Downcoding is a denial without a reviewer

Downcoding is a denial without a reviewer. When a payer downcodes a visit, it pays for a lower level of service than the clinician billed. The payer has decided that the visit was less complex than the record says. In the programs described here, that decision is made from claim data before anyone at the payer reads the record.

The practice has spread quickly. Cigna began adjusting higher-level office visit claims in October 2025. Blue Cross Blue Shield plans in Illinois and Texas began their own programs on July 1, 2026. In Illinois, 72% of medical practices surveyed by the state medical society said they had experienced automatic downcoding, according to WBEZ.

States have started to answer. Indiana and Virginia now require a person to review a downcode before it stands. Illinois will bar automated downcoding from 2028. Downcoding escapes the rules that govern other coverage decisions, and the reasons why point to what a fair process would look like.

Key takeaways

• Downcoding is spreading. Cigna’s policy took effect in October 2025, and 2 large Blue Cross plans followed in July 2026.
• Practices feel it. In Illinois, 72% of practices in a state medical society survey reported automatic downcoding.
• It sits outside review rules. A downcode reduces payment based on how complex a visit was, yet it is usually handled as claim editing rather than clinical review.
• States are drawing lines. Indiana and Virginia now require human review of downcodes. Illinois will bar automated downcoding from 2028.
• The burden falls on appeals. A practice that disagrees must contest each claim, which few can do at volume.
• Accuracy is unpublished. No payer has released how often its downcodes are reversed on appeal.
• A fair process starts with a reviewer. We think software should flag claims and a person should decide, with the reason stated in terms of the visit’s documentation.

How downcoding works

Office visits are billed in levels. A new patient visit runs from 99202 to 99205, and an established patient visit runs from 99212 to 99215. Higher levels pay more. The level has depended since 2021 on the complexity of the clinician’s medical decision-making or on the total time spent that day. That rule comes from the AMA’s coding guidelines.

Medical decision-making has 3 parts, and the record has to support them.

• Problems. The number and complexity of the problems the clinician addressed.
• Data. The tests and outside records the clinician reviewed or ordered.
• Risk. The risk of complications from the management the clinician chose.

Payers have long audited visit levels by reading records. Downcoding programs work differently. Software compares what was billed with other data on the claim, such as the diagnosis codes. When the claim does not appear to support the billed level, the payer pays at a lower level. The clinician can appeal, usually by sending the record.

The result is a payment decision made on claim data alone. The record that would show the visit’s complexity is not part of the decision. It enters only if the practice chooses to fight.

This table describes the general pattern reported across programs. Each payer’s process differs in its details.

Why claim data cannot see complexity

A claim carries codes, not reasoning. Diagnosis codes say what conditions a patient has. They do not say how many results the clinician reviewed or how hard the management decision was. A visit for a common diagnosis can still be complex. The patient’s other conditions can make it so, and so can the treatment options.

Time-based billing makes the gap wider. A long visit for counseling can be billed at a high level on time alone. The diagnosis on the claim may look simple. A screen that reads the diagnosis will flag the visit, even though the time documented in the record supports the level.

That is why the cuts land where they do. The visits most likely to look wrong on claim data are the ones whose complexity lives in the note.

Who is doing it

Cigna’s policy, known as R49, took effect October 1, 2025, according to Becker’s. It covers the 2 highest levels of new patient visits and established patient visits, along with the 2 highest levels of office consultations. Cigna estimated that more than 97% of providers who often bill those levels would not be affected. Cigna excluded services for infants under 1 year old. It also excluded claims from behavioral health providers and some complex conditions.

Cigna paused the policy for California HMO plans while the state’s managed care regulator reviewed it. The California Medical Association had asked regulators whether the policy was legal, according to the CMA.

Blue Cross and Blue Shield of Texas began its program for commercial plans on July 1, 2026. When billed services do not support the reported level, the plan pays at a lower level. It said it would follow AMA guidelines on medical decision-making. Physicians who disagree can submit records, according to the Texas Medical Association.

Blue Cross and Blue Shield of Illinois started the same day. Its notice said that when billed services do not support the level, it will pay for a lower validated level. The plan told WBEZ that its claims review process is well designed and supports accurate payment. It did not say whether it uses an AI tool.

What the payers say

The payers frame these programs as coding accuracy. Visit levels drive a large share of what they pay physicians, and a level billed higher than the record supports is an overpayment. Cigna’s estimate that more than 97% of frequent high-level billers would be unaffected suggests a screen aimed at outliers. The Texas plan tied its review to the AMA’s own guidelines.

