What Is Evaluation and Management Coding? A Clear Guide

The code on your claim is not a visit summary. It is a financial classification, and a single wrong level, repeated across hundreds of encounters, is a structural revenue gap hiding in plain sight.

The Definition That Actually Matters for Your Revenue

Most outpatient practices submit hundreds of E&M claims every week without ever questioning whether the code on each claim accurately reflects the clinical work performed. That gap between submission volume and coding accuracy is where revenue quietly disappears, and it compounds across thousands of encounters before anyone notices. Evaluation and management coding is the process of selecting CPT codes that classify physician and non-physician practitioner visits for billing purposes. See our AI medical scribe for how this works in practice.

Under Medicare Part B, as tracked by CMS, E&M codes cover office visits, hospital encounters, and consultations, and they determine the allowed charges and payments a practice receives for those services. That last part is what textbooks tend to understate: the code selected is not a summary of the visit. It is the financial classification of the visit.

E&M services represent the dominant share of Medicare Part B allowed charges, making them the single largest reimbursement category most outpatient practices bill. For a practice running a high volume of annual visits, a consistent one-level coding error is not a rounding problem. It is a structural revenue gap.

The common assumption is that clinical documentation and billing are sequential tasks: the provider documents, the billing team codes, revenue follows. In practice, the code submitted is inseparable from what the note captured. Payers have been documented automatically downcoding 99214 to 99213 regardless of clinical complexity, a practice reported across multiple commercial plan types, cutting reimbursement on qualifying encounters.

When the documentation does not clearly support the higher level, there is no defense against that downcode. The classification decision was effectively made at the point of care, not in the billing queue. A significant share of E&M visits are inaccurately documented, with undercoding as the dominant error direction rather than overcoding.

Both errors carry cost, but neither triggers an automatic alert. Undercoding never generates a denial. Overcoding only surfaces when a payer pulls records.

The code selected is not a summary of the visit. It is the financial classification of the visit.

Key takeaways

• E&M coding is the process that ties every outpatient encounter to a reimbursement level, and most practices are getting it wrong on roughly half their claims without knowing it.
• The 2021 AMA/CMS overhaul eliminated the old three-component framework for office visits; code level now runs on medical decision-making complexity or total time, and practices still mentally running the old model are systematically miscoding.
• MDM is where the real revenue gap lives, physicians perform complex cognitive work that never makes it into the finalized note because documentation habits and time pressure strip it out before the claim is built.
• A clean remittance is not evidence of accurate coding; it means the audit hasn't happened yet. Payers pay first and audit later, so low denial rates are a lagging indicator, not a clean bill of health.
• One real-world orthopedic group discovered a 33% overcoding rate only after running a structured review, the kind of liability that compounds silently across thousands of encounters.
• Manual coding review can't close this gap structurally: by the time a coder reads a finalized note, the clinical reasoning is gone and nothing can recover what wasn't captured in the room.
• iScribe Health's E&M Coding Intelligence generates coding recommendations from the complete clinical narrative as it unfolds, not from the compressed summary left in the finalized note, so the complexity that actually happened gets reflected in the code that goes out the door.

Key Components of E&M Coding, and the One Component That Now Drives Most Code Selections

Before the 2021 guideline overhaul took effect, billing teams could at least argue that the three-component framework was intuitive: document the history, record the exam, assess the complexity, and the code more or less selected itself. That framework is gone for office and outpatient visits, and practices that haven't fully absorbed the shift are making code-level decisions on a model that no longer governs their claims, a gap that hits hardest in high-volume practices where clinicians are already spending significant time outside of patient care hours just to keep pace. The 2021 shift to MDM-or-time coding didn't simplify E&M selection; it created a hidden decision fork that most practices never consciously make.

Providers who never explicitly chose a pathway are defaulting to whichever method their documentation habits accidentally support, almost certainly leaving significant revenue unclaimed or creating audit exposure on the visits where time would have justified a higher level.

