By Chetan Parikh, Founder and CEO, RAAPID
Since 2010, I've operated at the intersection of healthcare technology, building NLP platforms, collaborating with coders, and diving deep into the mechanics of risk adjustment. The ground beneath the retrospective chart review has fundamentally shifted. This is not speculation - it's firsthand insight from risk adjustment leaders and is now unmistakably reflected in recent OIG updates.
Retrospective programs were once a staple in risk adjustment - plans relied on them to recover undocumented diagnoses and close critical gaps. That era is over. Federal regulators have drawn a definitive line: programs that solely add diagnoses without removing unsupported ones are no longer considered compliance tools. They are now clear signals of intent to inflate payments, and regulators are responding accordingly.
The OIG Has Made Its Position Unmistakable
In the 2024 MA Industry Compliance Program Guidance, the OIG explicitly identifies conduct that federal investigations now deem potentially fraudulent. One pattern is unmistakable: failing to remove previously submitted diagnosis codes from CMS, even after chart reviews reveal they were unsupported or invalid.
A review process that uncovers unsupported code but leaves it unaddressed is not a minor oversight - it is a clear compliance failure. When this lapse is systemic, it forms the foundation for regulatory enforcement.
The OIG also called out the use of AI that encourages providers to add diagnoses patients never had. These are not theoretical concerns; these are the exact patterns drawing regulatory scrutiny.
The DOJ Showed Us What Enforcement Looks Like
In January 2026, Kaiser Permanente affiliates paid a historic $556 million to resolve False Claims Act allegations - the largest Medicare Advantage settlement to date. The DOJ made its case clear: Kaiser systematically mined historical records for unsubmitted diagnoses, then pressured providers to add them through addenda, sometimes long after the patient visit. This approach fueled nearly $1 billion in unsupported payments linked to almost 500,000 diagnoses.
Cigna’s $172 million settlement in September 2023 followed a similar playbook - leveraging chart reviews to secure additional payments, while purposefully ignoring overpayments revealed by those same reviews.
The common denominator in both cases was not a documentation lapse - it was the program’s architecture. These one-way systems acted as a ratchet, always increasing risk scores and never correcting downward. In the Kaiser case, the DOJ asserted the company was fully aware of the widespread, unlawful nature of its addenda practices and chose to overlook repeated warning signs. When a system is engineered to only move in one direction, regulators target the design itself as the root problem.
The $9.2 Billion Wake-Up Call
Long before the Kaiser settlement, the OIG issued a clear warning. Its September 2021 evaluation (OEI-03-17-00474) revealed that 162 MA organizations generated $9.2 billion in risk-adjusted payments from diagnoses appearing exclusively in chart reviews and health risk assessments - without any supporting service records. Twenty companies accounted for a disproportionate share. The OIG’s message was unambiguous: some organizations were using chart reviews and HRAs to maximize payment, not to ensure enrollees received necessary care.
On January 26, 2026, CMS signaled its intent to close this loophole. The CY 2027 Advance Notice proposes excluding diagnoses from unlinked chart review records from risk score calculations entirely, a policy CMS estimates would reduce MA payments by $7.2 billion. The value of add-only retrospective work isn’t just diminishing - CMS is preparing to eliminate it altogether.
The DOJ Has Set a New Standard for Enforcement
If your retrospective program only adds diagnoses and never removes them, you’re operating a system federal regulators have directly targeted. This isn’t conjecture - it’s the OIG’s formal stance, reinforced by DOJ settlements worth hundreds of millions of dollars.
The real question isn’t, “Did we find more HCCs?” It’s this: “Can we stand behind every diagnosis we’ve submitted, can we defend our codes?” Your retrospective process must be bidirectional - adding supported codes, removing unsupported ones, and documenting every decision with an evidence trail strong enough to withstand scrutiny from any direction.
What Steps Should You Take Going Forward?
I talk to risk adjustment leaders every week who know their programs need to change.
The first step is to look at your retrospective workflow and ask one question: Are your codes defensible? Does it delete? If every chart review can only add a diagnosis and never remove one, you have a structural problem that no accuracy rate will fix. The OIG spelled it out. The DOJ put a dollar figure on it.
Then look at your technology. If your AI suggests an HCC and can't show which clinical evidence supports it, that's not decision support. It's a liability. I've watched coders struggle for years with tools that gave them answers but no reasoning. Every recommendation your system makes needs a transparent path back to the clinical note, one that an auditor can follow without asking questions.
Finally, shift investment toward the point of care. The safest diagnosis is one that a provider confirms during a visit and documents in real time. Retrospective still has a role, but it's a supporting one: validating, cleaning, preparing for audits. The growth path runs through the encounter.
The federal government is no longer asking whether your coding is accurate. It's asking whether it's defensible. Add-only programs fail that test by design.
The editorial staff had no role in this post's creation.