Last week, a Texas federal court upheld a Medicare Appeals Council (MAC) decision finding that a Medicare contractor properly used statistical sampling to calculate a $773,967 Medicare overpayment to a home healthcare provider, Maxmed.

During its sampling review, the Medicare contractor found that 39 of the 40 reviewed claims were not eligible for reimbursement. The Medicare contractor then extrapolated that sample to a universe of 130 claims to calculate an overpayment of $773,967 to Maxmed.

On appeal, the administrative law judge (ALJ) invalidated the statistical sampling methodology used by the Medicare contractor, citing expert testimony that there were “multiple deficiencies” in the statistical sampling methodology, including: (1) failure to keep a record of the random numbers actually used in the sample; (2) failure to properly define sampling units; (3) failure to demonstrate an independence of sampling units; and (4) failure to demonstrate the average overpayment was normally distributed.

The MAC reversed the ALJ, finding that the Medicare contractor did properly adhere to statistical sampling guidelines outlined in the Medicare Program Integrity Manual (MPIM) and in previous agency rulings.

According to the MAC, the expert testimony cited by the ALJ improperly incorporated “academic standards that are not contemplated in CMS guidance or consistent with real-world Medicare practices.” The MAC stated “it is the standards found in CMS Ruling 86-1 and the MPIM that govern Medicare sampling and overpayment estimation; not those found in statistics texts and references.”

The Western District of Texas agreed and granted summary judgment in favor of the Department of Health and Human Services.

Rejecting the “multiple deficiencies” identified by the ALJ, the court determined that:

  1. The Medicare contractor’s failure to provide a list of the actual random numbers used in the sample did not invalidate the sample because the contractor was able to replicate the sample using an alternative approach allowed under the MPIM. The court explained the purpose of keeping a record of the random numbers is so that the sample can be replicated if the methodology is challenged. The court pointed out that Maxmed’s experts never testified that they were unable to replicate the sample based on the documentation provided by the Medicare contractor.
  2. The Medicare contractor could properly extrapolate the overpayment using a sample that included multiple claims from the same beneficiary because the MPIM does not require an independence of sampling units as Maxmed’s experts claimed.
  3. There is no requirement in the MPIM that the average overpayments in the sample must be normally distributed. According to the court, the MPIM does not require a specific level of precision; per MPIM §8.4.5.1, a “one-sided 90 percent confidence interval” is statistically valid for overpayment extrapolation purposes.
  4. The case MaxMed Healthcare, Inc. v. Burwell, 2016 BL 15612, W.D. Tex., No. 5:14-cv-00988-DAE, 1/20/16.

*Admitted only in Texas. Practice supervised by principals of the firm admitted in the District of Columbia.