The Three Legacy Components and Why Two No Longer Determine Your Code Level

Prior to 2021, E&M code levels for office and outpatient visits were determined by three components: History, Physical Examination, and Medical Decision-Making (MDM). For established patients, meeting or exceeding the threshold on two of the three components set the code level; new patients required all three. What the 2021 changes make clear across the board is that History and Physical Examination still need to be documented to the degree medically appropriate, but they no longer drive code selection. Treating them as active levers in the coding decision is one of the most persistent errors in outpatient billing today, and one that becomes exponentially more expensive when a practice is processing hundreds of encounters per week.

The 2021 AMA/CMS Pivot: MDM and Total Time as the Only Two Pathways

Effective January 1, 2021, the AMA and CMS replaced the three-component model with a binary choice: select the code level based on MDM complexity or on total time on the date of the encounter. No hybrid calculation, no two-of-three weighting. As Jonathan Rubenstein and Mark Painter noted in Urology Times at the time of the final rule, the intent was to reduce documentation burden and focus code selection on clinical work actually performed.

The practical problem is that reducing documentation burden in policy does not reduce it in practice, not without the right infrastructure. Clinicians we work with at iScribe Health who see high patient volumes routinely spend significant time after clinic ends completing notes, which means the cognitive bandwidth required to consciously apply the MDM-or-time fork to every encounter simply isn't there. Accurate, compliant code selection across that kind of volume requires that the E&M intelligence be embedded in the documentation workflow itself, surfacing the right code at the point of note completion, not hours later during a billing review.

That is precisely where iScribe Health's E&M Coding Intelligence operates: after the AI drafts the encounter summary from ambient listening, the system applies the current MDM and time frameworks automatically, selecting the defensible code level before the note ever reaches the billing queue. For practices already running a supported EHR, this happens inside the existing workflow with no separate tool to open. The result is accurate, compliant coding at high patient volumes, which is also the condition under which payer-level downcoding policies like the BCBS of IL and Texas 99214-to-99213 blanket reduction do the most damage.

MDM's Three Elements - What Each One Actually Measures

MDM is evaluated across three distinct elements:

• The number and complexity of problems addressed
• The amount and complexity of data reviewed and analyzed
• The risk of complications or morbidity tied to patient management decisions

The overall MDM level is set by the highest two of the three elements, not an average. A provider managing a new, undiagnosed condition with uncertain prognosis, reviewing external records, and initiating a prescription drug regimen is generating Moderate MDM material on all three axes.

The problem is that none of that clinical reasoning makes it into the note automatically; it has to be written in. When a clinician is already behind on documentation and a payer like BCBS of IL is set to downcode the visit regardless of what the note says, the temptation to abbreviate is understandable and the financial consequence is compounding. iScribe Health's Ambient AI Documentation captures the clinical reasoning as it happens, the differential, the data reviewed, the management decision and its rationale, and structures it into the note in the format that supports the code the encounter actually warrants.

As a foundational principle of the 2021 guidance, the documentation must reflect the complexity to support the level. iScribe Health ensures it does, across every encounter, every day of clinical practice.

Total Time as an Alternative Pathway - What Counts and What Doesn't

Total time, as defined under the 2021 framework, means all time personally spent by the billing provider on the date of the encounter, including pre-visit chart review, the face-to-face visit itself, ordering and reviewing tests, counseling the patient or family, completing documentation, and coordinating care with other providers. What it explicitly excludes is time spent by clinical staff and time on a different calendar date. The distinction matters because providers who mentally equate "total time" with "time in the room" are systematically undercoding visits where significant pre- or post-encounter work occurred, and providers who are counting staff time or carry-over documentation from the following morning are creating audit exposure on the claims they do submit at a higher level.

The time thresholds that govern code level under this pathway are specific: for established patients, 99213 requires 20–29 minutes, 99214 requires 30–39 minutes, and 99215 requires 40–54 minutes of total provider time on the date of service. For new patients, the thresholds shift upward across the same codes. A provider who spends 22 minutes face-to-face but another 12 minutes that morning reviewing a specialist's records and another 8 minutes after the visit coordinating a referral has 42 total minutes, enough to support a 99215 for an established patient, but only if the documentation captures and aggregates that time explicitly.

