Chapter 15: Evaluation and Management I: The Office Visit and the 2021 Rewrite

"For twenty-five years we counted bullets. Then one January we stopped, and the sky did not fall." — constructed

Overview

This is the most consequential chapter in this book for most of the people reading it.

Not the most difficult — Chapter 21 is harder, and Chapter 23 has more arithmetic. The most consequential, because evaluation and management services are the highest-volume, highest-dollar category of professional billing in the United States, because nearly every physician practice in the country bills them every day, and because they are audited more than anything else a coder touches.

A coder who is excellent at surgical coding and mediocre at E/M is a coder whose organization is exposed. A coder who is excellent at E/M is valuable everywhere.

And the rules changed. In January 2021 the office and other outpatient visit codes were rebuilt from the ground up, and in January 2023 the same framework was extended to nearly all the remaining E/M categories — hospital, observation, consultations, emergency department, nursing facility, home and residence. Two decades of accumulated practice were retired in twenty-four months.

That matters to you in a way it does not matter for most code-set changes. The people training you may have learned the old system. Reference material published before 2021 describes a framework that no longer governs. Study guides, practice tests, workplace cheat sheets, and the confident advice of a senior coder who has not read the current guidelines are all, in this one area, actively dangerous.

So this chapter does something the others do not. It teaches the current rules in detail, and it names the old ones specifically — not out of nostalgia, but so that you can recognize obsolete advice when someone gives it to you with total confidence.

In this chapter, you will learn to:

• Say what an E/M code is actually measuring
• Apply the new-versus-established test correctly, including its two qualifications
• Explain what 2021 changed and why
• Score medical decision making element by element
• Classify problems, count data, and determine risk
• Select by time instead, and know what counts
• Report prolonged services, including where Medicare differs
• Apply split/shared and incident-to, and know what breaks each
• Recognize documentation that survives an audit

And you will finally level Account 10-4471. The March 14 visit has been carrying a 99214 since Chapter 4, asserted and not defended. §15.13 defends it.

15.1 What an E/M code is actually measuring

Start with a question most coders cannot answer cleanly: what is the difference between a 99213 and a 99214?

The wrong answers are instructive.

"A 99214 is a longer visit." Sometimes. Time is one of two selection methods, and when the other method is used, duration is irrelevant.

"A 99214 is a sicker patient." Closer, but not it. A patient can be extremely sick and generate a low-level visit — a stable, well-controlled, complex patient seen for one straightforward issue.

"A 99214 has more documentation." This is the answer twenty years of training produced, and it is now wrong as a matter of rule. Volume of documentation is not the measure and has not been since 2021.

The actual answer:

An E/M code measures the cognitive work of the physician or other qualified health care professional in evaluating and managing the patient's problems — either by the complexity of the medical decision making, or by the total time spent on the date of the encounter.

Read that twice. Three things follow.

First, it measures the professional's work, not the patient's condition and not the practice's effort. Nursing time, staff time, the difficulty of the schedule, and the length of the chart are not in the definition.

Second, there are exactly two roads to a level, and you may take either one on any given encounter: medical decision making, or total time. They are alternatives, not a combination. You do not average them and you do not require both. Choose the one that supports the higher level and that the documentation actually supports.

Third, and most importantly for anyone trained before 2021: history and examination are no longer part of the level selection.

They are still required. The guidelines call for a medically appropriate history and/or examination — appropriate as determined by the treating professional. But "medically appropriate" is a clinical judgment, not a coding count. The coder does not score the history. The coder does not count examination bullets. Those numbers no longer exist.

⚠️ Where Claims Die

The single most common E/M error in the current environment is a coder applying a pre-2021 framework.

It shows up as: counting HPI elements. Counting review-of-systems items. Counting examination bullets or body areas. Asking whether the history is "comprehensive" or "detailed." Downcoding a visit because the physician's note is short.

Every one of those is scoring something the rules no longer measure. And the error runs in both directions — coders undercode well-documented moderate-complexity visits because the note lacks a ten-system review of systems, and coders overcode trivial visits because the electronic record auto-populated a full review of systems that nobody performed.

The second, which follows from the first: template-generated documentation that no longer buys anything. Many organizations still run note templates built to satisfy the 1995 and 1997 guidelines. They generate pages of history and examination that no longer affect the level and that a physician did not personally verify. That is not just wasted effort. It is an assertion in the medical record, and Chapter 5 §5.4 has already established what an assertion in the record is.

15.2 New versus established, and the three-year rule

Office visit codes come in two families, and choosing the wrong one is choosing the wrong code.

The distinction reflects real work: a professional seeing someone for the first time is building a picture from nothing.

The rule:

A new patient is one who has not received any professional services from the physician or other qualified health care professional, or another physician or QHP of the exact same specialty and subspecialty who belongs to the same group practice, within the past three years.

Everyone else is established.

Four load-bearing phrases, each of which produces errors.

"Any professional services." Not "an office visit." Any face-to-face service reported with a specific CPT code counts — a hospital visit, a procedure, a home visit. A patient the physician operated on two years ago in a hospital is established at the office, even though the office has never seen them. Conversely, some things do not count: a service where the professional never saw the patient face-to-face, such as reading an electrocardiogram, generally does not establish the patient.

"The exact same specialty and subspecialty." Not "the same practice." A multispecialty group's cardiologist and its dermatologist are different specialties, and a patient established with one is new to the other. And subspecialty matters — an interventional cardiologist and a general cardiologist in the same group may be distinct, depending on how the payer enrolls them.

"Who belongs to the same group practice." Same tax identification number, in practice. Two physicians of the same specialty in the same group: a patient established with either is established with both. Two physicians of the same specialty in different groups: independent.

"Within the past three years." Three years to the day, not three calendar years. A patient last seen on March 10 three years ago is new on March 14. A patient last seen on March 20 three years ago is established on March 14 — by six days.