What the payers have not published is how accurate the screens are. None of the programs described here has released the share of adjusted claims that were later reversed on appeal. Without that number, a claim of accuracy rests on the payer’s word.

What it does to practices

The effect falls hardest on visits that are long or complex. Practices in mental health and pediatrics reported the sharpest cuts in Illinois, according to WBEZ. A pediatrician said about 40% of his practice’s claims for higher-level sick visits had been downcoded. A psychiatric practice had 2,400 claims reduced in July alone.

Those are the visits that take the most time. A child with a complicated illness or a patient in a mental health crisis needs a longer visit, and the level billed reflects that. A screen built on claim data may not see the complexity that the record shows.

Appeals do not scale. A practice can contest each claim with records, but the work costs staff time that small practices do not have. Many practices told WBEZ the appeal burden was too heavy at the volume they faced.

A downcode that nobody appeals becomes the final answer.

The money moves even when the error rate is low. If a payer’s screen is wrong on a small share of claims, the practices on the other side of those claims still absorb the loss unless they appeal. The payer bears no cost for a wrong downcode that goes uncontested.

The incentive follows from that. Each downcode is small, but the volume is large. A payer that adjusts thousands of claims and sees few appeals keeps most of the difference. A screen tuned to catch more claims will also catch more correct ones, and the payer has little reason to measure how many.

Practices can prepare. A note that states the problems addressed and the data reviewed gives an appeal something to stand on. Tracking downcodes by payer and by visit type shows whether a pattern is a screen’s error or a coding habit worth fixing. A practice that sees the same visit type cut again and again can ask the payer which claim features triggered the adjustment.

Why it is utilization review

A downcode is a judgment about care. It says the visit did not involve the decision-making or the time the clinician recorded. That is a judgment about what happened clinically, and it changes what the payer pays.

Coverage decisions of that kind usually come with rules. A Medicare Advantage denial based on medical necessity must be reviewed by a physician or other appropriate professional. Our piece on talking peer to peer before a denial described that rule. State utilization review laws set similar requirements for many commercial plans. Our review of state utilization review laws in 2026 found states adding rules about who reviews and how.

Downcoding programs are generally run as claim editing rather than utilization review. That label matters. Claim edits are treated as payment accuracy, not as clinical decisions, so the reviewer requirements that apply to denials may not apply. The result is a reduction in payment that no clinician decided.

The same pattern shows up in other places where software makes the first decision. Our piece on prepayment review argued that a screen should be treated like a diagnostic test. A test has known error rates and a confirmatory step. A downcode screen without a reviewer skips the confirmatory step.

Some claim edits do not need a clinician. Medicare’s National Correct Coding Initiative edits, for example, check whether 2 codes may be billed together. That is a question about codes, and software can answer it. A downcode asks a different question, which is how hard the clinician had to think. Only the record can answer that.

The state response

States moved in 2026. The laws differ in how far they go.

Indiana’s law took effect July 1, 2026. It bars insurers from relying on AI alone to downcode claims and requires human review, according to the Indiana State Medical Association. It also requires more transparency when AI is used in prior authorization or downcoding. It sets new appeal requirements and limits retrospective audits and recoupments.

Virginia’s law bars downcoding unless a licensed physician or other licensed clinician reviews the decision, according to a bill summary. The clinicians who may review include nurse practitioners and physician assistants. A party that downcodes a claim must notify the submitter of the reason and the process to appeal.

Illinois went furthest. Its Transparency in Downcoding Act bars insurers from using automated processes to reduce claims, according to WBEZ. The law names AI and algorithms as examples. It requires notice to the clinician and bars downcoding that discriminates against clinicians who treat patients with complex or chronic conditions. It takes effect in 2028. Illinois practices face more than a year of the current program before then.

What the laws leave open

The laws reach only plans that states regulate. Self-funded employer plans fall under federal law, and the state rules do not reach them. A practice in Illinois may see downcoding on some patients’ claims and not on others for that reason.

None of the laws requires payers to publish how often they downcode or how often appeals succeed. A clinician review requirement improves each decision. It does not tell a state whether a payer’s screen is accurate across thousands of claims. A state could close that gap by asking regulated plans to report downcode and reversal counts each year.

The laws also differ on what review means. Indiana requires a human. Virginia requires a licensed clinician. Illinois removes the automated step altogether. A payer operating in several states will need a process that meets the strictest of them, or separate processes for each.

What a fair downcode would look like

A payer has a real interest in accurate coding. Visit levels drive a large share of physician payment, and some claims are coded too high. The question is how a payer acts on that interest without shifting the cost of its errors to practices.

