What does FY2025 CERT data say about Medicare coding errors?

Written by Amanda McGurk · Published 9 November 2026 · Last reviewed 9 November 2026

CMS's FY2025 CERT data attributes 11.1% of Medicare fee-for-service improper payments, about $3.2 billion, to incorrect coding. Insufficient documentation accounts for 53.0%, about $15.3 billion. Coding errors cluster in E/M visits, and they run in both directions, so audits should test code accuracy and record support separately.

Sources: CMS FY2025 Improper Payments Fact Sheet for the national rate, and the CMS 2025 Medicare FFS Supplemental Improper Payment Data report (Tables A3, A5, F4 and F5) for the category and service figures.

What is CERT, and what counts as an improper payment?

The Comprehensive Error Rate Testing (CERT) program is how CMS measures improper payments in Medicare fee-for-service. According to Chapter 12 of the Medicare Program Integrity Manual , CERT reviews a stratified random sample of claims to decide whether Medicare paid or denied them correctly under coverage, coding and billing rules.

CMS counts any payment that should have been denied, or that was paid in the wrong amount, as improper. That includes overpayments and underpayments. The CMS fact sheet states that the improper payment rate is not a measure of fraud. A claim can fail CERT because a chart was never sent, even when the care was appropriate and the code was right.

The FY2025 rate covers claims submitted from July 1, 2023 through June 30, 2024, per the supplemental report . The report year and the claim year differ, so date your internal comparisons with care.

Why does a coding-only audit miss most CERT findings?

Picture a hospital HIM director who runs a quarterly coding audit. The audit recodes a sample of charts, reports an accuracy figure and sends education to the coders who missed. That program tests whether the coder picked the code the record supports.

CERT tests whether the record supports the payment. CMS sorts every error into one of five categories: no documentation, insufficient documentation, medical necessity, incorrect coding, and other. In the FY2025 adjusted figures, incorrect coding is 11.1% of improper payment dollars. Insufficient documentation is 53.0%, and medical necessity is 15.3%.

An audit built only around recoding can report strong coder accuracy while the organization still carries most of the exposure CERT measures. If a CERT contractor requests a chart and the signed order sits in a referring clinic's file, a clean recoding result will not save that payment. The missing pieces sit in signatures, orders, certifications and records held by other providers. Those belong to HIM, clinical documentation and compliance, and they need a test of their own. Our guide on why correctly coded claims still get denied covers the same gap from the payer side.

Where do CERT coding errors concentrate in FY2025?

Table F4 of the supplemental report lists the 20 service types with the most incorrect-coding improper payments. E/M visits hold 7 of the top 10 spots.

Source: CMS 2025 Medicare FFS Supplemental Improper Payment Data, Table F4.

The table supports two readings. Established office visits carry the most dollars at a 3.5% rate because Medicare pays for so many of them. Emergency room and critical care visits carry higher rates on smaller dollar bases. Your audit plan needs both views. Dollars show where exposure sits, and rates show where coders and providers miss the supported level most often.

Inpatient DRG lines also appear in Table F4, including simple pneumonia and pleurisy (DRGs 193, 194, 195) at $44,829,382 and GI hemorrhage (DRGs 377, 378, 379) at $37,725,102. Hospital leaders should read those rows next to the setting view in Table A5, where acute inpatient hospitals show about $0.9 billion in incorrect coding and $2.9 billion in medical necessity errors.

Is incorrect coding the same as insufficient documentation?

No. Incorrect coding means the record supports a different code, level, unit or DRG than the one billed. Insufficient documentation means the reviewer cannot tell from the record whether the service was provided, provided at the level billed, or medically necessary. CMS gives that definition in MLN909160 .

The remedy differs. An incorrect-coding finding goes to coder and provider education. An insufficient-documentation finding goes to record retrieval, authentication and documentation practice. MLN909160 lists incomplete progress notes, missing signatures, illegible signatures without a signature log or attestation, and missing orders among the common insufficient-documentation errors. It also says the billing provider must obtain supporting records from referring physicians and facilities.

Is CERT only looking for overcoding?

No. CERT counts downcoding as an improper payment, and Table F5 projects $891,887,256 in downcoding errors across services. For established office visits, Table 7 of the report lists 120 sampled claims where documentation supported a lower E/M level than billed, and 7 where it supported a higher level. A defensible audit tests both directions and reports them separately. For the billing-side view, see is downcoding fraud in medical billing? .

What does a diagnosis validation failure look like?

The HHS Office of Inspector General audit A-03-17-00010 reviewed a random sample of 200 inpatient claims carrying severe malnutrition codes E41 or E43. Hospitals billed 27 correctly. For 164 claims, the record supported another form of malnutrition or none, producing $914,128 in net overpayments and an estimated $1 billion across FY2016 and FY2017. CMS noted the estimate represented less than .5 percent of inpatient payments in that period, so treat it as a targeted case, not a national DRG error rate. For more on this failure mode, see our guide to clinical validation challenges .

How should you build an internal audit that mirrors CERT categories?

