In briefMDM has three columns, problems, data, and risk, and the visit level is whatever two of the three support.

What feels routine to you is often high complexity

Reading your own emergency film, weighing surgery in a diabetic patient, deciding against an intervention: these feel ordinary because you do them daily. CMS scores them as high-level decision making.

Definition: medical decision making
Medical decision making, or MDM, is the measure of the cognitive work a visit required: how many problems you addressed and how complex they were, how much data you reviewed and analyzed, and how much risk your decisions carried. Since 2021 it is one of the two ways to select an office visit level, the other being total time. It is not a count of how much you wrote.

The Four MDM Levels and the Codes They Correspond To

MDM LevelNew Patient CPTEstablished Patient CPTwRVUs (New / Est. Patient)
Straightforward99202992120.93 / 0.70
Low99203992131.60 / 1.30
Moderate99204992142.60 / 1.92
High99205992153.50 / 2.80

The Three Columns, and Why You Only Need Two

Two columns at a level carry the visit: a complex problem plus a prescription is moderate with no data review, and heavy data can pull up a simple problem.

The intimidating table, and the plain-language version
The official MDM table is a dense grid of qualifiers. Three questions cover almost every visit: how many problems did I deal with, and how sick is the patient; how much data did I dig into, labs, images, outside records, other clinicians; how risky was my plan, mainly what I prescribed or decided? Take the level at least two reach, and document those two.

Column 1: Problems Addressed

The key word is addressed: evaluated, treated, or managed at this visit, not merely noted. Diabetes in the past medical history does nothing; "reviewed glucose control, stable on current therapy, continued" counts.

Practical tip: two stable chronic illnesses reach level 4
A stable chronic illness is a long-term condition (expected to last at least a year, or until death) at its treatment goal and not worsening: controlled hypertension, type 2 diabetes at target, stable hypothyroidism. Two addressed at the same visit meet moderate and support a level 4 (99214 established, 99204 new) with no acute problem. A common cold alone is a level 2; uncontrolled type 2 diabetes needing a change in therapy can support a level 4 by itself.

Three categories get scored too low. Stable versus exacerbated chronic illness: "Type 2 diabetes, stable at goal" is low; "Type 2 diabetes with worsening control, adjusting therapy" is moderate. The new problem with uncertain prognosis: a new breast lump, unexplained weight loss, or a new neurologic symptom under workup is moderate before any diagnosis. The acute illness with systemic symptoms: fever or malaise makes it moderate.

One line does it: "Managed hypertension and type 2 diabetes; both at goal on current therapy. Adjusted metformin dosing given recent weight change." Two stable chronic illnesses, moderate. Could a coder tell from your assessment how many problems you addressed and how sick the patient is?

Column 2: Data Reviewed

Nothing here scores unless it is documented, which is why this column is most often left blank.

Three kinds of effort are tallied. Tests and records: each unique test ordered or reviewed, outside records reviewed, or history taken from someone other than the patient. Ordering a test includes its later review; the same result earns nothing next visit. Discussion with another clinician: a genuine two-way exchange about this patient's management with an external physician or other qualified professional, not a curbside formality; document who and about what. Independent interpretation: personally interpreting a study someone else ordered or performed, and that you are not separately coding. Most clinicians get this one wrong.

The classic case is an electrocardiogram formally read by another physician who codes for that interpretation; your own documented read counts because the order and the coded interpretation belong to someone else. A routine lab result has no professional read to interpret and counts under tests reviewed. No credit for a test you ordered and read yourself, or one you code separately. Record who ordered the study, the date, and your read; a formal report is not required.

A common orthopaedic example
A patient arrives with an MRI ordered by their primary care physician and read by a radiologist. You pull up the images, confirm the tear and its location, and plan surgery on your own read. "Personally reviewed the MRI images; full-thickness supraspinatus tear with retraction, consistent with the report" records an independent interpretation: the order was someone else's and you are not coding the read. Most surgeons do this and never write it down.

Name the data. "Reviewed today's in-office HbA1c and basic metabolic panel. Reviewed outside cardiology note from three months ago" is test review plus external records, moderate. "Labs reviewed" leaves the credit on the table.

Column 3: Risk

Prescription drug management is broad: starting, refilling, adjusting, or discontinuing a prescription, or continuing one with documented reasoning. It is the most common path to a level 4. The trap auditors look for: an over-the-counter drug written on a prescription pad is still non-prescription, low risk, a level 3, unless the strength is higher than what is sold over the counter. And a prescription is one column; the level 4 needs a second, usually the problem addressed.

A note on prescription drugs
Managing or adjusting any prescription medication, even a straightforward one, meets moderate risk and supports a level 4 (99214 or 99204). High risk requires clinically intensive monitoring for toxicity, in the AMA sense of monitoring for serious adverse effects rather than efficacy: warfarin with INR management, methotrexate, and lithium are the classic examples. The drug name alone is not enough; routine annual labs or efficacy checks do not qualify.
Minor versus major sets the baseline; the patient sets the level
Risk keys off minor versus major and patient risk factors, not how difficult the procedure felt. The global period is a proxy: a 0 or 10-day global is generally minor, a 90-day global major. A minor procedure with no identified risk factors, such as a skin tag removal or a single trigger-point injection, is low risk, moderate when patient risk factors complicate it. Elective major surgery without identified risk factors is moderate, a level 4; the same decision in a patient with significant comorbidities, or emergency major surgery, is high, a level 5. The decision itself carries the credit: documented surgical counseling earns it at the visit where you only decided.