🎓 Exam Watch

99201 no longer exists. It was deleted effective January 1, 2021, because with history and examination removed from level selection there was nothing to distinguish it from 99202. If a study resource lists 99201, the resource is out of date and you should be suspicious of everything else it tells you about E/M.

The new-patient family therefore has four levels and the established family has five. That asymmetry is a favorite exam item and it looks like a typographical error until you know why.

99211 is the other asymmetry. It is an established-patient visit that may not require the presence of a physician or other qualified health care professional — the classic nurse visit. It has no MDM requirement and no time requirement in the current structure, and it is the only E/M office code like that.

📞 On the Phone

"Your system says this patient is new. Ours says established. Which of us is wrong?"

Frequently neither, and the conversation is worth having correctly.

The payer's determination is driven by claims history under your group's tax identification number and the enrolled specialty of the rendering provider. Yours is driven by your practice management system's record of encounters.

They diverge for three ordinary reasons. The patient was seen by a physician in your group who enrolls under a different specialty. The prior service was in a different setting and your office record does not show it. Or the prior service was billed under a group the payer does not associate with yours because of a tax identification number change.

Ask which claim and which date the payer is looking at, and ask under which provider and specialty. Then check your own record for any professional service, not just office visits, in the last three years. Most of these resolve in one call — and the ones that do not are usually an enrollment problem, which is Chapter 25 §25.7's subject, not a coding problem.

15.3 What changed in 2021 and why

You need this section for two reasons: to recognize obsolete advice, and because the reasoning behind the change tells you what the current rules are trying to reward.

The old framework, in one paragraph. From the mid-1990s until the end of 2020, an office visit level was determined by three "key components" — history, examination, and medical decision making. Each was scored on its own scale. History was scored by counting elements of the history of present illness, systems in the review of systems, and items of past, family, and social history. Examination was scored by counting body areas and organ systems under the 1995 guidelines, or by counting individual bullets under the 1997 guidelines, which the coder could choose between per encounter. For an established patient, two of the three key components determined the level; for a new patient, all three.

What that produced. Physicians learned quickly that the cheapest way to raise a level was to document more history and more examination, because those were countable and medical decision making was not. Electronic records made the countable parts nearly free — a single click could populate a complete review of systems and a full multisystem examination.

The result was note bloat: documentation that grew and grew while conveying less and less. Records in which the clinically meaningful content occupied four sentences buried inside four pages. Copied text propagated forward through encounters for months. Reviewers, auditors, and — worse — the next treating physician had to excavate.

Meanwhile the actual cognitive work of the visit, the part that distinguishes a hard patient from an easy one, was scored by a component nobody could define consistently.

What changed effective January 1, 2021, for office and other outpatient services:

1. History and examination were removed from level selection. The requirement became "a medically appropriate history and/or examination," determined by the treating professional.
2. 99201 was deleted.
3. Medical decision making was redefined with explicit definitions and a published grid, so that two coders scoring the same note would more often agree.
4. Time was redefined as total time on the date of the encounter, including non-face-to-face work, and the counseling-dominance requirement was eliminated. Under the old rules, time could only be used when more than half the visit was counseling or coordination of care. Now time may be used on any encounter.
5. A new prolonged services code was created for the office setting.

What changed effective January 1, 2023 — the same framework extended to most other E/M families: inpatient and observation care (merged into a single set of codes), consultations, emergency department, nursing facility, and home or residence services. Several code families were deleted and several were merged. Chapter 16 is that story, and the fact that this chapter's framework applies there is the reason Chapter 16 can be shorter than it would otherwise need to be.

⚖️ Compliance Check

A widespread misreading of the 2021 change: "history and exam don't matter anymore."

They do not matter for level selection. They matter enormously for everything else.

The record is a clinical document first. Medical necessity is still established by the clinical content, and a note with an impeccable medical decision making grid and no discernible clinical reason for the visit will not survive review — Chapter 22 §22.2 makes this point at length, and Chapter 5 §5.2 made it about the certification.

There is also a subtler exposure. Some organizations, reading "history and exam no longer count," turned off documentation their clinicians relied on. The coding consequence was zero and the clinical consequence was not. The change removed a counting requirement. It did not remove the obligation to document the encounter.

15.4 Medical decision making: the three elements and the two-of-three rule

Medical decision making is scored across three elements:

1. The number and complexity of problems addressed at the encounter
2. The amount and/or complexity of data to be reviewed and analyzed
3. The risk of complications and/or morbidity or mortality of patient management

Each element is assigned one of four levels. The four levels map to codes:

The rule that ties them together:

Two of the three elements must meet or exceed a level for the encounter to reach that level.

This is the single most important mechanical rule in the chapter, and it has a consequence people consistently miss:

One element can be low — or absent entirely — and the visit can still be moderate.

A visit with no data reviewed at all can be a 99214, if problems and risk are both moderate. A visit with extensive data can be a 99213 if problems and risk are only low. The lowest element does not cap the level. It simply does not participate.

Coders who came from the old two-of-three-key-components world usually find this comfortable. Coders trained on the idea that "you need all three" will undercode systematically until someone corrects them.

   THE TWO-OF-THREE RULE

   Element              this visit
   ─────────────────────────────────────────
   Problems addressed   MODERATE  ✓
   Data                 limited (low)
   Risk                 MODERATE  ✓
   ─────────────────────────────────────────
   Two elements at moderate  ►  MODERATE MDM
                             ►  99214 established
                             ►  99204 new

   The low data element does not reduce the level.
   It is simply not one of the two.

🔍 Check Your Understanding

An established patient encounter scores: problems high, data minimal, risk low.

What is the MDM level, and what code?