Borrow the CERT structure without claiming CERT precision. The supplemental report describes a service-level stratification plan with about 100 strata per claim type, plus sub-strata for high-payment services with uneven provider error rates, such as home health, SNF and hospice. Your audit will be smaller, and its results will not equal a national rate.

1. Define the population. Use paid Medicare FFS claims for a closed period and keep the original claim version, codes, units, modifiers and payment.
2. Stratify by your own service mix. Separate E/M levels, inpatient DRGs, outpatient procedures and any SNF or DME lines you bill. Do not size strata from CERT's national shares, because those are dollar shares, not claim counts.
3. Request the full record first. Pull orders, signatures, attestations and third-party records before a coder recodes. A missing order is a documentation finding, not a coding finding.
4. Classify each error into one CERT category. Record one primary category and one operational root cause, such as retrieval, authentication or code selection.
5. Recode and reprice. For incorrect-coding findings, capture the supported code and the payment difference in both directions.
6. Route the fix by owner. Coding owns code selection, HIM owns record completeness, and clinical or compliance reviewers resolve necessity and diagnosis support questions.

Our post on medical coding audit sample size covers how many charts to pull. If an audit points to Medicare overpayments, involve compliance and qualified statistical support before any projection. The OIG report above cites the 60-day rule for identifying, reporting and returning overpayments.

Where does MediCodio AI fit in a CERT-informed audit program?

MediCodio AI applies AI medical coding to the coding half of this problem and gives HIM a clean record of the evidence behind each coding decision. Nothing runs autonomous on day one. Everything starts in review, and the autonomous lane widens one specialty or case type at a time as your auditors confirm first-pass accuracy on your charts. 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.

That routing matters for CERT categories. A chart with a missing order or an unsigned note is a documentation question, and it belongs with a person who can fix the record. MediCodio AI keeps a 100% code-level audit trail. Every code ships with its documentation passage and the compliance rule behind it. When a CERT request or an internal auditor asks why a level was billed, your team opens the code and reads the supporting passage instead of rebuilding the logic from memory.

MediCodio AI reports 98%+ coding accuracy as 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. The platform connects to the EHR you already run by API or RPA. Learn more about our medical coding audit services and how the code-level audit trail works for hospitals and health systems .

What has not changed?

The documentation standard has not moved. CMS pays when the record supports coverage, coding and billing requirements, and MLN909160 says CMS may treat paid claims without sufficient documentation as overpayments. The FY2025 national rate fell from 7.66% to 6.55%, yet the category mix still places most improper payment dollars on documentation rather than code choice.

Keep the FY2025 figures labeled by year. CMS publishes a new supplemental report each reporting cycle, and the service rankings shift between years.

Frequently asked questions

Can we use the 6.55% CERT rate to estimate our own coding exposure?

No. The 6.55% figure is the national FY2025 Medicare FFS improper payment rate across all five error categories. It is not a provider-level coding liability. Estimate your own exposure from your paid claims, your service mix and the payment difference your auditors find, using a sampling design your compliance team approves.

Which CERT error category is the largest?

Insufficient documentation is the largest category in the FY2025 adjusted data, at 53.0% of improper payments, or about $15.3 billion. Medical necessity follows at 15.3%, no documentation at 12.0%, and incorrect coding at 11.1%. Your own mix depends on your services, so classify internal findings with the same categories.

Which services have the most CERT coding errors?

In Table F4 of the FY2025 supplemental report, established office visits carry the most incorrect-coding dollars at $556,500,710. Subsequent hospital visits follow at $361,156,694. Emergency room visits show a higher coding rate, 9.7%, on a smaller dollar base. E/M services fill most of the top of the list.

Does CERT count undercoding as an error?

Yes. CERT treats underpayments as improper payments. Table F5 projects $891,887,256 in downcoding errors across services. For established office visits, CMS lists 7 sampled claims where documentation supported a higher E/M level than billed. An internal audit should report overcoding and undercoding as separate findings with separate education plans.

Should coding or HIM own CERT readiness?

Both, with separate accountability. Coding should own code selection, units, levels and the payment difference on recoded claims. HIM should own record retrieval, signatures, orders and third-party records, since insufficient documentation drives the largest share of improper payments. Clinical and compliance reviewers should settle medical necessity and diagnosis support questions.

Book a Demo to run an accuracy review on your own charts and see the documentation passage and rule behind every code. Keep CERT responses, overpayment decisions and statistical projections with your compliance team.

Sources

• CMS: Fiscal Year 2025 Improper Payments Fact Sheet, January 15, 2026 .
• CMS: 2025 Medicare Fee-for-Service Supplemental Improper Payment Data .
• CMS: Medicare Program Integrity Manual, Chapter 12, The CERT Program .
• CMS MLN909160: Complying with Medical Record Documentation Requirements, December 2024 .
• HHS OIG: Hospitals Overbilled Medicare $1 Billion by Incorrectly Assigning Severe Malnutrition Diagnosis Codes, A-03-17-00010, July 13, 2020 .