A surgical claim is not a list of codes. It is a set of decisions about what was separate, what was already included, and what the note can prove. The audit question is never whether a code exists for the service the surgeon performed. It is whether that code belonged on the same claim as the others.
What this covers
- The global surgical package already pays for the approach, the closure and the routine postoperative visits. Every separately billed service has to sit outside it.
- The highest yield audit query is not a code. It is two providers billing the same procedure for the same patient on the same day.
- Modifier 59 is the last resort. XS, XE, XP and XU describe the reason for separation more precisely, and payers read them differently.
- A modifier 22 carries no fee schedule percentage. The MAC can raise payment only after reviewing records, and most claims arrive without them.
- Sample by risk. A hundred charts drawn from four queries tell you more than a thousand drawn at random.
A bundled pair is not a coding error until you can name the code that should have absorbed the work.Bundling is a payment rule. The finding has to point at the procedure that already covered the service.
This is written for the billing manager or compliance lead who has been handed a year of surgical claims and told to find the problems. The instinct is to pull a random sample and read. That instinct is expensive and slow. Surgical errors cluster in a handful of patterns, and each pattern is visible in the claim data before anyone opens a chart.
Why the unit of audit is the claim, not the code
A single code can be perfectly supported and still be wrong on the claim. Medicare’s global surgical package pays for the procedure plus the services normally furnished around it, and the global period assigned to the code sets the boundary. Procedures carry a period of 000, 010 or 090 days. Codes marked XXX have no global period at all. Codes marked ZZZ are add-on codes, which are always billed with another service and never stand alone.
That structure is what makes surgery different from most other audit work. On an office visit claim, each line is a separate encounter with its own note. On a surgical claim, the lines interact. One procedure can absorb another, a modifier can claim a separation that the note does not support, and a payment rule can reduce a line to half its fee schedule amount before anyone asks whether it should have been there.
For every line after the first, ask what makes it separate. If the answer is that the surgeon did more work, the code is usually wrong. Separation is anatomic, temporal or by encounter, and the note has to name which one.
So the auditor reads a claim as a set of relationships. Which code is the primary? Which are add-ons? Which modifier asserts a separation, and what in the operative report establishes it? Most claims resolve on those three questions, and the answers sit in medical coding rules your team already has.
Four queries that find the cases worth reviewing
A random sample measures your luck. These four queries target the patterns that produce recoverable errors, and each one is a filter you can run against claims you already have. The first is the most productive.
| Query | What it surfaces | Where the error sits |
|---|---|---|
| Same procedure code, same patient, same date, two providers | Co-surgery and assistant-at-surgery billed without the modifier that reduces payment | Claim |
| More than one line with 59 or an X modifier in one operative session | Unbundled components reported as separate procedures | Coding |
| Modifier 22 above a dollar threshold or above a rate per surgeon | Increased work claimed without the record that proves it | Documentation |
| Add-on code billed without its primary, or with a primary from another anatomic family | A code that was never separately payable | Coding |
The Office of Inspector General ran the first query across calendar years 2017 through 2019 and reviewed a stratified sample of 100 services. Sixty-nine did not comply with federal requirements. Forty-nine were billed without the co-surgery modifier, 14 without an assistant-at-surgery modifier and six were duplicates. The estimated improper payments came to $4.9 million, and OIG attributed the pattern to CMS system controls rather than to individual coder error.
That finding matters for your audit design. Two providers billing the identical code for the same patient on the same day is a data pattern, not a clinical judgment. You can pull every one of those claims in an afternoon, and the modifier question has a single right answer per claim. Start there before you spend a week on documentation review.
The modifier 22 query needs a threshold to be useful. Set it in dollars, or set it as a rate per surgeon per quarter, and review everything above it. A surgeon who appends 22 to a quarter of their cases is describing something other than unusual work.
Reading the operative report against the codes billed
The note answers a fixed set of questions, and each one maps to a coding decision. Read it before the claim, or you hunt for support for codes chosen for other reasons.
Open, laparoscopic, arthroscopic or percutaneous. The approach decides which code family applies. An arthroscopy converted to an open procedure is not both procedures.
Name every organ, joint, nerve, vessel and lesion, with laterality. A separate structure is that supports an XS modifier, and it has to appear in the note in those terms.
