If your site has never conducted a clinical research billing audit, or even if you’ve done several, you might be wondering just how big your sample size should be. What is the typical sample size for an audit?
How large your sample size is will depend on the purpose of your audit. If it’s a classic internal audit looking at controls and processes and you’re not necessarily looking for whether your organization has received an overpayment, five to ten units may be sufficient for you to map the handoffs and processes. This sample can be spread across subjects or studies. Can you do more? Certainly. However, you should be able to get a good look into billing processes with this sample size – especially if spread out across different types of studies. I want to emphasize, however, that this sample would look at processes – not the accuracy of billing. Looking at accuracy of billing puts you into the world of compliance audits.
So, what is the sample size if you’re doing a compliance billing audit?
This is a little different than a process and controls review because your goal here is to determine whether the bill to the subject’s health insurer was accurate. You’re looking at end-result accuracy in the sample in order to assess risk. For this type of review, I typically recommend a sample size of 20-40, which is the standard many government agencies use for a “probe sample.”
In a probe audit, the results don’t usually have enough statistical significance to extrapolate, but the sample is large enough to get a sense of risk. In a probe sample, the government often looks for a “net financial error rate” of more than 5% as an indicator of a need to do a larger, statistically significant sample. If you have ever reviewed an OIG Corporate Integrity Agreement (CIA), you will see a similar concept called a “Discovery Sample,” though a CIA usually sets the Discovery Sample size at 50 units.
What makes up the unit that you are auditing?
This is a big question in a research billing compliance audit. In other words, 20 units of what? As far as I’m aware, there’s no guidance from the government regarding the right approach to selecting the units. If your organization has several hundred studies, you might want to consider a sample of 20 studies, and then audit one or two subjects from each of those studies. Or perhaps you might consider 10 studies and three subjects each. Usually, two to three subjects are enough to learn if something is off within that study and if the research team is communicating the right information. If your organization has a small number of studies, perhaps 30, it is going to be difficult to do a probe sample that is statistically significant, and you might want to consider a systematic review of all your studies over the course of a period of time.
This may not always hold true across the board, but generally, if the study team is doing it wrong for one subject, it signals that the team is not communicating information properly and will tend to replicate the activity for all subjects on that study. Conversely, if they’re doing it right, they tend to do it right for all subjects within that study.
Additional cautionary note
Although it’s possible that the probe audit results are not a sufficient size to be able to extrapolate error rates and overpayments, it is often the case that, within a single study, the audit results for two to three subjects can put an institution on notice of what is going on inside that study. Once an organization is on notice of a systemic error inside one study, there will be obligations to correct the bills for the other government health program subjects that would have had the same services in that study.