Your Billing Reports Look Fine. Here’s Why You’re Still Losing Money

On October 1, 2025, Cigna rolled out a policy called R49. 

Here is what it did. Every claim you submitted with CPT codes 99204, 99205, 99214, or 99215 got scanned by an AI algorithm. If the algorithm decided your claim did not look complex enough, it dropped the code one level and paid you the lower amount. No notification. No denial. No appeal notice in most states. Just a smaller check that showed up in your EOB looking like a normal adjustment.

On March 13, 2026, Maryland’s Insurance Administration fined Cigna $80,000 and told them to stop. Regulators found the policy broke state law because Cigna was cutting payments without first disputing the claim or asking the provider for documentation. 

Three weeks after that, Maryland extended the same ban to every insurer in the state through Bulletin 26-9. But from October 2025 until that order, thousands of practices took those reduced payments and never knew anything was wrong.

And Cigna was not the only one doing this.

Karen Zupko and Associates’ August 2025 white paper found that UnitedHealthcare, Aetna, and regional BCBS plans had been running AI-driven downcoding programs at scale since late 2022. 

The federal government has taken notice too. 

Both the OIG’s prior authorization denial reports and the February 2026 OIG Medicare Advantage Compliance Guidance call out the risk of using algorithms to make coverage decisions without reviewing the individual patient’s medical records or clinical notes.

This is what your claims are running into right now.

What Algorithmic Downcoding Looks Like on Your EOB

When a payer downcodes your claim, it goes one of two ways.

Here is what that costs at 2026 CMS national average non-facility rates. 

CPT 99215 pays approximately $192. CPT 99213 pays approximately $95. That is a $97 difference per visit. Between 99214 and 99213 the gap is roughly $40. If you are a primary care provider seeing 250 established patients a month and even 20% of those are getting dropped one level, you are losing between $2,000 and $4,850 every month. Over a year, that is $24,000 to $58,200 per provider. Not a single one of those losses shows up on your denial report.

To find out if this is happening to you, run these steps:

MBC’s 2026 RCM industry analysis puts the average annual cost of systematic downcoding at $40,000 to $180,000 for a multi-specialty practice. The same analysis found roughly 68% of those downcoded claims can be recovered on appeal when your documentation is mapped clearly to AMA 2021 MDM criteria.

Whether that appeal works depends heavily on what is inside your documentation. Which is exactly where the next problem starts.

Why Your Own Notes Are Making the Problem Worse

Payer algorithms have an easier time downcoding your claims when your documentation does not spell out the complexity of the visit. And most notes do not.

Under the AMA 2021 E/M guidelines, how you select a code level has nothing to do with how long your history section is or how detailed the physical exam looks. The level is set by Medical Decision Making (MDM) or total time. MDM has three parts:

• Number and complexity of problems you addressed
• Amount and complexity of data you reviewed and analyzed
• Risk of complications or harm from the treatment decisions you made

You need to meet or exceed two of the three to justify a code level. For a 99214, you need moderate MDM: at minimum, one chronic illness that is getting worse, or two stable chronic conditions. For a 99215, you need high MDM: a chronic illness with severe flare-up, or something that could threaten life or physical function.

Here is the problem in plain terms. 

Say you see a patient with uncontrolled diabetes, hypertension that needs a medication change, and early CKD. That visit clearly supports a 99214 or 99215. But if your note says “diabetes stable, hypertension managed, CKD monitored, continue current meds,” a payer algorithm reads zero MDM. It does not see which problems you addressed. It does not see what data you reviewed. It does not see the risk behind your management decisions. So it drops the code.

A note like that is not longer. It is more targeted. And a payer algorithm cannot drop a code that is backed by explicitly stated MDM criteria. That same documentation habit also closes a completely separate revenue gap, one that starts before a claim is ever submitted.

The Revenue You Are Losing Before the Claim Even Goes Out

AAPC Audit Services audited E/M visits nationwide after the 2021 guideline changes and found that up to 19% of visits were being undercoded. Physicians were picking a lower code than their own documentation supported. 

AAPC puts the cost at $37 per visit where a level 3 gets billed instead of the supported level 4, based on the national Medicare Physician Fee Schedule. If you see 18 patients a day and just four of those visits are being coded one level low, that adds up to $35,000 to $70,000 in lost revenue every year per provider.

There is also a compounding problem here. MGMA’s advisory on payer AI points out that payer algorithms look at your historical billing patterns, not just the individual claim. If you have been consistently billing level 3 codes for years, the algorithm reads that as normal for your practice and flags higher-level claims as outliers. Undercoding over time trains the algorithm to expect low codes from you, which makes future downcoding easier for the payer to justify.

What Your Write-Off Column Is Saying

Even if your denial management is tight and your documentation is clean, there is still a third place revenue disappears: the write-off column.

Most practices treat all write-offs the same. They are not. A contractual adjustment is money you were never going to collect anyway. It is the difference between what you charged and what your contract allows. Writing that off is just standard accounting.

An avoidable write-off is different. That is money you earned, for care you delivered, that got abandoned somewhere in the billing process.

These avoidable losses come from specific, findable failures:

• A claim missed the payer’s filing deadline because it sat in a work queue too long. Once that deadline passes, there is no appeal path. The money is gone for good.
• A new provider saw patients before their credentialing was complete with a specific payer. Every claim from that window is unrecoverable, no matter how clean the coding is.
• A patient balance got written off at 90 days because nobody followed up, not because the patient could not pay. That loss shows up exactly like a charity write-off in most reports, even though it was not.
• A denied claim got coded as a contractual adjustment to close out the account. This is the most damaging one because it makes your net collection rate look better than it is. The denial vanishes from your dashboard, the NCR improves on paper, and the revenue loss is completely invisible.

Three Reports Worth Pulling This Month

You do not need new software to check any of this. These three reports come out of systems most practices already have.

Report #1: Submitted code vs. paid code, filtered to E/M CPT codes 99202 through 99215, for the past 90 days. 

Look for any row where what the payer paid does not match what you submitted and there is no denial code attached. Sort by how often it happens and how much it costs. One or two payers consistently paying 99213 on your 99214 and 99215 claims is enough to file a formal dispute backed by AMA 2021 MDM documentation.

Report #2: Your E/M code distribution per provider, compared to AAPC national benchmarks for your specialty. 

Pull how many times each provider billed each E/M level in the past 90 days, then calculate each level as a share of total visits. If a provider is running 15 or more percentage points above the national level 3 average for your specialty, and they are seeing complex patients, that is a conversation about documentation coaching, not a compliance investigation.

Report #3: Write-off reason codes for the past two quarters, split into categories. 

You want CO-45 contractual, timely filing, uncredentialed provider, patient balance, and other denial write-off separated out. Your avoidable loss sits in that last bucket. Any CO-45 write-off you cannot tie to a specific fee schedule amount is worth pulling five to ten records to see whether it was actually a denied claim that got reclassified to close the account.

Final Thoughts

OmniMD’s RCM platform runs all three of these reports in real time, with submitted-versus-paid code variance flagged by payer and CPT code, E/M distribution benchmarked against national specialty norms, and write-off reason codes broken out and searchable. The revenue that is currently settling silently into your financials gets surfaced before it turns into a quarter-end shortfall. Request a demo now to see what it looks like in your specific payer mix.