Harvard Law Review Harvard Law Review Harvard Law Review

False Claims Act

United States ex rel. Martin v. Life Care Centers of America, Inc.

Eastern District of Tennessee Rules that Statistical Extrapolation May Suffice to Prove Liability.

The False Claims Act1×1. 31 U.S.C. §§ 3729–3733 (2012). (FCA), originally enacted to stymie fraud during wartime, is now one of the government’s most effective tools for policing health care fraud.2×2. See generally Patricia Meador & Elizabeth S. Warren, The False Claims Act: A Civil War Relic Evolves into a Modern Weapon, 65 Tenn. L. Rev. 455 (1998). The act enables recovery of substantial civil penalties and damages3×3. The FCA allows for $5500 to $11,000 in penalties per claim plus treble damages. See 31 U.S.C. § 3729(a)(1); 28 C.F.R. § 85.3(a)(9) (2014). Because of the large number of claims at issue in medical fraud cases, the total amounts in controversy can be astounding. For example, in 2014, Johnson & Johnson agreed to pay $1.1 billion to settle FCA allegations relating to the promotion of three prescription drugs. Press Release, Dept. of Justice, Justice Department Recovers Nearly $6 Billion from False Claims Act Cases in Fiscal Year 2014 (Nov. 20, 2014), http://www.justice.gov/opa/pr/justice-department-recovers-nearly-6-billion-false-claims-act-cases-fiscal-year-2014[https://perma.cc/BNA3-7RG3?type=source]. against one who, among other things, “knowingly presents . . . a false or fraudulent claim for payment” to the federal government.4×4. 31 U.S.C. § 3729(a)(1)(A). The FCA attaches liability to each claim individually,5×5. See, e.g., United States ex rel. Nathan v. Takeda Pharms. N. Am., Inc., 707 F.3d 451, 456 (4th Cir. 2013). but some cases involve such a large number of claims at issue that claim-by-claim review is impractical.6×6. See, e.g., Andrew J. Silver, Tennessee Federal Court’s Ruling Approving Use of Statistical Sampling in False Claims Act Cases Will Help Stamp Out Large-Scale Fraud, Nat’l L. Rev. (Oct. 17, 2014), http://www.natlawreview.com/article/tennessee-federal-court-s-ruling-approving-use-statistical-sampling-false-claims-act [http://perma.cc/J4V4-MXBT]. Recently, in United States ex rel. Martin v. Life Care Centers of America, Inc.,7×7. Nos. 1:08-cv-251, 1:12-cv-64, 2014 U.S. Dist. LEXIS 142660 (E.D. Tenn. Sept. 29, 2014). the Eastern District of Tennessee ruled that the government could use statistical sampling and extrapolation,8×8. Statistical sampling in this context involves selecting a representative portion of all the claims at issue, evaluating how many of those claims were fraudulent, and from this information estimating the total number of fraudulent claims. See United States ex rel. Martin v. Life Care Ctrs. of Am., Inc., Nos. 1:08-cv-251, 1:12-cv-64, 2014 WL 4816006, at *1, *6–7 (E.D. Tenn. Sept. 29, 2014) (discussing, in denying Life Care’s motion to exclude expert testimony, the government’s statistical expert’s proposed methodology). rather than claim-by-claim litigation, to prove liability under the FCA.9×9. Life Care, 2014 U.S. Dist. LEXIS 142660, at *60–61. Since Life Care was decided, another court has ruled that statistical extrapolation may be used to prove liability under the FCA. See United States v. AseraCare Inc., No. 2:12-CV-245-KOB, 2014 U.S. Dist. LEXIS 167970, at *24–26 (N.D. Ala. Dec. 4, 2014). The court supported important parts of its opinion with the fact that courts commonly allow statistical extrapolation in other contexts, particularly in determining damages or amount of loss.10×10. See Life Care, 2014 U.S. Dist. LEXIS 142660, at *49, *60. However, the court did not critically examine precedential cases that draw a clear distinction between damages and liability, and between the kinds of evidence considered sufficient to prove each. Had the court done so, it would have found that the reasoning underlying these decisions surprisingly buttresses its holding.

