By the last week of February, most practices have posted a full month of Medicare remits for January dates of service. Those remits are the first real test of the 2026 fee schedule, and in the practices we have reviewed so far the posting teams have been doing what posting teams do: posting the allowed amount the MAC sent as the contractual allowed, and moving on. Nobody has compared it to what the allowed should be.
In an ordinary year that is a small risk, because the MAC pricing is usually right. This year it is a larger risk, because expected reimbursement changed for almost every code and in more than one direction at once. Practices that have not rebuilt their expected amounts cannot tell a correct 2026 payment from a wrong one. Here is how to build the expected amount and what to look for.
Key takeaways
- Posting the MAC's allowed amount as the contractual allowed makes underpayments invisible by definition; the expected amount has to be loaded from the fee schedule file.
- Every input to the Medicare formula changed for 2026: two conversion factors, a 2.5 percent work RVU cut on most procedures, new practice expense allocation and updated GPCIs.
- Apply the adjustments the fee schedule file does not include (sequestration, MPPR, bilateral, assistant at surgery, non-physician practitioner, PC/TC) before comparing.
- Systematic variance on a code is a setup error on one side and repeats every month; scattered variance is claim-specific and worth a reopening if the dollars justify it.
- Commercial contracts that pay a percentage of current Medicare had to load 2026 too, and some are still paying on 2025.
The formula
Medicare pays a physician service as the sum of three components, each adjusted for geography, multiplied by the conversion factor:
Allowed = [(work RVU × work GPCI) + (practice expense RVU × PE GPCI) + (malpractice RVU × malpractice GPCI)] × conversion factor
Every input changed for 2026. The conversion factor is $33.57 for qualifying APM participants and $33.40 for everyone else. Work RVUs fell 2.5 percent on most non-time-based services through the efficiency adjustment. Practice expense RVUs shifted between facility and non-facility settings under the new indirect cost allocation. The GPCIs were updated as part of the periodic review. The MAC fee schedule file for your locality has already done this arithmetic; the reason to understand it is so you can tell which component explains a variance.
Step one: load the right expected amount
Download your MAC's 2026 fee schedule file for your locality and load the non-facility and facility amounts for every code you bill. Then apply the adjustments the file does not include:
| Adjustment | Effect on expected payment | How to check |
|---|---|---|
| Sequestration | 2 percent reduction to the Medicare payment (not the allowed) after deductible and coinsurance | Appears as CO-253 on the remit; verify it is 2 percent of the payment amount |
| Conversion factor | Fee schedule file may be published for one factor; confirm which applies to your TIN | QP status in the QPP lookup; the difference is about 0.5 percent |
| Multiple procedure payment reduction | Second and subsequent surgical procedures on the same day paid at 50 percent; imaging and therapy have their own reductions | Modifier 51 or MAC-applied ranking; check the order the MAC ranked them |
| Bilateral procedures | Modifier 50 pays 150 percent of the fee schedule for eligible codes | Bilateral indicator in the RVU file |
| Assistant at surgery | Modifier 80, 81, 82 pays 16 percent; AS for non-physician practitioners pays 85 percent of that | Assistant surgery indicator in the RVU file |
| Non-physician practitioner | Services billed under an NP or PA NPI pay 85 percent of the fee schedule | Rendering NPI on the claim |
| Professional and technical components | Modifier 26 and TC split the global amount | PC/TC indicator in the RVU file |
Step two: run the variance report
For every paid line in the February remits, compute expected allowed minus actual allowed. Set a threshold; we use one dollar per line to filter rounding, and then sort by total variance per code. In a typical practice the report shows three kinds of results.
No variance on most lines. Good. This confirms the fee schedule load, and you now have a baseline for the rest of the year.
Systematic variance on a code or a group of codes. This is a setup problem on one side or the other. If the MAC allowed is consistently higher than your expected, your load is stale or you applied the wrong conversion factor. If the MAC allowed is consistently lower, check whether the code was subject to the efficiency adjustment (you may have loaded a 2025 RVU), whether the place of service on the claim pulled the facility rate when you expected non-facility, or whether the MAC applied a reduction you did not model.
