Payers run automated checks on your claims. Run yours first.

Before a person at the payer reads your note, software reads your codes. It checks them against rules. It can lower a visit level or deny a claim line before anyone opens the chart.

That matters because denials are the revenue leak practices name most. In a January 2026 MGMA Stat poll, 48% of respondents named denials and appeals as their biggest revenue cycle leak. Coding came in at 13%.

Here are three payer rules worth knowing, and what they have in common.

Cigna judged visit levels from the claim

Cigna's R49 policy took effect October 1, 2025. It lets Cigna lower 99204-99205, 99214-99215, and 99244-99245 by one level when "the encounter criteria on the claim" don't support the code. A 99214 can be paid as a 99213. If you later send records that support the original level, Cigna pays that level.

The AAFP reported that the trigger looks at the primary diagnosis and other claim-based criteria. Cigna's example diagnoses were "earache" and "sore throat." The AAFP's response: it takes the medical record to judge a visit level, and "the primary diagnosis on a claim is not sufficient."

The policy drew pushback. The AAFP reported that Cigna paused it in California. In March 2026, Maryland fined Cigna $80,000 and ordered it to stop automatically lowering E/M codes. In April, Maryland told every health insurer in the state that this kind of downcoding isn't allowed.

The lesson is not to stack diagnoses on a claim. It's that your codes should tell the same story as your note. Code what you managed today as specifically as the note supports. If the note supports a 99213, bill a 99213.

UnitedHealthcare now enforces Excludes1 on professional claims

Excludes1 is the ICD-10-CM rule for codes that can't be billed together. The official guidelines describe it as "NOT CODED HERE!" The two codes should never be used at the same time.

UnitedHealthcare Medicare Advantage used to apply this rule only to inpatient claims. Its February 2026 bulletin set March 1, 2026 as the start of enforcing Excludes1 across all claim types, including professional claims. Claims that break the rule "may be subject to edits or denials."

Two pairs of codes that coincide in primary care, per the FY2027 ICD-10-CM tabular list:

• R73.03 (prediabetes) with E11.9 (type 2 diabetes). The category that holds prediabetes (R73) excludes diabetes codes E08 through E13. This can happen when the problem list still carries prediabetes after the diagnosis changed.
• Z87.891 (history of nicotine dependence) with F17.210 (nicotine dependence, cigarettes). The patient either quit or still smokes. The claim can't say both.

In both cases, pick the code that's true today. Then clean up the problem list so the old code doesn't return next visit.

Anthem said it would deny claim lines missing a left or right

Many ICD-10-CM codes name a side. H61.20 is impacted cerumen, unspecified ear. H61.21, H61.22, and H61.23 name the right ear, left ear, or both. M25.569, pain in unspecified knee, has right and left versions too.

Anthem told its California Medicare Advantage providers in 2023 that professional claims using an unspecified code when a sided code exists "will be denied." Its example was H60.339, swimmer's ear, unspecified ear, billed with a right-side procedure. Anthem said that claim line would be denied. The same notice said Anthem would keep enhancing its editing system "to automate edits."

One caution: don't guess a side. The guidelines say to use the unspecified code when the record doesn't name one. The real fix is a note that says which ear.

Check your claims the way the payer will

These rules share a pattern. They run automatically, they read the codes, and they act before a person looks. Physicians are wary of payer AI too. In a February 2025 AMA release, 61% of physicians surveyed said they worry health plans' use of AI is increasing prior authorization denials.

You don't need the payer's software. You need a check before submission that asks the same questions:

• Do any two diagnosis codes coincide under Excludes1?
• Does every code that has a side name one, and does the note say which?
• Do the codes match what was managed today, and does the note support the level, up or down?

A safe check, whether a person or software runs it, flags problems in both directions. It shows the rule behind each flag. And a person makes every change.

Cortex Lens is one way to run these checks inside Practice Fusion. It checks the full Excludes1 rule set and missing laterality, with optional visit-level support. It's read-only, and its flags go both ways. See how billers use it.

Start this week with last quarter's denials. Sort them by reason. If codes that coincide or a missing side show up, add those two checks before your next batch of claims goes out.