Life Care Centers of America (Life Care) owns and operates more than 200 nursing homes across the United States.11×11. Id. at *6. These facilities provide a variety of services, including assisted living and both inpatient and outpatient rehabilitation therapy.12×12. Our Services, Life Care Ctrs. of Am., http://www.lcca.com/services (last visited Mar. 29, 2015) [http://perma.cc/TU7U-59TP]. Life Care bills many of its services to Medicare; it received $4.2 billion from the program from 2006 through 2011.13×13. Life Care, 2014 U.S. Dist. LEXIS 142660, at *6. The amount that Medicare pays Life Care (and nursing facilities generally) for any particular patient depends upon the patient’s Resource Utilization Group (RUG) classification.14×14. Id. at *8–11. The higher the RUG level assigned to a patient, the more Medicare pays for care of that patient.15×15. Id. For example, 2015 reimbursement rates for urban skilled nursing facilities range from $194.59 to $776.81 per day, depending on RUG category. Prospective Payment System and Consolidated Billing for Skilled Nursing Facilities for FY 2015, 79 Fed. Reg. 45,628, 45,634–35 tbl.4 (Aug. 5, 2014) (to be codified at 42 C.F.R. pt. 488).

In 2012, the government intervened in and successfully sought consolidation of two qui tam actions16×16. The FCA allows whistleblowers, called “relators,” to bring suits in the government’s name in exchange for a portion of the recovery; such suits are qui tam actions. See Meador & Warren, supra note 2, at 466–67. The government may choose to intervene in the suit, at which point the relator forgoes control over the litigation. Id. at 467–68 (citing 31 U.S.C. § 3730(b)(2) (2012)). brought against Life Care by former employees.17×17. Life Care, 2014 U.S. Dist. LEXIS 142660, at *5–6. In its consolidated complaint, the government alleged that Life Care had repeatedly violated the FCA by inflating RUG classifications in order to charge Medicare for unnecessary services.18×18. Id. at *12–18. Because the FCA “attaches liability, not to the underlying fraudulent activity . . . , but to the ‘claim for payment,’”19×19. Sanderson v. HCA–The Healthcare Co., 447 F.3d 873, 877–78 (6th Cir. 2006) (quoting United States v. Rivera, 55 F.3d 703, 709 (1st Cir. 1995)) (internal quotation mark omitted); accord United States ex rel. Nathan v. Takeda Pharms. N. Am., Inc., 707 F.3d 451, 456 (4th Cir. 2013). the government bore the burden of proving liability for each claim it contended was false.20×20. See Life Care, 2014 U.S. Dist. LEXIS 142660, at *22–23. However, the case involved too many claims to litigate practicably on a claim-by-claim basis.21×21. Id. at *45–46. As a result, the government sought to use a random sample of 400 patient admissions to determine how many of the 154,621 claims at issue were fraudulent.22×22. Id. at *20.

Life Care moved for summary judgment on those claims that the government sought to prove were fraudulent solely through statistical extrapolation.23×23. Id. at *4–5. It argued that statistical extrapolation could not suffice to meet the government’s burden of proving each element of liability.24×24. Id. at *22–23. Life Care also argued that imposing liability based on statistical extrapolation would violate the Due Process Clause and impermissibly “shift the burden of proof onto [Life Care].”25×25. Id. at *59.

Judge Mattice denied the motion for summary judgment and held that statistical extrapolation from a random sample of claims may suffice to prove liability.26×26. See id. at *4, *60–61. He began by providing an overview of the FCA and the historical use of statistical extrapolation in FCA cases and other contexts.27×27. See id. at *23–31. The court then decided that existing case law accepting the use of extrapolation to prove facts like amount of damages did not control whether extrapolation could prove liability.28×28. See id. at *43. The court conducted an exhaustive discussion of similar cases before concluding that none were “on point.” Id. It determined that cases like United States v. Friedman, No. 86-0610-MA, 1993 U.S. Dist. LEXIS 21496 (D. Mass. July 23, 1993), in which courts declined to use extrapolation, could readily be distinguished on their facts. Life Care, 2014 U.S. Dist. LEXIS 142660, at *32–35. It also distinguished the use of statistical sampling in damages calculations and in appeals from administrative decisions. Id. at *35–39. Finally, the court distinguished two cases in which statistical extrapolation had been used to prove liability because in the first the defendant had failed to appear and the second involved unusual circumstances like use of a bellwether trial. Id. at *39–43. Considering itself left with little direct guidance, the court decided it would further consider “the facts . . . of the instant case” and the statute’s text and purpose in making its ruling.29×29. Life Care, 2014 U.S. Dist. LEXIS 142660, at *43; see also id. at *60–61.