Staged, related or unrelated decides 58, 78 or 79, and each carries a different payment consequence.
A harvested graft, a sent specimen or a placed implant may carry its own code, or may be included in the primary. The note tells you which.
Layered closure, drains and imaging guidance are usually included. The distinction is rarely worth a finding on its own.
Co-surgeon, assistant or team, and whether the second surgeon’s role appears in the note at all. A signature is not a role.
One habit makes this faster. Highlight the operative narrative in three colors: one for the primary procedure, one for anything the surgeon describes as additional, and one for the words that describe difficulty. Most disputed claims resolve on the third color, because a note that says a case was difficult without saying how is a note that cannot support modifier 22.
Notes describe what was done and omit where. Two procedures on the same organ, described in sequence, are bundled unless the note names a different site, a separate incision or a different joint. An auditor cannot infer the separation the surgeon did not write.
The modifier decisions that fail most often
59, XS, XE, XP and XU
CMS asks for the most specific modifier available. 59 identifies a procedure that was distinct from other non-E/M services on the same day, and it exists for the cases where nothing more precise fits. The X modifiers say why the service was distinct. XS means a separate organ or structure, XE a separate encounter on the same date, XP a different practitioner, and XU an unusual service that does not overlap the usual components of the primary.
A different diagnosis is not a reason to use any of them. The codes stay bundled unless the procedure was performed at a different anatomic site, during a separate encounter, or in one of the other defined situations. Payers read the modifier as an assertion about anatomy or timing, and they deny the line when the note describes neither.
Arthroscopy makes the rule concrete. 29827 with 29820 on the same shoulder in the same session stays bundled, and no modifier changes that. On different shoulders, the answer is RT and LT, not 59. Coders reach for 59 because it is familiar, and the finding is usually the same: the separation was laterality.
58, 78 and 79
These three describe a procedure during or after another one. A 58 is a staged or related procedure during the postoperative period, and it starts a new global period. A 78 is a return to the operating room for a related complication, and it pays on the intraoperative portion rather than the full package. A 79 is an unrelated procedure during the period, and it carries its own global period. Choosing between them is a clinical question that the note should answer directly. When the note does not, the claim usually lands on 78 because it pays something and does not require a new diagnosis.
50 and 62
Bilateral work goes out with 50 at 150 percent of the fee schedule. Co-surgery goes out with 62, at 62.5 percent for each surgeon. Team surgery uses 66 and is paid by report. Billing two full lines for the same procedure is the pattern OIG found most often, and a claim edit can block an identical code and date from a second provider without one of these modifiers.
22
Modifier 22 has no percentage attached to it. The MAC may increase or decrease payment only under very unusual circumstances, and only after reviewing medical records and other documentation. That makes 22 a documentation finding rather than a coding finding. The audit test is whether the note names the specific difficulty, quantifies it where possible, and reaches the payer with the claim. A 22 without the operative report is a claim asking for a review it has made impossible.
What the payment rules do to a multi-procedure claim
Knowing the arithmetic tells you what an unbundled pair actually cost. Under the standard multiple surgery rules, the highest valued procedure pays 100 percent of the fee schedule, the second through the fifth pay 50 percent, and anything beyond the fifth goes to manual review. A bilateral procedure is treated as 150 percent of the fee schedule and then ranked with the rest of the claim.
Multiple endoscopies follow a different rule. The payer takes the full value of the highest valued endoscopy, then adds the difference between the next endoscopy and the base diagnostic endoscopy, because the diagnostic procedure is already built into the therapeutic ones.
| Scenario | Payment | What it means for the audit |
|---|---|---|
| One procedure on the claim | 100 percent of fee schedule | The baseline every other line is measured against |
| Second procedure, same session, properly supported | 50 percent of fee schedule | Removing a second line recovers half the fee, not the whole fee |
| Bilateral procedure with modifier 50 | 150 percent of fee schedule | Two full lines overpay by half the fee |
| Co-surgery with modifier 62 | 62.5 percent to each surgeon | Two full payments is a duplicate, not a co-surgery claim |
| Sixth and later procedures | By report | Volume alone flags the claim for review |
Two conclusions follow. Report findings in dollars, not in chart counts, because a compliance committee responds to exposure, and a general surgery billing review that reports only chart counts rarely moves a budget. And separate your findings into overpayments already made and revenue at risk that was never paid. They need different fixes: the first needs a refund decision and a documented rationale, the second needs an edit in the scrubber before the next claim goes out.