Answer: One element is high, one low, one minimal. No level is met by two elements at that level or above. Working down: is low met or exceeded by two? Problems (high) exceeds low, risk (low) meets low — that is two. Low MDM, 99213. The high problem element gets you nothing without a second element to support it, which is exactly the point of the rule and exactly why the physician in this scenario will be unhappy.

15.5 Problems addressed

The first element counts problems addressed at the encounter, and both words are doing work.

A problem is addressed when it is evaluated or treated at the encounter by the professional reporting the service. It includes consideration of further testing or treatment even when that option is declined — by the patient, by another professional, or by the physician's own judgment.

Notation in the record of a problem being managed by someone else, without additional assessment or care coordination, is NOT "addressed." Neither is a referral without evaluation.

That last sentence is the whole game. A problem list is not a set of addressed problems. A condition mentioned is not a condition addressed. Chapter 14 §14.4 made the identical point about modifier 25 — mention is not management — and it is not a coincidence that the same distinction governs both. It is the same underlying question: did the professional do cognitive work about this thing?

The categories, from the published definitions:

The definition of "stable" is the trap in this element, and it catches nearly everyone.

"Stable" for MDM purposes means the patient is at their treatment goal.

It does not mean "unchanged since the last visit."

A patient whose hemoglobin A1c has been 10.2 for two years is not stable. Nothing has changed, and nothing being at goal either. That patient has a chronic illness that is not at treatment goal, which pushes toward the exacerbation-or-progression language rather than the stable language.

Read the consequence carefully, because it runs against intuition: the poorly controlled patient generates a higher problem score than the well-controlled one, which is correct, because managing a patient who is not responding is harder work.

📋 Read the Chart

Source: office note, assessment and plan Encounter: established patient, office What it says:

"1. Type 2 diabetes mellitus — A1c 6.8, at goal. Continue metformin 1000 mg BID. 2. Essential hypertension — 132/84, at goal. Continue lisinopril 20 mg daily. 3. Hyperlipidemia — continue atorvastatin 40 mg daily; lipid panel ordered today. 4. Chronic kidney disease, stage 3a — followed by nephrology."

What it means: Items 1, 2, and 3 are stable chronic illnesses, addressed — each is assessed, each has a stated status, each has a plan. Three stable chronic illnesses meets "2 or more," which is MODERATE.

Item 4 is not addressed. "Followed by nephrology" is notation of a problem managed by someone else with no additional assessment and no documented care coordination. It is a real part of the patient's picture, and it contributes nothing to this element.

What to do about it: Nothing, as a coder. You do not add it and you do not query to have it added, because a query asking a physician to document work they did not do is exactly what Chapter 5 §5.6 forbids. If the physician did coordinate care with nephrology, the note should say so — and that is a documentation improvement conversation, not a coding decision. Chapter 38 §38.3 is about having that conversation.

Where it appears: the problems element of nearly every primary care visit in the country.

Two more definitional notes that produce real errors.

"1 undiagnosed new problem with uncertain prognosis" is moderate, and it is a category coders underuse badly. A new complaint that might be several things, where the workup has not yet resolved it, is moderate on this element by itself — no chronic conditions required. Coders who default to "it's just one problem, so it's low" are misreading the grid.

"Acute illness with systemic symptoms" is moderate, and "systemic" is the operative word. A local infection is not systemic. Fever, chills, myalgias, and malaise are. The definition explicitly warns against treating a general statement of feeling unwell as systemic involvement.

15.6 Data reviewed and analyzed

The second element is the one coders score inconsistently, because it has the most structure.

Data is organized into three categories, and the level is reached by satisfying combinations of them.

Category 1 — tests, documents, or independent historian. Each item counts as one:

• Review of each unique test
• Order of each unique test
• Review of each unique prior external note from each unique source
• Assessment requiring an independent historian

Category 2 — independent interpretation of a test performed by another professional, where that interpretation is not separately reported.

Category 3 — discussion of management or test interpretation with an external physician, other QHP, or appropriate source, not separately reported.

The levels:

Now the counting rules that decide most disputes.

A "unique test" is defined by its CPT code. Two tests with the same code are one unique test, no matter how many times they are run. A panel is one test, not a test for each analyte — a comprehensive metabolic panel is one item, not fourteen. This is the single largest source of inflated data scores.

Ordering and reviewing the same test is one item, not two. The guidelines are explicit: the review of the result is included in the order. You cannot count the order today and the review next week as two items on this encounter.

"External" means a different group practice or a different specialty or subspecialty — not merely a different individual. A note from the physician down the hall in the same practice and specialty is not external.

An independent historian is someone who provides history in addition to the patient — a parent, a spouse, a caregiver, a translator of circumstances the patient cannot supply — because the patient is unable to provide a complete or reliable history. A family member who happens to be in the room is not automatically an independent historian. The necessity is the point.

Category 2 requires an actual interpretation, documented, of a test performed by someone else, and requires that you are not billing for that interpretation separately. If you report the professional component with modifier 26 — Chapter 14 §14.7 — you cannot also count the interpretation here. You may be paid for it or you may count it. Not both.

Category 3 requires discussion, which means an interactive exchange. It is not a note sent, a report received, or a message left. The guidelines say so directly, and the exchange must be documented. It also must be with someone external, and it must not be separately reported — which excludes it when an interprofessional consultation code is billed.

🔢 Code It

Score the data element:

Scenario A. The physician orders a hemoglobin A1c and a lipid panel. No prior records reviewed, no independent historian, no outside interpretation, no discussion.

Category 1: two unique tests ordered = 2 items. That satisfies "any combination of 2." → LIMITED (low). Two orders is not three, and a lipid panel is one test regardless of how many analytes it contains.

Scenario B. The physician orders a complete blood count, reviews a cardiology note from an outside group, and personally interprets an electrocardiogram performed at a hospital last week — interpretation documented, not separately billed.

