Kaiser's $556M FCA Settlement: What the Addenda and Query Allegations Mean for Working Coders

What changed at your desk

The DOJ's announcement isn't a new enforcement theory. It's a settlement that puts specific dollar figures and specific alleged mechanics on the public record. The question for working coders isn't whether Kaiser did what DOJ alleged. It's whether any part of your own addenda and query workflow looks like what DOJ described.

On January 14, 2026, DOJ announced that five Kaiser Permanente affiliates agreed to pay $556 million to resolve False Claims Act allegations related to Medicare Advantage risk-adjustment submissions. The settlement resolves civil claims only; there's no admission of liability. The cases originated from whistleblower suits filed by former Kaiser employees, with relators collectively receiving approximately $95 million.

The alleged mechanics

The DOJ's theory, as reported by KFF Health News from the allegations, breaks into three parts.

First: Kaiser allegedly mined past medical histories for diagnoses that hadn't been submitted, then used a structured query-and-addendum program to encourage physicians to add them to visit records, sometimes months after the encounter, sometimes more than a year later.

Second: Kaiser allegedly set aggressive internal targets for securing those addenda and tied physician performance evaluations and financial incentives to risk-adjustment objectives.

Third: physicians were allegedly pressured to add diagnoses "regardless of whether these conditions were actually considered or addressed by the physician during the patient visits" (DOJ language, as reported by KFF Health News). And when internal compliance personnel and physicians flagged codes as lacking clinical support, Kaiser allegedly failed to remove them.

The span: 2009 to 2018. The scale: nearly 500,000 diagnoses via addenda, allegedly generating around $1 billion in improper CMS payments.

What the settlement does NOT say

This was a civil settlement, not a criminal conviction, and not an admission of wrongdoing. The legal standard for FCA liability is "knowingly" submitting false claims. That's a specific legal element, not a general statement about retrospective coding.

Retrospective chart review is standard practice and is not inherently fraudulent. Physician queries are a legitimate documentation improvement tool. Addenda that add clinical information a provider genuinely considered but didn't initially document are appropriate; the ICD-10-CM Official Guidelines allow coding diagnoses documented as part of the record.

The line DOJ drew here was between two things: diagnoses the provider actually considered and addressed, added to the record later for accuracy, versus diagnoses the provider was pressured to add to hit a target, regardless of clinical basis. That distinction is the center of this case.

The comparison that matters: legitimate vs. what DOJ alleged

Two categories worth extra scrutiny

The DOJ allegations against Kaiser, as reported, didn't single out specific diagnosis categories. But the same KFF Health News coverage flagged two conditions that reporting and a recent Senate report on MA overbilling have repeatedly tied to upcoding across the industry. Both are HCC-generating, and both are exactly the kind of dx a query-and-addendum program can lean on, so they're worth a second look in your own queue.

Opioid dependence disorder (F11.2x category): The pattern flagged in reporting is codes in this category submitted for patients taking opioids as prescribed for pain who did not have the disorder. F11.20, F11.21, F11.23, and the rest of the F11.2x group require meeting the clinical criteria for opioid use disorder. A patient on a legitimate opioid prescription doesn't automatically meet those criteria. The distinction requires explicit physician documentation, not a medication list, not a presumption from the Rx. Coders can't make that call unilaterally; the provider has to document it.

Dementia (F03.xx, F01.xx, F02.xx): These codes require documented cognitive decline affecting function, a finding the physician actually assessed at the encounter. An unaddressed problem-list entry or a diagnosis carried forward from a prior note without fresh documentation doesn't support a current-year code. The F03 category carries severity specifiers in the current code set (F03.A mild, F03.B moderate, F03.C severe), and dementia codes are HCC-generating, so the documentation bar for each of them is real, and the incentive to code them is real too.

The pattern in both categories is the same: an HCC-generating dx that a medication list or a prior note might suggest, but that requires specific physician documentation to code. If the physician didn't document the actual disorder, as distinct from a related medication or a historical flag, there's no code.

Where this lands for coders who do retrospective review

The settlement doesn't change what good coding looks like. It does clarify what DOJ considers the fraud boundary. Three things distinguish legitimate retrospective review from what was alleged here:

The condition was addressed. Not mentioned, not inferred, not mined from history. Addressed at the encounter the addendum references. No MEAT, no code is still the right frame: the condition has to have been evaluated or managed for the dx to be codable from that visit. See the MEAT Criteria Guide for the elements.

The query is asking for clarification, not a result. A well-structured query asks "did you evaluate or treat this condition at this visit?" It doesn't presuppose the answer. When a query program's goal is to get a specific dx added, and physicians know their evaluations are tracked against how often they say yes, that's where the fraud theory starts.

Unsupported codes get deleted. This one is the simplest and the sharpest. If a code can't survive your own compliance review, it has to come out. Leaving it in when your own team flagged it as unsupported is a specific element of what DOJ described.

For coders: you don't control the query program your organization runs. You do control whether the dx you code is supported in the documentation in front of you. The addendum that adds a diagnosis still has to describe the physician actually addressing the condition at the DOS. If the addendum just adds a name to a list, it doesn't create codability. Use the Evidence Checker to work through the documentation elements for a specific dx before submitting.

The settlement in context

Prior to this settlement, the largest MA risk-adjustment FCA settlements were Cigna at $172 million in 2023 and Independent Health at $100 million in 2024. The $556 million settlement is more than three times Cigna's. The scale reflects both the alleged duration (nine years) and volume (approximately 500,000 diagnoses).

The case originated from whistleblower suits; the relator awards were approximately $95 million total. FCA cases in MA risk adjustment frequently originate from insiders who understand the query and addendum workflows from the inside.