A four-physician internal medicine group asked us to look at a letter they had received from a Medicare Administrative Contractor. It was not a demand for records. It was a comparative billing report, a few pages of charts showing that one of their physicians billed 99215 on 31 percent of established patient visits while the state average for internal medicine was closer to 8 percent. The practice manager had never run that comparison herself. She had the data. Nobody had asked the question.
That is the whole point of an E/M distribution analysis by provider. Before a payer or a contractor tells you a physician looks unusual, you should already know, and you should already know why. Sometimes the answer is that the physician runs the practice's complex care panel and the curve is exactly right. Sometimes the answer is a template that defaults every visit to the same level. The report cannot tell the difference. Opening ten notes can.
A glossary line for physicians reading this: E/M stands for evaluation and management, the office visit codes. Established patient visits run from 99211 (minimal, no physician required) through 99212, 99213, 99214 and 99215, in order of increasing medical decision making or time. New patient visits run 99202 to 99205. A distribution analysis simply counts how many visits each provider billed at each level and expresses the result as percentages.
Key takeaways
- An E/M distribution is a count of visits by code level per provider over a period, turned into percentages so providers of different volume can be compared.
- Since the 2021 office visit rules moved leveling to medical decision making or total time, most primary care curves peak at 99214, not 99213, so an old benchmark will mislead you.
- The comparison that matters is against the same specialty and the same patient mix, and the fairest first comparison is against the provider's own colleagues.
- An outlier is a reason to sample notes, not a conclusion; both over-coding and under-coding show up as skewed curves.
- Run the report quarterly, keep the results, and document what you did about them; that record is itself compliance evidence.
Pulling the report from your practice management system
Every practice management system can produce this. The report is usually called a procedure frequency report, a CPT utilization report or a charges by provider by code report. Filter to date of service in the period, rendering provider, and CPT codes 99202 to 99205 and 99211 to 99215. Count units, not dollars. Use rendering provider, not billing provider, or every visit in a group that bills under the owner will land on one person.
We run new and established visits as two separate curves. They have different shapes and different reasons to be skewed, and mixing them hides both. We also exclude telehealth visits into a third bucket when the volume is meaningful, because telehealth visits in most practices lean toward 99213 and pull the whole curve down if they are blended in.
A quarter is the right window. A month is too small for a part-time provider; 60 visits do not make a curve. A year hides a change in behavior that started in March. Twelve months of data is still useful as a baseline, then quarterly after that.
Export to a spreadsheet and compute, for each provider, the percentage at each level. Then compute the same for the practice as a whole. That practice-level row is your first benchmark, and for a single-specialty group it is often the most relevant one you will find.
What a normal curve looks like after 2021
Before January 1, 2021, office visit levels were driven by history, exam and medical decision making, and the national Medicare curve for established patients peaked at 99213. The 2021 revision to the office visit guidelines dropped history and exam from level selection and let providers choose by medical decision making (MDM) alone or by total time on the date of the encounter. Medicare utilization data in the years since shows the peak moving to 99214, and in our experience most primary care and medical specialty practices now sit with 99214 as the single largest bar.
So an old chart that says "99213 should be your most common code" is out of date, and a consultant who tells a physician to move visits down to match it is giving bad advice. The honest benchmark is current specialty data. CMS publishes the Medicare Physician and Other Practitioners by Provider and Service dataset every year, with counts by NPI and by HCPCS code, and it can be filtered by specialty and state. Several specialty societies publish their own distributions. Comparative billing reports, when you receive one, contain the contractor's own peer numbers, which is the comparison an auditor would use.
Two things are true at the same time. The distribution has shifted up because the rules changed. And some providers have shifted further than the rules justify. The analysis exists to tell those two apart.
A worked example with four providers
Here is a fictional quarter for the internal medicine group in the opening, established patient visits only, 99211 excluded because it is rarely billed by physicians.
| Provider | Visits | 99212 | 99213 | 99214 | 99215 |
|---|---|---|---|---|---|
| Dr. A | 1,180 | 3% | 28% | 58% | 11% |
| Dr. B | 1,050 | 2% | 24% | 43% | 31% |
| Dr. C | 940 | 9% | 61% | 28% | 2% |
| Dr. D | 1,210 | 4% | 30% | 55% | 11% |
| Practice | 4,380 | 4% | 35% | 47% | 14% |
Dr. A and Dr. D look like each other and like what we expect from general internal medicine in 2026. Dr. B is the physician in the letter: 31 percent at 99215 is roughly three times his colleagues. Dr. C is the outlier nobody writes letters about. Sixty-one percent at 99213 and 2 percent at 99215 in an internal medicine panel with the same scheduling template as the others almost always means under-coding.
