The 30-Chart Internal Audit: Error Rate, Offsets, and What Leadership Should Do
Riverbend Primary Care has 1,200 Medicare Advantage patients. In June the compliance officer audits 30 randomly selected 2024 charts covering 32 submitted HCCs and finds 7 unsupported: 3 rest on 'history of' language (history of lung cancer, history of stroke), 2 rest on 'probable' or 'cannot be ruled out' statements in outpatient notes, and 2 have no mention of the condition anywhere in the 2024 record. The audit also finds 5 documented HCCs that were never submitted, including 2 with clear specialist notes from the same year. The practice's 2024 RAF is 1.21 against a projected 1.08 for its risk profile. Leadership wants to know whether to keep the current query process, whether the 7 findings are all equally serious, and whether the 5 misses offset the 7 errors.
1. Classify the 7 unsupported HCCs by severity and explain what each tells you about the process that produced it.
Read the worked answer
Treat them as three different failures with three different fixes.
- 'History of' codes (3). These are the most serious, because the record affirmatively says the condition is not current. Someone converted a history statement into an active diagnosis, which is fabrication rather than a specificity lapse. Look at who did it: if it came from a vendor coder or a templated note, that is a systemic control failure.
- 'Probable' / 'cannot rule out' in outpatient notes (2). Under ICD-10-CM outpatient guidelines, uncertain diagnoses are not coded as confirmed; you code the signs and symptoms instead. These are typically caused by coders treating inpatient conventions (where 'probable' may be coded) as universal. Fix with targeted education, not discipline.
- No documentation at all (2). The condition appears nowhere in the year. Either the code came from an old problem list, a prior-year note carried forward, or a data-entry error. These indicate a recapture process that is pulling stale data.
Severity matters for the correction plan: the two zero-documentation charts and the three history-of charts suggest the submission file is not being checked against the source at all, while the two 'probable' charts suggest a rule-application error. A single 'retrain everyone' response will fix neither.
2. Do the 5 missed HCCs offset the 7 errors? What should leadership conclude about exposure and remediation?
Read the worked answer
No offset. Unsupported HCCs removed in an audit create overpayment exposure; HCCs the practice failed to capture are a missed-revenue opportunity that does not cancel the overpayment. Regulators do not net the two, and a plan will not accept 'we also under-coded' as a defense. Report them as two separate lines.
Exposure. A 7-in-30 chart error rate is roughly 23% of sampled charts, which is high enough that a payer or CMS contractor could reasonably extrapolate a sample finding across the population rather than accept a chart-by-chart correction. Extrapolation policy is set by the auditing body, but the practice should plan on the assumption that the sample rate, not the sample count, is what matters. The 1.21 RAF against a 1.08 projection is consistent with over-capture and should be treated as a red flag, not a marketing point.
What leadership should do
- Immediately remove the 7 and re-run the sample with a second reviewer to confirm the classification.
- Add the 5 missed HCCs where the specialist notes are qualifying 2024 encounters - these are legitimate.
- Replace annual retrospective 'add HCCs' campaigns with pre-submission validation: every submitted HCC must trace to a dated encounter note by a qualifying provider.
- Audit again in 90 days on a new sample; improving a process you cannot measure is guesswork.