Without a workflow that prompts the provider to record all contributing time segments at the point of note completion, that visit almost certainly closes as a 99214 or lower. iScribe Health's E&M Coding Intelligence addresses this directly: because the ambient AI is present across the encounter and the system is designed to surface total time inputs at documentation close, the full picture of provider time is captured and applied to the coding decision before the note is finalized, ensuring that the time pathway is exercised when it produces a higher defensible code level than MDM alone, and never applied when the documentation doesn't support it.

Medical Decision-Making (MDM) Explained, and Why It's Where Revenue Gets Left on the Table

MDM scoring sits at the center of that gap. Physicians perform genuinely complex cognitive work during every clinic visit, weighing competing diagnoses, recalling a patient's medication history, scanning outside records on a second monitor, and mentally calculating the risk of adding a new prescription. The problem is that MDM scoring rewards only what the note explicitly records, and time pressure during a full clinic day means a significant share of that reasoning never makes it into the documentation.

What makes this especially costly is that the lost revenue is not theoretical. Revenue cycle managers working with high-volume practices describe the gap as potentially substantial, not a rounding error, but a structural drain that compounds silently across every claim submitted. Chief Medical Officers, practice administrators, and individual physicians who close that gap get paid faster and lower the operational costs associated with medical scribing or transcription services.

Both outcomes hinge on the same root fix: documentation that actually captures what the provider's reasoning already earned.

The Three MDM Elements That Determine Your Code Level

"Providers are leaving significant revenue on the table by not fully understanding or leveraging MDM coding, with one Revenue Cycle Manager explicitly stating the lost revenue 'could be substantial'."

MDM is built on three distinct elements:

• The number and complexity of problems addressed
• The amount and complexity of data reviewed
• The risk of complications tied to management decisions

To qualify for a given MDM level, a provider must meet or exceed the threshold in at least two of the three elements, according to current AMA guidelines. That two-of-three requirement matters because a provider who thoroughly reviews external records but documents neither the review nor the reasoning behind a treatment decision may only satisfy one element on paper, regardless of what actually happened in the room.

A compounding real-world friction point: some billing platforms block providers from submitting certain E&M codes when a visit runs under ten minutes, even when MDM criteria are fully and demonstrably met. That kind of system-level constraint, which prioritizes time over MDM as the determining factor, effectively overrides the clinical complexity the provider already addressed. It is one reason iScribe Health's E&M Coding Intelligence applies automated E&M code assignment at the point of note completion, after the AI drafts the encounter summary and the clinical reasoning is already captured, rather than leaving code selection to a downstream billing step where visit-length proxies can quietly displace MDM evidence.

The Four MDM Levels Mapped to Office Visit Codes 99202 to 99215

The four MDM levels, Straightforward, Low, Moderate, and High, map directly onto the E&M code range for office visits. Straightforward MDM supports 99202 (new patient) and 99212 (established patient); Low supports 99203 and 99213; Moderate supports 99204 and 99214; High supports 99205 and 99215. Each step up the ladder carries meaningfully higher reimbursement, which means a consistent one-level undercode across thousands of annual visits is not a rounding error.

It is a structural revenue gap that compounds quietly with every claim submitted. iScribe Health's ambient listening and conversational AI captures the clinical narrative in real time during the encounter, so the documented record reflects the full complexity of the visit before the code is ever assigned, making each step of the MDM ladder accessible based on what the provider actually did, not what time permitted them to type.

Why Undercoding Dominates Inaccurate E&M Documentation

Only ~55% of E&M visits accurately documented

Industry data consistently indicates that only approximately half of E&M visits are accurately documented, meaning roughly 45% contain coding errors, a figure corroborated across multiple coding compliance analyses. Undercoding is the dominant direction for most practices, not overcoding. That framing matters because the common compliance fear runs the other way: providers habitually code conservatively to avoid audit scrutiny, which feels safe but silently erodes revenue.

Providers leave significant revenue on the table by not fully understanding or applying MDM coding, and that underutilization tends to be worst in high-volume practices or health systems where clinicians regularly spend significant time outside of patient care hours, simply because documentation fatigue degrades note completeness before MDM reasoning can be recorded. iScribe Health's ambient AI documentation addresses that directly: by reducing the charting burden that drives conservative, incomplete notes, it makes accurate MDM capture the path of least resistance rather than an additional task.

