E/M Undercoding: How to Tell If Your Practice Is Billing Too Low

Why undercoding stays invisible

Every other revenue problem announces itself. A denial produces a remittance code. A rejection bounces back from the clearinghouse. An underpayment shows as a variance against the contracted rate. Each one lands in somebody’s work queue.

Undercoding produces none of that. The claim is clean. It pays on the first pass, at the full allowed amount for the code submitted. The only evidence is a note in the chart describing more work than the code reflects, and nobody in the billing workflow is reading the note.

That is why it accumulates. A practice can undercode consistently for years while every operational metric — clean claim rate, days in A/R, denial rate — looks healthy.

How office visit levels are actually chosen

Since the 2021 AMA revision to office and outpatient E/M, you select the level of a 99202–99215 visit one of two ways: by medical decision making, or by total time on the date of the encounter. History and exam no longer drive the level. You perform and document them as clinically appropriate, but they do not decide the code.

Most practices adapted their documentation to the change. Fewer adapted their code selection. That gap is where the money sits.

Route one: medical decision making

Medical decision making has three elements. The level is set by whichever two of the three are met:

• Problems addressed.The number and complexity of the problems you managed at that visit — not the problems on the chart, the ones you actually addressed.
• Data reviewed and analyzed. Unique tests ordered or reviewed, notes obtained from other clinicians, your own independent interpretation of a test somebody else billed, and discussion of the case with an external professional.
• Risk. The risk of complications or morbidity from the management decisions you made, including the decision to do nothing.

Route two: total time

Total time is your own time on the date of the encounter. It is broader than the time in the room. It includes reviewing the record beforehand, ordering tests and medications, counseling, coordinating care, and documenting the visit that day. It excludes clinical staff time and any service you bill separately.

You may use whichever route gives the accurate answer for that visit. A long counseling visit for a straightforward problem is a time visit. A short visit for a patient on four interacting medications is an MDM visit. Practices that only ever use one route leave the other on the table.

A worked example

Two chronic illnesses with exacerbation is generally a moderate problem set. Prescription drug management is generally moderate risk. That is two of three elements at moderate, which supports a 99214 on medical decision making alone, before you consider time or the data element.

Visits like this are routinely filed as 99213. Not because anyone decided to, but because 99213 is the default the practice drifted into. The landing page for this platform puts that gap at roughly $30 to $80 per encounter. One such visit a day, per provider, compounds into real money over a year.

Five things that cause it

1. Audit fear. Someone was burned once, or heard about someone who was, and the practice quietly agreed that 99213 is the safe choice. It is not safe. It is inaccurate in the direction that costs you money.
2. Time is never counted. Providers record the time in the room, if they record time at all, and never count the twenty minutes of chart review, ordering, and documentation around it.
3. The data element is under-counted.Independent interpretation of an outside test, or a real discussion with the patient’s cardiologist, both count and both are usually invisible to the coder because they sit in free text.
4. The EHR calculator is trusted too far. Level calculators read structured fields. Work in narrative text does not reach them.
5. Nobody owns the question. Coders code what the note says. Providers write the note. No one in the loop is asked whether the note describes more than the code claims.

How to check your own practice

You can do this yourself with an export from your practice management system. No software purchase required.

1. Export twelve months of E/M line items with the CPT code, the rendering provider, and the date of service.
2. For each provider, calculate the share of visits at each level. Do established and new patients separately — mixing them hides the pattern.
3. Compare each provider against the others in the same specialty in your own practice first. An outlier inside one building, seeing the same case mix, is the strongest signal you will find.
4. Then compare against your specialty nationally. CMS publishes Medicare Part B utilization by specialty and HCPCS code, which is a public and free reference point.
5. Where a gap appears, pull 20 to 30 charts for that provider and read the notes against the code billed. You are answering one question: does the note describe work the code does not claim?

What to fix first

Fix the documentation before you fix the codes. If you raise coded levels without changing what the notes say, you have converted a revenue problem into a compliance problem. The order matters.

• Get total time recorded as a single explicit statement on visits where time is the better route.
• Get independent interpretation and external discussion written down where they happen. Both count and both are usually invisible.
• Make the problems addressed and their status explicit — stable, worsening, exacerbated. “Diabetes” on a problem list is not the same as a problem addressed at that visit.
• Re-run the distribution after a quarter. A correction supported by better notes moves gradually and is defensible. A step change is not.

Related leaks

Undercoding is one of four. If a provider’s levels look low, check whether modifier 25 is being omitted on the same visits, and whether the note supports the level at all. The three tend to travel together, because they share a root cause: nobody is reading the note against the claim.