Admissibility of statistical sampling to prove liability in FCA suit is fact dependent.

In a February 14, 2017 decision, the Fourth Circuit declined to rule on the question of whether statistical sampling can be used to prove liability and damages in False Claims Act (FCA) lawsuits, concluding that it should not have agreed to hear the appeal in the first instance. U.S. ex rel. Michaels v. Agape Senior Community, Inc., Case No. 15-2145 (4th Cir. Feb. 14, 2017). Because the question presented turned on the facts and circumstances of the case and was therefore not a pure question of law, the court found the question inappropriate for review before the trial court issued its final ruling. This decision was disappointing for those who had hoped for some appellate-level resolution on the question of whether using statistical sampling to prove FCA liability is permissible.

The interim appeal was taken from United States ex rel. Michaels v. Agape Senior Community, Inc., No. 0:12-cv-03466 (D.S.C. Mar. 16, 2015), a declined FCA qui tam suit filed by two relators alleging that Agape had provided medically unnecessary hospice services to federal beneficiaries. Joseph F. Anderson, Jr., a senior district judge for the District of South Carolina, rejected the relators’ statistical sampling model, holding that the relators would be required to prove the falsity of each and every claim at trial. The district court reasoned that Agape was a case where it was possible to prove damages on a claim-by-claim basis—statistical sampling was not necessary as all of the underlying medical records for the claims at issue were “intact and available for review by either party.” The district court also concluded that where falsity turns on medical necessity, an individualized assessment is required to determine the truth or falsity of each claim:

[E]ach claim asserted here presents the question of whether certain services furnished to nursing home patients were medically necessary [and] [a]nswering that question for each of the patients involved . . . is a highly fact-intensive inquiry involving medical testimony after a thorough review of the detailed medical chart of each individual patient.

The district court suggests that reliance on statistical sampling is perhaps less likely to be appropriate in the context of a medical necessity assessment. However, in the decision granting interlocutory appeal, the district court observed that whether statistical sampling is appropriate depends on the unique circumstances of the case: “[I]t is [the] Court’s responsibility to determine the fairest course of action based upon the facts presented and the claims asserted in that case.”

This holding is consistent with the subsequent ruling by the Supreme Court in Tyson Foods, Inc. v. Bouaphakeo, 136 S. Ct. 1036 (2016), holding that the permissibility of a representative or statistical example as a means to establish or defend against liability “turns not on the form a proceeding takes—be it a class or individual action—but on the degree to which the evidence is reliable in improving or disproving the elements of the relevant cause of action.”

Following the Fourth Circuit’s decision, the relators in Agape will be required to prove the falsity of every single claim at trial, unless the district court reconsiders its initial holding. Significantly, the case involves between 53,000 and 61,000 individual claims, with medical charts for 10,000–20,000 nursing home patients. It is conceivable that the district court’s ruling could effectively end the qui tam law suit, as it may be too costly for the relators to pursue relative to their potential recoveries.

Most district courts that have recently weighed in on the issue have allowed statistical sampling to prove liability or damages, or at least acknowledged that there is no ban on relying on statistical sampling to prove FCA liability. See, e.g., U.S. ex rel. Martin v. Life Care Ctrs. of Am., Inc., 2014 WL 4816006 (E.D. Tenn. 2014); United States v. Robinson, 2015 WL 1479396 (E.D. Ky. 2015); United States v. AseraCare, Inc., 2014 WL 6879254 (N.D. Ala. Dec. 4, 2014); U.S. ex rel. Ruckh v. Genoia Healthcare, LLC, 2015 WL 1926417 (M.D. Fla. 2015).

But to be admissible, a sample must still be reliable, which means that defendants still have tools available to challenge the use of statistical sampling, particularly where there is better evidence available (i.e., where the claims themselves can be individually reviewed). Defendants should assess—and may attack—the sample size (is the sample is large enough to be significant?); whether the sampled claims are truly representative of the broader universe of claims the plaintiff proposes to extrapolate (comparing apples to apples?); the standards applied in reviewing the records (were the correct criteria applied?); and the subjective experience and qualifications of the expert performing the review (does this guy really know what he’s talking about?).

The likelihood of a party’s success at trial may turn on the admissibility or persuasiveness of the statistical evidence. Relators and the government may be able to take a shortcut, establishing the falsity of a broad universe of claims by demonstrating the falsity of a few with potential for obtaining substantial damages. If, on the other hand, plaintiffs are required to prove the falsity of each and every claim because of the particular facts and circumstances of a case (or because their statistical sample was found unreliable and therefore inadmissible), they may be willing to settle for a portion of what they believe is the true value of damages, just to avoid potentially significant litigation costs. Both plaintiffs and defendants in FCA cases should begin thinking early on about the admissibility, strength, and reliability of any sampling evidence.