Point-of-Care Labs, EKG Reads, and External Records - What Actually Counts as Data Reviewed

The Data element of MDM is where documentation gaps are most concrete and most costly. Reviewing and independently interpreting a point-of-care lab result, reading an EKG, or personally reviewing external records from another facility all count toward the data threshold. Each of those actions takes place in the room, often while the provider is still speaking with the patient, which is precisely where iScribe Health's ambient listening captures it. When the encounter summary is drafted, those data-review actions are already embedded in the note, and the platform's E&M Coding Intelligence maps them to the correct MDM data element automatically. Real-Time Denial Alerts then surface any downstream claim issues before they result in lost reimbursement, closing the loop between what the provider did clinically and what the practice is ultimately paid for it.

E&M Code Levels and Selection Criteria - What Each Level Requires and What Moves You Between Them

Selecting the right E&M code level is not a judgment call made at the end of a visit; it is the direct output of a structured decision process that either supports the claim or quietly undermines it. Since the 2021 AMA/CMS overhaul, that process runs on two branches only, MDM complexity or total time, and practices still applying the old three-variable logic are building a selection error into every claim they submit. What follows breaks down exactly what each pathway requires, where documentation commonly fails to support the chosen branch, and how real-time ambient capture changes the equation for clinicians navigating this daily.

The Distribution Signal Auditors See First

A practice's E&M code distribution is a reflection of its documentation habits, not its patient population complexity. CMS publishes specialty-level utilization benchmarks, and any practice whose distribution deviates significantly from specialty norms is most likely revealing a systematic documentation behavior pattern. That deviation is the precise signal payers use to trigger audits, and it becomes visible to them before it becomes visible to the practice.

iScribe Health's Real-Time Denial Alerts are designed to surface that signal internally, at the claim level, before it accumulates into an auditable pattern. When documentation consistently supports the code submitted, and when the code family and level are validated before submission, the distribution naturally reflects the actual complexity of care being delivered. That alignment is what coding precision means in practice: not upcoding, not undercoding, but documentation that captures reality accurately enough that the code writes itself.

1. Level 2 E&M (99212/99202): Minimal MDM Defines the Floor

Level 2 visits require straightforward medical decision-making or minimal total time, making them appropriate for simple, self-limited problems with minimal data review and low-risk management. Practices managing high volumes of minor acute visits rely on this level as a baseline. The real tradeoff: chronic disease patients with even one stable condition often qualify higher, so defaulting to Level 2 risks systematic undercoding and lost revenue.

Documentation Requirements for E&M Visits - What Has to Be in the Note to Defend the Code You Submitted

A paid claim and a defensible claim are structurally different objects, and practices that conflate them are accumulating retrospective audit liability in real time. Payers routinely process claims without verifying documentation specificity, which means a clean remittance is not evidence of coding accuracy. It is evidence that the audit has not happened yet.

What "Supports the Code" Actually Means

The evidentiary standard payers apply during audit.

Payers do not evaluate whether a note is long or thorough. They evaluate whether each Medical Decision-Making element is explicitly supported with the required language and structured evidence. According to audit guidance and widely published billing compliance analysis, a note can be detailed and still fail audit if it lacks the specific terminology, data source identification, and risk justification that auditors use as their evidentiary checklist.

The code gets downgraded or clawed back not because the care did not happen, but because the documentation does not prove it did. This is precisely where ambient AI documentation changes the risk profile. iScribe Health's Ambient Listening and Conversational AI captures the full clinical encounter hands-free, delivering the greatest value when physicians want a completely hands-free documentation experience during the visit, and then, at the point of note completion, the AI drafts an encounter summary structured around the language and elements auditors actually check.

The output is designed to create more defensible documentation, not simply longer documentation.

MDM Documentation Specificity

How to record problem complexity, data sources, and risk.

Specificity is the operative word in every MDM element. Documenting "reviewed prior records" without naming which records and what clinical finding they informed does not satisfy the data complexity element for a Moderate MDM level. A provider who genuinely reviewed outside imaging, cross-referenced lab trends, and weighed competing diagnoses has done the cognitive work, but none of that work is auditable unless the note names the sources, describes what was found, and connects the findings to the clinical decision.

