Roughly two-thirds of Medicare’s improper payments1 trace back to missing or insufficient documentation. For a decade, the industry’s answer has been to fix that downstream in the form of better queries, faster attestations, and bigger appeal teams. Now that CMS excludes the diagnoses those queries produce and the DOJ is prosecuting the programs that lean on them, that quick fix has become the weak point.
Inside most risk-adjustment programs, that weak point looks the same way: a query lands on a physician’s phone, a few taps trigger an auto-generated note, and a diagnosis gets support it didn’t have an hour earlier. The software can be accurate and the physician’s answer correct, and neither matters much to a RADV auditor or an OIG investigator pulling the claim later. What they check is whether the documentation reflects independent clinical judgment, applied during a real chart review, not a workflow built to produce the right answer on demand. Physicians and coders using these tools aren’t cutting corners; they’re operating inside a workflow regulators no longer tolerate.
Retrospective review is now both risky and beside the point
Two forces are closing in on the same weak point. CMS’s latest rate-setting excluded diagnoses drawn from unlinked chart-review records2, and V28, fully in effect since 20263, stripped more than 2,000 codes4 while raising the specificity bar. DOJ settlements over invalid HCC submissions run into the hundreds of millions of dollars, with programs adding diagnoses without evidence5 and failing to remove irrelevant ones6. Auditors now want documentation that’s real, traceable, and current.
Even with setting the compliance risk aside, retrospective review still solves the wrong problem, even without regulators involved. A query can only act on what’s already written down, pulling a diagnosis forward from somewhere in the chart. If a condition is that buried, the patient may not be benefiting much from the recapture either.
Picture it as an iceberg. Documented HCCs sit at the surface. Just below is the recapture layer: the zone retrospective tools were built to reach, and the one regulators are now most suspicious of. Underneath is the mass nobody built a tool for: the creatinine trend, the medication list, evidence one layer too deep for any query to reach. In the gap analysis below, that layer worked out to 2.4 additional HCCs per patient, worth roughly $4,850 per patient in annualized revenue, across thousands of at-risk patients.
The fix sits upstream, not downstream
Organizations positioned to come through this cleanly capture the right diagnosis during the encounter, evidence attached from the start. A condition caught during the visit gets treated during the visit, not surfaced months later in a query aimed at a patient who’s moved on. The evidence comes from the encounter itself, so an auditor finds what a RADV review wants, and the revenue that follows is the form CMS and the DOJ have no interest in dismantling, because it was never separate from the care that produced it.
What a gap analysis actually shows
Most health system leaders suspect this gap, but few can quantify it. Regard ran an HCC gap analysis at a large academic health system, reviewing a 500-patient sample against documented conditions, scoring the gap with CMS’s V28 risk model, and extrapolating across roughly 3,700 at-risk patients.
The result is 2.4 net-new HCCs per patient, and $18.2 million in projected annualized revenue from those conditions alone, drawn from the kind of buried evidence described above, not recaptured diagnoses or specificity corrections. Counted alongside conditions already in the chart, the total rose to an estimated $29.3 million, or roughly $7,825 per patient.
The underlying argument
An undiagnosed condition isn’t a coding gap. It’s a patient not getting the treatment they need. For decades, no one had time to read every chart, so conditions went unaddressed. That constraint is starting to lift.
The health systems still optimizing the query are optimizing the wrong half of the process. The diagnosis was always the point where the risk score, the audit trail, and the patient’s care converged. Fixing it after the fact was never enough. Now it isn’t allowed to be.
Sizing your own gap takes weeks
The analysis above is a repeatable engagement: a standard BAA, roughly 500 patients, and about a month for Regard to score them against the V28 model. What comes back is the same view, run on your own population. Before rebuilding risk adjustment around the encounter, it’s worth knowing what the encounter would have caught.
Sources
Reference list for the numbered citation links. Delete this section before publishing.
1. CMS, 2025 Medicare Fee-for-Service Supplemental Improper Payment Data, Table A3 (released Jan 24, 2026) — insufficient documentation 53.0% plus no documentation 12.0% = 65% of Medicare FFS improper payments. https://www.cms.gov/files/document/nov-2025-medicare-ffs-supplemental-improper-payment-data-2025922.pdf
2. CMS press release, CY2027 Medicare Advantage and Part D Rate Announcement (April 6, 2026) — exclusion of diagnoses from unlinked chart review records starting CY2027, with an exception for beneficiaries switching between MA organizations. https://www.cms.gov/newsroom/press-releases/cms-finalizes-2027-medicare-advantage-part-d-payment-policies-strengthen-accountability-long-term
3. CMS fact sheet, CY2027 Rate Announcement — confirms the 2024 model (V28) was fully implemented in CY2026; impact of the exclusion without the switcher exception estimated at −1.78%. https://www.cms.gov/newsroom/fact-sheets/2027-medicare-advantage-part-d-rate-announcement
4. AAFP Family Practice Management (November 2023) — CMS removed more than 2,000 diagnostic codes from the V24 model in developing V28. https://www.aafp.org/pubs/fpm/issues/2023/1100/hcc-update.html
5. U.S. Department of Justice press release (January 14, 2026) — Kaiser Permanente affiliates pay $556 million to resolve False Claims Act allegations of adding diagnoses after visits regardless of whether they were addressed. Largest MA risk-adjustment FCA settlement to date; also substantiates “hundreds of millions.” https://www.justice.gov/opa/pr/kaiser-permanente-affiliates-pay-556m-resolve-false-claims-act-allegations
6. U.S. Department of Justice press release (March 2026) — Aetna pays $117.7 million to resolve allegations it submitted inaccurate diagnosis data, failed to withdraw it, and falsely certified it was correct. https://www.justice.gov/opa/pr/aetna-agrees-pay-1177-million-resolve-false-claims-act-allegations
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.
