They expected 17,000. They got 2.6 million.

I have had one of these.

You go to a hospital that takes your insurance. You check that first, because you are careful. You get treated and you go home.

Six weeks later an envelope arrives from a doctor you never chose, at that same hospital, for an amount that makes no sense at all.

Congress made that illegal. The No Surprises Act took effect in January 2022, and for patients it worked. You cannot be sent that bill anymore.

But the money did not stop existing. It moved.

The law set up an arbitration process to settle what the insurer owes the doctor. Each side names one number. An arbitrator picks one of them. Congress expected almost everything to settle before it got that far.

Federal officials projected about 17,000 of those cases a year.

In 2025 there were 2.6 million.

The surprise bill did not go away. It changed who receives it.

That is about 153 times the forecast. Volume rose 77% in a single year, and it is still climbing.

Researchers at Georgetown's Center on Health Insurance Reforms published the arithmetic in Health Affairs last week. They put the running cost of the arbitration system at $22.4 billion.

Of that, $15.6 billion is money paid to providers above what an in-network doctor would have been paid for the same work. Another $4.2 billion is administration. A further $2.7 billion goes to the arbitration firms themselves.

None of it reaches the patient. It goes to the health plan. For most working Americans the health plan is the employer, and the employer funds it out of the same pot as the payroll.

So a law written to protect people from one bill is producing a much larger one, in a place where nobody can see their own share of it.

I find that interesting. Not in a good way.

We'll be following the quarterly filings from here.

This connects directly to the incentive underneath the whole system. Filing is cheap. Winning is not rare. So the question is only how large a number you can name.

Between 2024 and 2025, the number of medical billing disputes filed rose 77%. Total payouts through the process rose 264%.

So the average case is getting much bigger, not just more common. Providers are filing on higher-value claims and asking for more when they do.

The benchmark rate in these cases is called the qualifying payment amount. It is meant to stand in for the local in-network price. In 2025 the median award came in at more than four times that figure.

For emergency medicine, the median award was 315% of the benchmark. For breast reduction surgery, the median 2025 award was more than eighty times what Medicare would have paid.

Surgery accounted for only 5% of all disputes between 2023 and 2025. It accounted for $3.8 billion of the awards.

And here is where it spreads. This is not millions of individual doctors sending in paperwork.

In the first two quarters of 2025, the top four filers accounted for 56% of every dispute submitted nationally.

Radiology Partners won 92% of its cases in the first quarter of 2025 and 95% in the second.

The Georgetown researchers describe the disputes as primarily initiated and won by private-equity-backed provider groups. Around them sits a service industry of revenue-cycle firms and law firms that file arbitration claims at volume for a share of the award.

Some of those firms advertise win rates above 90%.

A dispute process becomes a business the moment somebody can file at scale.

The arbitration firms are paid per case, and the parties have handed them $2.7 billion in fees so far. Nobody in that chain is paid to reduce the number of cases.

Meanwhile, the invoice for all of this arrives in October, when employers set next year's plans.

Marsh released preliminary results from its national survey of employer-sponsored health plans on September 2. It covers more than 1,800 employers.

Cost per employee is projected to rise 8.2% in 2027. That is the largest increase in twenty-four years, and it already assumes employers take action to hold it down. Without those steps they put the number at 11%.

It follows 6.0% in 2025 and 6.7% in 2026. Average cost per employee passed $17,496 last year and is expected above $18,500 this year. Aon puts the 2027 rise higher still, at 9.5%.

Marsh's actuaries attribute roughly one full percentage point of the 2027 increase to the arbitration system alone.

Nearly half of large employers say they will raise deductibles or copays next year. So the cost completes the circuit and comes back to the patient, just through a different door.

Three more things worth keeping track of.

Retail investors are asking for their money back from semi-liquid private credit funds faster than the funds will return it. Robert A. Stanger puts second-quarter repurchase requests at 12.4% of net asset value, the highest reading it has ever recorded and up from 10.4% in the first quarter. Sponsors met 38% of them and paid out about $5.9 billion. Fitch counts ten of the sixteen non-traded business development companies it tracks breaching their quarterly caps. The unfilled requests do not disappear. They join the queue for next quarter.

Rate futures spent this week pricing the Federal Reserve's next move as an increase rather than a cut, with the ten-year Treasury trading near 4.8%. The European Central Bank is widely expected to raise its rate this month as well. For anyone who has spent two years planning around cheaper money arriving eventually, both sides of the Atlantic are now pointing the other way.

Vegetable oil is getting tight in a way that reaches the grocery shelf. Soybean-oil stocks are historically low, Black Sea sunflower-oil flows are under threat, and palm production carries its own risks into next season. New-crop American soybean export sales are running at more than double last year's pace. Cooking oil is one of those inputs that appears in almost everything and gets noticed by almost nobody until it moves. We'll see.

In 1966 an economist named Carl Stevens wrote a paper about arbitration.

His complaint was that the ordinary kind ruins bargaining. If both sides expect an arbitrator to split the difference, both sides name an extreme number. Neither side negotiates.

So he proposed a different rule. Each side submits one final figure. The arbitrator picks one of the two and may not choose anything in between.

Stevens argued this drives both sides toward a sensible offer, because a greedy number simply loses. He called it well designed to encourage genuine negotiation before the hearing.

The whole point of the mechanism was that it would hardly ever be used.

Major League Baseball took it up in the 1973-74 offseason to settle player salaries. The first hearings ran in 1974.

It behaved exactly as Stevens said it would. Between 1974 and 1993, only 9% of eligible players ever got as far as a hearing. In 2011 the figure was 2.5%. In 2013 not one player reached a hearing at all. The busiest year on record was 1990, with 162 cases filed.

Across the entire life of the system, arbitrators have decided 476 cases.

Four hundred and seventy-six hearings in fifty years. That was the design working.

Congress dropped the same mechanism into American medical billing.

In the first six months of this year, providers filed 1.75 million cases.

Baseball works because thirty clubs and a few hundred eligible players meet again every winter. Salaries are public. Comparisons are obvious. Everybody involved has a reputation to protect and will be back at the table next year.

Medical billing has none of that. Millions of separate claims, no comparison price both sides accept, and firms built specifically to submit at volume for a cut of what comes back.

Take away the repeat game and final-offer arbitration stops being a way to avoid hearings. It becomes a way to hold them.

The mechanism did not fail. It was installed somewhere it was never built for. We'll see.