The same principle applies to risk: "prescription drug management" must be stated explicitly, not implied by a medication list sitting elsewhere in the note. The structural gap between what happened clinically and what the note proves happened is where audit liability quietly accumulates. iScribe Health's E&M Coding Intelligence layers onto the AI-drafted note to flag exactly these gaps, identifying where MDM elements lack the explicit language payers require, so the provider can confirm or correct before the note is finalized and the claim is submitted.

The result is documentation that standardizes clinical documentation quality across the practice, not just for the providers who already write with auditor language in mind. This is most impactful in high-volume practices or health systems where clinicians are regularly charting two or more hours outside of patient care time, because the volume of encounters means MDM specificity gaps compound silently across hundreds of claims before any audit surfaces them.

Documenting Time Correctly

What counts, what does not, and the scribe time trap.

For time-based E&M selection, CMS specifies that only the billing provider's total time on the date of service counts. That includes pre-encounter record review, face-to-face time, and post-encounter work such as ordering, coordinating care, and completing documentation. Time performed solely by a scribe or other clinical staff does not count toward the billable total.

This distinction creates a compliance trap that surfaces frequently in practices using scribing support: if the provider documents a total time that includes scribe activity, the time figure in the note overstates the billable total and the selected code level becomes indefensible under audit. iScribe Health's ambient model is built around the billing provider's own encounter. Because the conversational AI is capturing the physician's spoken clinical reasoning in real time, hands-free, without delegating the cognitive or verbal work to a separate scribe, the documented time more accurately reflects the provider's direct participation.

The Automated E&M Coding feature then applies coding logic at the point of note completion, after the AI drafts the encounter summary, so the selected code level is grounded in what the note actually supports, not what the provider assumed it supported.

Complete Notes vs. Audit-Proof Notes

Thoroughness is not enough.

The honest limitation here is that thoroughness and defensibility measure different things. A provider can write four paragraphs of clinical narrative and still leave the MDM elements structurally unsupported because the language does not map to the audit checklist payers apply. Thoroughness describes volume. Defensibility describes structure.

iScribe Health's Real-Time Denial Alerts address this gap prospectively, surfacing documentation and coding issues at the encounter level, before a claim is ever submitted, rather than after a denial or audit demand forces a retrospective review. For practices already running a supported EHR, the ambient documentation and E&M intelligence integrate seamlessly into the existing workflow, so the compliance layer arrives without adding a separate tool or a separate step.

The value is realized across every patient encounter and every day of clinical practice, not as a one-time intervention, but as a standing structural improvement to how documentation exits the visit and enters the billing cycle.

Related Reading

• Orthopedic Coding Guidelines
• Medical Coding Automation
• E&m Coding Cheat Sheet
• Orthopedic Medical Coding
• Urology Coding Guidelines

E&M Coding's Impact on Reimbursement - The Revenue Math Most Practices Have Never Actually Run

The common assumption among practice administrators and billing decision-makers is that if providers are documenting every visit and claims are going out the door, E&M coding accuracy is probably fine. Revenue leakage from E&M coding inaccuracy rarely looks like a problem. Claims go out, payments come in, denial rates stay low, and the billing dashboard shows nothing alarming. That is precisely the trap. The financial damage from systematic coding inaccuracy does not announce itself; it accumulates quietly across every visit, every day, compounding into a number most practice administrators have never actually calculated.

The Per-Visit Dollar Gap Between Adjacent E&M Levels

The Medicare Physician Fee Schedule makes this concrete: the allowed amount difference between CPT 99213 and 99214 for an established patient visit is meaningful, typically several dozen dollars per claim. That gap feels small in isolation. It is not small in aggregate.

A single provider generating several thousand established-patient visits per year who consistently codes one level below what the documentation supports is forgoing a material share of annual Medicare revenue before commercial payer rates are factored in. The math is rarely run; the gap is rarely seen. The same dynamic plays out in behavioral health in a way that directly collapses reimbursement integrity.

CPT code 90837 covers therapy sessions of 53 minutes or more, and the Medicare Physician Fee Schedule reimburses it at a higher rate than CPT 90834, which covers 37–52 minute sessions. When payer-side coding errors or documentation failures result in 90837 being reimbursed at the 90834 rate, the entire financial incentive for providing extended sessions disappears, not because the clinician stopped delivering that care, but because the documentation or the submitted code failed to capture it. That is not a marginal rounding error.

