Six defects that change the money without ever denying a claim
A denial gets worked. Somebody owns it, and it shows up on a report. These six do not deny — every line stays payable and every modifier stays correct — so nothing routes them to anyone. They are found in a regression deck or they are not found.
1. Bilateral before or after the endoscopic base subtraction
CMS applies the bilateral adjustment before the multiple-procedure rules, so the family’s diagnostic base code subtracts from the 150% amount, not from the unilateral one. Reverse the two and you subtract first, then inflate the smaller number.
On 31256 at a facility rate of $158.80: bilateral first gives $238.20 − $56.23 = $181.97. Base first gives ($158.80 − $56.23) × 1.5 = $153.86. The order is worth $28.12 on one line, and nothing about the claim looks wrong either way.
2. One line taking two reductions
The endoscopic reduction is the family’s reduction. A catch-all multiple-surgery pass that cuts the same line a second time, because it is not the highest-valued procedure on the claim, is the most expensive sequencing defect on a mixed endoscopic claim: $157.84 becomes $78.92.
Nothing denies. The line is payable, the modifier is right, and the remittance shows a paid line at half the amount it was owed.
3. Treating MAI 2 and MAI 3 as the same edit
| Indicator | What happens to the excess units |
|---|---|
| MAI 1 | A per-line edit. Genuinely furnished units above the limit can go on a separate line. This is 106 of 33,354 codes — 0.3%. |
| MAI 2 | A date-of-service edit and absolute. Not payable, not appealable. Bill the maximum. |
| MAI 3 | A date-of-service edit set from a clinical benchmark. The excess is payable where the units are correctly coded and medical necessity is documented. |
Neither MAI 2 nor MAI 3 clears by splitting units across lines — the edit sums the date, so two lines of four are the same eight units. But collapsing them into one rule costs real money in the other direction: treat an MAI 3 like an MAI 2 and you write off an amount that documentation would have recovered. See unit limits and the MAI indicator.
4. Reading the modifier indicator from the wrong file
The correct coding modifier indicator is not a property of a code pair. It is a property of the pair, the file and the program, and it changes quarterly. The same two codes carry different answers at the same moment:
| 00100 with 94761 | Modifier indicator |
|---|---|
| Medicare, practitioner file | 0 — no NCCI-associated modifier bypasses it |
| Medicare, outpatient hospital file | 1 — a modifier may bypass it where the circumstances support one |
| Medicaid, practitioner | 1 |
A system that holds one indicator per pair is right about two-thirds of that table. Indicator 0 is also scoped to the same provider, same beneficiary and same date of service, so it is not a statement that the two codes can never both be paid. See where the two programs differ.
5. Wage index vintage
A facility pricer that resolves a wage index from the provider’s ID has to bind it to a rate year. Pair a new year’s rates with last year’s index and the result looks entirely plausible — it is a dollar figure in the right range, and no edit inspects a wage index.
Between FY2026 and FY2027, 2 of 468 SNF payment areas had an identical wage index. The other 99% mispay, by a median of about 1.4% and by more than 10% at the tail. The two that did not move are pinned by a floor rather than by their area’s wages, which is why they are the exception.
This is an estimator, repricing and contract-modelling problem rather than a claims-adjudication one: Medicare’s own pricer is selected by the discharge date, so it cannot happen in fee-for-service. It also is not a single “same year” rule — SNF, IRF, IPF, LTCH and hospice run on the federal fiscal year while home health, ESRD and OPPS run on the calendar year, so a correctly configured system holds different years at once by design.
6. Ranking that moves lines but not totals
This one is the opposite of the others, and it is the one most often got wrong in both directions. Within an endoscopic family the reduction is a flat subtraction of the base code, and flat subtraction is commutative:
A + (B − base) = B + (A − base)
So ranking the family the wrong way round does not cost the claim anything. The total reconciles to the cent. What changes is which line carries the subtraction — and that decides the patient’s coinsurance, what a secondary payer sees on crossover, per-line appeal rights, and any analysis that ranks codes by what they actually paid.
A vendor telling you a ranking defect cost you money is describing something that did not happen. A vendor telling you it does not matter is describing something that did. See how the reductions stack.
Why none of these reach a report
Denial reporting answers “what did we refuse to pay, and why”. Four of the six above never refuse anything, and the two that do — the unit limits — deny correctly and get worked as though the answer were final. Nothing in the remittance distinguishes a line paid at the right amount from one paid at half of it.
What surfaces them is a regression deck: a fixed set of claims whose correct answer is known independently, re-run whenever a rate file, an edit file or a configuration changes. If the deck has no claim that tests sequencing, sequencing is untested.
Figures are facility-setting CY2026 amounts, NCCI July 2026 files, and the FY2026 and FY2027 SNF wage index tables. Every one is reproducible from the tables this site runs on.
Check it against your own claim
Reading about the rule is not the same as knowing whether it fires on the codes in front of you. The checker answers that against the policy in force, and says what would clear it.
Check a claim against all six