Case Study 2 — The Specificity That Nobody Uses

A documented, widely studied phenomenon. Tier 1 for the structural facts; qualitative for magnitudes, which vary by setting, specialty, and study and should be taken from current sources.

Background

Section 7.9 said unspecified codes cost something and listed four costs. This case study is about the gap between what the classification can express and what actually appears on claims — and it is the best available illustration of why §7.1 insisted that a classification's design is a series of compromises with what gets recorded.

The transition to ICD-10-CM (Chapter 6, Case Study 1) was justified substantially on specificity. ICD-9-CM had roughly fourteen thousand diagnosis codes; ICD-10-CM has somewhere near seventy thousand. The expansion was not arbitrary. It added laterality, episode of care, stage, type, and a great deal of anatomical precision, and the argument for it was that better data would follow.

The transition succeeded technically. The specificity has been used far less than its design contemplated, and understanding why is more instructive than the fact itself.

The finding

Across settings and specialties, studies and payer analyses have consistently found substantial use of unspecified codes where more specific options exist. The pattern is well documented; the magnitudes vary widely — by specialty, by setting, by category, and by how the study defines "unspecified" — and anyone quoting a figure should quote it from a source rather than from a textbook.

The pattern is what matters, and it has three distinct causes that are frequently conflated.

Cause 1: the documentation does not support more

This is the largest cause and it is not a coding failure.

A note that says "knee pain" without laterality supports M25.569, and a coder who assigns it has done exactly the right thing. Chapter 4 §4.7 forbids inference; §7.9 says the unspecified code is correct when the record does not carry more.

The specificity ICD-10-CM added is only available if clinicians document it, and a great deal of it — laterality especially — is clinically irrelevant to the person writing the note. A physician examining a knee knows which knee it is and has no clinical reason to write it down, because the patient is in front of them and the treatment does not depend on recording the side in prose.

That is not carelessness. It is a mismatch between what documentation is for clinically and what it is used for administratively, and it is structural rather than fixable by exhortation.

Cause 2: the record supports more and nobody looked

This one is a coding failure, and §7.9's ⚠️ Where Claims Die describes it exactly. The assessment says "knee pain"; the examination two paragraphs above describes the right knee; the coder codes from the assessment alone.

This cause is invisible in aggregate data, because an unspecified code produced by cause 1 and one produced by cause 2 look identical on a claim. Only an audit that reads the whole record can distinguish them — which is why an internal audit that only checks whether the code matches the assessment will never find this category.

Cause 3: the unspecified code is safer

The defensive habit. A coder uncertain between two specific codes assigns the unspecified one, and experiences it as the conservative choice.

Chapter 5 §5.8 is the answer: it is downcoding, it is an inaccuracy, and it is not a defense. But the habit persists because it is genuinely never punished — no auditor has ever cited a practice for insufficient specificity, and the costs in §7.9 are all either invisible or land somewhere other than on the coder.

What it shows

First, the classification's expressiveness is bounded by the record, not by the code set. ICD-10-CM can express laterality on tens of thousands of codes. Whether laterality reaches a claim depends on whether a clinician wrote a word — and the design of the classification has no leverage over that at all.

This is the single most important structural fact for a newcomer to absorb about the specificity argument: adding a code does not add information. It adds a place to put information that somebody else has to produce.

Second, the three causes require three entirely different responses, and organizations that do not distinguish them waste effort:

An organization that responds to an unspecified-code finding with clinician education, when the actual cause was cause 2, has spent a great deal of political capital to fix the wrong thing.

Third, the cost landed somewhere the transition's designers did not fully anticipate. The justification for specificity was largely statistical and clinical. The place where unspecified coding now costs the most money is risk adjustment (Chapter 36), which grew enormously in importance after the transition was designed. An unspecified diabetes code and a specified one may pay identically on a fee-for-service claim and differ substantially in what a plan is paid for that member for the following year.

Fourth, and most honestly: some of the specificity was never going to be used, and that is a design question rather than a compliance one. A classification that offers distinctions clinicians have no reason to record has added categories that will be populated by "unspecified" indefinitely. Whether that was worth the transition cost is a real question, and reasonable people in this field disagree about it.

The lesson

Specificity is produced by documentation and merely recorded by codes.

Three carry-forwards:

Before assigning an unspecified code, search the rest of the encounter. Thirty seconds. It is the only one of the three causes a coder can fix alone, and it is the fastest correction in diagnosis coding.

When you find a pattern of unspecified coding, diagnose the cause before proposing a fix. Pull twenty charts and determine, for each, whether the record supported more. The answer determines whether this is a coder problem or a clinician problem, and getting that backwards is expensive in ways that are not financial.

And do not treat the unspecified code as shameful. It is correct when the record does not carry more, it is always valid, and a coder who feels bad about assigning it will eventually assign something they cannot defend. The rule is exactly as specific as the documentation carries you — Chapter 4 §4.7's ladder, in both directions.

Reported rates of unspecified code use vary widely by specialty, setting, category, and study methodology. Take figures from current sources, and check what the source counted before comparing it to anything.

Discussion questions

1. The case study says adding a code does not add information. Restate that as a general principle about classifications, then name one place in Part I where the same principle appeared in a different form.

2. Cause 1 and cause 2 produce identical claims. Design the audit that distinguishes them. What is the sample, what does the auditor read, and what does the finding look like?

3. A physician tells you that recording laterality in the assessment is "documentation for billing, not for patients." Is that true? Answer honestly, and then say what you would say to them.

4. §7.9 lists four costs of unspecified coding, and three of them land on someone other than the coder or the practice. What does that predict about how much effort organizations will put into fixing it, and is the prediction borne out by anything in this case study?

5. Was the ICD-10 transition's specificity argument worth it? Take a position, and be specific about what evidence would change your mind.