It is a structural revenue differential that evaporates claim by claim, silently, with no denial and no alert. Ensuring reimbursement integrity across time-based codes like these is precisely the problem iScribe Health's E&M Coding Intelligence is designed to solve, surfacing the correct level at the point of note completion, after the AI drafts the encounter summary, before a claim is ever submitted.

Why Undercoding Is a Silent Revenue Leak

No denial, no alert, and no visibility make undercoding particularly damaging. Overcoding generates a denial, a recoupment letter, or a payer audit. Undercoding generates nothing except a lower payment. Per the CMS Physician Fee Schedule, a claim submitted at 99213 when 99214 was supported pays out at the lower rate without any flag, any remittance explanation, or any downstream alert.

A practice can submit a high volume of claims per year with a significant undercoding rate and never receive a single signal that anything is wrong. The only way to see it is to run the math no one has run. iScribe Health's Automated E&M Coding addresses exactly this visibility gap.

By analyzing the completed encounter note in real time, through Ambient AI Documentation that captures the conversation as it happens, the system surfaces the defensible E&M level before the claim leaves the practice. In high-volume practices where clinicians regularly chart well beyond patient care hours, documentation habits formed under that time pressure are the direct upstream cause of the undercoding pattern. Simplifying the documentation workflow is not an ergonomic nicety; it is a reimbursement integrity intervention realized across every patient encounter and every day of clinical practice.

The Overcoding Side of the Equation - Audit Exposure, Recoupment Demands, and False Claims Act Risk

The risk runs in both directions. Overcoding exposes practices to payer audits, recoupment demands, and potential False Claims Act liability, even when the pattern is unintentional. Data published through CMS and the OIG consistently shows that upcoding patterns, once identified through a Targeted Probe and Educate review or a RAC audit, result in repayment demands covering multiple years of claims.

One real-world orthopedic audit surfaced a substantial overcoding rate the group had no visibility into until the review was triggered. iScribe Health's Real-Time Denial Alerts are designed to interrupt this exposure before it compounds. Rather than discovering a systematic overcoding pattern during a payer-initiated review, practices running iScribe Health through an integrated EHR workflow receive signals at the claim level, not months later during an audit cycle.

The platform delivers the most value when a practice is already running a supported EHR, because the E&M coding intelligence layer operates within the existing workflow rather than requiring a parallel process.

How Documentation Habits Compound Across Every Claim

The root cause of both problems, undercoding and overcoding, is documentation that does not accurately reflect what happened in the room. When clinicians are charting under time pressure, the note captures less than the encounter delivered, or defaults to a prior-visit template that no longer reflects the patient's current complexity. Neither pattern produces a denial.

Both patterns produce a coding record that diverges from clinical reality, either leaving revenue on the table or creating audit exposure that grows silently with every submitted claim. iScribe Health's Ambient Listening and Conversational AI address this at the source. By capturing the encounter as it unfolds and drafting the note automatically, the documentation reflects the actual visit rather than the provider's end-of-day recollection of it.

E&M Coding Intelligence then applies to that richer note, producing a defensible code level grounded in what the CMS Physician Fee Schedule and CPT guidelines actually require, not what a fatigued clinician estimated before moving to the next patient.

Why Manual E&M Coding Review Doesn't Scale, and What AI-Powered Coding Intelligence Changes

Audit workflows feel like a safety net. The problem is that by the time a coder reviews a finalized note, the encounter is closed, the clinical reasoning is gone, and the only thing left to evaluate is a compressed summary of what actually happened. That structural gap, not effort or intent, is why manual review reliably misses the accuracy problems that cost practices the most.

The Finalized Note Is Already a Lossy Compression of What Actually Happened Clinically

Every finalized note is an edited version of the encounter, not a transcript. During a visit, a provider considers differentials, weighs data sources, and mentally assesses risk, but time pressure means only a fraction of that reasoning reaches the note. As coding compliance analyses of high-volume clinical settings have consistently found, volume of documentation does not equate to accuracy of E&M level assignment. A note can be thorough in length and still fail to capture the clinical complexity that determines the correct code.

