A physician we spoke with described a new-patient visit that went the way the hard ones do. The patient was frustrated and in real pain. An insurance change had left them without treatment for months, and they arrived with a complicated history that had to be untangled before anything else could happen. The visit ran very long.
Weeks later, a complaint from the insurer arrived, and answering it meant sending in a very long SOAP note and the labs. He had billed the visit as a 99204, as usual. Then he looked again at what he'd sent and thought about how difficult that note had been to write, even though this kind of visit is something he does all the time.
Does that sound familiar?
Most independent physicians have some version of that visit every week. It is the most cognitively demanding work in the schedule, and it is very often coded at the same level as a much simpler visit, because getting the full reasoning onto the page at 7 p.m. is the part that doesn't happen. That is the real hidden revenue in complex encounters. It is not hidden in a checklist. It is hidden in the gap between what you did and what you wrote down.
First, the shortcut that doesn't work: more review of systems
A common assumption, and one that circulates in plenty of AI-generated coding advice, is that adding a fuller review of systems or documenting more body systems on exam will lift a 99204 to a 99205. For office visits, that hasn't been true since 2021.
When the AMA revised the office and outpatient E/M codes (99202 through 99215) effective January 1, 2021, it removed history and examination as factors in choosing the level. The American College of Cardiology's E/M summary puts it plainly: the level now rests on medical decision making or on total practitioner time on the day of the visit. You still document a medically appropriate history and exam, because the next clinician needs it. But a ten-system ROS doesn't move the code, and the CMS Evaluation and Management Services booklet (MLN006764) is clear that the volume of documentation shouldn't be what drives the billed level.
Why this matters more with AI in the room. An AI tool that pads every note with templated systems review produces longer notes, not better-supported codes. Bloated, near-identical notes are exactly what auditors and payer algorithms flag. If a vendor's pitch is that its scribe "adds" documentation to raise your level, treat that as a red flag, not a feature.
So if ROS isn't the lever, what is? There are exactly two, and the visit in that story may already have qualified under either one.
Lever one: total time on the date of service
Under the current CPT E/M guidelines, you can choose the level purely on the total time you personally spent on the patient's care on the date of the encounter (DOS). That includes face-to-face time, and it also includes reviewing outside records and labs before the visit, counseling, ordering tests and prescriptions, coordinating with other clinicians, and documenting in the EHR, as long as it happens that same calendar day.
| Code | Patient | Total time on the date of the encounter |
|---|---|---|
| 99204 | New | 45–59 minutes |
| 99205 | New | 60–74 minutes |
| 99214 | Established | 30–39 minutes |
| 99215 | Established | 40–54 minutes |
Time ranges per the AMA CPT office/outpatient E/M descriptors as reproduced in CMS guidance.
A visit that "ran very long" with a patient who had gone months without care, plus the record review it took to understand their history, can easily cross 60 minutes. If it did, 99205 may have been supportable on time alone, and no clinical-complexity argument was needed. But only if the minutes are in the note. "Spent a long time with the patient" supports nothing. A specific total does.
A time statement can be short and in your own words, for example:
I personally spent 68 minutes on this patient's care today, including review of outside records and prior labs, history and examination, counseling on the treatment plan, and documentation.
Two rules keep that statement honest. Only your own time counts, so staff time doesn't, and neither does any time a scribe, human or software, spends drafting. And only time on that date counts, so finishing the chart the next morning is off the clock. If a visit goes well past the top of the range, the CPT prolonged-services add-on 99417 begins at 75 minutes for a new patient, while Medicare uses HCPCS G2212 instead, starting at 89 minutes, as Noridian's prolonged-service guidance lays out.
Lever two: medical decision making (MDM)
The other pathway ignores the clock entirely. The level is set by three elements of medical decision making, and you need two of the three at a given level. The AMA's E/M revisions FAQ walks through how the elements are applied; the table below summarizes the moderate-versus-high distinction in plain terms.
| MDM element | Moderate (99204 / 99214) | High (99205 / 99215) |
|---|---|---|
| Problems addressed | A chronic illness that's progressing or worsening; two or more stable chronic illnesses; a new problem with an uncertain prognosis | A chronic illness with severe exacerbation or progression; an illness or injury posing a threat to life or bodily function |
| Data reviewed and analyzed | One category met, such as a combination of outside records reviewed, unique tests ordered, and unique results reviewed | Two categories met, such as that combination plus an independent interpretation of a test, or a management discussion with an outside clinician |
| Risk of patient management | Prescription drug management | Drug therapy requiring intensive monitoring for toxicity; a decision about hospitalization or escalation of care |
Paraphrased from the AMA MDM framework. Consult the full CPT definitions before coding.
Read that table honestly and two things follow. Many complex visits are legitimately 99204s: a tangled history managed with prescription changes lands at moderate risk, and that is the right code. But some are not. A patient whose chronic condition has severely progressed after months without treatment, whose outside records and labs you reviewed, and for whom you ordered and interpreted further workup, may meet high complexity on problems and data together, regardless of risk.