How many charts, and which ones
You do not need to read every chart. You need enough charts per query to know whether a pattern is systemic, plus every chart in the highest dollar cells. A defensible starting point is 10 to 15 charts per query per surgeon, every claim above a dollar threshold you set with your compliance officer, and 100 percent of the claims where two providers billed the same code for the same patient on the same day.
Record the error rate per query, never as a single blended number. The method behind comprehensive chart auditing applies here on a narrower population. A 12 percent error rate that comes entirely from modifier 22 tells you where the next training hour goes. A blended 4 percent tells you nothing you can act on.
- Freeze the date range and the claim population before pulling anything, so the sample can be repeated later
- Pull the operative report, the claim, the remittance and the coding summary for every chart in the sample
- Score each finding as documentation, coding or payment, and record which one
- Total the dollars at risk and split them into paid and unpaid
- Keep a list of every rule you had to look up. That list is the training plan
The list of rules you had to look up is more useful than the findings. It names the two or three rules your coders do not hold in memory, and fixing those makes the next audit shorter.
What to do with a finding
An audit that ends in a spreadsheet changes nothing. Every finding needs an owner in three places: the coder who applied the rule, the surgeon whose note failed to support the claim, and the person who owns the claim edit that should have caught it.
For documentation findings, a one page query template per specialty works better than a policy memo. Surgeons answer a specific question about a specific note. For coding findings, change the scrubber edit, because a coder working under time pressure will repeat a rule they remember rather than one they were told. For payment findings, decide whether to refund and document the decision either way, since the 60 day clock runs from the date you identified the overpayment.
Then re-run the same four queries in 90 days. The measure of the audit is whether the error rate moved, not whether the training happened.
Questions we get about surgical coding audits
For a single specialty group, two to three weeks is realistic: one week to pull the population and run the queries, one to read the charts, and one to price the findings and write them up. Groups with several specialties and multiple locations should expect longer, mostly because operative reports arrive from more places and in more formats.
No. Audit by risk: every surgeon in a high dollar specialty at least once, plus any surgeon whose modifier 22 or 59 rate sits above your threshold, plus any surgeon named in a payer edit or denial pattern. Low volume surgeons with clean histories can be sampled every second or third year without much loss of assurance.
A coding audit asks whether the codes match the record. A documentation audit asks whether the record supports the service at all. In surgery the two overlap constantly, because most coding findings turn out to be documentation gaps: the note describes the work but not the site, the time or the difficulty that would have justified the code.
Use them as a screen, not as the finding. The Procedure to Procedure edits tell you which code pairs are usually not payable together, and each edit carries a modifier indicator that says whether a modifier can override the pair at all. When it cannot, the second code is denied regardless of documentation. When it can, the finding depends on the operative report, not the edit.
Twelve months is the usual working range, because it covers a full modifier cycle and stays inside most payer appeal windows. Anything older is usually worth reviewing only when a payer has already raised it, since the recovery effort exceeds the finding. Some practices run a two year lookback once, to establish a baseline, then move to annual samples.
The operative report, sent with the claim, describing the specific difficulty and how it increased the work. Terms like difficult or extensive on their own are not enough. The note should identify what was unexpected, how long the additional work took where time is relevant, and why the procedure could not be performed as typically described.
The bottom line
Surgical coding errors are rarely a matter of a coder not knowing a rule. They come from claims built line by line without anyone asking what made each line separate. A short, targeted audit tells you whether that is happening in your practice, and the four queries above tell you where to look first.
Not sure which surgical claims to pull first?
We audit a sample of your surgical claims against the operative reports, price the findings, and hand back a worklist with the dollar exposure attached. You also get the queries we used, so your team can run them again next quarter.
Request a free surgical coding auditThis article describes Medicare rules as published at the time of writing. Commercial payers and state Medicaid programs may apply different policies, and nothing here substitutes for a review of your own contracts and fee schedules.
