Evaluation and Management Codes: A Complete Reference Guide

Silent undercoding costs practices thousands annually without a single denial to flag it. Here is how E/M codes actually work, and where the logic breaks down by care setting.

The common assumption among practice administrators and billing decision-makers is that if claims are processing and denials aren't spiking, E/M coding must be close enough. That assumption is expensive. Coding decisions that shape reimbursement happen in seconds, but their financial consequences accumulate over months of claims processing, often without a single denial to signal that anything went wrong.

Coding decisions that shape reimbursement happen in seconds, but their financial consequences accumulate over months of claims processing, often without a single denial to signal that anything went wrong.

See our AI medical scribe for how this works in practice.

Evaluation and Management (E/M) codes are the CPT code family that translates the narrative of a patient encounter into a standardised payment tariff. As Laycock et al. note in the Future Hospital Journal, clinical coding converts history, examination, diagnoses, and procedures into standardised codes that determine the payment assigned to that encounter, establishing a direct link between documentation quality and reimbursement level. The E/M family spans CPT codes 99202 through 99499, covering office and outpatient visits, hospital care, emergency encounters, nursing facility visits, and preventive medicine services.

The dollar gap between adjacent code levels is concrete, and systematic undercoding at a single level quietly erases substantial revenue annually, with no denial, no flag, and no operational signal. Laycock et al. confirm that undercoding assigns a lower-complexity code than the encounter warrants, resulting in lost revenue that does not generate a denial or an audit flag, so the problem persists silently. Compounding the risk: providers who learned E/M coding before the landmark 2021 restructuring, which eliminated history and physical examination components in favour of medical decision making or total time, may be applying outdated logic to current encounters, and that coding drift goes undetected without systematic review.

Key takeaways

• E/M codes are not a billing formality, they are the financial architecture of every outpatient encounter, and the wrong code on a processed claim is still the wrong code when an auditor pulls it six months later.
• Only ~55% of E/M visits are accurately documented, and a single encounter analysis at one ortho group surfaced a 33% overcoding rate no one inside the practice had detected.
• Five distinct code families govern E/M billing by care setting, and treating them as interchangeable is one of the most common structural billing errors, one that rarely surfaces until a payer flags it.
• Code level selection follows two defined pathways, Medical Decision Making or total time, and the errors that cost practices real reimbursement cluster at predictable decision points in both.
• MDM scoring requires meeting or exceeding the benchmark on at least two of three elements; missing that threshold by a single tier on high-volume encounters adds up to significant undercoding or overcoding across a year of claims.
• New-versus-established patient misclassification is not an isolated error, it creates a pattern across encounters that payers recognize, and patterns trigger targeted audits.
• Total E/M time includes pre-visit record review, care coordination, and post-visit orders, not just face-to-face minutes, and most practices leave real reimbursement on the table by never documenting it.
• iScribe Health's E/M Coding Intelligence closes the gap by generating coding recommendations from the complete clinical narrative in real time, before the note is finalized, so the reasoning that determines code level is captured while it still can be.

The Main Categories of E/M Codes Organized by Care Setting

Five distinct code families govern E/M billing, and the care setting determines which family applies before a provider ever considers code level. Treating these families as interchangeable is one of the most common structural billing errors in practice, and it rarely surfaces until a payer flags the claim, a costly discovery point in any high-volume practice where hundreds of encounters have already moved through the same flawed workflow.

The table below maps each setting to its governing code range and the billing logic that controls level selection.

• Care Setting

Code Range

Level Selection Basis

• Office / Outpatient

99202-99215

MDM or total time (provider chooses)

• Hospital Inpatient and Observation

99221-99239

MDM or total time (unified in 2023)

• Emergency Department

99281-99285

MDM only; time does not apply

• Consultation

99242-99255

MDM or time; requires Request, Recommendation, and Report

• Nursing Facility

99304-99316

MDM or time

• Home or Residence

99341-99350

MDM or time

• Preventive Medicine

99381-99429

Age and patient status; MDM and time do not apply

Source: Novitas Solutions Medicare Part B E/M Services guidance, 2024.

