Brandt Welker
← Essays

Healthcare

How the Sausage Is Made in Healthcare

Most people picture a careful doctor weighing their claim. What actually happens is software, a rulebook the payer owns, and a medical director signing denials in about a second each.

·12 min read

I don’t think many people understand how doctors and insurers actually work. With each other, or against each other, depending on the day.

The version most of us carry around goes something like this. You go to the doctor and the doctor orders something, maybe a scan or a drug. A claim goes to your insurer. Somebody there pulls your file and looks at your history, weighs it carefully and decides whether to pay. Pretty straightforward, right?

Yep. And all of that is complete and utter bullshit. It doesn’t happen like that at all.

The insurer knows you’re coming

The machinery starts the moment you book the appointment. Your provider’s system sends the insurer an ASC X12 270 transaction, and the insurer answers with a 271. Under HIPAA, those are the standard transactions CMS requires for checking eligibility. That exchange tells the office whether you’re covered today and what you’ll owe at the desk, copay and deductible included. It also tells the payer a claim is probably on its way.

So the little card in your wallet matters less than you think. The front desk may still ask for it, but they already have the answer straight from the payer.

The first real “no” comes from a different ASC X12 transaction. When your doctor wants something expensive, like an MRI, the office often has to send a 278 prior-authorization request. That’s where your scan goes to wait. And after years of filing these, the office already has a pretty good idea what your plan will approve and what it won’t. Yes, I know these transaction numbers by heart. This info has never once helped me at a party.

Advil first, MRI later

Say you’ve got back pain and your doctor thinks you need an MRI. What usually happens instead is you’re told to take some Advil and come back in a few weeks. Next visit, you get an x-ray. If you’re still hurting and the x-ray is clean, then you get the MRI.

There’s a reason for that order, and it’s written down. In my experience, a lot of plans’ prior-authorization criteria want weeks of conservative treatment and an x-ray before they’ll approve a back MRI. So the doctor is building a case for your insurer, one visit at a time.

To be fair, for a lot of people that path is also good medicine. The American College of Radiology’s Appropriateness Criteria for low back pain say uncomplicated pain can usually wait through about six weeks of conservative care unless there are red flags, and red flags like suspected cancer or a fracture call for prompt imaging. Plenty of backs get better on their own.

What bothers me is who sets the order and who gets told. The payer sets it. The patient almost never hears that. My wife is a physician, and she describes the worst version of this: sometimes the doctor can already tell the long path is a waste of time, and still has to walk the patient down it.

I’ll use myself. I started on a GLP-1 when they were new, when most plans flat-out denied them for weight. So I paid out of pocket. Back then I was obese. My labs had me pre-diabetic and pre-hypertensive, my cholesterol was high, my joints hurt, and I was on the road to heart disease. We already knew the weight was feeding every one of those. That’s exactly the patient these drugs are for. Even then, I’m sure I’d have gotten the full treatment: step through the cheaper stuff first and document the failures. Then come back in three months. Hurdles, every one of them labeled “evidence-based,” and every one a chance for me to quit while the diseases kept progressing. Sixty pounds later, if I asked them to pay for maintenance, I’d get a no and a long back-and-forth before a maybe. The drug worked, and that’s now the reason not to cover it, because my labs and my weight no longer call for it. It’s a giant waste of my time and my doctor’s, and in the end the insurer gets what it wanted. It doesn’t pay. To me this is annoying but I can afford it. Many can’t.

My read, and it’s only my read, is that the plan is happy either way. Either the Advil works, or you get tired of coming back and drop it. I can’t see inside anyone’s head. But the appeal numbers I’ll get to below don’t make me feel any better about it.

More than 3 billion claims, mostly read by machines

After every visit, the provider sends the payer an ASC X12 837, which is the claim itself. The CAQH Index counts more than 3 billion medical claims a year, and that’s only from the plans that report to it.

Nobody is carefully reading 3 billion files. From what I’ve seen, the small, cheap claims mostly go through software. Is this person covered on this date? Is there a rule that says deny it? No? Pay it and send back the ASC X12 835 remittance advice. There’s a good chance no human ever looks at it.

As far as I can tell, most claim denials are automated too, and plenty of them are boring: coding errors or duplicate claims. The interesting ones are the expensive or unusual claims. Those get flagged for a human reviewer, who checks for a prior authorization and maybe notices your two rounds of Advil, maybe doesn’t. If the reviewer approves it, that’s the end of the story.

The medical director’s queue

When the reason for a “no” is medical necessity, a physician has to sign off on it. That physician is the insurer’s medical director.

Look at how the flow works. From everything I’ve seen, an approval never needs a doctor. A denial does. So the doctor on the payer side spends the day looking at a queue that is mostly denials waiting for a signature. I really don’t see how anyone can be a medical director for an insurance company. You went to medical school to take care of patients, and now your day is signing no’s for people you’ll never meet, faster than anyone could actually read their charts.

And “review” may be a generous word. ProPublica reported in March 2023 that Cigna medical directors denied more than 300,000 claims over two months in 2022 using a system called PXDX, spending an average of about 1.2 seconds on each. One former Cigna doctor told ProPublica, “It takes all of 10 seconds to do 50 at a time.” Cigna called the report “riddled with factual errors and gross mischaracterizations.” I can’t read a patient’s name in 1.2 seconds, and to me a signature at that speed is a rubber stamp with a medical license attached.

And that’s the whole group. ProPublica also reported that one Cigna medical director denied roughly 60,000 claims in a single month. At 1.2 seconds each, that’s about 20 hours of signing.