Category 1: one order + one external note = 2 items. Category 2: one independent interpretation. Two categories are satisfied. → EXTENSIVE (high). Two of three categories is the high requirement, and this encounter meets it without a large volume of anything.

Scenario C. The physician orders a comprehensive metabolic panel, a complete blood count, and a thyroid-stimulating hormone.

Category 1: three unique tests ordered = 3 items. → MODERATE. Three items in Category 1 is the moderate threshold. Note that this is three tests, not three panels' worth of analytes.

Scenario D. The physician orders a comprehensive metabolic panel and reviews the results of the comprehensive metabolic panel ordered at the last visit two weeks ago.

One unique test ordered. The prior panel is the same CPT code — not unique — and in any case, review is included in the order. → 1 item. MINIMAL. This scenario is on certification exams because it looks like two things and is one.

⚠️ Where Claims Die

The data element is where auditors find inflation most reliably, because it is the most countable and therefore the most automatable.

Four patterns, all of which an auditor can detect from claims and a chart sample:

Panels counted as multiple tests. A comprehensive metabolic panel counted as fourteen items turns a minimal data element into an extensive one, single-handedly.

Orders and reviews double-counted. The same test counted once when ordered and again when the result comes back.

Internal notes counted as external. Records from within the same group and specialty, counted as unique external notes.

Category 3 claimed for one-way communication. A letter sent to a specialist counted as a discussion. It is not, and it is trivially disprovable — there is no return.

None of these is exotic. All four appear in the same audit findings, over and over, and the reason is almost never fraud. It is a coder who learned the element from a summary sheet rather than from the definitions.

15.7 Risk of complications and morbidity

The third element:

The risk of complications and/or morbidity or mortality of patient management decisions made at the visit, associated with the patient's problems, the diagnostic procedures, and the treatment.

This includes the management options selected AND those considered but not selected.

That second sentence is worth money and gets ignored. Considering and rejecting a higher-risk option counts — if it is documented. A physician who considered hospitalizing a patient and elected to manage them at home has made a high-risk decision, and the record should say so. A record that says only "sent home" documents nothing.

The four levels:

Four of these need explanation.

"Prescription drug management" is the most-used and most-misused phrase in the element. It means managing prescription medication: prescribing a new one, adjusting a dose, discontinuing one, or continuing one with a documented decision to continue.

It does not mean that a prescription drug appears on the medication list. A medication list is a list. Management is a decision.

The distinction is visible in the note. "Metformin 1000 mg BID" under Medications is a list. "Type 2 diabetes, at goal — continue metformin 1000 mg BID" under Assessment and Plan is management. Same drug, same dose; one is documentation of a fact and the other is documentation of a decision.

Because prescription drug management alone reaches moderate risk, a very large fraction of primary care and specialty visits have a moderate risk element. That is not a loophole; it reflects that prescribing carries real risk. But it means the risk element is frequently the easiest of the three to satisfy at moderate, which makes the problems element the one that usually decides the visit.

"Drug therapy requiring intensive monitoring for toxicity" is high, and it is narrower than it sounds. The intent is monitoring for toxicity, not for therapeutic effect, and it generally means laboratory or physiologic monitoring at a frequency driven by the danger of the drug. Checking a hemoglobin A1c twice a year is not intensive monitoring for toxicity. Long-term monitoring at quarterly or longer intervals generally is not either.

"Minor surgery with identified patient or procedure risk factors" is moderate; without them it is low. The risk factors must be identified in the documentation. A note that says "injection performed" documents no risk factors. A note that documents anticoagulation, immunosuppression, or a prior adverse reaction identifies one.

"Social determinants of health" reaching moderate is the newest and least-used entry, and it is worth knowing. When a diagnosis or treatment is significantly limited by, for example, housing instability, food insecurity, inability to afford medication, or lack of transportation, the risk element reaches moderate. The limitation must be documented and must actually affect the management plan. Chapter 12 §12.9 covers the Z codes that describe these circumstances, and this is one of the places where coding them is not merely descriptive — it corresponds to a real element of the visit.

⚖️ Compliance Check

A risk element cannot be documented after the fact by the coder, and the "considered but not selected" language is where that temptation lives.

A coder who reads a note, recognizes that hospitalization was probably considered, and scores high risk on that basis has scored something that is not in the record. A coder who queries the physician — "Was hospitalization considered at this encounter?" — has asked a legitimate question, provided it is genuinely open and not leading. Chapter 5 §5.6 and Chapter 38 §38.3 both govern this.

The line: you may ask what happened. You may not decide what happened.

And the practical version, which is the one that keeps people out of trouble: if the level depends on it, and it is not written, you do not have it yet.

15.8 Selecting by time instead, and what counts

Time is the second road, and since 2021 it is available on every office encounter — no counseling dominance required.

Total time on the date of the encounter is the time personally spent by the reporting physician or other qualified health care professional on that patient on that calendar date, including both face-to-face and non-face-to-face time.

What counts:

• Preparing to see the patient — reviewing tests, reviewing prior records
• Obtaining and reviewing a separately obtained history
• Performing a medically appropriate examination or evaluation
• Counseling and educating the patient, family, or caregiver
• Ordering medications, tests, or procedures
• Referring and communicating with other health care professionals when not separately reported
• Documenting clinical information in the health record
• Independently interpreting results not separately reported and communicating them
• Care coordination not separately reported

What does NOT count:

• Clinical staff time. A nurse's forty minutes are not the physician's time. This is the largest single error in time-based selection.
• Time on any other calendar date. Documentation completed the following morning is not on the date of the encounter.
• Time performing other services that are reported separately. The knee injection's time is the procedure's time, not the visit's.
• Travel time.
• General teaching not related to management of this specific patient.

