V24 habits, catch both directions of error before a payer’s vendor does.

Retrieval speed matters as much as accuracy. A practice that cannot locate and produce a signed, complete note within a plan’s requested window loses the HCC on a technicality, regardless of whether the underlying diagnosis was correct. Organizations that validate documentation before submission, rather than scrambling after an audit notice, consistently show far lower error rates when their charts are finally pulled.

The practices that come through V28-era RADV audits cleanly are not the ones with the most diagnoses on file. They are the ones whose notes already answer the question an auditor will ask before that auditor ever asks it.