About Solutions Your data Your connections How it works Fractional Services Innovation Lab Insights Contact Client Login Start a Conversation
← All insights
Innovation & Future Models

The Industry FHIR Was Supposed to Shrink

3 min read

There is a large, mature, profitable industry that sits out in front of the healthcare payment pipeline, reviewing what already happened: what was coded wrong, what was paid at a rate that didn’t match, what needs to be recovered. Call it payment integrity, claims auditing, or recovery review. It’s a real business, built on real errors. But it should be getting smaller. If interoperability standards did what they say on the label, the reason it exists, messy, inconsistent, error-prone data leaving the building in the first place, would mostly go away. It hasn’t. That’s worth asking why.

The answer isn’t that these companies are doing anything wrong. They exist because claims data has historically been produced downstream of the actual event, coded and submitted after the fact, and checked for correctness only once the claim has been paid. Somebody has to catch what’s wrong, and catching it after the money moves is better than not catching it at all. It’s a legitimate business. It’s just one whose size directly measures how much correction has to happen after the point where the data should already have been right.

Part of the reason is what “FHIR-compliant” tends to mean. Ask most vendors and the answer is yes, because somewhere in the pipeline the data eventually gets converted into a FHIR bundle. What that answer leaves out is where. For most systems it happens after the request has already been received and stored, waiting in whatever form it arrived in: a scanned fax, a flat file, a nightly batch export. The FHIR version is a translation, produced later, of a record that already exists somewhere else. That makes two records, not one, and they can drift apart the moment either changes. By then, the cheapest moment to catch an error, when the request first arrives, has already passed.

The next few months will make this more common. CMS-0057-F requires impacted payers to have standardized FHIR APIs, including a dedicated prior authorization API, live by January 1, 2027. The rule sets a deadline for the endpoint to exist. It doesn’t say where in the pipeline the conversion has to happen. Many payers will meet it by putting a vendor in front of their API, a weigh station: the vendor takes the provider’s fax and converts it upstream, so the payer is compliant and the provider still has no direct path in. A system can meet the letter of that requirement by bolting a FHIR export onto the end of the same slow intake process it has always had, checking the box without touching the actual bottleneck, and without shrinking the market for anyone whose business is catching what that old process still gets wrong.

Converting when the request first arrives is a better design. The structured version isn’t produced from the record; it is the record. Validation (does this bundle conform to the standard, is anything missing) happens before the request moves anywhere, not as a cleanup pass after something downstream chokes on bad data, or after an audit catches it a quarter later. There is one representation of the request, not two to reconcile.

That reaches the correction industry because a claim is built from the request and the encounter behind it. An error that is never caught at the start carries all the way through to payment. This won’t end every dispute, since some are about rates and policy rather than data. But it removes one class of them, the missing and mismatched data, at the point where fixing it costs the least.

A pipeline where the record is validated before anything downstream can act on it doesn’t need as large an industry standing in front of it. That was supposed to be the point of a shared, structured standard. Whether it delivers depends on where in the pipeline the conversion happens, and for most systems today that is after the point where it would have mattered. It’s the layer we build at Orion: conversion at intake, so the request is already right when it moves.

More from Insights

Let's connect you to what your data already knows.

Tell us what is getting in the way. Let’s work on it together.

Start a conversation
What's underneath