The thresholds:

You must meet the minimum of the range. Twenty-nine minutes on an established patient is a 99213, not a 99214, and there is no rounding. The most common documentation failure in time-based selection is a note that says "approximately 30 minutes" — which is either exactly at the threshold or a guess, and an auditor will read it as a guess.

🧮 Run the Numbers

When is time the better road?

Consider an established patient with one stable chronic illness (problems: low), no data (minimal), and prescription drug management (risk: moderate).

By MDM: two of three at low — problems meets low, risk exceeds it. Low MDM → 99213.

Now suppose the professional documents 34 minutes of total time on the date of the encounter, because the patient needed extensive counseling about a medication change and a long discussion of options.

By time: 30–39 minutes for an established patient. → 99214.

Same encounter, one level apart, and both are correct. The rules permit you to select by either method, and nothing requires you to choose the lower one.

Two conditions, and they are absolute. The time must be documented — a stated total, on that date, by the reporting professional. And it must be true. A statement of total time in a medical record is a factual assertion, and Chapter 5 §5.3's definition of "knowingly" applies to it exactly as it applies to a code.

The practical guidance most compliance programs give: document the time on every encounter, then select by whichever method the documentation supports at the higher level. That is not gaming anything. It is using both of the two methods the rules provide.

🎓 Exam Watch

Three time questions that appear constantly:

"The nurse spent 20 minutes and the physician spent 15. What is the total time?" — 15 minutes. Clinical staff time never counts.

"The physician spent 25 minutes with the patient and finished documentation the next morning, spending 10 more minutes. Total?" — 25 minutes. Time on the date of the encounter only.

"The physician spent 40 minutes, of which 12 were performing a separately reported procedure. Total?" — 28 minutes. Time spent on a separately reported service is excluded. An established patient at 28 minutes is a 99213 by time, which is why this question is asked.

15.9 Prolonged services

When an encounter runs past the top of the highest level's time range, prolonged service time may be reported — but only after the highest level's time is fully met.

For office and other outpatient services, CPT provides 99417, reported for each additional 15 minutes of total time beyond the minimum required time of the highest-level code (99205 or 99215). It is an add-on code, so Chapter 13 §13.7's rules apply — it is never reported alone and never carries modifier 51.

Three rules that produce errors:

You must reach the highest level first. 99417 cannot be added to a 99214. It attaches only to 99205 and 99215.

Full increments only. Fourteen additional minutes is not a unit. The increment must be complete.

It is total time, not face-to-face time, on the date of the encounter — the same definition as §15.8.

And now the complication. Medicare does not use 99417 for these services. CMS created its own HCPCS Level II code, G2212, with a different starting threshold — CMS's threshold begins after the maximum time of the highest level rather than the minimum, which means the prolonged service starts later and fewer encounters qualify.

You do not need to memorize the arithmetic. You need to know that the threshold differs and that the payer determines which code applies, because a practice that reports 99417 to Medicare and G2212 to a commercial payer has them exactly backwards, and both claims will be wrong.

⚠️ Where Claims Die

Prolonged services are a documented audit target, and the reason is arithmetic.

To report the office prolonged service code you must have already documented at least 40 minutes (established) or 60 minutes (new) of total time, and then 15 more. A practice reporting prolonged services at any volume is asserting that its clinicians routinely spend over an hour with patients.

That assertion is checkable against the schedule. It is checkable against the number of encounters per day. And in the cases that make the news, it was checkable against the total number of hours in the day — which is how the outlier reports are generated, and it requires no chart at all.

This is Chapter 14 §14.5's point about modifier 59 in a different costume: when a code's appropriateness is measurable from claims data alone, the audit does not need your records to find you.

15.9a The visit complexity add-on

One more office add-on deserves a section of its own, because it is recent, it is widely under-reported, and it rewards something this chapter has been circling the whole way through.

G2211 is a HCPCS Level II add-on code describing the visit complexity inherent to evaluation and management associated with medical care services that serve as the continuing focal point for all needed health care services, or with medical care services that are part of ongoing care related to a patient's single, serious condition or a complex condition.

That is a long descriptor for a short idea: the code pays for the relationship, not the visit.

Everything else in this chapter measures what happened on one date. G2211 measures something that exists only across dates — that this professional is the patient's continuing point of care, or is managing a serious or complex condition over time. The same office visit, performed identically by a physician seeing the patient once and by the physician who has managed them for eleven years, is not the same service, and this is the code that says so.

Three things a coder needs to know about it.

It is an add-on code, so Chapter 13 §13.7 governs: never reported alone, never with modifier 51. It attaches to office and other outpatient E/M services.

Its payment status and its restrictions have moved. It was established, then not paid for a period, then activated, and its interaction with modifier 25 in particular has been the subject of specific policy. Verify current status and current restrictions before reporting it, and verify them by payer — the code is Medicare's, and commercial adoption varies.

It is reported far less often than it applies. The under-reporting is not mysterious: it is a G-code, it does not appear in CPT where most coders look, its descriptor is opaque, and no edit fires when it is omitted. It is Chapter 14's silent underpayment in a different form — money that is not denied, because it was never asked for.

⚠️ Where Claims Die

A code you never report cannot be denied, and nothing in your revenue cycle will ever mention it.

Every control described in this book so far — scrubbers, edits, denial work queues, remittance review — is triggered by something that happened. Nothing is triggered by something that did not.

The only mechanism that finds an omitted code is a person who knows the code exists. That is Chapter 20's argument for reading the HCPCS Level II updates, Chapter 28 §28.8's argument for comparing payments to contracts, and this section's argument for knowing what G2211 is.

Account 10-4471 is a case in point. It is a long-standing patient of this practice, seen for three chronic conditions the practice manages continuously. Whether G2211 applies is a question the practice should be asking, and this book has no record that anyone did.

