Written by Amanda McGurk · Published 12 November 2026 · Last reviewed 12 November 2026
Periodic coding audits measure CDI effectiveness when each finding is classified as a documentation gap, a query problem, or a code-selection error, then routed to a compliant query, physician education, or coder feedback. The next audit checks the same category. Fewer repeat findings show the program is working.
| Figure | Number | What it means for a CDI director |
|---|---|---|
| Hospital IPPS improper payments tied to medical necessity (CMS FY2025, Table J4) | 47.6% | Clinical review of medical necessity belongs in the audit, not only code checks |
| Hospital IPPS improper payments tied to insufficient documentation (Table J4) | 29.0% | The record did not support the claim. That is CDI's lane |
| Hospital IPPS improper payments tied to incorrect coding (Table J4) | 19.5% | Code selection is real exposure, but the smallest of the three |
| Hospital IPPS improper payment rate (CMS FY2025, Table J4, All Provider Types) | 3.9% | The unadjusted rate the error-type split above applies to |
| ACDIS 2025 respondents who do not track their query rate | 10.33% (4.82% in 2024) | Some programs cannot yet show the first number a CFO asks for |
| Same audit scored per code vs per record (AHIMA archive example) | 97% vs 83.3% | Write down your scoring method before you set a target |
CMS published the hospital figures in its 2025 Medicare FFS Supplemental Improper Payment Data . The query-rate figures come from the ACDIS 2025 CDI Week Industry Overview Survey .
What does it mean to measure CDI effectiveness with coding audits?
A periodic coding audit is a scheduled review of coded records against the documentation and the official coding guidelines. Measuring CDI effectiveness with those audits means tracking whether the documentation problems the audit finds get smaller over time, in the categories CDI was asked to fix.
If you need the basics of the program itself, start with our guide to the clinical documentation improvement (CDI) program .
Why does a recoding-only audit miss most hospital exposure?
A recoding audit asks one question: did the coder pick the code the record supports? CMS's improper payment program asks a wider one: does the record support the payment at all?
For hospital inpatient claims paid under IPPS, CMS's FY2025 CERT data split improper payments this way: medical necessity 47.6%, insufficient documentation 29.0%, and incorrect coding 19.5% (Table J4 of the 2025 supplemental data ).
Separately, CMS put the adjusted Part A hospital IPPS improper payment rate at 3.1% in Table A4 of the same report.
Across all Medicare fee-for-service claims, hospital inpatient included, the split looks different. In Table A3, insufficient documentation is 53.0% of improper payments and incorrect coding is 11.1%. Our breakdown of what FY2025 CERT data says about Medicare coding errors covers the service-level tables.
The hospital split carries a direct lesson for audit design. Medical necessity and documentation outweigh code selection. An audit that only rescores codes will report a healthy accuracy number while the larger exposure sits untouched.
HHS OIG makes the same point in its voluntary General Compliance Program Guidance (November 2023). It says claims reviews and audits should include a review of medical necessity by an appropriately credentialed clinician, and that entities that skip it may miss important compliance concerns.
What is the difference between a coding audit and a CDI audit?
A coding audit checks whether the assigned codes match the documentation and the coding guidelines. It scores the coder.
A CDI audit checks whether CDI found the right documentation opportunities and wrote compliant queries. It scores the query process and the documentation it produced.
Run both on the same records and you can see where the chain broke. A missed diagnosis that was documented points to coding. A diagnosis that was never documented, despite clinical indicators in the chart, points to CDI.
Which audit finding goes to which owner?
