Case Study 1 — The Category That Was Not There: A Composite

DataField.Dev

Case Study 1 — The Category That Was Not There: A Composite

Constructed. The practice, the payer, and the figures are not real. The failure — a root-cause list with no row for the thing that was actually happening — is ordinary, and it is the first failure in this book whose principal cost was borne by people rather than by an account.

Background

Section 29.4 gave a fifteen-item root-cause list and then attached a rule that reads like an afterthought:

"Payer error" must be available. If your list assumes every denial is your fault, §28.8's findings have nowhere to go and they will be miscategorized as coding problems.

This is a practice that had no such row, and the consequences of that omission ran for two years and were mostly not financial.

The composite

Constructed.

An eleven-provider group with a genuinely good denial operation. A log, a category list, a monthly report, a manager who read it. By the standards of this chapter they were doing almost everything right, and that is why the composite is worth studying.

Their root-cause list had eight categories, built years earlier by somebody sensible:

   REGISTRATION · AUTHORIZATION · CODING · DOCUMENTATION ·
   TIMELY FILING · COVERAGE · DUPLICATE · OTHER

Seven of the eight describe something the practice did.

"Other" was discouraged, because a catch-all collects everything and everyone knew it. The manager had, quite reasonably, asked staff to pick a real category wherever possible.

What was actually happening

(Constructed.)

A large payer's enrollment file ran behind.

When a member changed plans, or an employer's enrollment updated, the payer's eligibility system reflected it on its own schedule — Chapter 27 §27.2's 834, the transaction almost nobody in a billing office ever thinks about. The practice would verify eligibility at check-in and get an accurate answer, then submit the claim two days later and have it deny for coverage not in effect.

The front desk had done the verification. The verification had been correct. The claim denied anyway.

And there was nowhere to put it.

   THE DENIAL SAID ............ coverage not in effect
   THE NEAREST CATEGORY ....... REGISTRATION
   THE ACTUAL CAUSE ........... the payer's enrollment file
   THE AVAILABLE CATEGORIES ... did not include that

Staff categorized it as registration, which is what a conscientious person does when the list has no better row and "other" is discouraged.

Two years of the wrong conclusion

(Constructed.)

Registration became, and stayed, the practice's largest denial category.

The monthly report said so. The manager acted on it, correctly, by the logic available:

The front desk was retrained. Twice.

A verification checklist was introduced, and it was a good checklist.

A second verification step was added for certain visit types, which cost time at the desk.

And the number did not move, which was read as a training and diligence problem rather than as evidence that the diagnosis was wrong.

Every intervention was competent, well-intentioned, and aimed at a cause that did not exist.

The cost that mattered was not the denials. (Constructed.) It was two years of a department being told, monthly, with data, that it was failing at something it was doing correctly — and a supervisor who eventually stopped arguing about it because the report was not on her side.

How it surfaced

The front desk manager asked to see the denials.

(Constructed.) Not the report — the actual denied claims. She had been retrained twice, she did not believe the finding, and she wanted to look at the underlying items rather than the summary.

She pulled forty of them and read the eligibility verifications against the denial dates.

In a large majority, the verification had returned active coverage, on the date of service, from the payer that later denied the claim. She had the 271 responses. They were unambiguous.

The person the report was blaming was the person who found it, and she found it by refusing the summary and asking for the rows.

What it cost, and what it recovered

(Constructed.)

Financially, the denials were mostly recoverable and mostly had been recovered — they were soft, and the practice worked them. The direct loss was modest.

What was lost was two years of misdirected improvement effort, a verification step at the front desk that had never addressed the actual cause, and a working relationship between two departments that took longer to repair than the process did to fix.

What was gained, once the category existed:

The pattern became reportable. A category called PAYER — ELIGIBILITY FILE produced a number, and a number produced a conversation with the payer.

The conversation had an outcome. (Constructed.) Not a dramatic one — a documented process for resubmitting affected claims and a contact for the pattern — which is the ordinary result and is worth considerably more than nothing.

And the front desk's actual registration denial rate turned out to be low, which nobody had known, including the front desk.

What it shows

First, and this is the transferable sentence:

**A measurement system with a missing category does not report "unknown." It reports the nearest

available answer.**

It does not fail loudly. It produces a plausible number, in a real category, that a competent manager will act on. The absence of a row is invisible in every output of the system.

Second, the list's shape encoded an assumption. Seven of eight categories described something the practice did. That is not neutral — it means the system could only ever conclude that a denial was the practice's fault, and it was a design decision nobody remembers making.

Third, discouraging "other" made it worse, and the instruction was correct. §29.4 says a catch-all collects the interesting third; this manager had acted on exactly that advice. The lesson is not that the advice was wrong. It is that discouraging "other" only works if the list is complete, and nothing in the system could reveal that it was not.

Fourth, the finding required someone to refuse the summary. The report was accurate about what it measured. Every number in it was correct. The only way to see the problem was to read forty underlying claims, and the person motivated to do that was the one the report was blaming.

And fifth, this is the first failure in this book whose principal cost was borne by people. Chapter 26's clerk was never told her field mattered; this department was told repeatedly that it was failing at something it was doing right, with data, monthly, for two years. The claims were recovered. That is not.

The lesson

A root-cause list is a hypothesis about what can go wrong. If it is incomplete, your data will be confidently wrong in the direction of whoever is on the list.

Four carry-forwards:

Put "payer error" on the list, and a row for external causes generally. §29.4. A list that cannot express "this was not us" will never say it, no matter how many times it is true.

Review the categories, not just the numbers. Once a year, ask: what have we seen this year that had no good row? The people who work the queue know the answer immediately and are almost never asked.

When an intervention does not move a number, question the diagnosis before questioning the effort. Two retrainings that change nothing are evidence. The practice read them as evidence about the front desk; they were evidence about the category.

And read the underlying rows before acting on a summary about a department. Forty claims. It is an afternoon, and in this composite it was the only thing that worked.

Discussion questions

The category list was built by somebody sensible and had eight reasonable rows. What question would have revealed the gap at the time it was built?
The manager discouraged "other," following exactly the advice in §29.4. Was that wrong? State the condition under which the advice holds.
Seven of eight categories described something the practice did. Is that a natural way to build such a list? What does it suggest about how these lists get made?
The person the data was blaming is the person who found the error. What does that predict about who is motivated to audit a measurement, and is that a reliable control?
The direct financial loss was modest and the case study calls the cost serious anyway. Do you agree? Say what was actually lost and whether it can be priced.
Compare with Chapter 28's Case Study 1, where a number moved and was explained away. Here a number was accurate and pointed at the wrong department. Which failure is harder to detect?