15.10 Leveling the same note four ways

The best way to understand what actually drives an E/M level is to hold a note constant and change one thing at a time.

The note: an established patient office visit. Three chronic conditions, each assessed with a plan, all at goal. Three medications reviewed and continued. Two laboratory tests ordered. One new complaint evaluated. A minor procedure performed and separately reported. Time not documented.

(This is Account 10-4471's March 14 note. §15.13 codes it formally. Here it is a laboratory.)

Reading 1 — as written

Two of three at moderate → MODERATE MDM → 99214.

Reading 2 — change one thing: the chronic conditions are listed but not assessed

Suppose the assessment reads only: "Chronic conditions stable, medications refilled." The problem list still shows three conditions. Nothing else changes.

Two of three at low → LOW MDM → 99213.

One sentence of documentation, one level, and the physician did the same work in both versions. This is the case for documentation improvement in its purest form, and it is why Chapter 33 exists.

Reading 3 — change one thing: add a documented total time of 35 minutes

Everything else as written.

By MDM: 99214. By time: 30–39 minutes established = 99214. Both roads arrive at the same place, which is common and reassuring. Report 99214 and note that you have two independent supports for it — which is a materially better position in an audit than having one.

Reading 4 — change one thing: the knee is the only reason for the visit

No chronic conditions addressed. No medications managed. One new complaint, evaluated, injected.

Only one element reaches moderate. Two of three at low? Problems exceeds low, risk meets low. → LOW MDM → 99213.

And this is the reading that matters most for Chapter 14. In this version there is no separately identifiable E/M above and beyond the procedure at all — the entire visit is the evaluation leading to the injection. Modifier 25 would not be supportable here, and the correct claim is the procedure alone.

🔍 Check Your Understanding

Reading 1 and Reading 4 are the same patient, the same knee, the same injection, and the same physician. One is a 99214 with modifier 25 plus a procedure; the other is a procedure with no E/M at all.

State, in one sentence, what makes the difference.

Answer: Work independent of the procedure — three chronic conditions individually assessed with plans, prescription drug management, and two tests ordered with stated reasons. Remove that work and both the E/M level and the separate reportability of the E/M disappear together, because they are answering the same question from two directions.

15.11 Split/shared and incident-to

Two mechanisms let a service involving a non-physician practitioner be reported in a particular way, and they are constantly confused. They apply in different settings and their rules are different.

Split/shared visits — the facility setting

A split (or shared) visit is an E/M service performed jointly by a physician and a non-physician practitioner in the same group, in a facility setting — a hospital, an emergency department, a skilled nursing facility. The service is reported by whichever of them performed the substantive portion.

The definition of "substantive portion" has moved, and this is where currency matters. It has been defined by CMS in terms of history, examination, or medical decision making, and separately in terms of more than half of the total time, with the policy transitioning between them over several years. Verify the current definition against current CMS guidance before you rely on this paragraph. It is one of the few areas in this book where the rule has changed more than once in a short period.

Split/shared services are identified on the claim with a modifier designating the service as split or shared. Chapter 16 §16.9 addresses the facility side.

Incident-to — the office setting

Incident-to is a Medicare billing provision that permits services furnished by auxiliary personnel — including non-physician practitioners — to be billed under the physician's national provider identifier, and therefore paid at 100% of the fee schedule rather than the 85% that applies when a nurse practitioner or physician assistant bills under their own.

Every one of these conditions must be met:

1. The setting is a physician's office, not a facility
2. The patient is established
3. There is an established plan of care for the problem, created by the physician
4. No new problems are addressed at the encounter
5. The physician provides direct supervision — present in the office suite and immediately available, though not in the room
6. The physician remains actively involved in the course of treatment
7. The auxiliary personnel are employed or contracted appropriately by the practice

Conditions 2, 3, and 4 are the ones that break. Any of the following disqualifies incident-to billing for the encounter:

• The patient is new
• The problem is new
• The plan of care is new
• An established patient presents with a new problem — even alongside established ones

⚠️ Where Claims Die

The new-problem disqualification is the one that costs organizations the most, and Account 10-4471 illustrates it exactly.

Suppose the March 14 visit had been conducted by a nurse practitioner rather than the physician.

Three chronic conditions with established plans of care: incident-to would be available for those.

The right knee is a new problem. It has no established plan of care. It is being evaluated for the first time.

The encounter cannot be billed incident-to. It must be billed under the nurse practitioner's own national provider identifier, at 85%.

The failure mode is systematic rather than occasional: a practice configures incident-to billing as a default for all established patients and never re-evaluates per encounter. Every visit in which an established patient brings up something new is then billed at 100% when it was entitled to 85%.

That is an overpayment, it accrues on every affected encounter, and Chapter 5 §5.1's sixty-day rule attaches to it once it is identified. It is also detectable — the diagnosis codes on the claim will show problems with no prior history in the record.

And it has no financial signal. The claims pay. Nothing denies. Nobody calls.

📋 Read the Chart

Source: encounter header and signature block Encounter: established patient, office What it says:

"Rendering provider: [nurse practitioner], NP. Supervising physician: [physician], MD. Billed under: [physician], MD."

What it means: an incident-to claim was submitted. The header does not tell you whether it was permissible.

What to do about it: check the four questions the header cannot answer. Was the patient established? Was every problem addressed an established problem with an existing plan? Was a physician physically in the office suite at the time? Is the supervising physician's presence documented for that date, not merely asserted by the template?

Where it appears: in every practice that employs non-physician practitioners, on a very large fraction of encounters, usually configured once and never revisited.

15.12 The documentation that survives an audit

Everything in this chapter converges here. An E/M level is a conclusion, and the record either supports it or does not.

Six characteristics of documentation that holds up.

