Auditing and Data Accuracy
This domain tests whether you can find, quantify and prevent coding errors. Auditing is where the coding rules, the payment model and compliance meet, and the exam expects you to compute and interpret accuracy rather than just describe it.
Audit mechanics
Define the population and draw a random or stratified sample; auditing only suspicious charts biases the result. Report two different numbers: chart-level accuracy (records without error divided by records reviewed) and payment-impact or HCC-change rate (only errors that move the RAF). CMS RADV audits of Medicare Advantage contracts can extrapolate a sample error rate across the whole population for audited payment years from 2018 onward, so a small sample error becomes a large refund.
Worked example
An internal audit reviews 120 randomly selected members. Nine records contain at least one unsupported or misspecified diagnosis, giving a chart-level error rate of 7.5%. Of those nine, four involve a diagnosis that was the only one in its hierarchy, so deleting it lowers the member's RAF; the other five were trumped by a higher-ranked HCC, so payment does not change. The payment-impact rate is 4/120 = 3.3%. Under RADV-style extrapolation across a 20,000-member contract, that 3.3% is the figure that creates financial exposure - and the figure the corrective action plan must attack first.
Root cause and corrective action
- Trend findings by cause: copy-forward, unspecified codes, unclosed queries, missing CKD stages, unlinked conditions.
- Feed findings back at the code level to the specific providers and coders involved.
- Decide between prospective review before submission and retrospective review after payment.
- Retract or delete unsupported diagnoses before the submission deadline and confirm the correction posted.
Traps that cost marks
- Counting every documentation discrepancy as a payment error.
- Ignoring hierarchy when deciding whether an error changed payment.
- Auditing a convenience sample and treating the result as a contract-wide rate.
- Writing a corrective action plan that is not measurable or not re-audited.
- Assuming a correction was submitted when no confirmation exists.