1. Office and Outpatient E/M Codes (99202-99215) - The Most Frequently Used Category

Office and outpatient codes cover new patients (that same figure-99205) and established patients (99211-99215), and they are the dominant E/M category by encounter volume. Solutions Medicare Part B E/M Services guidance, 2024, this family offers two compliant level-selection pathways: MDM or total time on the date of service. That dual-pathway flexibility is specific to this setting. Assuming it transfers to the ED or inpatient context is a structural cross-setting error, and it is the kind that does not show up until a payer flags the claim.

In high-volume outpatient practices, where clinicians regularly chart two or more hours outside of patient care time, the cognitive overhead of consistently applying setting-specific billing logic across a full schedule is the root condition that produces these errors. iScribe Health's Ambient AI Documentation and E&M Coding Intelligence reduce that overhead by handling encounter capture and code-level mapping automatically, applied at the point of note completion after the AI drafts the encounter summary. The result is accurate, compliant medical coding sustained across high patient volumes without adding documentation burden to the clinical team.

2. Hospital Inpatient and Observation E/M Codes (99221-99239) - Admission, Subsequent, and Discharge Visits

The 2023 CPT updates unified inpatient and observation care under a single code range, consolidating two previously separate billing tracks into that same figure-99239. Initial care codes run from that same figure (straightforward or low MDM) through 99223 (high MDM), each with defined minimum time thresholds; subsequent care codes follow the same MDM structure at lower time thresholds. Practices that carried pre-that same figure documentation habits into the unified system introduced a transition-period error window. Many have not fully closed it, particularly when providers rotate across inpatient and outpatient settings on the same day.

Streamlining clinical workflows so care teams can maintain documentation quality even during peak census periods is a care delivery infrastructure problem. iScribe Health's Ambient Listening and Conversational AI address it at the encounter level, reducing the charting burden that compresses the clinical detail required for correct MDM-level selection under the unified range. For health systems already running a supported EHR, EHR Integration means this capability materializes as a seamless ambient documentation experience without disrupting existing workflows.

3. Emergency Department E/M Codes (99281-99285) - MDM-Only, No Time-Based Option

Emergency department codes are selected on MDM alone. according to industry data Solutions Medicare Part B E/M Services guidance, time-based selection does not apply because ED care is unscheduled and continuous, making total time difficult to define the way office visits allow. This is the sharpest cross-setting compliance exposure in the E/M code map, and it is invisible inside documentation workflows that treat all settings the same.

iScribe Health's E&M Coding Intelligence is applied at the point of note completion, after the AI drafts the encounter summary, and maps each note against the specific billing logic of its care setting. In the ED, that means the system identifies when time has been used as a level-selection signal where the setting does not permit it and surfaces the mismatch through Real-Time Denial Alerts before the claim moves to submission. For emergency departments running high volumes of unscheduled encounters, this is the review step that catches what manual workflows miss.

4. Nursing Facility and Home or Residence E/M Codes (99304-99316 and 99341-99350) - Post-Acute and Domiciliary Care

Nursing facility codes (99304-99316) cover initial and subsequent care in skilled nursing and long-term care settings, while home and residence codes (99341-99350) apply to visits in private homes, assisted living, and domiciliary facilities. Both categories follow MDM or time-based selection similar to office codes. The practical challenge is documentation burden, these settings often lack robust EHR infrastructure, increasing undercoding risk for high-complexity patients.

5. Preventive Medicine E/M Codes (99381-99429) - Where 99395 Lives, Separate from Standard Office Visit Levels

Preventive medicine codes operate on entirely different selection logic from every other E/M family. A code like 99395, covering an established patient aged 18 to 39, is chosen by age and patient status, not MDM complexity or time. MDM and time are irrelevant to level selection here. The compliance risk surfaces when a provider bills a same-day problem-oriented visit alongside a preventive code without appending modifier 25. Without it, payers routinely bundle the two claims and deny the problem-oriented service.

iScribe Health's E&M Coding Intelligence recognizes the structural separation between preventive and problem-oriented services and applies it at the point of note completion. When both services occur on the same date, the system identifies the modifier 25 requirement and surfaces it before the claim is submitted, turning an avoidable denial into a correctly billed encounter. Across high patient volumes, that catch, applied consistently on every encounter and every day of clinical practice, directly reduces claim denials and protects reimbursement for services that were clinically justified and correctly documented.