Why Retrospective Coding Audits Cannot Fix E&M Accuracy Problems

The core issue is structural, not procedural. Coding errors are created in real time, at the moment clinical reasoning goes undocumented, not at the moment a code is selected. Retrospective chart review can identify that a code looks unsupported, but it cannot recover the differentials that were considered and never written down. Auditing the output of a lossy process does not fix the process. It only confirms that something was lost.

How Real-Time Documentation Captures the Clinical Complexity That Manual Notes Routinely Drop

The accuracy gap is a documentation problem, not a billing problem. Closing it requires intervening at the point of capture, during the encounter, before reasoning is filtered out. When a provider verbally rules out a secondary diagnosis or references a prior imaging result, an AI medical scribe captures that reasoning in real time.

That captured narrative, including differentials considered, data reviewed, and risk language, becomes the source material for coding recommendations rather than the stripped-down finalized note. iScribe Health's E&M Coding Intelligence generates automated coding recommendations from the complete clinical narrative, meaning the code reflects what the provider actually did, not just what survived the documentation process. In internal testing across a sample of outpatient encounters, coding recommendations drawn from AI-scribed narratives frequently surfaced MDM complexity elements that were absent from the corresponding manually finalized notes, reducing the documentation gap that drives undercoding.

Coding from a richer source document reduces undercoding without inflating codes beyond what the clinical record supports. Because the recommendation is drawn from the full clinical narrative rather than a compressed manual note, the code reflects actual complexity rather than documentation habit.

Next steps

If your practice is submitting claims without ever seeing where documentation habits leave revenue behind, the path forward starts with capturing clinical reasoning at the moment it occurs, not reconstructing it after the note is saved. Start with our AI medical scribe.

The 2021 shift to MDM-or-time coding created a decision fork most practices never consciously made, meaning providers are defaulting to whichever code their documentation habits accidentally support rather than what the encounter actually warrants. And because systematic undercoding never triggers a denial, a practice billing 99213 on visits that support 99214 across thousands of annual encounters loses six figures annually with no remittance flag, no alert, and no signal that anything is wrong. Together, these realities point to one intervention: closing the gap at the point of capture, during the encounter, before clinical complexity is filtered out by time pressure.

Start with an AI medical scribe built for clinical-grade documentation. The encounter narrative gets captured in real time, MDM elements are recorded as care is delivered, and the code reflects what actually happened in the room.

Frequently Asked Questions

What exactly are E&M codes?

E&M codes are CPT codes that classify physician and non-physician practitioner visits for billing purposes. Under Medicare Part B, they cover office visits, hospital encounters, and consultations, and they determine the allowed charges and payments a practice receives for those services, making them the single largest reimbursement category most outpatient practices bill.

Which CPT codes correspond to E&M office visits?

Office visit E&M codes run from 99202 to 99215. New patient visits use 99202–99205 and established patient visits use 99212–99215, with each step up the range tied to a higher level of Medical Decision-Making complexity or total time, Straightforward MDM supports 99202/99212, Low supports 99203/99213, Moderate supports 99204/99214, and High supports 99205/99215.

Can I use total time instead of MDM to pick the E&M code level?

Yes, since the 2021 AMA/CMS guideline overhaul, code level selection for office and outpatient visits runs on exactly two branches: Medical Decision-Making complexity or total time on the date of service. Total time covers all clinician time on that date, including pre-visit chart review, the face-to-face encounter, and post-visit documentation.

How does MDM scoring actually work, what are the three elements and how do they combine?

MDM is evaluated across three elements: the number and complexity of problems addressed, the amount and complexity of data reviewed and analyzed, and the risk of complications or morbidity tied to patient management decisions. To qualify for a given MDM level, a provider must meet or exceed the threshold in at least two of the three elements, the overall level is set by the highest two, not an average.

If my providers are busy and keeping notes short, are they more likely to be overcoding or undercoding?

Undercoding is the dominant direction for most practices, not overcoding. Industry data indicates that roughly 45% of E&M visits contain coding errors, and the problem tends to be worst in high-volume settings where documentation fatigue causes providers to write abbreviated notes, meaning clinical reasoning that would justify a higher MDM level simply never makes it into the record.