This analysis began by addressing whether the government could meet its burden of proof on each element of liability under the FCA: (1) identification of specific claims or statements, (2) knowledge, (3) falsity, and (4) materiality.30×30. See id. at *44–59. Judge Mattice rejected Life Care’s arguments on the first two elements for case-specific reasons. Life Care’s sole argument with regard to the identification prong was that it would be impossible for the government to specifically identify each claim that it alleged to be false.31×31. See id. at *44–46. Because the court believed it would merely be impractical, but not impossible, to identify the specific claims, the court rejected this argument and moved to the next prong.32×32. See id. Similarly, Judge Mattice rejected Life Care’s arguments that extrapolation could not prove knowledge, reasoning that the government did not plan to prove knowledge through extrapolation but rather through evidence of corporate practices tending to show that Life Care knew it was submitting false claims.33×33. Id. at *54.

While Judge Mattice analyzed the first two prongs with reasoning particular to the case, the remainder of his analysis applied generally to the use of statistical extrapolation to prove liability. This analysis was based largely on the fact that statistical extrapolation is well established as a method of proof in some other contexts. For example, Life Care had argued that the uniqueness of each claim created a degree of uncertainty that precluded proving falsity through extrapolation.34×34. See id. at *46–50. The court responded that “[s]tatistical 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.”35×35. Id. at *49 (citing Georgia v. Califano, 446 F. Supp. 404, 409 (N.D. Ga. 1977)). The court decided the materiality element on similar grounds, incorporating its reasoning on the falsity prong and concluding that concerns over individuality were “best left to the finder of fact . . . in determining how much weight should be attributed to the extrapolated evidence.”36×36. Id. at *58–59 (citing Yorktown Med. Lab., Inc. v. Perales, 948 F.2d 84, 90 (2d Cir. 1991)). Similarly, the court rejected Life Care’s due process arguments, noting that “courts that have considered the issue of statistical extrapolation to calculate overpayment have found that it . . . does not violate a defendant’s due process rights.”37×37. Id. at *60. Judge Mattice concluded that the opportunity to challenge and rebut the government expert’s testimony would afford Life Care sufficient due process.38×38. Id.

Having dispensed with Life Care’s affirmative arguments in favor of its motion, the court further supported its ruling by turning to the purpose of the FCA: “to protect the treasury [against] the hungry and unscrupulous host that encompasses it on every side.”39×39. Id. at *63 (quoting United States v. Griswold, 24 F. 361, 366 (D. Or. 1885)) (internal quotation marks omitted); see also id. at *23–27, *61–63. In the court’s view, to disallow the use of statistical extrapolation to prove liability under the FCA “would materially limit the efficacy of the FCA as a tool to combat fraud against the government.”40×40. Id. at *62–63. The court feared that doing so would cause “perpetrators of fraud [to] be emboldened by the fact that a claim-by-claim review is often impractical.”41×41. Id. at *63.

While the Life Care court did not rely exclusively on the general acceptance of statistical sampling to prove damages, it did use this fact to undergird significant portions of its opinion. In the course of doing so, the court did not engage with case law that distinguishes proof of damages from proof of liability. In particular, the court did not meaningfully address a longstanding contrast between judicial acceptance of uncertain or probabilistic evidence to establish damages and judicial reluctance to allow this evidence to prove liability. But while this traditional distinction between damages and liability may appear to cast doubt on the Life Care court’s judgment, the rationale underlying that precedent actually bolsters the court’s opinion.

The Supreme Court has held that uncertain or probabilistic evidence may suffice to prove the extent of damages, but generally not other elements — including liability. This understanding began with Eastman Kodak Co. v. Southern Photo Materials Co.,42×42. 273 U.S. 359 (1927). which held that there was no requirement for “exactness and precision” in evidence of damages.43×43. Id. at 379. Four years later, the Court clarified in Story Parchment Co. v. Paterson Parchment Paper Co.44×44. 282 U.S. 555 (1931). that this relaxed certainty requirement applied only to the extent — but not to the fact — of damages.45×45. Id. at 562–63. The Court stated: “[T]here is a clear distinction between the measure of proof necessary to establish the fact that petitioner had sustained some damage, and the measure of proof necessary to enable the jury to fix the amount.”46×46. Id. at 562.

