Insurer Downcoding: How to Catch What the Revenue Payers Are Quietly Taking Back

When a payer denies a claim, at least it has the decency to tell you. Downcoding does not. The remittance says “paid.” The math on the page looks clean. Your biller posts it and moves to the next one. And a level-four visit just got paid at a level-three rate, with nothing but a quiet remark code to mark the spot.

That is the whole trick. Payers reimburse a lower E/M level than the one you billed, without requesting documentation and without issuing a denial.  On paper the claim looks paid while in reality it is underpaid. No fight, no fax, no flag. Just a little less than you earned, over and over, until it becomes a number that would make you sick to total.

What downcoding actually is

Downcoding is the systematic reduction of a billed code to a lower-paying one, most often an E/M visit knocked down a level, or a hospital stay dropped to a lower severity. Sometimes the payer changes the code. Sometimes they do not change the code at all, they simply pay less, and you only catch it by comparing what you expected against what actually landed.

Here is why it stings more than a denial. A denial is visible and immediate; you know to work it. An underpayment is hidden inside numbers that look correct until you compare them line by line against your contracted fee schedule. If your billing system is not built to flag it, there is no alert, no cue, no reason for anyone to push back. It is designed to slip past a busy front office; and it does far too often.

Who’s doing it, and how

This is where it should make you angry, because the biggest names in the industry have made downcoding a policy, not an accident.

Start with Cigna. Its E/M Coding Accuracy Policy, internally, Policy R49, was built to automatically downcode level 4 and 5 E/M claims by one level when the submitted diagnosis did not appear to support the complexity of the visit. Sit with the logic there: the system judges complexity off the diagnosis code on the claim, not the medical record, the time, or the decision-making. Two of the example diagnoses Cigna offered were “earache” and “sore throat:” as if a sore throat can’t be the visit where you catch the thing that matters. Physician groups pushed back hard, pointing out that coding rules require the medical record, not the primary diagnosis, to determine whether a visit was coded correctly.

The good news, and proof that pushback works: Cigna paused R49 in the face of advocacy from the AMA and others after critics noted it never disclosed the specific algorithms or data it used to decide. 

Then there’s Aetna. Providers reported something quieter and arguably worse: Aetna was not even disputing some claims, just automatically paying them at a lower rate. Aetna’s defense is that the program is narrow, applying only to the roughly 3% of in-network providers who submit higher-intensity codes than CMS or the AMA would expect, and it has moved toward a severity-based inpatient payment policy for its Medicare Advantage plans as well.

UnitedHealthcare has framed the same goal in the language of stewardship: its leadership says the company is focused on identifying waste and abuse in outlier coding, citing billions in costs it attributes to aggressive physician billing. On the post-acute side, the scrutiny has reached federal court. In Estate of Lokken v. UnitedHealth Group, families of deceased Medicare Advantage members allege the insurer used an AI tool called nH Predict, built by its Optum subsidiary naviHealth, to cut off skilled-nursing and rehabilitation coverage over the objections of treating physicians. On March 9, 2026, a federal magistrate judge in Minnesota ordered UnitedHealth to turn over a broad set of internal records on how it used the algorithm, granting six of the plaintiffs' seven categories of requests, though the court declined to compel the tool's actual source code. The judge pointed to a 2024 U.S. Senate investigation that found UnitedHealthcare's post-acute denial rate more than doubled after it adopted nH Predict. UnitedHealth disputes the allegations, maintaining the tool is only a guide and that coverage decisions are made by medical directors, not AI. Its production deadline passed at the end of April 2026, and the case is now moving toward class certification.

And it does not stop with three names. Humana and some Blue Cross and Blue Shield plans have taken similar approaches, and an oncology association asked federal regulators to examine how UnitedHealthcare, Elevance Health, Centene, and Molina handle E/M claims. The common thread: these are data models comparing your billing patterns against your peers and downgrading anything that looks like an outlier, inside automated systems that rarely show their reasoning. An algorithm scanning for a keyword is not a clinician reading your note. It cannot see the exacerbation, the comorbidity, the forty-five minutes you actually spent. It sees “sore throat” and pays you accordingly.

Why this lands hardest on independents

The physician bodies have said the quiet part plainly. The AMA formally resolved to vigorously oppose unilateral downcoding of E/M services, holding that it is never acceptable to downcode a claim without reviewing the medical record. The insurers’ trade group defends the practice as a necessary check on waste, pointing to “coding intensity” studies and lobbying against federal bills that would limit it. You can hold both ideas at once: some upcoding exists, and a blanket algorithm that pays everyone less without reading a single chart is not the cure.

Here is what we keep coming back to. A large system has a revenue-integrity department to model this, appeal it, and absorb the float. An independent practice has you, and maybe one overextended biller. The tactic works precisely because it is betting you are too busy to notice and too small to fight. That bet is what we exist to beat.

How to catch it

You cannot recover what you cannot see. Here is where we would start:

1. Compare expected to actual on every paid claim, not just the denials. Build a reference table of your top CPT codes by payer and their contracted allowed amounts, and check each remittance against it instead of memorizing fee schedules. If you billed a 99214 and the dollars match a 99213, you have found one.
2. Read the remark codes you’ve been skimming past. Watch for short phrases such as “service level adjusted per policy,” and for codes like CO-97, that often mark a silent downgrade rather than a routine note.
3. Don’t let CO-45 lull you. When the allowed amount on a CO-45 line doesn’t match your contracted rate, that is an appealable error, and practices without a structured process miss these entirely.
4. Track the pattern by payer. One downcode is noise; the same payer doing it to your 99214s every week is a policy. The pattern is your evidence, and your leverage.
5. Appeal with the record, and cite the standard. Pair the corrected claim with the documentation and the AMA’s position that complexity is determined by the chart, not the diagnosis code.

Here’s our part

This is exactly what our revenue-cycle KPIs and denials tools were built to surface. Insight monitors your paid claims the way the big systems monitor theirs; comparing billed codes to allowed amounts, flagging downcode patterns by payer, and turning “looks paid” into “actually owed.”  The money stops leaking while you are busy seeing patients. We tailor it to your specialty and your contracts, because a sore throat in one patient and a sore throat in patient are not the same claim, and your data should know the difference.

You earned the level you billed. Let’s go get it back. Request a KPI & Denials Review at www.insightmedconsults.com and we will show you, in your own numbers, exactly where the quiet adjustments are adding up.

‍