A medical coding audit is a measurement, not an accusation. The point is to find out whether your billed codes match your documentation before a payer, a Medicare contractor or the OIG decides to find out for you. Done right, it also finds the visits you have been under-billing, which is the more common result.
What a coding audit covers
A sample of charts is read against the claims that were submitted for them: the CPT and HCPCS codes, the units, the modifiers, the diagnosis codes and the place of service. Findings are reported in both directions, codes billed above what the documentation supports and codes billed below it. Under-coding is the more common finding in small practices and the one that quietly costs the most money, usually a provider billing 99213 for visits whose MDM supports 99214. Over-coding is rarer and more dangerous, because it is what a payer audit is looking for.
E/M level distribution
Your distribution of evaluation and management levels is compared against your own documentation under the 2021 office visit rules, where the level rests on medical decision making or total time and nothing else. A distribution skewed hard to one level is worth understanding whichever direction it leans, because it usually reflects a habit or a template rather than the actual case mix. A provider whose visits are 90 percent 99213 in a panel full of diabetics on insulin is under-coding. A provider whose 99215 rate is triple the rest of the group needs a note-by-note look before someone else takes one.
Modifiers, units and edit pairs
Every modifier in the sample is checked against the note and the NCCI table for that date of service. Modifier 25 needs a separately identifiable E/M in the record, not just a procedure note with vitals attached. Modifier 59 needs a distinct site, session or encounter, and the payer will want to see which. Units are checked against the MUE limit, so a J3301 billed at 4 units carries a dosage that supports 40 mg. Imaging split with 26 and TC is checked against who owned the equipment. Where a pattern appears, we pull the CO-4 and CO-97 denials from the same period to show whether the payers have already noticed.
How a coding audit is scoped and run
Scoping fixes four things in writing: which providers, which date range, which service lines and how many charts per provider. The sample is drawn at random from the claims in that range, not chosen by the practice, because a hand-picked sample measures your best work rather than your typical work. Each chart is read by a credentialed coder against the codes, modifiers and units that were billed for it, using the documentation guidelines in force on that date of service. Every variance is recorded with the chart reference, the code billed, the code supported and the reason. Findings are written up, checked by a second coder and presented to you on a call before the report is final.
Baseline, focused and follow-up audits
A baseline audit reads a sample across all providers and service lines to find out where you stand before anything changes. It is also what the OIG compliance guidance for small physician practices asks for, and its suggested starting point of five to ten records per physician is smaller than most practices expect. A focused audit narrows to one thing, such as one provider's E/M levels or one procedure that has been attracting denials, and goes deeper on it. A follow-up audit repeats a previous sample design a few months after a change so you can see whether the change actually took. Practices moving their medical coding to us usually run a baseline first, because it separates what the old process was doing from what the new one does. A focused audit is often the right response to a denial management report showing a single modifier or code recurring.
What separates a useful coding audit from a score
A useful audit tells you which charts, which pattern and what to change, in that order. It reports under-coding and over-coding with equal weight, because an audit that only finds one direction is answering the question it was asked rather than the question that matters. It quotes the documentation guideline each finding rests on, so a provider can check it rather than take it on trust. It distinguishes documentation problems, where the work was done but not recorded, from coding problems, where the record was read wrongly, because the fixes are different: a template change for the first, coder training for the second. And it ends with a remediation plan that names the training, template or workflow change that addresses each pattern, with a date for the follow-up sample. A single accuracy percentage with none of that behind it is not something you can act on.
What Is Included
- Random chart sample, agreed per provider and date range
- E/M level distribution against the 2021 MDM and time rules
- Modifier 25, 59 and 26/TC use checked against NCCI edits
- Under-coding and over-coding reported with equal weight
- Written findings, chart references and remediation plan
| Edit standard applied | NCCI procedure-to-procedure pairs and medically unlikely edits, updated quarterlySource: CMS, National Correct Coding Initiative |
|---|---|
| OIG guidance on physician self-audits | The OIG compliance guidance for small physician practices recommends a baseline audit and suggests five to ten records per physician as a starting sampleSource: HHS OIG, Compliance Program Guidance for Individual and Small Group Physician Practices |
| Modifier-related denial code | CARC 4: the procedure code is inconsistent with the modifier usedSource: X12, Claim Adjustment Reason Codes |
Coding & E/M Audits: Common Questions
Sample size is agreed up front as part of scoping and depends on provider count and how many service lines you want covered. Ten to twenty charts per provider is a common baseline, drawn at random from a defined date range.
No. Audits are most useful as a baseline before there is a problem, and periodically afterwards to confirm changes actually took. The OIG guidance for small practices recommends exactly that pattern.
You get told plainly, along with what the exposure looks like and what to correct going forward. Deciding what to do about past claims, including any refund, is a conversation for you and your compliance advisor.
You decide. Findings are delivered to the practice owner or administrator you name, under the signed BAA, and are not shared with anyone else. Many practices ask for a per-provider version so each provider sees only their own charts.
It depends on sample size and how quickly we get chart access. The timeline is agreed at scoping along with the sample, and the presentation call is scheduled at the same time so the findings do not sit unread.
Yes. A coding audit is a fixed-scope project and does not require any change to who does your billing or coding. Some practices audit annually and never move anything else.