• Software flags, a person decides. A screen can identify claims that look out of line. A coder or clinician should read the record before the payment changes.
• The reason names the documentation. A notice should say which element of the visit the record did not support. The complexity of the problems is one such element. The data the clinician reviewed is another.
• Records come first where possible. A payer that suspects a pattern can ask a practice for records before adjusting payment, rather than after.
• Complex patients are protected. Screens should be tested for bias against practices that treat complex patients, as Illinois will require.
• Rates are reported. Payers should publish how often they downcode by specialty and how often appeals succeed.

A notice that met that standard might look like this.

This is an illustrative example, not a real notice.

Reporting rates would tell everyone whether the screens work. A high reversal rate on appeal means the screen is wrong too often. A low rate means it is catching real errors. Today neither payers nor practices can see that number.

Where Redlake stands

Redlake builds tools that health plans use in utilization review. Our rule is simple. A machine can say yes, but only a person can say no. A downcode is a partial no, and it should meet the same standard as any other.

Treat downcoding as review. A payment reduction based on the complexity of a visit is a clinical judgment. It should follow the reviewer rules that apply to other coverage decisions, whatever the payer calls it.

Show the reason. Every recommendation our tools make shows its reasons. Those reasons come from the criteria and the record. A downcode notice should do the same, in terms a coder at the practice can check.

Keep the record of the decision. Every decision should be possible to reconstruct, as our piece on denial audit trails argued. A downcode should carry the screen’s output and the name of the person who reviewed it. It should also point to the documentation that person relied on.

Test the screen. A payer should check a sample of flagged claims against the full record before it adjusts any payment. The share that the record supports is the screen’s false alarm rate.

Publish reversal rates. A payer confident in its screen should be willing to show how often appeals overturn it.

Recover what review confirms. A payer that suspects a pattern could pay as billed and review a sample of records. It would then recover only what the review confirms. The practice would contest findings instead of screens.

What to watch in the rest of 2026

• California. Watch whether the state’s managed care regulator lets Cigna apply its policy to HMO plans.
• Other Blue Cross plans. Watch whether more plans adopt programs like those in Illinois and Texas.
• 2027 sessions. Watch for downcoding bills modeled on the Indiana and Virginia laws in states that have not yet acted.
• Federal action. Watch whether any federal proposal reaches self-funded plans, which state laws cannot.
• Appeal data. Watch whether any payer publishes how often its downcodes are reversed.
• Medicare Advantage. Watch whether Medicare Advantage plans adopt similar programs, where federal reviewer rules for denials already apply.

About this analysis

Method. The programs covered are the payer downcoding programs announced from 2025 through mid-2026 and the state laws enacted in response, taken from payer notices as reported by medical associations and trade press and from state legislative summaries. No estimate is made of how much payment downcoding has shifted, because no payer has published that figure.

Data notes. The 72% figure comes from an Illinois State Medical Society survey of practices, as reported by WBEZ. Practice-level figures in this analysis are individual accounts, not representative data. Payer program details may have changed since they were reported.

Sources. Becker’s Payer Issues (October 2, 2025). California Medical Association (October 1, 2025). Texas Medical Association on the Texas Blue Cross policy (2026). WBEZ (August 26, 2026). Indiana State Medical Association (July 2026). Bill summary of Virginia SB 164 (2026).

Frequently asked questions

What is downcoding?

Downcoding is when a payer pays a claim at a lower level of service than the clinician billed. For office visits it usually means paying a level 4 or 5 visit as a lower level.

What decides an office visit’s level?

The level depends on the complexity of the clinician’s medical decision-making or on the total time spent that day. The record has to support whichever basis the clinician used.

Is downcoding legal?

Most states have not banned it. Indiana and Virginia now require human review of downcodes. Illinois will bar automated downcoding starting in 2028.

Which payers downcode visit claims?

Cigna began a program in October 2025. Blue Cross and Blue Shield plans in Illinois and Texas began programs on July 1, 2026. Other payers audit visit levels through record review.

How can a practice contest a downcode?

A practice can appeal and send the record that supports the billed level. Some payers also accept records when a practice disagrees with the adjustment.

Can AI downcode claims?

Indiana bars insurers from relying on AI alone to downcode a claim. Illinois will bar automated downcoding from 2028. Elsewhere, payers have not been required to say whether software made the decision.

Why are mental health visits hit hardest?

Mental health visits are often long and billed on time. The diagnosis on the claim can look simple even when the visit was complex. A screen that reads claim data alone is likely to flag those visits.

Do state downcoding laws apply to employer plans?

They apply to plans the state regulates. Self-funded employer plans fall under federal law, so state downcoding rules generally do not reach them.