Case Study 1 — The Rule That Stopped Being Right: A Composite

Constructed. The practice, the payer, and the figures are not real. The failure — a posting rule that was correct when it was written and became wrong when a payer changed its behavior — is ordinary, and it is the twelfth configuration finding in this book with a mechanism none of the other eleven had.

Background

Section 28.6 named the most expensive single posting error in the chapter:

A CO-97 for \$128.40 posted as "contractual adjustment" produces an account that looks paid in full. The claim balances. Nothing appears on an aging report, in a denial log, or in a work queue.

The money is gone and the account looks correct.

This is a practice where that happened at volume, for about two years, because of a rule nobody had any reason to revisit.

The composite

Constructed.

A fourteen-provider primary care group. Autoposting since well before anyone currently employed there started. It works, and it works well — the practice posts a large volume accurately and quickly, which is exactly what §28.7 says autoposting is for.

When the rules were configured, somebody made a reasonable decision.

"Any adjustment with a CO group code posts to contractual adjustments."

At the time, that was very nearly true. (Constructed.) The practice's payer mix produced overwhelmingly CO-45 — the ordinary charge-exceeds-contracted-amount adjustment — and the handful of other CO codes were rare enough that nobody built a branch for them.

The person who wrote the rule was right. Nobody has ever been able to identify who it was.

What changed, and it was not the practice

(Constructed.)

A major payer revised its edit configuration. Not dramatically, and not in a way it announced to providers as a change — it tightened a set of bundling edits, of the kind Chapter 21 covers, which began firing on a common pairing the practice bills constantly.

The result was a steady flow of CO-97 adjustments where there had previously been almost none.

Every one of them posted to contractual adjustments.

   THE PAYER              began producing CO-97 at volume
              ↓
   THE RULE               "any CO → contractual adjustment"
                          did exactly what it was told
              ↓
   THE ACCOUNTS           balanced. Zeroed. Aged out of nothing.
              ↓
   THE DENIAL LOG         showed no increase, because these
                          denials never entered it

Nothing was misconfigured. The configuration became wrong.

The number that moved, and the story that explained it

This practice's failure is different from an invisible one, and that is what makes it worth studying.

(Constructed.) Contractual adjustments as a percentage of gross charges rose by several points over about eighteen months.

That number was on a monthly report. The practice manager reviewed it. It was discussed.

And it was explained.

"Payer mix has shifted. We're seeing more of the plans with the deeper discounts."

Which is a completely reasonable explanation, is the correct explanation the large majority of the time a practice sees that number move, and was wrong.

Nobody tested it. Testing it would have meant breaking the contractual adjustment total down by reason code — a query nobody had ever needed to run, against data the posting rule had already homogenized.

How it surfaced

A physician.

(Constructed.) One of the fourteen asked why his collections per visit had drifted down over two years while his visit volume and his coding profile had not changed. He was not making an accusation; he was asking a question about his own production report.

The answer required somebody to look at the adjustments on his claims specifically, and the person who looked found CO-97 adjustments on office visits — modifier-25 visits billed alongside minor procedures, exactly the pattern Chapter 21 predicts and exactly the pattern Account 10-4471 produced.

She pulled a month. Then a year.

(Constructed.) A substantial and rising volume of bundling denials, none of which had ever been worked, appealed, or counted.

What it cost

(Constructed.)

Two years of a recurring denial nobody saw.

And here is the part that hurts, which §28.6 predicted: most of it was not recoverable.

The appeal deadlines had passed on the great majority of the affected claims. Chapter 30's levels and windows all run from the date of the determination, and the determination was made on the day each remittance posted.

A defined subset was still inside the window and was appealed. (Constructed.) The overturn rate on those was high — because the denials were wrong, the notes documented separately identifiable services, and the modifier was properly appended. Which is the most painful possible finding: the practice would have won nearly all of them.

What was recovered going forward was larger than what was recovered backward. The rule was fixed, the pattern was worked prospectively, and — because the denials were now visible — the practice was able to raise the underlying edit behavior with the payer, which it could not have done while it had no data.

What it shows

First, and this is the whole case study: a correct configuration decayed because the world changed around it. Chapter 27's Case Study 2 had a configuration that was wrong from the moment it was made and was masked. This one was right when it was made and stopped being right, and there was no event inside the practice to trigger a review. Nobody had done anything.

Second, the metric moved and was explained away. This is a different failure from Chapter 27's Case Study 1, where the metric moved in the flattering direction. Here the number went the wrong way, was seen, was discussed, and was assigned a plausible cause nobody tested. That is arguably worse: the organization used its attention and got the wrong answer, which is more discouraging than not looking.

A number with a story attached stops being a question. The story does not have to be self-serving to do the damage — this one was routine, professional, and true in most practices most of the time.

Third, the posting rule destroyed the evidence needed to test the story. By the time the number reached the report, every CO adjustment looked the same, because the rule had made them the same. The query that would have answered the question in ten minutes was impossible against the practice's own data. A posting decision made years earlier had removed a category of question from the organization.

Fourth, the finding came from outside the business office. A physician looking at his own production report asked a question nobody in billing had asked. Chapter 26's Case Study 1 was detectable from outside the hospital before it was detectable from inside; this one was detectable from a different department. Both are versions of the same uncomfortable point: the people closest to a process are frequently the least able to see it.

And fifth, the practice would have won the appeals. (Constructed, and it is the ordinary case.) The claims were correct. The loss was not a coding failure, a documentation failure, or a payer being right. It was a routing decision made by software, and every one of those denials was a claim the practice was entitled to be paid for.

The lesson

A posting rule is a standing decision about what your organization is allowed to notice.

Four carry-forwards:

Never map a group code to a destination. Map the CARC. §28.3 and §28.6. CO-45 is a contractual adjustment. CO-97, CO-50, CO-151, and CO-16 are denials that happen to carry a CO group code, and the distinction is the entire case study. This is a configuration change measured in hours.

Break contractual adjustments down by reason code, monthly. One report. If the practice in this composite had ever run it, the CO-97 line would have appeared out of nowhere and grown, and somebody would have asked. The report is trivial and almost nobody produces it.

When a number moves and somebody explains it, ask what would have to be true. "Payer mix shifted" is testable in about ten minutes if the data survived. The explanation was not the problem; accepting it without a test was.

And review posting rules on a schedule, not on an event. There will be no event. The rules in this composite were never wrong on any day anybody was looking at them — they were correct when written and correct at every review that never happened, and they were wrong for two years.

Discussion questions

1. The rule was correct when it was written. Who owned it afterward? Name a role and say what its review cadence should be.

2. "Payer mix has shifted" is usually the right answer. What would have made someone test it here? Write the specific check.

3. The posting rule destroyed the data needed to test the explanation. Name another decision in this book that removed a category of question from an organization.

4. The practice would have won most of the appeals and could not file them. Is the appeal deadline running from the remittance date a fair rule? Argue both sides.

5. A physician asked the question that found it. What does that suggest about where to look for revenue-cycle problems, and is that suggestion practical?

6. Compare with Chapter 27's Case Study 2. Both are configuration failures. State the difference in mechanism in one sentence each, and say which is harder to prevent.