Clinical coding audits

The clinical coding audit process: a practical guide

The clinical coding audit process works like this: an auditor selects a sample of admitted patient records, re-codes each one from the full medical record, compares the new codes with the codes originally assigned, then classifies and weights the errors, feeds the results back to the coders, has the codes amended once auditor and coder agree and reports the findings. The eight steps come from one published Queensland guide from 2017, not from a national standard.

Diagnosis related groups (DRGs), which are used to calculate public hospital funding on an activity basis, are assigned largely from the codes clinical coders allocate, together with other routinely collected variables such as age and length of stay (Independent Health and Aged Care Pricing Authority (IHACPA, formerly IHPA), 2026). When a coder assigns the wrong code, the hospital can be funded on the wrong DRG. Audits are how hospitals check that coded data still matches the records behind it, and the wider context sits in our guide to clinical coding in Australia.

In a coding audit, selected records are independently re-coded and the resulting codes are compared with the codes originally assigned (Australian Institute of Health and Welfare (AIHW), 2026). The auditor does not accept the coded data at face value: they go back to the full record, code it again, and measure the distance between the two versions.

Australia has no national audit method to follow. The AIHW states that there are no national standards for this auditing, so audit results cannot be used to make quantitative assessments of coded data quality on a national basis (AIHW, 2026). Because the auditing is not standardised, no national error rate can be compiled from the results. Who audits what around the country is covered in what a clinical coding audit is and who runs one.

The eight steps are to identify the scope of the audit, choose the sample, schedule the time, notify the coders, classify errors consistently, give feedback, amend the codes and report the findings. They come from Suggestions for Quality Assurance, a 2017 guide from Queensland’s Coding Audit and Education Committee (QCAEC), hosted by the Clinical Coders’ Society of Australia (CCSA). It is one published state guide with a review date that has passed, not a current national standard, so treat it as a starting point to adapt to your service’s own audit protocol rather than as a set of rules.

1. Scope. Identify the scope or extent of each episode of care you will audit. Not every code always needs auditing, so decide deliberately which codes the audit will check (QCAEC, 2017).
2. Audit sample. Choose the records to review; a low-volume targeted audit may cover 100% of cases (QCAEC, 2017).
3. Time. Schedule the audit, the coder feedback and the report writing; audits finish on time, and with greater accuracy, when time is scheduled in advance (QCAEC, 2017).
4. Notification. Tell coders that an audit is under way, what type of audit it is and how feedback will be given (QCAEC, 2017).
5. Consistency. Classify errors consistently, using documented error types and reasons (QCAEC, 2017).
6. Feedback. Give coders constructive, evidence-based feedback, and pass relevant documentation findings back to clinicians (QCAEC, 2017).
7. Amending. Amend the codes in the audited records once agreement on code changes has been reached between auditor and coder, preferably by the coder (QCAEC, 2017).
8. Reporting. Produce a report covering the audit findings, including areas of good practice, analysis of errors and recommendations (QCAEC, 2017).

Two obligations run through every step. For auditors, the source document made available must be the full medical record, paper, electronic or both, because information in the record is the standard against which data quality is measured (QCAEC, 2017). And the auditor must provide evidence for why a code is incorrect, rather than simply asserting that it is (QCAEC, 2017).

Disagreements are handled in two different ways. Differences in clinical interpretation between coder and auditor are best resolved with a documentation query to the treating clinician. Disputes about coding rules go to arbitration, and to prevent bias the independent third party is not told the auditor’s or coder’s opinion (QCAEC, 2017).

Sampling decides which records the audit reviews, and the method depends on the audit’s purpose. For most audits a simple random sample is sufficient, the guide says, but stratified sampling may be more useful for some audits, for example sampling across clinical specialities when the audit examines an individual coder’s work (QCAEC, 2017).

Sample size follows the same logic. Larger samples give more confidence in the outcome being assessed (QCAEC, 2017). Small, targeted internal audits, which the guide says remain the domain of hospital-based coding services, suit focused questions and double as coder education. External audits suit larger samples that aim to measure overall coding service quality; because they are independent and structured, they reduce bias and allow benchmarking over time (QCAEC, 2017). Targeted audits, the guide adds, cannot provide a definitive measure of overall coding service quality (QCAEC, 2017).