Put a dollar figure on Dr. C so the conversation with her is concrete. Suppose the practice's blended allowed amount is about $92 for 99213 and $130 for 99214. If a note review shows that a third of her 99213 visits, about 190 in the quarter, met 99214 by MDM, that is about $7,200 of documented work per quarter that was never billed. Nobody will audit her for it. The practice still loses it, and it compounds every year.
What an outlier pattern usually means when we open the notes
A skewed curve has a handful of common causes, and the note sample tells you which one you have. We pull 10 to 20 notes per outlier provider, weighted toward the level in question, and score them against the MDM table: number and complexity of problems, amount and complexity of data, and risk of patient management.
For a high curve like Dr. B's, the usual findings are these. A template that pre-populates "chronic illness with severe exacerbation" or lists every stable chronic condition as if it were addressed, which inflates the problems column. Time statements that include staff time or time on another day, which is not allowed under the 2021 rules. Counting each lab in a panel as a separate data point when CPT counts the panel once. Or a genuinely heavier panel: a physician who sees the practice's transplant, oncology follow-up and complex geriatric patients will legitimately sit high, and the notes will show it.
For a low curve like Dr. C's, we usually find a physician who still levels by exam habit from before 2021, a note that addresses three chronic conditions with medication changes but only lists one in the assessment, or a belief that 99214 requires a long visit. Prescription drug management alone is moderate risk. Two stable chronic illnesses alone are moderate problems. Those two facts together are 99214, and many physicians do not know it.
New patient curves have their own patterns. A provider at 70 percent 99204 with almost no 99203 is worth a look; so is a provider whose new patient visits are all 99203 in a specialty where a first visit routinely includes ordering and reviewing multiple tests.
Turning the analysis into a standing process
The report is only useful if someone acts on it and writes down what they did. We recommend a one-page quarterly summary: the distribution table, the practice row, the external benchmark used and its source, which providers were sampled and why, the sample results as an agreement rate, and the education or template change that followed. Keep it with the compliance program documents. The Office of Inspector General's compliance program guidance for physician practices, first published in October 2000, lists auditing and monitoring as a core element, and a dated record of this exercise is exactly what that looks like in a small practice.
Share results with providers as a group first, anonymized, then individually. Physicians respond well to seeing their own bar next to their colleagues' bars; they respond badly to being told a payer average says they are wrong. If you want help setting up the review or the note sample, that is what our coding and revenue audit does, and our coder training covers MDM scoring for the staff who will run it after us.
Questions we hear
Our EHR shows a suggested level. Doesn't that make the distribution automatically right?
No. The suggested level is only as good as what the template pre-populates and what the provider clicks. Coding assistance tools tend to produce very tight distributions, because every provider is being pushed by the same logic, and a suspiciously uniform curve across providers with different panels is itself a pattern reviewers notice.
How different does a provider have to be before we worry?
There is no regulatory threshold. As a working rule, a provider more than about 10 percentage points off the practice at any one level, or more than double the specialty benchmark at 99215 or 99205, gets a note sample. Below that, we watch the trend across quarters rather than react to one report.
Should we count visits billed with modifier 25 separately?
Include them in the main curve, but flag the count. A provider whose curve looks fine but whose modifier 25 rate is twice the group's has a different problem, and it is one that payers audit on its own.
What to do this week
- Run a procedure frequency report for the last full quarter, by rendering provider, codes 99202 to 99205 and 99211 to 99215, units not dollars.
- Build the percentage table in a spreadsheet with separate curves for new and established visits and a practice-level row.
- Find one external benchmark for your specialty and write down where it came from and what year it covers.
- Pick the providers who are furthest from the practice row in either direction and pull 10 to 20 notes each at the level in question.
- Score the notes against the MDM table, record the agreement rate, and schedule a 20-minute conversation with each sampled provider.
- Put the quarterly review on the compliance calendar and file this quarter's one-page summary.
