Three things can go wrong when somebody keys a medical bill by hand. They are not equally expensive, and the order is not the one people expect.

The bill can land on the wrong claim. The bill can land on the wrong provider. Or the provider’s specialty can be wrong.

The first two eventually surface. The third one doesn’t, and it is the one worth your attention.

The two that announce themselves

A bill on the wrong claim becomes a reconciliation problem. It gets paid against the wrong file, the claim reserves look wrong, and somebody notices in a month or a quarter when the numbers are reviewed. Annoying, expensive in staff time, but visible.

A bill on the wrong provider becomes a payment problem. Same provider name, different NPI, different tax ID, sometimes a different contract entirely. This one usually surfaces faster, because someone eventually calls about a payment they did not receive, or a 1099 arrives somewhere it shouldn’t.

Both are real problems. Both have a moment where they become undeniable.

The one that doesn’t

Provider specialty decides which fee schedule applies.

That single sentence is the whole argument. In workers’ compensation the allowed amount is not a property of the bill — it is the output of applying a schedule to the billed service, and which schedule applies depends on what kind of provider rendered it. Get the specialty wrong and the bill still prices. It prices cleanly, against a real schedule, using correct arithmetic.

It just prices against the wrong schedule.

There is no exception to work. Nothing pends. The EOR looks entirely normal, because it is normal — every step downstream of the error executed correctly. The reduction is defensible on its face. The payment goes out. The savings get reported.

The wrong fee schedule was applied correctly. That is why nothing catches it.

It isn’t one bill, it’s every bill

This is the part that turns a small error into a structural one.

Claim errors and provider errors are per-bill. You mis-key one bill, one bill is wrong. Specialty is different, because specialty usually is not read off the bill at all — it comes from the provider record in your system, or from the taxonomy attached to the NPI.

Which means a wrong specialty is not a mistake on a bill. It is a mistake on a provider. And it silently misprices every bill that provider ever submits, from the day the record was created until the day somebody happens to look.

Provider records get set up once, usually quickly, often by whoever was available. They are rarely revisited. A practice changes what it does, a group adds a service line, a record gets created from a bill that was itself ambiguous — and nothing in the workflow ever goes back to check.

Why manual keying makes this worse

Not because people are careless. Because the person keying cannot see what they would need to see.

They have a document in front of them. They can read what is printed on it accurately — most people do, most of the time. What they cannot do, from a scanned CMS-1500 or UB-04, is answer the questions that actually matter:

  • Does this claim exist, and is this the right one?
  • Is this provider the provider already in the system, or a near-match with a different NPI?
  • Is the specialty on file the specialty this service was actually rendered under?

Those are lookups against other systems. They are not visible on the page. Asking somebody to catch them while typing two hundred keystrokes a bill is asking them to validate data they have no access to.

The throughput argument against manual keying is the obvious one and the least interesting. The real cost is that keying is a transcription step in a place where the work needed is verification.

Validate before import, not after

The fix is not better keying. It is moving the check earlier.

Everything needed to catch all three errors exists in systems you already have. The claim exists or it doesn’t. The NPI resolves against the registry or it doesn’t. The provider on the bill matches a record on more than a name, or it doesn’t. The specialty on file is consistent with the taxonomy and the services billed, or it isn’t.

None of that requires judgment. It requires the lookups to happen before the data lands in the bill review system, rather than after — because after import, the bill is already in workflow, and the only thing that will stop it is an exception nobody has a reason to raise.

That ordering is the whole design principle. Read the bill, then check what you read against the claim, the provider record, the NPI registry and the specialty. Import what passes. Hold what doesn’t, and say why.

What to do about it this week

You do not need a project to find out whether this is happening to you.

Pick your twenty highest-volume providers. For each one, compare the specialty on the provider record against the taxonomy registered to that NPI, and against what they actually bill for. You are looking for records where those three disagree.

If they all agree, that is genuinely good news and it cost you an afternoon. If any of them disagree, you have found a fee schedule that has been applied to every bill from that provider for as long as the record has existed — and you now know something about your program that no report was going to tell you.

Either way you will know, which is more than most programs can say about it.

We build automated bill intake that reads scanned CMS-1500 and UB-04 forms and validates the claim, the provider record, the NPI and the specialty before anything is imported. How BillCapture works.