The war of pricing bots

Accelerationism...but for downcoding

I’ve been trying to write a piece about AI in healthcare for weeks. My question: does AI change the dynamics in the healthcare system, or does it exacerbate existing issues?

So far, it seems like the latter is more likely.

I think this is probably a good thing insofar as we’re not ready for AI-as-doctors. I don’t have any particularly hot takes on AI-as-doctors — just a vague sense that the tech isn’t ready for it, and that we as humanity probably aren’t ready either.

But the use of AI to accelerate existing problems in healthcare is a bad thing insofar as…there’s a lot of bad stuff in the healthcare system that’s going to get amplified by the use of AI. We don’t need amplification of insurance denials, prior authorization paperwork, or automatic upcoding/downcoding…but that’s what we’re getting.

More downcoding, faster

Last month, Cigna announced that it was going to downcode (or downgrade the amount the company would reimburse doctors for a specific service) certain Evaluation and Management (E/M) service bills. From the press release:

Therefore, in an effort to promote accurate and consistent reimbursement and protect customers from potential overbilling, Cigna Healthcare will implement a new Evaluation and Management Coding Accuracy (R49) medical reimbursement policy on October 1, 2025.* This policy will review certain professional claims billed with CPT E/M codes 99204-99205, 99214-99215, and 99244-99245 for billing and coding accuracy.

In alignment with American Medical Association (AMA) E/M services guidelines, services may be adjusted by one level only for providers who our records indicate as having a consistent pattern of coding at a higher E/M level for routine services compared to their peers.

Cigna emphasized that it would only apply this policy to providers with a history of upcoding (or billing the service at a more advanced code than the service actually warranted, in hopes of getting reimbursed more), and that 99% of doctors would be unaffected. Nonetheless, this policy announcement, and a similar one by Aetna, caused such an uproar that Cigna temporarily paused implementation of the policy.

Is Cigna using AI for this downcoding? It’s unclear. Cigna told STAT News that the downcoding will not happen automatically and would not include use of AI or algorithms. Physician trade organizations said otherwise, telling their members that the policies would be automatic. My close read of the press release leads me to guess that Cigna was planning to use AI to flag certain bills, which would then be downcoded unilaterally by Cigna employees.

Guilty until proven innocent

In a real lesson for comms professionals in how to message your way to the top, Cigna frames the unilateral downcoding as a good thing.

Rather than delaying payment and requesting supporting documentation for the billed code upon receipt of the claim, Cigna Healthcare will issue payment promptly for the lower-level adjusted code, clearly indicating on the explanation of payment (EOP) if a service was adjusted to be reimbursed at a lower level.

Note, however, the assumption that doctors are, in the words of one physician, “guilty until proven innocent.” And it requires that doctors notice the downcoding, then initiate and manage the appeal process.

This is a much bigger burden for the physicians who are trying to stay independent. As one doctor wrote in an op-ed for STAT News:

At first glance, downcoding might sound like a technical tweak. In reality, it creates a crushing administrative burden. Each downcoded visit will surely trigger a full appeal: pulling charts, re-coding, resubmitting, and waiting. Large health systems may absorb this with billing staff. Smaller independent practices cannot. Appeals already clog the system, and a new approach to downcoding risks further backlogs.

Cigna is not the only insurer that is doing this. Aetna, Anthem Blue Cross and Blue Shield, Humana, and Molina all do some version of this, likely with AI involvement. (And it’s worth noting that the tech involvement itself isn’t new; there are many companies that sell algorithms to upcode from the physician side, downcode from the insurer side, and “risk adjust” for practices submitting to Medicare.)

The prices are still fake

And with AI involvement — are we not just sending AI bots off to war over prices that were fake to begin with? The whole thing is enough to turn anyone, including Wall Street Journal opinion editor Matthew Hennessey, into a radical, per his recent op-ed screed against an Evaluation and Management fee:

A fun game to play is to ask your most politically doctrinaire friends if there’s a policy they secretly support that cuts against their usual views. I’m a free-market guy. You can look it up. But my increasing irritation with the American healthcare system could make me grab a pitchfork before too long. Another email like that and I might join the movement to tear it all down. Getting nickel-and-dimed is one thing. Getting treated like an idiot is something else.

Except now we’re all getting treated like idiots while AI tools battle over what the optimal (fake) price for a service should be. Literally no one is winning, including the companies that have to use AI tools to engage in the billing war.

Source: https://x.com/matthewstoller/status/1981039207570149564

At least some of the money, to answer my colleague Matt’s (somewhat rhetorical) question above, is going to a costly war between pricing bots!