1. The assessment names each problem individually and states its status and plan. Not "chronic conditions stable." Diabetes — at goal — continue metformin. Reading 2 in §15.10 is the whole argument, and the difference is one sentence per problem.

2. Decisions are visible as decisions. "Continue lisinopril 20 mg" in an assessment is a decision. "Lisinopril 20 mg" in a medication list is an inventory. Prescription drug management is documented in the plan, not in the list.

3. Orders state their reason. "A1c ordered to assess glycemic control" is worth more than "A1c." It supports the data element, it supports medical necessity, and it supports the diagnosis linkage on the claim — three things from four extra words.

4. Considered-and-rejected options are written down. The risk element counts them explicitly and they exist only if documented. "Discussed referral to orthopedics; patient prefers to try injection first" is an option considered and not selected, and it is invisible unless typed.

5. If time is used, the total is stated as a total, on the date, by the reporting professional. "Total time on the date of the encounter: 35 minutes" is a defensible statement. "Spent about half an hour with the patient" is not, and neither is a system-generated duration derived from how long a record was open.

6. The note is about this patient on this day. Copied-forward text is the single most damaging thing in a modern medical record. It inflates volume, it removes the ability to tell what happened today, and when an auditor finds an examination finding that was carried forward from a visit six months earlier, every other assertion in the note becomes questionable. Chapter 4 §4.7 covers copy-forward at length and Chapter 38 §38.1 covers what to do about it.

⚖️ Compliance Check

What an E/M audit actually looks like, so it is not a surprise.

A payer, contractor, or internal auditor pulls a sample — commonly ten to thirty encounters — and levels each one independently from the documentation. The finding is expressed as an error rate and frequently as an extrapolated overpayment, in which the error rate found in the sample is projected across the full population of similar claims in the period. That extrapolation is why a thirty-chart audit can produce a six-figure demand.

The most common finding is not fraud. It is a level unsupported by the assessment — usually because the problems element was scored from a problem list rather than from addressed problems.

The second most common is a data element that does not survive counting — panels, double-counted orders, internal notes counted as external.

Both are correctable by reading the definitions, which is a genuinely encouraging fact: the most common audit findings in the highest-volume service category in American medicine are fixed by reading four pages of guidelines carefully.

15.13 🗂️ The Encounter — leveling Account 10-4471

The claim has carried 99214-25 since Chapter 4. Chapter 14 defended the modifier. This section defends the level.

From the March 14 note (Figure 4.2), the note states explicitly: "Time was not used for level selection on this encounter." So the level must stand on medical decision making, and the time alternative is unavailable — not because the visit was short, but because the total was never documented.

Problems addressed

The assessment contains four items, each individually assessed with a stated status and a plan:

1. Type 2 diabetes mellitus — stable, metformin continued, A1c ordered
2. Essential hypertension — at goal, lisinopril continued, no change
3. Hyperlipidemia — atorvastatin continued, lipid panel ordered
4. Right knee pain — new complaint, consistent with a degenerative process, no definitive diagnosis established today

Items 1 through 3 are three stable chronic illnesses, each addressed. The threshold for moderate is two or more. Moderate.

Item 4 is independently interesting: a new complaint, evaluated, without a definitive diagnosis. That is one undiagnosed new problem with uncertain prognosis, which is also moderate on its own.

Problems: MODERATE. Reached two independent ways, which is a comfortable place to be.

The chronic kidney disease entry on the problem list is not addressed — the assessment does not reference it and there is no documented care coordination. It contributes nothing here. That entry is Chapter 36's, and this chapter does not touch it.

Data reviewed and analyzed

Two unique tests ordered: hemoglobin A1c, lipid panel.

Category 1: 2 items. The threshold for limited is any combination of 2. The threshold for moderate is 3.

Data: LIMITED (LOW).

This element does not reach moderate, and it does not need to. The lipid panel is one test, not several. No external records were reviewed, no independent historian was needed, no outside test was independently interpreted, and there was no discussion with an external professional.

Risk

Three medications reviewed and continued with documented decisions — metformin, lisinopril, atorvastatin. That is prescription drug management.

Risk: MODERATE.

(The injection is reported separately as 20610, so its own work is not what carries this element — and it does not need to. Prescription drug management reaches moderate independently, which is the cleanest position to argue from and the one this book will hold to.)

The determination

   ACCOUNT 10-4471 — March 14, established patient

   Problems addressed .......... MODERATE   ✓
   Data reviewed ............... limited (low)
   Risk ........................ MODERATE   ✓
   ─────────────────────────────────────────────
   Two of three at moderate  ►  MODERATE MDM
                             ►  99214  (established)

   Time: not documented — the time road is unavailable.

99214 is correct, and this is the textbook illustration of the two-of-three rule: the data element is a full level below and changes nothing.

Two things this section deliberately does not do.

It does not code osteoarthritis. The note says no definitive diagnosis was established and no prior imaging is available. M25.561 — pain in right knee — is the correct diagnosis for March 14, and Question 5 stays open until Chapter 22, when imaging exists.

And it does not revisit the modifier. Chapter 14 closed that. What this section adds is that the same documentation supports both conclusions — the three chronic conditions and the prescription drug management are simultaneously the moderate problems element, the moderate risk element, and the evidence that the E/M was separately identifiable from the injection.

One set of sentences in an assessment is doing three separate jobs. Remove them, as Reading 2 and Reading 4 in §15.10 did, and the level falls to 99213 and the modifier becomes unsupportable in the same stroke.

🎓 Exam Watch

The counterfactual that certification exams love: what if the physician had documented 28 minutes of total time?

By MDM: 99214. By time: 28 minutes is in the 20–29 range for an established patient, which is 99213.

Report 99214. The two methods are alternatives and you select by the one that supports the service. Documenting time does not obligate you to use it, and time never caps an MDM-based level.