Classify every finding by root cause, give it one owner, and name the check the next audit will run.
| Audit finding | Root cause | Owner | Action | Re-audit check |
|---|---|---|---|---|
| Diagnosis documented but not coded | Coding | Coding manager | Coder feedback and education, not a physician query | Same diagnosis and DRG family next cycle |
| Clinical indicators present, diagnosis never documented | Documentation gap | CDI | Compliant, non-leading query and a physician education topic | Query rate and agreement rate for that topic |
| Diagnosis coded without clinical support | Clinical validation | CDI and physician advisor | Clinical validation query or review | DRG change rate on that diagnosis |
| Query is leading, lacks clinical indicators, or shows impact in its title | Query compliance | CDI educator | Retrain against the AHIMA-ACDIS query guidelines | Query compliance audit |
| Present-on-admission indicator wrong | Process | HIM and CDI together | Joint process review | POA change rate on audited records |
| Service may not meet medical necessity | Medical necessity | Clinician reviewer | Clinical review, per OIG guidance | Same service line next cycle |
For the clinical validation row, our article on clinical validation challenges walks through the review itself.
The query compliance row has a written standard. The AHIMA and ACDIS Guidelines for Achieving a Compliant Query Practice (2022 Update) , published with a 2023 addendum, call for a non-leading query statement, clinical indicators specific to the patient and episode of care, and provider-visible titles that leave out impact such as reimbursement or quality indicators.
Two routing rules keep physician queries for the cases that need them:
- A diagnosis the physician already wrote down is a coding miss. Route it to coder feedback, not to the physician.
- Give every finding an owner and a re-audit check, or nobody can tell next quarter whether it was fixed.
Which CDI metrics show the loop is working?
No single number proves a CDI program works. AHIMA's CDI Key Performance Indicators Practice Brief (August 2022) gives standard equations for the core set. They are industry definitions, not MediCodio AI benchmarks. Write your own definitions into policy.
| Metric | Equation (AHIMA 2022 unless noted) | What it can and cannot prove |
|---|---|---|
| Review rate | Records reviewed / records assigned for review | Coverage, not quality |
| Query rate | Records that required a query / records reviewed | AHIMA says there is no industry standard for it |
| Response rate | Queries with a response / queries sent, in a set time frame | Provider participation |
| Agreement rate | Queries meeting your "agreed" definition / queries sent | No industry standard; a rate stuck near 100 percent is a warning sign |
| CC/MCC capture rate | Cases in DRGs titled "with CC", "with MCC" or "with CC/MCC" / total cases | Secondary diagnosis capture, after your defined exclusions |
| Case mix index | Sum of MS-DRG weights / number of MS-DRGs | Moves with surgeon schedules and service mix, not only CDI |
| MS-DRG reconciliation rate | Reconciled DRGs / DRGs reviewed | Whether CDI and coding land on the same DRG |
| Denial rate by type | Claims denied for a set reason / claims submitted in the period | Downstream outcome; define the reason categories first |
| Audit DRG change rate (audit convention) | Audited records where the auditor changed the MS-DRG / audited records | The audit's own outcome; define it in policy |
Two cautions from the AHIMA brief matter most for a CDI director. First, an agreement rate that stays at 100 percent over time may mean providers are not fully reading queries, and AHIMA suggests a compliance audit of the query response process. Second, many variables move case mix index, so a CMI change alone cannot carry the claim that CDI worked.
Many programs are still building this measurement base. In the ACDIS 2025 survey of 783 respondents, 10.33% said they do not track query rate, up from 4.82% in 2024. Asked how many of their inpatient claims are denied, 75.15% said they do not know. For provider education, the most common measure of success was improvement in CDI metrics, at 79.59%.
These are self-reported survey bands. They describe a measurement gap inside CDI programs, not a national rate.
How often should coding audits run, and what accuracy target is right?
No CMS or OIG rule sets a coding audit frequency or sample size. OIG's guidance says compliance risk assessments should be conducted at least annually. The Compliance Committee should then put a schedule of audits in the compliance work plan, based on the risks that assessment finds. OIG adds that internal or external auditors may do the work if they have expertise in federal and state health care program requirements.
Many hospitals layer monthly or quarterly internal reviews on top, plus a focused re-audit after each education round. For how many charts to pull, use our guide to how many charts a medical coding audit should review .
On accuracy, AAPC has written that "there is no federally sanctioned threshold of accuracy below 100 percent" ( AAPC, 1 November 2021 ). Any target you set is an internal policy choice.