Say you’re that medical director, and you’re really good. You’re only wrong 5% of the time. That’s 3,000 people a month. Say you’re the best in the business, the top 1%. That’s still 600 people a month, more than 7,000 a year. So much for do no harm. Even delays are harm. Great ethics. Every one is a real person stuck with a bill they shouldn’t owe, or care that gets put off. And you’re reviewing them at 1.2 seconds a claim, so you know your numbers aren’t that good. My bet is 5% is gracious and the real number is way higher. Honestly I think that might be worse than doing it on purpose, because the tools let you do it at a scale nobody could by hand.

I get why doctors take these jobs. Regular hours, no call, home for dinner, a chance to finally sleep at night. I just don’t know how you sleep at night after a month like that, when those 600 people a MONTH are your friends and neighbors and the folks you run into at the grocery store.

It shows up in prior authorization too. ProPublica and The Capitol Forum reported in October 2024 that EviCore, which Cigna owns and which runs prior auth for imaging and other services for many large plans, has what insiders describe as “the dial,” a setting that can be turned to deny more. The same reporting found that some EviCore contracts pay the company more the more it cuts spending. Cigna says the tool exists to speed approvals and that doctors can appeal.

The rulebook

When your insurer says the MRI isn’t “medically necessary,” they want you to picture a careful doctor weighing the science. Evidence-based medicine. And it was a great idea when doctors built it. The term came out of McMaster University in the early 1990s, and the landmark paper ran in JAMA in 1992, building on Archie Cochrane’s 1972 book on effectiveness in medicine. Physicians wrote it for physicians. No insurer was in the room.

Then somebody on the payer side looked at it and saw the best idea they’d ever heard. A reason to say no, with a citation on it. I wasn’t in that meeting. I’d bet the pitch was simple: we’ll enforce the practice of evidence-based medicine, all in the name of saving money.

So who writes the rules now? Most plans license a criteria product that turns “the evidence” into yes or no, and a King & Spalding summary of the 2024 Medicare Advantage rule names the two big ones as InterQual and MCG. Guess who owns InterQual.

UnitedHealth Group does, through its Optum unit, which got it by buying Change Healthcare in 2022 after the Justice Department sued to block the deal and lost. While the deal was pending, the American Hospital Association publicly warned that UnitedHealth could use those criteria to drive up denials.

Now follow the money around in a circle. UnitedHealth owns InterQual, the rulebook that defines medically necessary for a lot of plans. UnitedHealthcare, its insurance arm, pays the claim or doesn’t. The company has also sat on a big pile of the data the evidence gets built on. Boston University reported in 2014 that Optum Labs, founded by Optum and Mayo Clinic in 2013, opened de-identified claims and clinical records on millions of people to academic partners. Becker’s reported that Penn Medicine and Optum Labs co-funded ten research projects, and that UnitedHealthcare put $1 million toward University of California health research. That reporting runs from 2014 to 2020, so treat it as history.

And when the rulebook isn’t enough, there’s the algorithm. A class action filed in November 2023, as Healthcare Finance News reported, alleges that UnitedHealth’s naviHealth unit used a tool called nH Predict to cut Medicare Advantage patients off rehab against their doctors’ judgment. UnitedHealth is contesting the case, so it’s an allegation. Read the complaint anyway.

It gets worse, because your doctor usually can’t even see the rulebook. The criteria are proprietary. It got bad enough that CMS wrote a rule, which took effect in 2024, requiring Medicare Advantage plans to make their coverage criteria public and explain the evidence behind them. In January 2025 the American Hospital Association told CMS its members were still getting denials based on InterQual and MCG with no access to the criteria.

So how is your doctor supposed to know whether the evidence got bent somewhere on its way into that box? She can’t. It’s a black box and the payer owns it.

One company writes the test and grades it, and it keeps the answer key locked in its own desk. I can’t point to a single study that was bought, and I won’t pretend to. What I can say is that nobody outside the company can tell whether any were shaped by who held the data, and neither can your doctor, which is kind of my whole point. Somewhere along the way the evidence turned into a hurdle with a footnote on it.

Cigna’s right about one thing, though. You can appeal. The appeal numbers go a long way toward explaining why they’re so comfortable saying that.

The math only works if you quit

KFF reported in January 2026 that Medicare Advantage insurers made nearly 53 million prior-authorization decisions in 2024 and fully or partially denied 4.1 million of them. Only 11.5% of those denials were appealed, according to KFF. Of the ones that were appealed, KFF found 80.7% were overturned in whole or in part, and that share has stayed above eight in ten every year KFF looked at, back to 2019.

If I ran something where most of my decisions got reversed when somebody challenged them, I’d figure my decisions were broken.

That’s the business model as I see it. Every hurdle in this story costs the insurer almost nothing and costs you your time, and sometimes your health. Some people run out of patience first, and every person who quits is a claim that never gets paid.

What to do about it

If a prior auth gets denied, have your doctor’s office request a peer-to-peer, where your doctor gets the insurer’s physician on the phone. Make a doctor defend the denial in more than 1.2 seconds.

Then appeal. By KFF’s count, more than eight in ten appealed denials get overturned in whole or in part. The odds are on your side, and the whole system is betting you won’t bother.

Your doctor isn’t the enemy here. She’s stuck in the same queue you are.

Next time the doctor tells you to take some Advil and come back in three weeks, ask one question while she’s still in the room: is this your real advice, or what you know my insurance will approve?

Sources