Why do we do a Medical Claims Audit? This question was addressed in a previous post, highlighting that 3 to 5% of claims are erroneously paid out. Once a medical claims audit or Medical Claims Audit has been determined is required, the next question to consider is, “What sampling method for the health plan or TPA audit would provide the most accurate results: random or total?”
First, a definition of terms—random sampling is performed exactly as it sounds. A sample size is determined based on a specific statistical selection criteria (random or stratified) or dictated by contract, and those specific claims selected are reviewed for errors. Generally, the claims are chosen at random—hence, the name “random sampling.”
A total sample is an audit of every claim processed within a specific timeframe or some other categorization, which statistically is a 100% review. This is essential in the Payment Integrity review.
The Pros and Cons of Random Sampling
Imagine going to the grocery store and being charged for five apples when you only bought three, or buying an item on sale but the discount didn’t process at the register. If you catch the error, you can recoup your loss, but if you do not review your receipt, you’re at a loss. Now imagine if you never check your receipt or only look at a random selection, out of all the receipts you received over the course of a year. How do you know how many times you might have overpaid and how much money you lost?
Take the same scenario and apply the concept similarly to a random sampling of medical claims (your healthcare services receipt)—a common auditing method when access to the total or actual claims database is limited or unavailable. With only a percentage of claims being reviewed, the high potential exists for errors being missed in the larger portion not under review since the errors are not homogenous or evenly distributed.
For instance, if a Medical Claims Audit performs a 250-claim audit chosen or random out of 25k individual medical claims, this essentially means that the audit only performs a 1% review. If there are 1,000 errors, but you only take a sample of 250, you are missing a huge number of errors from being identified because your sample size is too small for all the potential errors. And if you assume, distribution of error being 4%, your statistical finding will be even less (250 x 4%).
See the table below for the simple logic calculation.
If the sample size is 250 claims only biasing at high dollars, essentially 997.5 errant claims are not reviewed or considered; this means money is being left on the table with unnecessary payouts to the health plan. There is no direct correlation between the dollar value and error, and high dollar claims tend to have the opportunity for high dollar identification.
However, the aggregated error of multiple claims may also result in equivalent or greater dollar identifications. However, this identification of errant payments is only a bonus (as identified dollars does not necessitate return payment).The real value is derived from the audit results of the overall performance of the health plan’s payment integrity.
So why do auditing firms choose the random sampling method? Short answer, it is an easier audit and less resource dependent. They will point to effort versus the dollar result as they are often focused on the potential dollar return as well as the dollars recouped and time saved by not auditing every claim.
However, they are operating under the assumption of even distribution of errors in relation to dollars lost to justify the minimization of resources to review every claim. This is a logical and statistical fallacy because medical claims are not discrete like data, and claim errors are not an even distribution. The dollar impact is further unknown since there is no direct correlation with error and claim value.
If only the top dollar claims are reviewed, there is no guarantee of the errors occurring in the selected bias claims set. Even if there is an error within the high-dollar medical claim that error does not necessarily result in a high-value error being identified or errant payment. Thus within unanalyzed claims are undiscovered errors, which are missed, not providing an accurate audit result of the health plan’s performance.
Further, the error rate is not correlated or related to cost, so therefore, stratifying medical claims by cost (high dollars) essentially does not identify all the errors and errant payments as a result.
This topic was supported by Consova’s industry experience and observations from multiple employers and health systems that have dared to ask: “Is my health plan performing in our best interest and ensuring cost containment?”
A study in the International Journal of Electronic Healthcare compared random and 100-percent-of-claims sampling. The study noted that, “When compared head-to-head, 100-percent auditing, of course, identifies more claims than the random-sampling approach, but, our study results show that the random sampling approach misses a significant proportion of claims paid amounts.” Even when the sample size was increased, over 90% of claim errors were still missed.
The ROI for Total Sampling
There’s no denying that total sampling is a time and labor-intensive process. And in some instances, random sampling may be appropriate. For example, taking a total measurement of an entire population, while ideal, is but costly or difficult to perform — hence the use of sampling. This type of sampling requires reducing variation and biasing significantly. These become degrees of freedom factored into the results as a confidence level in typical data analysis.
Random sampling may work in statistics and science experiments because of discrete data or continued observations that reduces variation, but it is far less reliable when auditing medical claims because the data is complex and has shades of variation, especially with coordination of benefits, change of benefits, types of eligibility, and so forth.
The greater the complexity of claims processing, the greater the increase in error rate, making it imperative that the correct sampling method to be applied to the claims being reviewed.
Unfortunately, medical bill audit firms use the method that will provide their value, meaning they are looking at mostly high dollar errors within their sampling and using the identified return as their value proposition. Health plan buyers deserve more.
Having a total picture will allow no money to be left on the table, which is the reason why Consova uses total sampling when conducting a Medical Claims Audit to ensure payment integrity. Since identifying errant payments do not guarantee return of those payments, the purpose of a Medical Claims Audit is often lost with many firms. The true goal for a Medical Claims Audit is to provide a true and accurate measurement of the performance of an organization’s selected health plan or TPA without bias.