Statistical Sampling Growing in Acceptance – and Importance – in False Claims Act Cases (Part 2 of 3)

Sampling Approved: United States ex rel. Martin v. Life Care Centers of Am., Inc.

United States ex rel. Martin v. Life Care Centers of Am., Inc., 114 F. Supp. 3d 549 (E.D. Tenn. 2014), involved allegations that the defendant nursing homes were providing medically unnecessary services to Medicare beneficiaries in order to inflate their bills to the government. Id. at 553–55. Faced with 54,396 patient admissions comprising 154,621 total claims, the Government sought to use a random sample of 400 admissions from 82 facilities and extrapolate from that sample to estimate the total number of false claims and the total amount of overpayments made by Medicare to the defendant. Id. at 556. The defendant challenged this use of statistical sampling in a motion for partial summary judgment.

The Martin court undertook a lengthy analysis of statistical sampling in general and FCA cases in particular, noting that “‘[c]ourts have routinely endorsed sampling and extrapolation as a viable method of proving damages in cases involving Medicare and Medicaid overpayments where a claim-by-claim review is not practical.’” 1  The court then turned to the question of whether using “statistical sampling to find liability for extrapolated claims could be in conflict with the Government’s evidentiary burden to establish the elements of a FCA claim,” id. at 563, and undertook an element-by-element analysis of proving FCA liability with statistical sampling.

First, addressing “Identification of Specific Claims Submitted or Statements Made,” the court rejected the common defense argument that statistical sampling is always insufficient to prove liability because it forgoes the individual identification of each false claim:

Considering the evidence and argument before it, the Court finds that the Government could specify in detail the specific claims for which it alleges are false, but in order to do so, it would require the devotion of more time and resources than would be practicable for any single case. However, as the Government has identified in its Response, the purpose of statistical sampling is precisely for these types of instances in which the number of claims makes it impracticable to identify and review each claim and statement. See Fadul, 2013 WL 781614, at *14. Thus, given the set of circumstances before the Court, the Court does not find Defendant’s argument that the Government cannot “specify with detail” all of the individual claims to be a compelling one. 2

Next, analyzing “Falsity,” the court evaluated the defendant’s arguments “that statistical sampling cannot be used in this context because of the ‘fact-intensive, subjective determinations by scores of different physicians, therapists, and other professionals as to whether individualized therapy treatments’ were medically necessary,” and that the Government cannot establish that therapy provided to each patient was medically unnecessary without evidence concerning individual patient’s condition. Martin, 114 F. Supp. 3d at 565–66.

The court rejected defendant’s arguments, finding that distinguishing factors unique to each patient do not necessarily preclude proving falsity with a statistical sample:

Statistical sampling has been used in litigation for decades, and Defendant’s argument regarding the individuality of each claim in the sample is not unique to this litigation. See State of Ga., Dep’t of Human Res., 446 F.Supp. at 409. In fact, Defendant’s argument highlights the very nature of statistical sampling: that a smaller portion of claims will be used to draw an inference about a larger, not entirely identical, population of claims. In re Countrywide Fin. Corp. Mortgage–Backed Sec. Litig., 984 F.Supp.2d at 1033. If all of the claims were exactly the same in every respect, there would be no need for statistical sampling and extrapolation in litigation because each individual unit would be identical, and it would be relatively simple to formulate a mathematical calculation for a large number of claims. 3

The court found the “large number of allegedly false claims at issue in this action leads to the natural effect that the claims are unique to one another in some respects,” and held that “as long as the statistical sample is a valid sample that is representative of the universe of claims, the natural disparity between the claims does not preclude using sampling and extrapolation as evidence of the total number of claims for non-covered services.” 4

The court then rejected any intent on the part of the defendant to characterize extrapolation from a statistical sample as sidestepping proof of Knowledge, finding that “the Government will attempt to meet scienter in each submitted claim and then extrapolate the total number of claims to the relevant universe. If Defendant’s intent is to challenge the extrapolation portion of this process and characterize it as ‘collective knowledge,’ it entirely misrepresents the purpose and procedure behind using statistical sampling.” 5 The court also found that establishing Materiality, applying the “natural tendency” to influence payment test, was not precluded by the Government’s use of a statistical sample and extrapolation. 6 Finally, the court rejected defendant’s Due Process argument, finding that “Defendant is not entitled to individually defend each claim brought against it under the FCA.” 7

The Martin court also outlined important policy reasons behind its ultimate holding that “statistical sampling may be used to prove claims brought under the FCA involving Medicare overpayment, but it does not and cannot control the weight that the fact finder may accord to the extrapolated evidence.” 8