A coder who downcodes to 99213 because "the time says 99213" has misunderstood the rule in a way that costs their organization money on a very large number of claims.

Summary

An E/M code measures the professional's cognitive work, by medical decision making or by total time on the date of the encounter — two roads, either one, never a blend.

New versus established turns on any professional service, from the same specialty and subspecialty in the same group, within three years. 99201 no longer exists.

2021 removed history and examination from level selection, deleted 99201, redefined MDM with a published grid, and redefined time as total time on the date including non-face-to-face work with no counseling-dominance requirement. 2023 extended the framework to hospital, observation, consultation, emergency department, nursing facility, and home services — Chapter 16.

MDM has three elements and a two-of-three rule. The lowest element does not cap the level.

Problems counts problems addressed — evaluated or treated, including options considered and declined. Notation of a problem managed by someone else is not addressed. "Stable" means at treatment goal, not unchanged.

Data has three categories with specific counting rules: a unique test is a CPT code, a panel is one test, ordering and reviewing the same test is one item, external means a different group or specialty, and Category 3 requires an actual two-way discussion.

Risk includes options considered but not selected. Prescription drug management is moderate and requires a decision, not a list. Social determinants that significantly limit diagnosis or treatment reach moderate.

Time counts the reporting professional's own time on that calendar date, including documentation and non-face-to-face work; it excludes staff time, other dates, separately reported services, and travel. Meet the minimum of the range; there is no rounding.

Prolonged services attach only to the highest level, in complete 15-minute increments — and Medicare uses G2212 with a different threshold than CPT's 99417.

G2211, the visit complexity add-on, pays for continuing care of a patient or ongoing management of a serious or complex condition — the relationship rather than the encounter. Verify its current status and restrictions by payer. It is under-reported, and a code you never report can never be denied.

Split/shared applies in facilities and is reported by whoever performed the substantive portion, whose definition has changed and must be verified. Incident-to applies in the office, requires an established patient with an established plan of care and no new problems, and pays 100% instead of 85%. A new problem breaks it, silently.

Account 10-4471 is a 99214 on moderate problems and moderate risk, with a limited data element that changes nothing — and the same three sentences in the assessment carry the level and the modifier both.

Key Terms

Evaluation and management (E/M) — the CPT category describing the cognitive work of evaluating and managing a patient, as opposed to performing a procedure. (Ch.15)

New patient — one who has not received any professional service from the physician or QHP, or another physician or QHP of the exact same specialty and subspecialty in the same group practice, within the past three years. (Ch.15)

Established patient — any patient who is not new by that test. (Ch.15)

Three-year rule — the lookback period defining new versus established, measured to the day. (Ch.15)

Medical decision making (MDM) — the complexity of establishing a diagnosis and selecting a management option, scored across problems addressed, data reviewed and analyzed, and risk, with two of three determining the level. (Ch.15)

Problems addressed — problems evaluated or treated at the encounter, including options considered and declined; excludes notation of problems managed by others without additional assessment or care coordination. (Ch.15)

Stable chronic illness — a chronic illness at its treatment goal. Not merely unchanged. (Ch.15)

Data reviewed and analyzed — the MDM element scored across three categories: tests, documents and independent historian (Category 1); independent interpretation (Category 2); and discussion with an external professional (Category 3). (Ch.15)

Unique test — a test identified by a distinct CPT code; a panel is one unique test. (Ch.15)

Independent interpretation — the professional's own documented interpretation of a test performed by another, not separately reported. (Ch.15)

Risk (MDM) — the risk of complications, morbidity, or mortality of patient management, including options selected and those considered but not selected. (Ch.15)

Prescription drug management — prescribing, adjusting, discontinuing, or documenting a decision to continue a prescription medication; reaches moderate risk. A medication list is not management. (Ch.15)

Total time on the date of the encounter — the reporting professional's own face-to-face and non-face-to-face time for that patient on that calendar date. (Ch.15)

Prolonged services — additional complete 15-minute increments beyond the highest-level code's time; CPT 99417, Medicare G2212 with a different threshold. (Ch.15)

Visit complexity add-on (G2211) — a HCPCS Level II add-on describing the complexity inherent to serving as the continuing focal point for a patient's care, or to ongoing care of a single serious or complex condition. Pays for the relationship rather than the encounter. (Ch.15)

Split/shared visit — an E/M performed jointly by a physician and NPP of the same group in a facility setting, reported by whoever performed the substantive portion. (Ch.15)

Substantive portion — the criterion determining who reports a split/shared visit; its definition has changed over time and must be verified against current guidance. (Ch.15)

Incident-to — a Medicare provision permitting qualifying office services by auxiliary personnel to be billed under the physician's NPI at 100%; requires an established patient, an established plan of care, no new problems, and direct supervision. (Ch.15)

Spaced Review

1. State the two roads to an E/M level and explain why they are alternatives rather than a combination.

2. A patient was last seen by a physician in your group, same specialty, three years and two days ago. New or established? What if that prior service was a hospital visit rather than an office visit?

3. An encounter scores problems moderate, data minimal, risk moderate. What level, and why does the minimal data element not reduce it?

4. (Chapter 14) What does modifier 25 require, and which sentences in Account 10-4471's assessment supply it? Name the overlap with this chapter's problems and risk elements.

5. Define "stable" for MDM purposes. A patient's A1c has been 10.2 for two years, unchanged. Is the diabetes stable? Explain the consequence for the problems element.

6. (Chapter 13) 99417 is an add-on code. State the two things Chapter 13 §13.7 forbids you to do with it, and then state the Medicare complication this chapter added.

7. An established patient with three established conditions presents to a nurse practitioner and also mentions a new rash. May the encounter be billed incident-to? State the rule and the financial consequence of getting it wrong.