Scattered variance on individual lines. These are claim-specific: an MPPR ranking that put the wrong procedure first, a modifier 50 that was not recognized, an assistant at surgery paid at the wrong percentage, a locality error because the service location on the claim does not match the enrolled location. Each is worth a reopening request if the dollars justify it, and each points to a claim setup fix that stops the pattern.
A worked example
Take an illustrative procedure with a 2025 non-facility fee schedule amount of $410.00 in your locality. In 2026 the work RVU fell 2.5 percent through the efficiency adjustment; suppose work was 60 percent of the total RVUs, so the total RVU fell about 1.5 percent. The conversion factor for a non-QP practice rose 3.26 percent. Net, the 2026 fee schedule amount is roughly $417.00. The remit shows allowed $417.00, deductible applied $0.00, coinsurance $83.40, payment before sequestration $333.60, sequestration $6.67, payment $326.93. That is a correct payment. If the remit instead shows allowed $404.00, the MAC priced the facility rate; check the place of service on the claim. If it shows $417.00 allowed but payment of $320.00, the sequestration or coinsurance math is off and the line needs a call.
The numbers in that example are illustrative, not from any fee schedule. The method is the point: know the expected amount to the dollar, and every deviation has one of a short list of causes.
Step three: extend to commercial payers
Many commercial contracts pay a percentage of the current Medicare fee schedule, and those payers also had to load 2026. In our experience some load it late, some load it wrong, and some quietly keep paying at the prior year's schedule until asked. Run the same variance report for each contracted payer against the contract percentage of the 2026 schedule. A payer whose contract says 120 percent of current Medicare and whose February remits match 120 percent of 2025 Medicare owes you the difference on every line, and the contract language on which year applies is what you will need to collect it.
The commercial version of the report needs one more column: the contract's definition of the schedule. Some contracts say "the Medicare fee schedule in effect on the date of service", which floats. Some say "the 2024 Medicare fee schedule", which is frozen and will not move in 2026 no matter what CMS did. Some say "Medicare" without a year, which is an argument waiting to happen, and worth settling in writing before the next renewal. Some pay a percentage of Medicare for E/M codes and a fixed schedule for procedures. Load the definition per contract, not per payer, because a large payer often has several contracts with a practice across product lines. In our experience the most common finding is not a payer that loaded the wrong year but a practice that loaded the wrong contract.
Underpayment review is one of the six checks in the RCM audit we run for practices, and for the practices we bill, the variance report is part of the monthly close. The earlier piece on identifying revenue leakage puts this check in context with the other five.
Questions we hear
Our posting software marks every Medicare payment as paid in full. Is the variance report still worth doing?
Yes, and that setting is the reason. Posting software that accepts the payer's allowed as the contractual allowed cannot detect an underpayment by definition. The variance report has to be built from a loaded expected amount, either in the practice management system's contract module or in a spreadsheet from the 835 data.
How long do we have to request a Medicare reopening for an underpayment?
A clerical error reopening can be requested within one year of the initial determination for any reason, and longer for good cause. Most pricing errors qualify as clerical. Start with the MAC's reopening process rather than a formal redetermination; it is faster for pricing issues.
The variance report shows we were overpaid on a few lines. Do we have to do anything?
Yes. Medicare overpayments must be reported and returned within 60 days of being identified, and the variance report is how you identified them. Confirm the overpayment is real (an MPPR the MAC failed to apply is the usual cause), then use the MAC's voluntary refund process. It is not a large amount of money in a typical month, but the 60-day rule is a compliance obligation with real consequences, and a practice that runs a variance report and acts only on the underpayments has a document showing it knew about the overpayments. Run the report, work both directions, and keep the log.
What to do this month
- Load the 2026 MAC fee schedule for your locality, facility and non-facility, and confirm which conversion factor applies to your TIN.
- Run the expected versus actual variance report for all January dates of service paid in February.
- Investigate systematic variances first; they are setup errors and they repeat.
- Submit reopening requests for claim-specific underpayments above your threshold, and refund any confirmed overpayments within 60 days.
- Run the same report for each commercial payer whose contract references the Medicare fee schedule, using the contract's own definition of the schedule.
- Schedule the report monthly; the fee schedule can change mid-year through corrections, and payers change their loads without notice.