How Providers Select the Correct E/M Code Level Using MDM or Time

Two Pathways, One Choice - MDM or Total Time Determines the Code Level

• Medical Decision Making (MDM) complexity
• Total Time spent on the date of the encounter

As the AMA CPT E/M Office Visit Guidelines established in that same reform, providers choose whichever pathway yields the more advantageous, fully supported code level for that visit.

The Context Variables That Frame Every Selection

• Place of service
• Type of service
• Patient status

These context errors are especially common in high-volume practices where clinicians are regularly charting two or more hours outside of patient care, the very conditions under which patient status and place-of-service details slip through unchecked. iScribe Health's ambient AI documentation captures clinical notes during or after the patient encounter and feeds them into its E&M Coding Intelligence layer, which evaluates these context variables at the point of note completion, after the AI drafts the encounter summary.

A 7-Step Decision Sequence That Turns Code Selection Into a Repeatable Process

The Gut and Liver study notes that EHR structured fields routinely fail to capture the full clinical reasoning underlying a visit.

The Three Elements of Medical Decision Making Every Coder Must Score Correctly

Three elements determine every MDM score: the number and complexity of problems addressed, the amount and complexity of data reviewed and analyzed, and the risk of complications or morbidity from patient management. Meeting or exceeding the benchmark on at least two of the three qualifies an encounter for a given complexity level: Straightforward, Low, Moderate, or High. That 2-of-3 threshold is where most coding errors originate, and most practices never see them coming.

1. Number and Complexity of Problems Addressed - The First MDM Element Coders Must Nail

This element scales from a single, self-limited problem at Straightforward through multiple chronic illnesses or a new problem with uncertain prognosis at Moderate, up to a severe exacerbation or threat to life at High. The scoring error here is subtler than it looks: a provider managing a diabetic patient presenting with new chest pain is addressing multiple chronic conditions plus an undiagnosed new problem, which can reach Moderate, but only if the note names and contextualizes each problem. A diagnosis list without clinical reasoning reads as lower complexity to any auditor or coder working from the finalized note.

2. Amount and Complexity of Data Reviewed - The Second MDM Element That Trips Up Even Experienced Coders

Data scoring requires documented evidence of specific actions:

• Ordering or reviewing tests
• Reviewing external records
• Performing independent interpretation of results

Independent interpretation is the most frequently under-documented sub-element in this category. A provider who personally reviews and interprets an EKG rather than relying on a cardiology read has met a meaningful data threshold, but if the note says only "EKG reviewed," that independent analysis is invisible to the coder. The data element collapses to a lower tier, and the overall MDM level follows.

3. Risk of Complications and the 2-of-3 Threshold Rule - Why High-Complexity Problems Alone Won't Justify a High-Level Code

Risk is the third MDM element, spanning Minimal through High based on patient management decisions such as prescription drug management or surgery with identified risk factors. Critically, Evaluation and Management Codes require meeting or exceeding two of the three MDM elements to qualify for a given level, Straightforward, Low, Moderate, or High. A provider may document a high-complexity problem, but if data and risk both fall short of threshold, the code level does not follow the problem complexity alone.

Why High-Complexity Problems Alone Don't Justify a High-Level E/M Code

Risk is scored on the management decisions made, not on diagnosis severity alone. Under the AMA CPT E/M Office Visit Guidelines, prescription drug management reaches Moderate risk, while drug therapy requiring intensive monitoring or a decision for hospitalization reaches High. The critical and most common documentation failure here is recording what was prescribed without capturing why that management decision carried meaningful risk. A note that lists a medication change but omits the clinical reasoning behind it gives the coder no basis to assign the higher risk tier. Combined with a data element that also under-scores, the 2-of-3 threshold goes unmet, and a clinically High encounter codes out as Moderate regardless of what actually happened in the room.

New Patient vs Established Patient Codes and the Office Visit Rules That Govern Both

The 3-Year Rule That Separates New Patient Codes from Established Ones

A new patient is anyone who has not received professional services from the provider, or from any provider of the same specialty within the same group practice, in the past three years. According to the AMA CPT Guidelines, that definition is group-wide and specialty-specific, not provider-specific.

