Case Study 2 — The Rule Nobody Remembered Configuring: A Composite

A composite constructed from the failure pattern described in §6.5 and documented enforcement themes. Tier 3; the practice and its figures are constructed. The mechanism is entirely ordinary.

Background

Section 6.5 asked three questions and asserted that most billing offices cannot answer any of them:

1. What automatic rules does our scrubber apply?
2. Who configured each one, when, and why?
3. When did anyone last review the output?

This case study is what happens when the answers are "we don't know," "someone who left in 2019," and "never." It is deliberately undramatic, because the point is that this is not an unusual situation. It is the normal situation.

The composite

Constructed. Not a real organization.

A nine-provider specialty practice uses a well-regarded practice management system with a configurable pre-submission edit engine. Over eleven years, staff have added rules to it — each one, at the time, a sensible response to a real problem.

The rules, as eventually inventoried:

Rule 7 is fine. It is documented, current, owned, and reviewed. It is included to make the point that automatic rules are not inherently wrong.

The other six are, collectively, a problem, and they are a problem in six distinct ways.

What each rule actually did

Rule 1 is Chapter 5's Figure 5.1 in miniature. It appends a modifier asserting a clinical distinction, on every claim, without any human confirming the distinction exists in the documentation. Some of those assertions were true. Nobody checked which.

Rule 2 changed a place of service code. Place of service affects payment — the site-of-service differential means the same service pays differently in a facility and a non-facility setting (Chapter 23 §23.5). The rule had been correcting a real problem in 2018; by the time it was found, it was misstating where services were furnished and, in some cases, causing them to be paid at the wrong rate.

Rule 3 silently discarded diagnosis pointers. Every claim with four supporting diagnoses lost the fourth, which meant that on any claim where the fourth pointer carried the diagnosis supporting medical necessity, the claim was submitted without the justification the payer needed — and denied for medical necessity, which the practice appealed, repeatedly, without ever understanding why.

This one is the most instructive. The rule was producing denials, the denials were being worked, and nobody connected the two. A scrubber rule can generate the work that hides it.

Rule 4 added a diagnosis code to claims. A rule that adds a diagnosis nobody documented is not a formatting convenience. It is the submission of a diagnosis that may not appear anywhere in the record, and it is difficult to characterize as anything other than a false statement, however innocently it began.

Rule 5 — the one nobody could explain — was suppressing a modifier, which meant claims were being submitted without an assertion the coder had made deliberately. The coder's work was being silently overridden by a rule from before their employment.

Rule 6 rounded units. Units on a drug code are a factual statement about how much was administered (Chapter 20 §20.3). Rounding them is misstating a quantity.

The inventory

The practice's new billing manager, four months into the job, asked §6.5's three questions and could not get answers. Rather than escalating, they spent a Friday afternoon on it.

The method was unremarkable:

1. Export the edit engine's rule configuration. (Half an hour. It had always been exportable.)
2. For each rule, identify what it changes, on which claims, and to which payers.
3. For each, find anyone who could explain why it exists. (Four of seven: nobody.)
4. For each, run a month of affected claims and read ten of them.

Step 4 is the one that mattered, and it is the one most inventories skip. A rule's description tells you what it was meant to do. Reading its output tells you what it does.

The finding, in one sentence: six of seven rules were changing the factual content of claims, none of them had been reviewed since configuration, and the oldest had been running for eleven years.

What it cost, and what it saved

Constructed.

The practice retained counsel, quantified the exposure for rules 1, 2, 4, and 6 — the four that affected the accuracy of what was represented — and self-reported. The overpayment was, in the end, modest: the constructed figure is \$61,400 across the affected period, most of it from rule 2's place-of-service effect.

Rule 3 went the other way. Once it was removed, the practice's medical necessity denial rate on the affected service fell substantially, and appeal volume dropped with it. The rule had been costing money for years in staff time and write-offs, and nobody had suspected the scrubber because the scrubber's job is to prevent denials.

And rule 5 was simply deleted, with no consequence anyone could identify, which is its own kind of finding: an automatic rule nobody can justify is probably not doing anything valuable, and it is certainly doing something.

What it shows

First, automatic rules are a category of risk that most organizations do not have on any risk register. They are invisible, they are cumulative, and they operate at exactly the volume that converts a small inaccuracy into a large one. Chapter 5 §5.3's reckless disregard is defined by the absence of review, and an unreviewed automatic rule is that absence made permanent and given administrative privileges.

Second, a rule can generate the work that conceals it. Rule 3 produced denials, the denials produced a work queue, the work queue produced appeals, and everyone was busy. Busy is not the same as correct, and a persistent denial category that nobody can explain deserves an upstream investigation, not a better appeal template.

Third, the inventory is cheap. Half a day. The rules were exportable the whole time. The barrier was not technical and it was not resources — it was that nobody had thought to ask, because a scrubber is understood as infrastructure rather than as a thing that makes assertions.

Fourth, the good rule proves the point. Rule 7 — the dollar-threshold hold — was documented, current, owned, and reviewed, and it was entirely appropriate. The problem is never automation. The problem is automation without review.

The lesson

Every automatic rule is a claim your organization makes on every affected claim, forever, without anyone reading it.

Three carry-forwards:

Inventory them. Half a day. Export the configuration, describe each rule, identify the owner, and read ten claims of output. The last step is the one that finds things.

Delete what nobody can justify. A rule whose purpose no one remembers is not neutral. It is changing claim content on the authority of a person who left.

And treat any persistent, unexplained denial category as an upstream question. If your practice has been appealing the same denial for two years and winning, the interesting question is not how to appeal faster. It is what is producing it — and one of the candidates is a rule intended to prevent exactly that denial.

Discussion questions

1. Rule 7 was appropriate and the other six were not. Write the four criteria that distinguish a legitimate automatic rule from a liability. Then apply them to rule 1.

2. Rule 3 produced denials that the practice successfully appealed for years. Nobody investigated upstream. What organizational features make that pattern likely, and what single report would have surfaced it?

3. The practice self-reported \$61,400. Using Chapter 5, argue both that this was obviously correct and that a rational owner might have hesitated. Which argument would you make to the owner, and in what order?

4. The inventory took half a day and had been possible for eleven years. Why do you think nobody did it? Give an explanation that does not rely on anyone being negligent.

5. Design the ongoing control. Not the one-time inventory — the recurring process that keeps this from accumulating again. Be specific about frequency, owner, and what "reviewed" means, because "reviewed annually" without a definition is how the next eleven years start.