In those visits, the difference between 99204 and 99205 usually isn't whether the work happened. It's whether the note says, in specific terms, how severe the progression was, which outside records and results you reviewed, and what you were weighing when you chose the plan. ICD-10-CM specificity helps here too, because a precise diagnosis code tells the reviewer more about problem complexity than a vague one.
What the difference is actually worth
The size of the gap is set by relative value units (RVUs) under the Medicare Physician Fee Schedule, and it's meaningful, though smaller than some marketing suggests. At 2026 national non-facility rates, before geographic adjustment:
- New patients: 99204 pays about $177 and 99205 about $237, a difference of roughly $59 per visit.
- Established patients: 99214 pays about $136 and 99215 about $192, a difference of roughly $57.
Your commercial allowed amounts depend on your contracts and can run higher or lower, so the number worth checking is the one on your own fee schedule. And the honest framing is that this revenue was never lost because the visit was over-documented or under-performed. It was lost because the documentation didn't match the visit. Coding a 99205 when the note doesn't support it isn't recovered revenue; it's a future denial, a takeback on the remittance advice (RA), or worse.
What about commercial plans like a BCBS PPO?
Most independent practices see more commercial patients than Medicare patients, and the gap between the two levels is usually wider on a commercial plan. Many commercial fee schedules, including many Blue Cross Blue Shield PPO contracts, are built on Medicare's resource-based relative value scale (RBRVS) and then multiplied by a contracted percentage. Because 99205 carries more RVUs than 99204 (3.50 work RVUs against 2.60), the multiplier widens the dollar gap along with it.
The table below applies a range of multipliers to the 2026 Medicare national rates. Treat it as an illustration, not a rate sheet:
| Contract rate | 99204 | 99205 | Difference per visit |
|---|---|---|---|
| Medicare (national) | $177 | $237 | $59 |
| 120% of Medicare | $213 | $284 | $71 |
| 135% of Medicare | $239 | $320 | $80 |
| 150% of Medicare | $266 | $355 | $89 |
| 180% of Medicare | $319 | $426 | $107 |
Calculated from 2026 Medicare national non-facility amounts ($177.36 and $236.81) before geographic adjustment. Your locality's GPCI moves the baseline up or down before any multiplier applies.
So on a well-paying commercial PPO, the difference between the two codes can plausibly reach $70 to $100 or more per visit. Three caveats keep that number honest:
- There is no single "BCBS rate." Blue Cross Blue Shield is a federation of more than 30 independent companies, and each negotiates its own contracts. Two practices across town can hold very different fee schedules with the same Blue plan.
- Office visits often get smaller multipliers than procedures. KFF's review of the research found private insurers paid 143% of Medicare for physician services on average, ranging from 118% to 179% across studies. It also notes that studies of office visits alone found lower ratios. The Health Care Cost Institute's claims analysis put the commercial average at 122% of Medicare, with some metro areas below Medicare and some above 200%.
- The patient may feel it too. On a high-deductible plan, a higher-level code can mean a larger bill for a patient who hasn't met their deductible. That's another reason the level has to be one the note clearly supports.
The reliable way to know your own number is to look up the allowed amount for 99204 and 99205 in your payer contract, or on a recent explanation of benefits (EOB) or remittance advice. Insurers also have to publish their negotiated rates under the federal Transparency in Coverage rule.
For scale: a practice that sees three new patients a week who genuinely meet high complexity but are billed as 99204s gives up roughly $11,500 a year at an $80 difference. That is real money for a solo office. It is only recoverable for visits where the work was actually done and written down.
Payers are paying attention to AI-assisted notes
This is the part most "AI scribe boosts your coding" articles leave out. In March 2026, Trilliant Health studied six large health systems that adopted AI scribing and found the share of high-intensity new-patient visits rose roughly 12 to 20 percentage points between 2018 and 2024, and 7 to 12 points for established patients. The same month, the Blue Cross Blue Shield Association published a report tying rising coding intensity to higher commercial spending.
Interpretations differ. Trilliant's own read is that the shift likely reflects more thorough documentation of work that was previously under-recorded, and the American Hospital Association argues that a sicker, older patient population explains much of the trend. But the practical consequence for a small practice is the same either way: a jump in your 99205 and 99215 share will be noticed, and each one needs a note that stands on its own. The insurer complaint in our opening story is what that scrutiny looks like from the inside.
Where an AI scribe genuinely helps
With all of that on the table, here's the real case for ambient documentation in complex visits, and it's a good one.
- It catches the reasoning you say out loud. In a long visit, you talk through the outside records, the prior labs, why one treatment is safer than another given the history. That reasoning is exactly what MDM rewards and exactly what disappears when the note is written from memory hours later.
