We work very closely in the payor and provider space to ensure that fraud, waste, abuse (FWA) and compliance are at the forefront of the payment integrity continuum. This is particularly the case when working to conduct medical record reviews and identify over and under payments.
When I was an OIG agent, I would often be the recipient of referrals from various Unified Program Integrity Contractors (UPICs) and Special Investigative Units (SIU), where overpayments were identified as either actual or extrapolated. I was always amazed, however, that there were no credits given to identify those instances where an underpayment was identified. Fully understanding that the goal of an SIU and the UPIC is to ferret out FWA, there must be equity in those analysis. One of the areas where there is always a fair amount of discussion is on sampling of medical records.
I am not a statistician and can only rely upon my one semester in graduate school in research methods, and my insatiable interest in learning about new things. One of the biggest areas of appeal and challenge in the UPIC world is the sampling methodology. As was mentioned several times in our OIG Roundtable podcasts, Matt Kochanski noted that appeals are often tied to the sampling. It is always important to ensure that methodologies are tested and backed by the science of statistics. While the goal of the sampling is to limit the amount of effort that is needed to identify the overpayment, the methodology must be such that it is also defensible. While it is possible to create a sample that is a “cafeteria” type selection (meaning, picking and choosing codes and payment thresholds), it is more important to understand that there will always be challenges at all appeal levels on that methodology.
If you are looking to conduct a sampling that includes claims that have a $0 pay, it is salient to know that there are potential challenges to including those claims, and the overall overpayment may be skewed. We will typically suggest a more conservative approach of only seeking to conduct sampling of those claims that are more than zero dollars. I’ve had numerous discussions with people on the pros and cons of including and excluding such payment amounts in the sampling, and there is a diverse set of opinions on the topic. In the end, it is most important that the sampling be defensible, with a statistician’s “sign off” of the methodology.
On the provider side, it can be an entirely different world, only because a provider may look to obtain feedback on billing and coding for internal purposes (hopefully as part of their internal compliance plan, and in accordance with the OIG Model Compliance Guidelines). Self-auditing can take many shapes and forms, and the use of various sampling methodologies should be an intrinsic part of the process. The bottom line is that with any sampling, the purpose and goal of the sampling is necessary to understand, and thus, the type of sampling will be more readily apparent.