Audit Planning, Sampling, and Risk Assessment
Audit Planning, Sampling, and Risk Assessment tests how you choose what to audit, how much to review, and how to defend the resulting conclusions.
Core rules
- Risk assessment should be documented before sampling. Consider claim volume and dollars, high-risk codes, new providers or services, prior findings, denial patterns, payer mix, contract changes, and regulatory priorities.
- Define the population and sampling frame precisely. A sample drawn from paid claims only cannot support conclusions about denied claims.
- Random selection supports statistical inference; stratified sampling improves coverage of high-risk groups; judgmental sampling is useful for investigation but does not support extrapolation.
- Sample size depends on confidence, precision, expected error rate, and population size. Document the method and preserve the random-number seed or selection trail.
- Extrapolate only when the sample design and policy permit it. Report stratified results stratum by stratum, not as one blended rate.
Worked example
A group has 5,000 claims. The auditor stratifies by modifier 25 claims with same-day procedures, a high-risk group of 400 claims, and randomly selects 50. Six fail, a 12% stratum error rate. The auditor reports the finding for that stratum and does not project 12% to all 5,000 claims, because the sample was not proportional and was not designed for whole-population extrapolation. A separate random sample would be needed for that conclusion.
Traps
- Choosing the first 25 charts because they are easy to pull.
- Changing the sample after seeing poor results.
- Using a sample size with no rationale and calling it statistical.
- Ignoring denied claims, underpayments, and non-claims data.
- Failing to document the population, random selection, or exclusions.