Case Study 1 — Forty-Two Claims and Eleven Regrets: Account 31-2245 in Full

Constructed. Ridgeview Orthopedic Surgery, the claims, the audit, and every figure are constructed. The pattern — a modifier applied by automation against a note that does not support it, discovered in volume — is one of the most frequently documented findings in surgical coding.

This is the book's designated audit cautionary file. Chapter 5 introduced it. §17.9 coded it. This is what happened.

The claim

An eight-surgeon orthopedic group. A shoulder arthroscopy with rotator cuff repair — the practice's highest-volume major procedure.

Lines 1 and 2 are correct. 29827 is the repair, carrying a 090-day global. 29826 is an add-on code with a ZZZ indicator — it lives inside 29827's global, it cannot be reported alone, and modifier 51 does not belong on it. The practice had all of that right.

Line 3 is the file.

29822 is limited debridement. Under NCCI it is bundled into 29827 when the debridement is in the same anatomic region as the repair. The operative note documents debridement of the same structures that were repaired. Modifier 59 overrode the edit, the claim paid, and it paid every time.

How the modifier got there

Nobody put it there. That is the whole point of this case study.

A billing macro appended modifier 59 to 29822 whenever it appeared on a claim with another arthroscopy code. The macro had been configured years earlier, by someone who no longer worked there, in response to a run of denials that the macro did in fact resolve.

It worked. The denials stopped. Everyone was pleased.

Chapter 14 §14.1 said that appending a modifier by rule is making a factual assertion at volume without checking. Chapter 15's prefilled time default and Chapter 16's reimplemented leveling logic are the same failure in different fields. A configuration made an assertion, on 42 claims, that nobody in the building had ever evaluated.

And the macro removed the only step that would have caught it. With modifier 59 appended automatically, nobody ever had to read the operative note to decide whether the debridement was distinct. The question stopped being asked because the answer stopped being needed.

The audit

A commercial payer's special investigations unit did not start with charts either. It started with a frequency report.

Modifier 59 on 29822 is exactly the kind of thing Chapter 14 §14.5 described: a modifier whose sole function is to defeat a control, applied by the party that benefits, and measurable from claims data alone. The practice appeared as an outlier because its rate of 59 on that code pairing was near 100%, and a rate near 100% on a modifier that describes an exception is a contradiction in terms.

The review covered 42 claims over 18 months. Average overpayment: \$612.40.

   42  ×  $612.40  =  $25,720.80

The eleven

Here is the part of this file that matters more than the number.

The practice's own internal review, conducted after the demand, found that eleven of the forty-two were, in fact, separately documented and defensible. In those eleven cases the debridement had been performed in a different anatomic region from the repair — genuinely distinct, genuinely separately reportable, genuinely modifier-59 territory.

The practice could not prove it.

The operative notes in those eleven cases said things like "debridement performed" and "the joint was debrided." True statements. They do not establish a distinct anatomic region, and nothing written eighteen months after the fact can make them.

The surgeons remembered the cases. Two of them remembered specific patients. Memory is not documentation, and a payer is under no obligation to accept a recollection in place of a contemporaneous record — nor should it be.

   $25,720.80   demanded
   −  6,736.40   the eleven defensible claims (11 × $612.40)
   ───────────
   $18,984.40   what it should have cost

   The difference:  $6,736.40, lost to sentences nobody dictated.

What one sentence would have been worth

"Debridement was performed in the subacromial space, anatomically distinct from the repaired supraspinatus insertion."

That sentence, dictated at the time, in the eleven cases where it was true, is \$6,736.40.

It takes four seconds to say. It requires no additional work, no additional care, and no change to the operation. It requires only that someone had told the surgeons what the coding depended on.

Nobody had. And nobody had, because the macro made it unnecessary to know.

What it shows

First, automation does not remove the judgment. It removes the person who was making it. The question "is this debridement distinct?" did not go away when the macro was configured. It went unasked, which is worse than being answered wrong, because a wrong answer can be found and an unasked question cannot.

Second, a modifier rate near 100% is a self-report. Modifier 59 describes an exception. A practice applying it to essentially every instance of a code pairing has published, in claims data, that it is not evaluating the exception. You can run this report on your own organization, and if a distinctness modifier is over 90% on any code pairing, you have found something before someone else does.

Third — and this is the sentence to carry out of Part III — the defense is contemporaneous documentation, and it cannot be built retroactively. Not by memory, not by an attestation, not by a surgeon's letter written in good faith eighteen months later. The record either said it at the time or it did not.

Fourth, the practice's loss was not the overpayment. The overpayment on the thirty-one wrong claims was money it was never entitled to; repaying it restored the correct state of the world. The loss was the \$6,736.40 on the eleven right ones — real work, correctly performed, correctly billable, and unprovable.

And fifth, the fix was not a compliance program. It was two things: turn off the macro, and tell the surgeons what one sentence is worth. Both were done in a week. Neither required a consultant.

What the practice did

(Constructed.)

Removed the macro. Modifier 59 now requires a coder to open the operative note. Claim volume through coding rose; nobody found this a hardship.

Wrote a one-page document for the surgeons, listing the four or five places in their common procedures where a single anatomic sentence determines whether a service is separately reportable. Not a coding lecture. A list of sentences worth dictating.

Started a modifier frequency report, monthly, by code pairing. It is one query and it is the control that would have caught this in month two rather than month eighteen.

**And repaid \$25,720.80.** Including the \$6,736.40 they had earned.

The lesson

A macro that resolves denials is a macro that has stopped anyone from reading the note. Both of those things are true at once, and only one of them is on the report you are looking at.

Three carry-forwards:

Before a distinctness modifier goes on a claim, someone reads the note. Chapter 14 §14.5 said it as a discipline. This is what it costs when the discipline is automated away.

Run a modifier frequency report by code pairing. Monthly. A distinctness modifier above 90% on any pairing is a finding, not a coincidence.

And tell the clinicians which sentences are load-bearing. They are not withholding them. They do not know which ones matter, because nobody has ever told them, and a one-page list is the highest-yield document a coding department can produce.

Discussion questions

1. The macro was configured to resolve a real run of denials, and it did. Was configuring it unreasonable at the time? What should have accompanied it?

2. Eleven claims were defensible and unprovable. Is it fair that the practice repaid them? Argue the payer's position, then the practice's, then say where you land.

3. §17.9 says the defense cannot be built retroactively. Is there any legitimate way to establish, a year later, that a documented service was distinct? What are the limits, and where does legitimate clarification become something else?

4. The practice's fix cost almost nothing and took a week. Why do you think it took an audit? What would have had to be true for someone to do it in advance?

5. Compare this with Chapter 15's Case Study 2 (the prefilled time default) and Chapter 16's Case Study 2 (the stale leveling criteria). All three are configurations that made assertions nobody chose. Write the single control that would catch all three. Can one control do it?