Why CPT 99211 Creates Specific Audit Exposure

iScribe Health's E&M Coding Intelligence and Automated E&M Coding surface these mismatches at the point of note completion, after the AI drafts the encounter summary and before the claim is submitted. At that moment, when the documentation is freshest and the encounter detail is most complete, the system's Real-Time Denial Alerts flag code-level inconsistencies, including scenarios where a documented physician judgment call is mapped to a nurse-level code, or where a higher-complexity note is assigned a code that understates the work performed.

Related Reading

• Cpt Code 99232
• Cpt Code 99244
• Cpt Code 99203
• Cpt Code 99221
• Cpt Code 99205
• Cpt Code 99223
• Cpt Code 99202
• Cpt Code 99222
• Cpt Code 99233

Total Time for E/M Coding and When Prolonged Services Apply

Every Qualifying Activity on the Date of Service Counts

Total time for time-based E/M coding includes both face-to-face and non-face-to-face work performed on the date of the encounter. According to the updated E/M guidelines, qualifying activities include reviewing prior records, ordering tests or medications, EHR documentation, care coordination, and communication with other providers.

How Add-On Code +99417 Keeps the Clock Running Past the Highest Level Threshold

A review of outpatient E/M guidelines confirms that prolonged services add-on code +99417 applies in 15-minute increments beyond that ceiling. Despite that straightforward structure, +99417 remains one of the most systematically missed codes in outpatient billing.

The Documentation Trap, Why Stating Total Time Explicitly Is Non-Negotiable

If the note does not state total time explicitly, the undocumented minutes cannot be counted, regardless of what actually occurred.

E/M Coding Compliance and the Documentation Requirements That Make or Break an Audit

The Three Documentation Requirements Auditors Actually Test Against

History and physical must appear in the record even though, under the AMA CPT E/M Guidelines, they no longer drive level selection.

Why Inconsistent Code Selection Across Providers Is the Audit Trigger No One Monitors

When one physician in a group consistently selects 99214 for encounters a colleague codes as 99213 for comparable complexity, that distribution pattern flags the practice for closer review. The CMS Recovery Audit Program recovered over $[1.3(https://www.lexology.com/library/detail.aspx?g=dbf2be19-590a-42d8-83a7-9aadbcb4bbe7) billion in improper Medicare payments in a recent reporting period, a figure that reflects exactly the kind of pattern-level detection practices cannot perform on themselves without systematic measurement.

iScribe Health's E&M Coding Intelligence applies Automated E&M Coding at the point of note completion, after the AI drafts the encounter summary, producing a consistent coding signal across every provider in the practice. By standardizing the moment and method of code selection, iScribe Health reduces the inter-provider variation that surfaces practices on payer radar, directly enhancing compliance integrity across the group.

Overcoding and Undercoding Are the Same Documentation Gap Pointing in Different Directions

An ortho group audited across 941 encounters found a 33% overcoding rate that no one in the building knew existed. Providers did not intend to upcode; no mechanism existed to verify that the documented note matched the code billed. As Lexology has detailed in its analysis of E/M audit exposure, both directions of miscoding carry real financial and legal consequences that practices routinely underestimate until a payer review forces the issue.

iScribe Health's EHR Integration applies when the practice or health system is already running a supported EHR and wants a seamless ambient documentation experience, with no parallel workflow, no duplicate data entry, and no gap between the clinical encounter and the coded note.

Why Accurate E/M Coding Stays Invisible Without Real-Time Documentation Intelligence

In a 2025 HAB Central review of AI-assisted billing across orthopaedic encounters, ~55% of E&M visits are accurately documented (industry baseline); nearly half of encounters carried a coding error under manual workflows, a finding that, while specific to that specialty context, points to a broader structural reality: in high-volume clinical settings, manual documentation workflows routinely leave coding errors undetected until an external review surfaces them. For any practice running comparable volumes, a meaningful share of visits may carry coding inaccuracies that no one inside the building will see until a payer or auditor runs the numbers.

Why Note Finalization Makes Coding Errors Structurally Invisible

In a manual documentation workflow, the provider finishes the encounter, writes the note from memory, and the note is finalized before anyone reviews the code selection. By that point, the clinical reasoning that would justify a higher or corrected level, what was ruled out, what alternatives were considered, what complexity the provider managed, is gone. A coder reviewing that finalized note isn't seeing the encounter; they're seeing a summary written after the thinking stopped.