The guide works through this part of the process in four moves: assess the coded data against three quality criteria, apply a set of error criteria to each audited record, weight every error found as Type 1, Type 2 or Type 3, then combine the weighted errors into an error discrepancy rate (QCAEC, 2017).

Which errors count in a coding audit

Three quality criteria apply to all audit types in the guide: accuracy, consistency and completeness of the coded data (QCAEC, 2017). In plain terms, the audit asks whether the codes are right, whether similar episodes are coded the same way, and whether anything that should have been coded has been left out.

Its example audit framework applies nine error criteria: the wrong condition selected as the principal diagnosis under ACS 0001, and diagnosis or procedure codes that are incorrect, missing, overcoded or unsubstantiated (QCAEC, 2017). An overcoded additional diagnosis is one that did not meet the ACS 0002 criteria; an overcoded procedure is one inherent in another procedure code (QCAEC, 2017). The guide records sequencing errors within these criteria rather than as a separate criterion, with the cause recorded as an ICD-10-AM convention or an ACS failure (QCAEC, 2017). Alongside the nine criteria, the guide reviews procedure date, the Condition Onset Flag and supplementary codes for accuracy and reports them separately; the Condition Onset Flag matters because hospital-acquired complications are identified from it (QCAEC, 2017).

Every error also gets a cause. The guide classifies causes as an ACS or coding rule failure, an ICD-10-AM convention or index/tabular failure, a documentation problem, or an abstraction error (QCAEC, 2017).

How errors are weighted: Type 1 to Type 3

Each error is then weighted Type 1 to 3, and the weighting decides whether it enters the error rate:

Weighting an error Type 1, Type 2 or Type 3, the guide states plainly, is a subjective task at the discretion of the auditor (QCAEC, 2017).

The error discrepancy formula

The rate itself is calculated with the formula the guide publishes:

In plain words, this is the error discrepancy rate: count the incorrect codes weighted Type 1 or Type 2, divide by the total auditor codes (the correct codes identified by the auditor and agreed to by the coder), and multiply by 100. Type 3 errors, most supplementary codes and external cause and morphology codes repeated because of Queensland’s QHAPDC sequencing requirements are excluded from the count. The guide’s percentages are an approximation and its example sample sizes are not statistically significant (QCAEC, 2017).

Documentation queries sit outside this count. Where the auditor notes that a documentation query might have assisted code selection, the query is noted, discussed with the coder and reported separately; it is not counted as an error (QCAEC, 2017).

The DRG change rate is a separate measure

A DRG change rate counts something different from the error discrepancy rate: audited episodes whose DRG changed after re-coding. In its national hospital morbidity data quality statement, the AIHW records that Victoria’s state-wide DRG change rate for audited records has been approximately 5%, which the state describes as indicating high-quality coded data, and quotes South Australia’s 2011 audit, which found a rate marginally above 10% for metropolitan hospitals (AIHW, 2026). Both figures are state statements with limited scope, not national benchmarks, and there is no national DRG change target.

Audit reports, in the guide, should contain the findings, including areas of good practice, analysis of errors and recommendations (QCAEC, 2017). Where the audit found DRG changes, the guide is direct: reports must explain the impact, because a DRG change can mean revenue loss or revenue gain for the hospital (QCAEC, 2017). How coded data shapes funding is covered in how clinical coding affects hospital funding.

The guide states that coded data will not improve until feedback is given, and that the feedback must be constructive and evidence based (QCAEC, 2017). Documentation findings can flow back to clinicians (QCAEC, 2017). Documentation is covered in more depth in clinical documentation and coding accuracy.

Education is the other lever. Significant audit findings on one or two coding concepts can be addressed in a structured education session with clinical coders (QCAEC, 2017). Individual coders’ audit results or performance should only be discussed with supervisors and other authorised recipients, a confidentiality rule worth knowing before you take part (QCAEC, 2017).

One expectation to set: auditing is senior work. CCSA’s position statement lists conducting coding and related audits among the higher-level duties normally undertaken by an advanced or senior clinical coder (CCSA, 2021).

This checklist is built only from the steps and criteria published in the QCAEC guide (2017). Most of the work belongs to the auditor; the coder’s part is to be notified, amend the code changes agreed with the auditor and take the feedback. Adapt it to your own service’s audit protocol.