Silent leaks: the downcoding problem hiding in your behavioral health remits

Prior authorization denials act like a smoke alarm. They're loud on purpose, generating a letter, a denial code, an appeal clock, and a task that sits in someone's queue until it's resolved. When one goes off, everyone knows the drill. The billing team for your outpatient behavioral health practice knows how to handle them.

Downcoding acts more like a pipe leaking under the sink. No alarm sounds when a payer's automated claim-editing software reads the submitted code, decides a level four visit was really a level three, and pays the lower rate before a human ever looks at the chart. The American Medical Association (AMA) has been blunt about how it works: an increasing number of payers downcode claims automatically using software algorithms, without first requesting or reviewing clinical records (AMA, 2022). Worse, some payers pay at the reduced rate without changing the billed code at all, which the AMA highlights make the reduction extremely difficult to spot in a routine coding audit. There's no document to appeal because on paper nothing was formally denied, the payment just comes back a little smaller.

The trouble with a slow leak is that you don't notice it until the damage is structural. By the time the damage begins to surface for your practice as a cash-flow gap, a budget that no longer reconciles, or a quarter that came in soft for reasons nobody can quite name, the water's already inside the wall. Your executive financial report tells you what was paid. It rarely tells you what was paid incorrectly, and that gap is exactly where the water collects.

Behavioral health sits directly in the path of these programs, because they target the exact codes your prescribers lean on most. In outpatient psychiatry and medication management, 99214 is the dominant evaluation and management (E/M) code, because active medication management routinely meets the moderate-complexity threshold the code requires. The AMA's own downcoding guidance lists prescription drug management as a moderate-risk example that supports a 99214 (AMA, 2022). The line between a 99213 and a 99214 is one of the most consequential coding distinctions in behavioral health, and it's precisely the line these payer programs move.

The downshift looks small per visit, often $30 to $40, but it doesn't feel small at volume. A prescriber running a full panel of medication-management visits, multiplied across a year, can quietly lose tens of thousands of dollars that never once showed up as a denial anyone could see.

Unfortunately this may not be fringe payer behavior, nor is it slowing down. In March 2026, Health Care Service Corporation, the parent of Blue Cross and Blue Shield of Illinois, Texas, and Oklahoma, told providers that for dates of service beginning July 1, 2026, it will review office, inpatient, and outpatient E/M claims and reimburse at a lower level when the documentation doesn't support the level billed, and if you disagree, the burden is on you to submit medical records afterward (Blue Cross and Blue Shield of Illinois, 2026). Aetna is moving on the same front a its March 2026 provider bulletin told practices to expect new claim edits under its Claim and Code Review Program beginning June 1, 2026, for commercial, Medicare, and Student Health members, with claims checked against Centers for Medicare and Medicaid Services and American Medical Association coding standards (Aetna, 2026).

This is already landing on behavioral health practices, not just in policy notices. When Blue Cross and Blue Shield of Montana began reassessing E/M levels on July 1, 2026, psychiatric and addiction-treatment providers in the state reported losing $60 to $110 per claim compared with what the plan had paid before (BHB, 2026). One Helena psychiatrist told Behavioral Health Business (2026) that he and a colleague spent close to 23 hours in a single month, outside of clinical work, appealing 40 to 50 downcoded claims (BHB, 2026). Providers noted the reductions hit claims that had already cleared prior authorization, which is what led many to believe the downcodes were automatic. The pattern is the same one showing up in Illinois, where mental health and pediatric practices reported the earliest and sharpest effects: small practices facing bulk reductions they lack the staff to appeal one claim at a time.

When the downcodes are this large and this widespread, a practice notices the way you notice a burst pipe. The harder problem is the smaller, steadier version: a single level knocked down here and there, never enough per claim to trigger alarm, adding up all the same. Both come back to the same defense. Montana's providers caught their downcodes only by spending 23 unpaid hours building, by hand, the exact worklist a system should have handed them.

The pushback is real, too, and worth naming honestly. Sustained advocacy from medical associations, along with state regulatory review, has already prompted several payers to pause or narrow these programs. However, a pause is not a reversal, and the pipe is still under the sink. The direction of travel is toward more automated review, not less, and the practices that come out ahead are the ones that can see the drip whether or not it's making the news.