The scoring method changes the answer. In an archived Journal of AHIMA article , Sarah Humbert, RHIA, describes a hypothetical department that requires coders to hold 95 percent accuracy. The same coder scored 97% per code (70 of 72 codes correct) and 83.3% per record (5 of 6 records correct). The coder passes under the per-code method and falls short under the per-record method.
Pick the method that matches what you need to know, then keep it fixed so trend lines stay comparable.
How does PEPPER help choose what to audit?
PEPPER, the Program for Evaluating Payment Patterns Electronic Report, summarizes a hospital's Medicare data for target areas often associated with improper payments due to billing, DRG coding or admission necessity issues. It compares your statistics with hospitals in the nation, your MAC jurisdiction and your state.
The AHIMA brief notes that PEPPER treats the 80th percentile as the high outlier threshold and the 20th percentile as the low one. A high or low outlier is a reason to audit that target area. It is not proof of an error, and PEPPER does not set your sample size.
How does MediCodio AI support the audit side of this loop?
The loop depends on auditors and CDI teams seeing why a code was assigned. MediCodio AI is built for that view.
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.
Every code ships with its documentation passage and the compliance rule behind it. That is a 100% code-level audit trail, so a reviewer can trace a disputed code back to the sentence in the chart. Nothing runs autonomous on day one. Everything starts in review.
If you want an outside team to run the audit and route findings, see our medical coding audit services . Hospital teams can also review how we work with hospitals .
What has not changed?
Audits do not make a claim payable, and no tool removes the need for clinical judgment. Compliant queries, clinical validation and medical-necessity review still need trained people.
A working loop changes where those people spend their time. Coders get feedback on coding misses, physicians get queries only where a query is warranted, and the next audit tells you whether the fix held.
Frequently asked questions
How do periodic coding audits improve CDI effectiveness?
They turn audit errors into targeted fixes. Each finding is classified as a documentation gap, a query problem, or a coding error, then routed to a compliant query, physician education, or coder feedback. The next audit checks the same category, so improvement shows up as fewer repeat findings in that area.
How often should a hospital run coding audits?
No CMS or OIG rule sets a frequency. OIG's voluntary 2023 General Compliance Program Guidance says to run a compliance risk assessment at least annually and to schedule audits based on the risks it finds. Many hospitals add monthly or quarterly internal reviews and focused re-audits after education.
How do you calculate query agreement rate?
Divide the number of queries that meet your definition of an agreed query by the number of queries sent. AHIMA says there is no industry standard for agreement rate. A rate that stays at 100 percent may mean providers are not fully reading queries, which calls for a review.
What is a good coding accuracy rate for a hospital?
There is no federally sanctioned accuracy threshold below 100 percent, according to AAPC. Hospitals set internal targets. State whether you score per code or per record, because one audit can pass a coder under the first method and fail the same coder under the second.
What is the difference between a coding audit and a CDI audit?
A coding audit checks whether assigned codes match the documentation and coding guidelines. A CDI audit checks whether CDI found the right documentation opportunities and wrote compliant, non-leading queries with clinical indicators. Running both on the same records shows whether a problem started in documentation or in coding.
Book a Demo to run an accuracy review on your own charts and see the documentation passage behind every code.
Sources
- CMS, 2025 Medicare FFS Supplemental Improper Payment Data (Tables A3, A4, J4) .
- CMS, Fiscal Year 2025 Improper Payments Fact Sheet .
- HHS OIG, General Compliance Program Guidance, November 2023 .
- AHIMA, Clinical Documentation Integrity Key Performance Indicators Practice Brief, August 2022 .
- AHIMA and ACDIS, Guidelines for Achieving a Compliant Query Practice (2022 Update, with 2023 addendum) .
- ACDIS, 2025 CDI Week Industry Overview Survey .
- AAPC, The Art of Setting E/M Pass Rate Thresholds, 1 November 2021 .
- Journal of AHIMA archive, How to Choose the Right Coding Audit Method (Sarah Humbert) .
- PEPPER, About .