What downcoding means
Downcoding is reporting or paying a code lower than the documentation and the applicable rules support. The word covers two events, and conflating them is where the confusion starts.
Payer downcoding happens after you submit. The insurer pays the lower level, and your route back is the documentation.
Provider undercoding happens before you submit. Your own team reports below what the record carries, and no appeal exists because nobody disputed anything.
Four things are not downcoding. A bundling edit that collapses two codes is the rule working. A coverage restriction is the policy working. Documentation that supports only a lower code produces the accurate code. And an unspecified code is often right: the ICD-10-CM guidelines state that sign, symptom and unspecified codes "have acceptable, even necessary, uses."
Is downcoding fraud?
Not on its own. The statute sets the test, and knowledge is only part of it.
Under 31 U.S.C. 3729(b)(1) , "knowing" and "knowingly" cover a person who "has actual knowledge of the information," "acts in deliberate ignorance of the truth or falsity of the information," or "acts in reckless disregard of the truth or falsity of the information." Those terms "require no proof of specific intent to defraud."
That cuts both ways. A coder who misreads one note has made an error. A department that knows its documentation supports more and reports less anyway is not protected by the absence of bad intent.
Knowledge is one element, not the whole test. The Act reaches a false or fraudulent claim for payment, a false record material to such a claim, or the improper avoidance of an obligation to the government, which the statute extends to retaining an overpayment. It defines "material" as "having a natural tendency to influence, or be capable of influencing, the payment or receipt of money or property." A deliberate pattern is therefore not automatically a violation, and a single knowing and material claim is not automatically outside one.
HHS-OIG has published no guidance calling undercoding fraud. It does list clustering among the billing risk areas it has investigated most often, defining it in the physician practice compliance guidance as coding one or two middle levels of service exclusively, noting that "in reality, this overcharges some patients while undercharging others." A habit that lands below the record appears on that list.
None of this is legal advice. The question for a compliance committee is narrower: could someone show we knew, or avoided knowing, that our codes sat below our records?
Why coding low to be safe still breaks the rules
Provision 4.8 of AHIMA’s Code of Ethics bars a health information management professional from "dishonesty, fraud and abuse, or deception," and lists "allowing patterns of optimizing or minimizing documentation and/or coding to impact payment" and "miscoding to avoid conflict with others" among its examples.
AHIMA names minimizing alongside optimizing, and names coding low to avoid an argument on its own.
For diagnoses, Section I.B.18 of the FY 2027 ICD-10-CM guidelines asks for "the level of certainty known for that encounter." Those guidelines govern diagnosis coding, where HIPAA requires adherence. Visit levels, procedures and separately reportable services answer to their own CPT, CMS and payer rules.
When does conservative coding become a legal risk?
The case. A cardiology service line bills level 4 office visits where the notes support level 5. It started with one coder who lost an appeal and decided to sit a level down. Two others picked up the habit. Nobody wrote the rule down, and no payer disputed a claim that paid.
One misread note is an error. The statute’s knowledge test reaches a standing practice the team can describe, and AHIMA lists "miscoding to avoid conflict with others" among its examples of dishonesty, fraud and abuse.
You will not find it in a denial report, because none of those claims were denied.
Can a provider be penalized for undercoding, not just upcoding?
Yes, though the exposure does not arrive as an overpayment demand.
Undercoding produces an inaccurate claim, the same defect as upcoding pointed the other way, and it feeds inaccurate data into payment programs sitting outside the claim. You end up with a distorted case mix, quality measures benchmarked against a population you did not treat, and reimbursement your documentation earned and nobody claimed.
Undercoding rarely produces a phone call. Your team does not appeal a claim that paid, and no payer writes to say you billed too little. Patterns run for years on that silence.
What it costs besides revenue
A low code costs the claim once. Those same codes then drive payment programs outside the claim, and the readmission penalty is the clearest example.
CMS calculates the Hospital Readmissions Reduction Program penalty from an excess readmission ratio. Its published methodology defines that ratio as the predicted readmission rate divided by the expected rate, both resting on the hospital’s "specific patient case mix."
Coded diagnoses feed that case mix. Underreporting a diagnosis inside the applicable risk-adjustment model can understate complexity and move the expected rate. Not every diagnosis affects the measure, and the models exclude some conditions, so treat this as a reason to code the record accurately rather than a formula for predicting a penalty.