Two situations are easy to miss. A decision about hospitalization, including sending home a patient who could reasonably have been admitted, is high risk. A diagnosis or treatment significantly limited by social determinants of health, such as a patient who cannot afford a medication or lacks transportation to follow a plan, is moderate.

Show the risk in the plan: "Continuing metformin 1000 mg twice daily; will monitor renal function given mild chronic kidney disease and recheck a basic metabolic panel in three months." For procedures, name the procedure, its global period, and any patient factors that raise the risk.

Two Notes, Same Patient

Both notes describe a 62-year-old returning with worsening knee osteoarthritis, started on a prescription anti-inflammatory. The long note is the old style; the short one earns more.

Note A. Exhaustive 99213 · 1.30 wRVU

HPI: 62-year-old returns with right knee pain, present for several years and gradually worsening. Medial, aching, about 6 out of 10. Worse with stairs and prolonged standing, better with rest and ice. No locking or giving way, no recent trauma. Partial relief with acetaminophen.

ROS: Constitutional, eyes, ear/nose/throat, cardiovascular, respiratory, gastrointestinal, genitourinary, musculoskeletal, skin, neurological, psychiatric, endocrine, hematologic, and allergic systems reviewed, negative except as noted above.

Exam: Well-appearing, no distress. Heart regular. Lungs clear. Abdomen soft. Right knee: no effusion, full range of motion, mild medial joint line tenderness, ligaments stable, no warmth. Left knee normal. Neurologic exam grossly intact.

Assessment: Knee pain.

Plan: Recommend NSAIDs and activity modification. Return as needed.

Note B. Decision-focused 99214 · 1.92 wRVU

HPI: 62-year-old with known right knee osteoarthritis, returning with a three-week flare. Now limiting walking to less than two blocks and interfering with work. Acetaminophen no longer controlling it.

Exam: Right knee with small effusion, medial joint line tenderness, crepitus with motion, stable. Antalgic gait.

Assessment: Osteoarthritis of the right knee with acute exacerbation, now limiting ambulation.

Plan: Personally reviewed today's weight-bearing radiographs, showing medial joint space narrowing without acute change. Started meloxicam 15 mg daily (prescription drug management), gastrointestinal precautions reviewed. Discussed activity modification, a future corticosteroid injection, and referral for surgical evaluation if symptoms progress. Follow up in six weeks.

Note A supports 99213, worth 1.30 wRVU; Note B supports 99214, worth 1.92, a gap of 0.62 wRVU per visit. The history and exam in Note A are not wrong; they no longer move the level. Naming the problem, the data, and the risk does.

The losing note is the longer one
Effort stopped being the currency in 2021; documented decision-making took its place. You are already doing the work in Note B. The only question is whether your note says so.

Choosing Your Level: Common Mistakes and How Auditors Think

The most widespread outpatient coding mistake is not overcoding; it is defaulting to 99213 for visits that meet 99214 criteria. It feels safe, but a coding distribution that does not match clinical reality is itself a compliance risk, and bloat has its own: a copied-forward fourteen-point review of systems or an exam that was not performed undermines credibility under audit.

Auditors, from your compliance department to a CMS Recovery Audit Contractor, look for statistical anomalies and documentation-to-code mismatches. Far more level-5 visits than your peer group is a flag, not because level 5 is wrong but because each must be justified note by note; high modifier 25 use or procedure-to-E/M ratios draw the same scrutiny.

Audit red flags
Coding 99215 for every established patient visit regardless of complexity. Using Modifier 25 on every visit that includes a procedure without documentation of a truly separate clinical evaluation. Applying time-based coding with no documented time in the note.

A self-check: take five recent notes you coded 99213. Did you address two or more problems, or one not fully controlled? Review labs, images, or outside records? Manage a prescription? If two of three are true, the visit was very likely an undercoded 99214. Check both routes at each encounter, MDM and total time (including pre-visit and post-visit work), and use whichever the documentation honestly supports.

New patient codes (99202-99205) carry higher wRVU values than their established equivalents at the same MDM level. A patient is new only if no provider in your group of the same specialty and subspecialty has seen them face to face in the past three years; a different office location does not reset it, and a same-specialty partner's visit last year makes the patient established to you. For payers that still recognize consultation codes, a referred new patient with no planned follow-up is often better coded as a consultation, which typically pays more; Medicare eliminated consult payment in 2010. The request-opinion-report requirement is in the consultation codes guide.

Three Visits, Worked Through

A straightforward upper respiratory infection

A healthy adult with three days of cough and congestion, no red flags: one self-limited problem, minimal data, low risk, supportive care and an over-the-counter decongestant. A level 3 established visit (99213); the over-the-counter recommendation does not raise it.

The same cough, in a patient with poorly controlled diabetes

Same cough, but the patient's type 2 diabetes is running high and you adjust the medication. An acute illness plus a chronic condition not at goal, and a prescription managed: problems and risk both moderate. A level 4 (99214), if the note states the diabetes assessment and the medication change.

A new patient with several active problems

Hypertension, hyperlipidemia, and new-onset atrial fibrillation; you order an ECG and labs, start an anticoagulant, and arrange cardiology follow-up. Multiple chronic conditions, one new and serious, data across more than one category, and high-risk medication management: a level 5 new patient (99205). The level is what the note shows you addressed, reviewed, and decided.

Sources and Further Reading for This Chapter

Put it to workRun a note through the AI Note Coding Assistant
PreviousCoding, Pay, and the 2021 E/M Rules