E/M Coding Automation in Your EMR: Capturing Revenue You Are Already Leaving Behind

The Under-Coding Problem

Studies consistently find that independent primary care physicians under-code their outpatient encounters at rates of 20–35%. The most common pattern: selecting 99213 (moderate complexity) for encounters that document 99214 (moderate-high complexity) or even 99215 (high complexity) level medical decision-making. Over a year of practice, this conservative coding pattern costs a typical solo physician $15,000–$35,000 in legitimate revenue.

Under-coding is not an ethics problem. It's a knowledge and time problem. The 2021 E/M coding changes (AMA revision effective January 2021) moved the coding framework from history-and-exam element counting to a medical decision-making (MDM) framework. Many physicians understood the change conceptually but do not apply it systematically at the point of care, under time pressure, with 14 more patients to see.

E/M coding automation addresses this: the EMR analyzes the encounter documentation in real time and suggests the appropriate E/M level based on the documented MDM, flagging when documentation supports a higher level than the physician initially selected.

How E/M Coding Automation Works

A well-implemented E/M coding module does the following:

1. Analyzes the note in real time: As the physician documents (or as the AI ambient scribe drafts the note), the system evaluates the clinical content against E/M MDM criteria: number and complexity of problems addressed, amount and complexity of data reviewed, and risk of complications or morbidity.
2. Suggests a code level: The system displays the supported E/M level alongside the note, before the physician signs or submits.
3. Explains the supporting documentation: Rather than simply outputting a code, the system shows which documentation elements support the suggested level. This is the educational component: physicians learn the MDM framework by seeing how their documentation maps to criteria.
4. Flags documentation gaps: If the clinical encounter appears to involve high-complexity MDM but the documentation does not explicitly capture the key elements (e.g., the risk assessment for a treatment decision is not documented), the system prompts the physician to add a sentence before signing.

The 2021 E/M MDM Framework: What the AI Is Evaluating

Under the current AMA MDM framework, E/M level is determined by the complexity of medical decision-making across three elements:

• Problems addressed: Number of problems (1 self-limited, 1+ stable chronic, 1 worsening chronic, 1 new undiagnosed, 1 with threat to life/function)
• Data reviewed and analyzed: Ordering/reviewing tests, reviewing external records, independent interpretation of results, independent historian
• Risk of complications and morbidity: Medication management risk, prescription drug management, diagnosis or treatment requiring intensive monitoring, decision regarding hospitalization

An E/M coding AI evaluates each of these elements in the note, determines the level for each element, and applies the "at least 2 of 3 elements" rule to assign the overall E/M level. It also checks whether modifier 25 is appropriate when a preventive visit and a problem-focused visit are documented on the same date.

Revenue Impact: What Practices Typically Find

Practices implementing E/M coding automation typically find that 18–28% of encounters were coded at a lower level than documentation supports. At average Medicare reimbursement differentials:

• 99213 → 99214 step: approximately $36 additional per encounter
• 99214 → 99215 step: approximately $52 additional per encounter

For a solo physician seeing 18 patients per day, 220 days per year, with 22% of encounters potentially under-coded by one level, the recoverable revenue is approximately: 18 × 220 × 0.22 × $36 = approximately $31,400 annually. This is revenue from encounters already worked. Documentation already written, revenue not yet captured.

Compliance Considerations

E/M coding automation that suggests a higher code level is appropriate only when documentation genuinely supports that level. The AI's role is to identify documentation that already supports a higher level, not to generate documentation to justify a code. The physician's responsibility is to verify that the suggested level accurately reflects the clinical complexity of the encounter before signing.

Properly implemented E/M coding automation is a compliance tool: it reduces both under-coding (a revenue problem) and over-coding (a compliance risk) by surfacing the documentation-supported level transparently.

Krasyn's Billing Intelligence

Krasyn includes real-time E/M coding analysis that surfaces the documentation-supported level before you sign each note, and explains exactly which MDM elements drive the suggestion. For DPC hybrid practices that do bill some fee-for-service services, this catches legitimate revenue before it is lost. See Krasyn's coding tools at krasyn.com/dpc.