  • “The purpose of the FCA as well as the development and expansion of government programs as to which it may be employed support the use of statistical sampling in complex FCA actions where a claim-by-claim review is impracticable.”
  • “Over time, the Medicare program has grown, dramatically changing the breadth of the landscape from which false claims may arise. Unlike when the FCA was originally enacted in the 1800s, those who commit fraud today have the aid of tools of technology and a relative unlikelihood of detection deriving from the sheer scale of the Medicare program itself.”
  • “Given the large number of claims that can be submitted by a single entity to be reimbursed by Medicare, it is often not practicable to do a claim-by-claim review of each allegedly false claim in a complex FCA action.”
  • “The language and the history of the FCA do not suggest that statistical sampling is an improper vehicle by which to litigate FCA claims.”
  • “Defendants position—that statistical sampling simply cannot be applied to an FCA case involving Medicare overpayment—is broad and potentially far-reaching. If accepted, it would materially limit the efficacy of the FCA as a tool to combat fraud against the government. The FCA is a remedial act, and it is intended ‘to protect the treasury from the hungry and unscrupulous host that encompasses it on every side.’”
  • If the Court were to reach the conclusion urged by the Defendant—that a claim-by-claim review is required in every FCA action and that statistical sampling is never permissible-potential perpetrators of fraud would be emboldened by the fact that a claim-by-claim review is often impractical. Armed with the knowledge that the government could not possibly pursue each individual false claim, large-scale perpetrators of fraud would reap the benefits of such a system.

United States ex rel. Martin v. Life Care Centers of Am., Inc., 114 F. Supp. 3d 549, 571 (E.D. Tenn. 2014) (citations omitted). Accord, United States v. Robinson, No. 13-CV-27-GFVT, 2015 WL 1479396, at *11 (E.D. Ky. Mar. 31, 2015) (“To require the United States to present individual evidence on each one of the 25,799 claims at issue would be unreasonable, likely impossible, and a waste of resources. Indeed, to take [defendant]’s argument to its logical conclusion would frustrate the purposes of the FCA because it would likely encourage anyone who fraudulently submitted claims to Medicare to do so in extremely large quantities so as to prevent the government from logistically being able to bring suit.).

The third and final Part of this blog post on statistical sampling in False Claims Act cases discusses a post-Escobar case that rejected evidence derived from a statistical sample, and offers some tips for litigators planning to use sampled evidence in their cases. For more information, or to discuss a potential case with us, please contact us.


  1.  United States ex rel. Martin v. Life Care Centers of Am., Inc., 114 F. Supp. 3d 549, 563 (E.D. Tenn. 2014) (quoting United States v. Fadul, No. CIV.A. DKC 11-0385, 2013 WL 781614, at *14 (D. Md. Feb. 28, 2013)). Accord United States v. Rogan, 517 F.3d 449, 453 (7th Cir.2008) (finding that “[s]tatistical analysis should suffice” rather than an individual review of claims).  The court also observed that when a statistical sample is admitted into evidence, the burden of evaluating the weight of that sample is on the fact finder: “While the proponent of sampling may argue that the sample permits the fact finder to draw an inference regarding the sample universe, the opposing party can challenge the sample through cross-examination of the proponent’s expert, presentation of its own expert, as well as other competing witnesses and evidence. See Michigan Dep’t of Educ. v. U.S. Dep’t of Educ., 875 F.2d 1196, 1206 (6th Cir.1989). The fact finder must then consider the evidence, including the risk of uncertainty and the size of the sample, in determining its weight.” Martin, 114 F. Supp. 3d at 560.
  2.  Martin, 114 F. Supp. at 565.
  3.  Id. at 566–67.
  4.  Id. at 567 (“If Defendant wishes to challenge the weight that a fact finder may attribute to the extrapolation, it can employ cross-examination and competing witnesses and testimony to highlight the disparity between claims.”).
  5. Id. at 568 (citing In re Chevron U.S.A., Inc., 109 F.3d at 1019–20 (“The essence of the science of inferential statistics is that one may confidently draw inferences about the whole from a representative sample of the whole.”)).
  6. Id. at 568–70.
  7.  Id. at 570.
  8.  United States ex rel. Martin v. Life Care Centers of Am., Inc., 114 F. Supp. 3d 549, 572 (E.D. Tenn. 2014). Accord United States v. Robinson, No. 13-CV-27-GFVT, 2015 WL 1479396, at *11 (E.D. Ky. Mar. 31, 2015) (“When presented with a challenge to the use of statistical samples similar to Dr. Robinson’s present challenge, the Sixth Circuit reasoned that using statistical methods and a random sampling technique as a basis for making arguments about the whole was not only valid but also necessary in certain complex situations, especially when the opposing party fails to demonstrate any other feasible way of making the necessary determination. In doing so, the Court further cautioned that this ‘is not to say that the statistical model will always be conclusive. The weight to be given to such statistical evidence is necessarily one which must be considered by the fact finder in light of the practical difficulties in obtaining a claim-by-claim review.’”) (citing and quoting Mich. Dep’t of Educ. v. U.S. Dept. of Education, 875 F.2d 1196, 1205 (6th Cir. 1989) (quotation and quotation marks omitted in Robinson).