This is why clean claims and low denial rates are unreliable proxies for coding accuracy. Undercoding never triggers a denial. A practice can sustain years of revenue loss while every internal metric signals that everything is working correctly. The absence of a warning is not evidence of accuracy.

What a 941-Encounter Audit Reveals About Undetected Overcoding

A 941-encounter audit at Center for Sports Medicine and Orthopaedics surfaced a 33% overcoding rate entirely invisible to providers and billing staff before AI-assisted review was applied. That figure carries two risks simultaneously: overcoded claims draw payer scrutiny and repayment demands, while undercoded claims leave earned reimbursement uncollected with no flag, no denial, and no signal that anything went wrong.

How Real-Time Narrative Capture Changes the Coding Equation

Addressing this requires moving the review upstream, before the documentation window closes, not adding a review step after the note is finalized.

iScribe Health's E&M Coding Intelligence is built on a different premise: by embedding AI-powered coding recommendations directly into the clinical narrative as it is captured in real time, the goal is to make the complete clinical picture, including what was considered and ruled out, available for review before the note is finalized and that justification disappears. Without that intervention, the gap between what was documented and what was managed accumulates silently, encounter by encounter, until a payer or auditor runs the numbers first.

Related Reading

• Cpt Code 99284
• Cpt Code 99396
• Cpt Code 99499
• Cpt Code 99386
• Cpt Code 99395
• Cpt Code 99385
• Cpt Code 99283
• Cpt Code 99024

Next steps

If your practice is processing claims cleanly but has never measured whether documented complexity matches what was actually delivered, the path forward starts with accepting that clean billing and accurate billing are not the same thing. The orthopaedic audit reviewed in this guide found a 33% overcoding rate that no one inside the building knew existed. Undercoding produces no denial at all. Both errors live in the same invisible gap between what the provider managed and what the finalized note captured. Start with our AI medical scribe.

The MDM two-of-three rule means a provider can deliver a genuinely high-complexity encounter and still code it lower because the note failed to independently evidence data reviewed and risk of management. Manual documentation means that reasoning is gone the moment the note closes, because providers reconstruct context from memory rather than capture it in real time. Together, they point to moving the coding review upstream, before the documentation window closes and before the clinical thinking that justifies the correct level becomes unrecoverable.

Start with an AI medical scribe that captures the full clinical narrative as the encounter unfolds and maps it to MDM element thresholds before the note is finalized. What follows is a coding signal grounded in what was actually said and decided in the room, not a summary written after the thinking stopped, applied consistently across every encounter and every day of clinical practice.

Frequently Asked Questions

Does it matter which care setting I'm billing from, or can I just pick the code level that fits the complexity?

Care setting determines which code family applies before you ever consider level selection, and treating those families as interchangeable is one of the most common structural billing errors in practice. For example, emergency department codes (99281-99285) are selected on MDM alone, time-based selection is not permitted there, while office and outpatient codes (99202-99215) allow either MDM or total time on the date of service.

Does it make a financial difference if a provider consistently picks the code one level below what the encounter actually supports?

Yes, and the damage is silent. Undercoding assigns a lower-complexity code than the encounter warrants, resulting in lost revenue that does not generate a denial or an audit flag, so the problem persists without any operational signal. Across a practice billing high volumes of visits monthly, systematic undercoding at even a single level can quietly erase substantial revenue annually.

We changed how we code after 2021, do the same rules still apply, or have there been more updates since then?

There have been additional changes since 2021. The 2023 CPT updates unified inpatient and observation care under a single code range (99221-99239), consolidating what had previously been two separate billing tracks. Practices that carried pre-2023 documentation habits into the unified system introduced a transition-period error window that many have not fully closed.

Does a patient being new versus established actually change which code I use for an office visit?

Yes. Office and outpatient codes are split by patient status: new patients fall under 99202-99205 and established patients under 99211-99215. Both groups use the same two level-selection pathways, MDM or total time on the date of service, but the correct code range depends on which patient status applies.

If I bill a preventive visit and also address a separate problem on the same day, what's the risk?

Billing a same-day problem-oriented visit alongside a preventive code without modifier 25 is a common bundling error with real denial consequences, payers routinely bundle the two claims and deny the problem-oriented service. This is avoidable, but only if the billing workflow treats preventive codes as a structurally separate family rather than a variant of the standard office visit.