E/M Level Selection in 2026: What Actually Decides the Code

Since the 2021 revision to office and outpatient visit coding, the level is chosen by medical decision making or by total time, and by nothing else. History and examination are still performed and still documented, and they no longer move the code. This guide explains the framework the MDM decision tree and the E/M calculator implement, so you can tell when their answer is right.

The rule almost everyone gets wrong

Medical decision making has three elements: Problems Addressed, Data Reviewed and Analysed, and Risk. The visit level is the level met or exceeded by two of those three.

Two of three means two separate elements. Meeting two categories inside the Data element is still one element. That single misreading is the most common reason a level 4 is billed, audited, and reduced.

The grid, in plain language

Established patient office visits. These summarise the structure of each level rather than reproducing the code descriptors, which are copyrighted by the American Medical Association.

Code	MDM	Problems	Data	Risk
99212	Straightforward	One self-limited or minor problem.	Minimal or none.	Minimal risk of morbidity from additional work-up or management.
99213	Low	Two or more self-limited problems, one stable chronic illness, or one acute uncomplicated illness.	Limited. One category met: a small combination of reviewed tests and notes, or an independent historian.	Low risk of morbidity.
99214	Moderate	One or more chronic illnesses with exacerbation or progression, two or more stable chronic illnesses, or an acute illness with systemic symptoms.	Moderate. One of three categories met, including independent interpretation of a test or discussion of management with an external clinician.	Moderate risk, which is where prescription medication management usually lands.
99215	High	One or more chronic illnesses with severe exacerbation, or an acute illness posing a threat to life or bodily function.	Extensive. Two of three categories met.	High risk of morbidity from additional work-up or management.

Three worked examples

Every vignette below is synthetic. No real patient and no real encounter is described.

The visit that looks like a 99213 and is not

Established patient, 61, seen for routine follow-up of type 2 diabetes and hypertension. Both are stable on current management. A1c and a basic metabolic panel from an outside lab are reviewed in the chart. Medications continued unchanged.

Problems:: Two stable chronic illnesses. That is the moderate row, not the low row.
Data:: Review of outside results. Limited.
Risk:: Continuing prescription medications is medication management, which sits at moderate.

99214. Problems and Risk both reach moderate, and two of three is the test.

The word "routine" does a lot of damage here. Nothing about this encounter felt like a level 4, and the coding does not care how it felt. Two stable chronic illnesses is moderate on the problems axis whether or not anything changed.

The visit that feels big and is not

Established patient, 34, acute low back pain after lifting. No red-flag features documented. Long discussion, physical examination documented across four systems, patient reassured, over-the-counter analgesia discussed. Thirty-four minutes of total time on the date of the encounter, not documented.

Problems:: One acute uncomplicated illness. Low.
Data:: None ordered, none reviewed. Minimal.
Risk:: Over-the-counter management. Low.

99213 on MDM.

The four-system examination is doing nothing. Since 2021 history and examination no longer drive the level; they are performed and documented as clinically appropriate and then ignored by the code. The thirty-four minutes, had it been documented, would have supported the next level up on time alone, because the established-patient time bands step at thirty minutes. That is the whole loss: the work happened and the note did not record the one fact that counted.

The visit where the data column looks like it carries the level

Established patient, 47, fatigue. One acute uncomplicated illness. The clinician personally reviews and interprets a previous radiology study rather than relying on the report, and documents that interpretation. A phone discussion with the referring gastroenterologist is documented, including what was decided.

Problems:: One acute uncomplicated illness. Low.
Data:: Independent interpretation of a test, plus discussion of management with an external clinician. Two separate moderate-level categories.
Risk:: Low.

99213. Data reaches moderate, Problems and Risk are both low, so exactly one of the three elements is at moderate and the two-of-three test is not met.

This is the one worth reading twice, because it looks like a level 4 and is not. Two of THREE elements must meet or exceed the level, and two qualifying categories inside the single Data element is still one element. Counting them as two is the most common way a level 4 is claimed and then reduced on audit.

Choosing on time instead

You may select the level on total time on the date of the encounter, and you pick whichever of time or MDM gives the supported level. Total time includes the non-face-to-face work you personally do that day: chart review before the visit, ordering, documenting, and communicating results.

It does not include time billed separately, and it does not include staff time. The practical failure is not miscounting. It is not writing the number down at all, which is what happens in the second vignette above.

Where this leaves documentation

Name every problem you addressed and say what you did about it. An unnamed problem cannot count toward the Problems element.
If you personally interpreted a study, say so, and record the interpretation. A report you merely read is a different category.
If you discussed management with another clinician, record who and what was decided.
If you are coding on time, record the total minutes. One sentence is the difference between two levels.

Work it through

MDM Decision TreeAnswer for Problems, Data and Risk and get the supported level.E/M CalculatorCompare the level supported by MDM against the level supported by time.

Explore the wider workflow

The tools here help you check the coding framework. When you want to see Krasyn in a broader outpatient workflow, choose a non-PHI demo or a guided walkthrough.

Explore demo paths

Sources and limits

The office visit framework described here is the 2021 revision to evaluation and management coding, which remains in force for 2026. Diagnosis code data on this site comes from the public-domain CMS and NCHS ICD-10-CM FY2026 releases recorded in the site source manifest.

CPT is copyrighted by the American Medical Association, so this page references codes by number and describes their structure rather than reproducing descriptor text. No payment amount, relative value or fee schedule figure appears anywhere on this page, because no fee schedule release is cited in the source manifest. This is a coding reference and not clinical, legal or billing advice; the vignettes are synthetic and are here to show how the elements combine.