CMS applies those reductions to all Medicare fee-for-service base operating DRG payments during the fiscal year, capped at 3 percent. Becker’s Hospital Review reported that CMS is penalizing 2,334 hospitals, about 80% of those evaluated, in fiscal 2027, with the reductions effective 1 October.
When is a lower code correct?
| Situation | Is the lower code appropriate? | What should happen |
|---|---|---|
| The documentation supports only the lower code | Yes | Report the supported code. This is accurate coding, not downcoding. |
| The documentation supports a higher or more specific code | No | Report the supported code where the applicable reporting requirements are met. |
| The documentation is incomplete, unclear or contradictory | Not automatically | Report what the record currently supports, or query the provider without suggesting the answer. |
| An NCCI edit bundles two services | Yes, where the edit applies | Follow the edit. Use a modifier only where the Correct Coding Modifier Indicator permits one. |
| An MUE flags the reported units | Not automatically | Check the units against the MUE Adjudication Indicator. Do not cut correctly documented units to clear the edit. |
| Payer policy does not cover the service | No | Keep the coding accurate and handle coverage, notice, denial and appeal as separate steps. |
| The coder is uncomfortable with the supported code | No | Recheck the documentation and the governing rule, then escalate to QA or compliance. |
On units, CMS states that where a denial on an MAI 1 or MAI 3 edit is appealed, contractors may pay units above the MUE value given adequate documentation of medical necessity. Trimming correctly documented units to clear an edit gives up a claim CMS says contractors may pay.
How to code to the documentation, and where MediCodio AI fits
CODIO AI reads the whole encounter instead of matching keywords, then assigns ICD-10-CM, ICD-10-PCS, CPT, HCPCS Level II and modifiers, coding toward the documentation with no thumb on the scale in either direction.
You can ask why a chart went low. The Prediction Trail keeps the reasoning behind every assigned code, down to the lines of the note that drove it and the rule applied. That is a 100% code-level audit trail, and it reads the same in both directions.
Thin documentation routes to a person. Complete charts flow through AutoPilot for fully autonomous coding. Charts where the documentation is thin or contradictory stay in CoPilot, where certified coders review and finalize the output.
MediCodio AI reports 98%+ coding accuracy. The basis is first-pass accuracy across deployments since 2023, measured at the code level on production charts before any correction, against a reference review by AAPC- and AHIMA-credentialed coders applying official guidelines and payer policy to the same documentation. Nothing runs autonomous on day one, and the autonomous lane widens as accuracy is proven on your charts, audited by your team. See hospitals and health systems and coding audit services for a read on which direction your own charts drift.
What this does not change
Accurate coding does not settle a compliance question in advance, and no platform can. An auditor can still disagree with a supported code.
The note still bounds the claim. Where the documentation omits laterality, no engine supplies it, and the answer is a query. CDI support sits upstream of that gap.
And a lower code is often correct. Aetna says as much to its own providers : "It is not medically necessary or appropriate to bill a higher level of E&M service when a lower level is reasonable." What the guidelines do not protect is a standing preference for it.
See which way your own charts drift
Book a Demo and we will run an accuracy check on a sample of your own charts, scored at the code level against your coders, with each code shown next to the documentation passage and the rule it was checked against.
Related: AI medical coding · Hospitals and health systems · Coding audit services · CDI support
Sources
1. 31 U.S.C. 3729, False Claims Act .
2. AHIMA, Code of Ethics, provision 4.8 .
3. Cigna Healthcare, Reimbursement Policy R49 , effective 1 July 2025.
4. Aetna, Evaluation and Management (E&M) Program Claim and Code Review .
5. HHS Office of Inspector General, Compliance Program for Individual and Small Group Physician Practices, 65 FR 59434 , 5 October 2000.
6. CMS and NCHS, ICD-10-CM Official Guidelines for Coding and Reporting, FY 2027 .
7. CMS, How to Use the Medicare NCCI Tools, MLN901346 .
8. CMS, Hospital Readmissions Reduction Program Payment Reduction Methodology, FY 2027 .
9. Becker’s Hospital Review, CMS penalizes 2,334 hospitals for high readmissions: 6 things to know , 7 October 2026.