Extrapolation.  If you are a Medicaid provider and have either received a Tentative Notice of Overpayment (TNO) or heard horror stories about TNOs, then the word, ‘extrapolation,’ most likely, will cause you to grimace.

So many providers come to me asking, “The audit amount was only for $1,500. How did that $1,500 become $850,000?” Or $1.5 million. Or $400,000? It does not seem to make sense.

But wouldn’t it be a beautiful thing if extrapolations worked in the opposite way? You put $1,500 in your bank account and the bank extrapolated the $1,500 to $850,000?  Or $1.5 million?  Or $400,000?  If opposite extrapolation existed, then maybe I would like extrapolation. Why? Because I would have a monetary incentive to like, accept, even agree with extrapolation.

Similarly, the Recovery Audit Contractors (RACs) have the monetary incentive to like, accept, even agree with extrapolation.  Approximately a 12.3%-contingency-fee incentive.

What exactly is an extrapolation?

Extrapolation is a statistical procedure in which the auditor takes findings from a small sample of Medicaid-paid claims and, using a mathematical formula, projects those results over a much larger “universe” of claims producing, in many cases, large dollar audit overpayments…sometimes even producing a overpayment amount over the amount the provider was actually paid.

Seem fair? No, at least, when these extrapolations are erroneously conducted.

But, sadly, I believe extrapolations are here to stay.  Even if I do not think Public Consulting Group (PCG) is here to stay in North Carolina, another RAC will take PCG’s place.

“Medical Practice Compliance Alert” agrees with me.  The July 8, 2013, issue of “Medical Practice Compliance Alert,” states “Like it or not, extrapolation audits are becoming the norm…”  See Volume 25, Issue 13 (by Lauren C. Williams).

Ms Williams also wrote that in one case she reviewed the overpayment was for only $10,000+, but the overpayment was extrapolated to $10 million.

Extrapolation is allowed by statute.  N.C. Gen. Stat. 108C-5 allows the RAC to extrapolate. But limits the audit period to 36 months from the date of payment of a provider’s claim.

“Except as required by federal agency, law, or regulation, or instances of credible allegation of fraud, the provider shall be subject to audits which result in the extrapolation of results for a time period of up to 36 months from date of payment of a provider’s claim.”

The North Carolina Administrative Code (NCAC) specifically sets forth the technique for such extrapolations.

 10A NCAC 22F .0606 TECHNIQUE FOR PROJECTING MEDICAID OVERPAYMENTS

(a) The Medicaid agency will seek restitution of overpayments made to providers by the Medicaid program.

(b) The agency may use a Disproportionate Stratified Random Sampling Technique in establishing provider overpayments.

(c) This technique is an extrapolation of a statistical sampling of claims used to determine the total overpayment for recoupment.

(emphasis added).

Disproportionate Stratified Random Sampling Technique.

What????

“Disproportionate” is defined as, “being out of proportion.”

“Stratified” is defined as, “the process of dividing members of the population into homogeneous subgroups before sampling. The strata should be mutually exclusive: every element in the population must be assigned to only one stratum. The strata should also be collectively exhaustive: no population element can be excluded.”

“Random” is defined as, “having no specific pattern, purpose, or objective.”

The definition of “Stratified Random Sampling” is, “a method of sampling that involves the division of a population into smaller groups known as strata. In stratified random sampling, the strata are formed based on members’ shared attributes or characteristics. A random sample from each stratum is taken in a number proportional to the stratum’s size when compared to the population. These subsets of the strata are then pooled to form a random sample.”

I had a tough time, as a double-major in English and Poli-Sci, with the whole “stratified” definition until I spoke, at length, with a statistician expert.  According to (we will call him Bill) Bill, two of the most important factors in a Disproportionate Stratified Random Sampling Technique are randomness and strata being mutually exclusive.  Meaning, each date of service (DOS) and each Medicaid recipient must be independent of one another, i.e. if Medicaid recipient X on DOS 1/1/10 is found to be incorrectly found noncompliant, then no other recipient and no other DOS should be contingent on finding of noncompliance from recipient X, DOS 1/1/10.  (As I explained to Bill, I had heard of strata before…plural for a layer of sedimentary rock or soil…but, apparently, this was a different strata).

Here’s a more specific and concrete example:

PCG audited 100 claims of my client, Samantha, the Medicaid provider.  Of the 100 claims, PCG audited 5 DOS for Medicaid recipient A: 2/3/10, 2/17/10, 3/1/10, 3/12/10, and 3/19/10.  The reason that PCG cited all 5 claims as noncompliant was that the auditor found no consent for services by the recipient.  In actuality, there was a consent for services.  The Medicaid recipient was only 10, so his mother signed the consent.  Problem? Mother’s last name was different from child’s last name? So PCG did not know that the signator was the mother. (Really, I say? Really?? In 2013? FYI: My last name is different from my daughter’s last name. But no private insurance has ever questioned the connection between my daughter and me).  Anyway, once the issue with the consent was cleared up (by explaining the child’s mother signed the consent), all 5 DOS were found compliant.

My statistician expert called this “a cluster.”  The importance of a cluster is that the strata is all messed-up.  A strata must be independent, one-element-affects-one-claim, otherwise the entire extrapolation is compromised.

Bill gave me this example:

If you want the statistics of flipping a coin and you flip a coin one time, then hit heads and just assume that heads would have been hit every time you flipped the coin…that’s a cluster.  No independent probability.

Or…tape 5 coins on a piece of paper, all showing heads. Flip the paper and determine, if it lands with all heads facing-up,  that, if independently thrown, all coins would have landed on heads.  Again, no independent probability.

Think about it…PCG (or whatever RAC) reviews 100 claims.  From those 100 claims, it extrapolates. But, what if, 5 or 10 or 20 claims are contingently dependent upon one another? Then the extrapolation is invalid.  Because, for an extrapolation to be valid, each claim must be independent; each element must affect one claim.

So, PCG (or whatever RAC) says you owe $1 million? Check for clusters!!!