Payers don't describe these new downcoding practices as revenue suppression. Their position is that some providers overcode, and that automated review corrects coding patterns that drift above same-specialty peers. Overcoding is a legitimate phenomenon, and good compliance means coding to the documentation, not above it.

But the AMA's principles draw a clear and reasonable line. Blanket downcoding isn't appropriate. Any program should target genuine outliers, not every claim. Downcoding a claim without reviewing the medical record is never appropriate, and the provider should have the chance to supply documentation before payment is reduced, not after (AMA, 2022). When a 99214 is documented to standard and paid as a 99213 by an algorithm that never opened the chart, the program has stopped correcting outliers and started taking revenue.

Automated downcoding has become a policy question. Indiana enacted a law in March 2026 which became effective July 1 that bars health plans from using an automated tool, including artificial intelligence, as the sole basis to downcode a claim without a human review of the medical record. It also requires that providers be notified, bars downcoding on diagnosis alone, and guarantees an appeals process (Indiana House Bill 1271, 2026). Regulators are acting on their own authority too. California's Department of Managed Health Care review was among the pressures that led one major insurer to pause its automated E/M program. Maryland went further and on March 13, 2026, the Maryland Insurance Administration fined Cigna $80,000, ordered it to stop automatic downcoding and reprocess affected claims, then three weeks later issued a bulletin extending the same prohibition to every health insurer and third-party payer operating in the state (Becker's Payer Issues, 2026; Maryland Insurance Administration, 2026). Payers face more pressure to justify these reductions than they did a year ago. That doesn't mean the downcoding practice is going away, which is exactly why you want to be able to see it yourself.

The good news is that the documentation supporting the original code is already in the chart. Much of what gets downcoded is recoverable on appeal, you just have to catch it first.

Catching it is a data problem, not a clinical one. The AMA's own advice to practices is to compare the payment for each billed Current Procedural Terminology (CPT) code against the amount you expected, and to work with your electronic health record or practice-management vendor to make that comparison routine (AMA, 2022). It surfaces in one report most in-house teams never run: a submitted-code-versus-paid-code variance report, broken out by payer. When that comparison runs automatically, the drip stops being invisible. It becomes a worklist, the same kind your team already knows how to clear.

This is exactly where an integrated platform changes the math. When your revenue cycle management (RCM) lives inside the same system as your clinical documentation and your clearinghouse, the submitted code and the paid code sit side by side. Variance gets flagged instead of absorbed. A downcode becomes a visible, trackable, appealable event with the supporting note one click away, instead of a number that quietly erodes your margin unnoticed every month.

That's the difference between billing software that records what happened and an RCM partner that protects what you earned. Practices running on ProsperityEHR see a 99.5% clean claim rate, a 97% collection rate, and a denial rate under four percent, and the reporting layer that produces those numbers is the same layer that makes silent denials loud.

ProsperityEHR compares the allowed amount on every paid line against your contracted fee schedule for that CPT code and payer, then flags the underpayments, grouped by payer, by code, above a discrepancy threshold you set.

Each exception exports with the patient and policy detail your team needs to file an appeal, and the whole report can be scheduled to run on its own and drop into a worklist, so catching the drip doesn't depend on someone remembering to look under the sink. For reductions that arrive with an adjustment reason code, such as a level of care change adjustment, you can build rules around those specific codes and route them to the same worklist instead of letting them dissolve into the remit.

The effect is simple: the underpayment stops being a number nobody sees and becomes a recoverable claim someone is assigned to work. That's the whole game with silent denials; the fix is being able to see them.

Financial sustainability is what lets a practice keep its doors open and keep saying yes to the next patient. You shouldn't have to lose a slow leak of revenue you never agreed to give up, on care you already delivered and documented.

Want to see your own submitted-versus-paid variance? Book a demo and we'll walk through how ProsperityEHR's embedded RCM surfaces downcoding before it becomes a quarter you can't get back. Request a demo →