- It keeps severity specific. "Worsening" and "severe progression after four months without treatment" support different levels. The draft keeps the words you actually used.
- It records the data review as it happens. Which outside records you read, which results you reviewed, what you ordered. Those are counted items under MDM, and they are the details most often compressed into "labs reviewed."
- It gives you back time to state your time. The scribe can't count your minutes, and it shouldn't. But with the drafting handled, closing the chart the same day, with an accurate time statement, becomes realistic.
As for how much time it saves: be wary of big round numbers. A 2026 JAMA study of 8,581 clinicians across five academic health systems found about 16 fewer minutes of documentation per eight hours of patient care, with larger savings for frequent users. That's real, but the bigger payoff in complex visits is completeness, not speed.
How MediAttune approaches this. MediAttune drafts the note from what was said in the room, and you review, edit, and sign it. It never adds a finding you didn't state, never fills in an exam element you didn't perform, and never suggests a billing code: coding stays with you and your biller. If it wasn't said, it isn't in the draft. The scribe is in early access and built for practices on Office Ally EHR 24/7 and Practice Mate.
Habits that close the gap, with or without AI
- Narrate your reasoning, briefly. "I reviewed her records from the previous practice and the March labs; given the kidney function, I'm avoiding NSAIDs and starting…" One sentence like that carries problems, data, and risk at once.
- Name the severity. If the condition has severely progressed, say so in those terms and say why.
- Decide your pathway before you close the chart. If the visit ran long, document the total time. If it was complex but efficient, make sure the MDM elements are explicit.
- Read the draft as an auditor would. Delete anything that wasn't said or done. A shorter accurate note beats a longer padded one every time.
- Watch your own distribution. If your share of level-five visits changes, you should be able to explain why from the notes themselves.
Clean claims start with notes that match the visit. For more on documenting complicated presentations accurately, see our guide to documenting mixed-etiology foot complications, or browse the rest of the MediAttune blog.
Frequently asked questions
Does adding a review of systems or examining more body systems raise a visit from 99204 to 99205?
No. Since January 1, 2021, history and examination are not used to select the level of an office or outpatient E/M visit (99202-99215). The note still needs a medically appropriate history and exam, but the level is chosen by medical decision making or by total physician time on the date of the encounter.
Can time an AI scribe spends writing the note count toward a time-based E/M code?
No. Time-based E/M coding counts only the billing physician's or qualified health professional's own time on the date of the encounter. Software time, like staff time, does not count. Your own time with the patient, reviewing records, ordering, and documenting on that date still counts, and you state it in the note.
How much more does Medicare pay for 99205 than 99204 in 2026?
At 2026 national non-facility rates, 99204 pays about $177 and 99205 about $237, a difference of roughly $59 before geographic adjustment. For established patients, 99214 pays about $136 and 99215 about $192. Commercial allowed amounts depend on your contracts.
How much more does a commercial plan like a BCBS PPO pay for 99205 than 99204?
It depends on your contract. Many commercial fee schedules are set as a percentage of Medicare, so at 120% to 150% of Medicare the difference is roughly $71 to $89 per visit, and at 180% it is about $107. Commercial multipliers for office visits are often lower than for procedures, so check the allowed amounts in your own contract or on a recent remittance advice.
Will an AI scribe automatically make my visits qualify for 99205?
No, and a tool that promises that is a compliance risk. An AI scribe can help the note capture the clinical reasoning, data review, and management risk you actually discussed. Whether that adds up to moderate or high medical decision making depends on the visit, and the coding decision stays with the physician.
When can prolonged services be added to 99205?
Only when the visit level was selected by time. Under CPT, 99417 applies starting at 75 minutes of total time for a new patient. Medicare uses HCPCS G2212 instead, which starts at 89 minutes. Check each payer's policy.
Sources and further reading
- AMA: CPT Evaluation and Management (E/M) revisions FAQs
- CMS: Evaluation and Management Services (MLN006764)
- American College of Cardiology: Evaluation and Management documentation
- Noridian Medicare: Prolonged service codes (G2212)
- KFF: How much more than Medicare do private insurers pay? A review of the literature
- Health Care Cost Institute findings on commercial physician prices vs. Medicare
- CMS: Transparency in Coverage (health plan price transparency)
- Trilliant Health: Outpatient coding intensity following adoption of AI-enabled scribing (March 2026)
- Blue Cross Blue Shield Association: Rising coding intensity and its impact on health care affordability
- American Hospital Association: Fact sheet on AI and coding intensity (August 2026)
- Summary of the 2026 JAMA multisite study on AI scribes and documentation time
Payment figures are 2026 Medicare national non-facility amounts, and commercial figures are illustrative multiples of them; both are for reference only. This article describes documentation practice and general coding rules; it is not coding, legal, or billing advice. Confirm code selection with your coder, your MAC, and your payer contracts.