Case Study 2 — The History Code That Wasn't: A Composite

A composite built from the Z85 rules in §10.5 and from documented risk-adjustment audit patterns. Tier 3; the organization and figures are constructed. The rule and the failure mode are real.

Background

Section 10.5 gave three conjunctive conditions for a personal history code: excised or eradicated, no further treatment directed to the site, and no evidence of remaining malignancy.

It also said something easy to skim past: coding an active malignancy for a patient in remission overstates their burden, coding history for a patient in treatment understates it, and the two errors are not symmetric in consequence.

This case study is about the first one, at scale.

The composite

Constructed. Not a real organization.

A large primary care group participates in a Medicare Advantage arrangement and, like most such groups, runs an annual review process to ensure that chronic conditions are captured for risk-adjustment purposes (Chapter 36). The process is not improper in concept — chronic conditions genuinely must be recaptured each year, and a condition that exists and is never reported is underreporting.

The problem is in the implementation. The group's process generates, for each patient, a list of conditions coded in prior years and prompts the clinician to address them at the annual visit.

Malignancies are on that list.

A patient who had a breast carcinoma excised in 2018, completed radiation in 2019, and has had no treatment and no evidence of recurrence since appears on the prompt list with the active malignancy code that was correctly assigned in 2018.

The clinician, working through a list at the end of a long visit, affirms it. The coder, working from a note that now names the malignancy, codes it.

And the patient has cancer again — on paper, for the fifth consecutive year.

Why nobody caught it

Four reasons, and each is ordinary.

The prompt list did not distinguish active from historical conditions. It was built from prior coded diagnoses and treated a 2018 malignancy code as a chronic condition needing annual recapture. A malignancy is not a chronic condition in that sense, and nothing in the tool knew that.

The clinician's affirmation was truthful in a narrow sense. The patient did have that cancer. The note said so. The clinician was not asserting active disease; they were confirming a history — but the note did not say "history," and the coder cannot supply it (Chapter 4 §4.7).

The coder followed the note. The assessment named a malignancy. §10.5's three conditions are determined from the record, and the record — as written — did not establish that treatment had ended or that no evidence remained. A coder reading only this note would have to code it as active, which is the correct behavior on the document in front of them.

And nothing rejected. This is Chapter 8's Case Study 2 again in a different costume. The claim was plausible, it paid, and no control the organization operated would flag a malignancy code on a patient who had, in fact, once had a malignancy.

What it cost

Constructed.

In risk-adjustment terms, an active malignancy and a personal history of malignancy are not equivalent — they describe different expected costs, and the group was paid on the difference for five years across an unknown number of patients.

In audit terms, this is precisely the fact pattern that risk-adjustment validation reviews are designed to find: a submitted diagnosis, and a medical record that does not support it. The record supported a history of the malignancy. It did not support active disease.

And in patient terms — which the organization discovered last and cared about most, once it was pointed out — a cohort of patients in long-term remission were carrying active cancer diagnoses on records that follow them.

What it shows

First, a recapture process is a coding process and needs coding discipline. The tool prompted; the clinician affirmed; the coder coded. At no point did anyone apply §10.5's three conditions, because the workflow was built around a different question — does this patient still have this condition? — rather than the classification's question, which is what does the record establish about it now?

Second, the error is invisible from the claim and visible from the record. As in Chapter 9's Case Study 2, no financial control detects this. Only reading records against the rules does.

Third, it demonstrates that "correct on this note" and "correct about this patient" can come apart. The coder's behavior was defensible on the document. The document was wrong. That is a documentation failure that presents as a coding outcome, and diagnosing it correctly determines whether the fix lands on the tool, the clinician, or the coder. Here it belongs on the tool.

Fourth, and most useful: the direction of the error matters. §10.5 said the two errors are not symmetric, and here is why. Overstating a patient's burden in a risk-adjusted arrangement produces a payment the organization was not entitled to, which is an overpayment with the sixty-day rule attached (Chapter 5 §5.1). Understating it produces a loss the organization absorbs. Both are inaccurate; only one of them is a repayment obligation and an audit finding.

The lesson

A condition that was true once is not thereby true now, and the record has to say which.

Three carry-forwards:

Apply §10.5's three conditions every time a malignancy appears. Excised or eradicated, no further treatment, no evidence remaining. All three. If the note does not establish them, the note describes active disease — and if that is not what the clinician meant, that is a documentation conversation, not a coding decision.

Audit any process that prompts clinicians from prior codes. These processes are widespread, they are not improper, and they have a specific failure mode: they treat every prior code as a chronic condition. Malignancies, acute conditions, and resolved conditions all need different handling, and a tool that does not distinguish them will generate exactly this error.

And watch the direction. Errors that overstate burden in a risk-bearing arrangement are the ones that become repayment obligations. That is not a reason to understate — Chapter 5 §5.8 is clear that downcoding is not safe — but it is a reason to look hardest at the codes that pay more.

Discussion questions

1. The coder coded what the note said and the note was wrong. Where does responsibility sit? Answer for the coder, the clinician, and whoever configured the prompt tool — and say which fix you would make first.

2. §10.5's three conditions are conjunctive. Design the query a coder should send when a note names a malignancy and the record suggests but does not establish remission.

3. Recapture processes exist because chronic conditions genuinely must be reported annually. Design a version of the prompt tool that gets chronic conditions right and does not produce this error. What does it need to know that the original did not?

4. The case study says the errors are not symmetric — overstating creates a repayment obligation and understating creates a loss. Is it therefore rational for an organization to bias toward understating? Answer using Chapter 5 §5.8, and be honest about the tension.

5. Compare this to Chapter 9's Case Study 2 (the ED coded as inpatient). Both produce records that misdescribe patients with no financial signal. What single control would have caught both?