In Bigelow v. RKO Radio Pictures, Inc.,47×47. 327 U.S. 251 (1946). the Court explained the reason behind this rule: once liability has been proven, the wrongdoer should “bear the risk of the uncertainty which his own wrong has created.”48×48. Id. at 265. Eastman Kodak, Story Parchment, and Bigelow were antitrust cases, in which damages could not be calculated with precision because doing so would require knowing an unknowable — the amount of profits that the plaintiffs would have earned absent the illegal activity.49×49. See id. at 263–64. Similar uncertainty arises in FCA cases, because damages are measured according to what the government would have paid absent the false statements. See, e.g., United States v. Killough, 848 F.2d 1523, 1530–31 (11th Cir. 1988) (applying the relaxed proof requirement for extent of damages where there was some uncertainty as to what the government would have paid absent the fraud). In these cases, “[t]he tortious acts had . . . precluded ascertainment of the amount of damages more precisely.”50×50. Bigelow, 327 U.S. at 264. Under such circumstances, to require that the extent of damages be precisely established “would enable the wrongdoer to profit by his wrongdoing at the expense of his victim. It would be an inducement to make wrongdoing so effective and complete in every case as to preclude any recovery . . . .”51×51. Id. Decades after Bigelow, federal courts continue to rely on its reasoning in allowing imprecise proofs of damages, including in FCA cases. For example, in United States ex rel. Miller v. Bill Harbert International Construction, Inc.,52×52. 608 F.3d 871 (D.C. Cir. 2010). the D.C. Circuit sustained a damages award for FCA violations over the defendant’s objection that the government had proved damages with insufficient precision: “[I]n determining what constitutes a ‘reasonable range’ for the jury’s award, we make allowances for the fact that the defendants’ own misconduct has foreclosed any exact calculation of what [the price absent the fraud] would have been . . . .”53×53. Id. at 905 (citing Bigelow, 327 U.S. at 264).

Notably, this reasoning is generally inapplicable to proof of liability because the defendant is not known to be a “wrongdoer” until after liability has been proven. Courts have accordingly been reluctant to allow probabilistic or statistical proofs of liability. As Life Care argued, allowing proof of liability through statistics could rob defendants of the chance to present every available defense, implicating due process concerns.54×54. See Life Care, 2014 U.S. Dist. LEXIS 142660, at *59; cf. McLaughlin v. Am. Tobacco Co., 522 F.3d 215, 231–32 (2d Cir. 2008) (holding that when in a class action “the right of defendants to challenge the allegations of individual plaintiffs is lost, [the result is] a due process violation,” id. at 232), abrogated on other grounds by Bridge v. Phoenix Bond & Indem. Co., 553 U.S. 639 (2008). Indeed, a number of courts have rejected proposed statistical proofs of liability, primarily in the context of class actions.55×55. See, e.g., McLaughlin, 522 F.3d at 231–32; Cimino v. Raymark Indus., Inc., 151 F.3d 297, 319–20 (5th Cir. 1998) (holding that a plan to extrapolate causation and damages from sample cases to the rest of the plaintiff class violated the Seventh Amendment and Texas state law); In re Wells Fargo Home Mortg. Overtime Pay Litig., 268 F.R.D. 604, 612 (N.D. Cal. 2010) (rejecting extrapolation of liability). But see Dilts v. Penske Logistics, LLC, 267 F.R.D. 625, 638 (S.D. Cal. 2010) (allowing extrapolation of liability to satisfy class certification requirements). The Supreme Court recently joined this line of decisions, criticizing a statistical approach to proving liability in Wal-Mart Stores, Inc. v. Dukes,56×56. 131 S. Ct. 2541 (2011). a sex discrimination class action. In that case, the Court chastised the Ninth Circuit for suggesting that the district court use statistical sampling to determine the number of valid claims.57×57. Id. at 2561. The Ninth Circuit had held that the class action could proceed by selecting a random sample of cases and allowing Wal-Mart to present its statutory defenses individually to each sampled case; the trial court would then take the results of these cases to estimate the total number of valid claims in the class action. See id. at 2550. The Court declared that the sampling plan would improperly replace Wal-Mart’s right to “litigate its statutory defenses to individual claims” with “Trial by Formula.”58×58. Id. at 2561.

The Life Care court recognized that “using extrapolation to establish damages when liability has been proven is different than using extrapolation to establish liability.”59×59. Life Care, 2014 U.S. Dist. LEXIS 142660, at *39. But the court nevertheless cited judicial acceptance of using sampling to prove damages in significant portions of its opinion. First, in deciding that the government had met its burden of proof on the elements of falsity and materiality, the court noted that “[s]tatistical sampling has been used in litigation for decades.”60×60. Id. at *49 (citing Georgia v. Califano, 446 F. Supp. 404, 409 (N.D. Ga. 1977) (holding that statistical sampling was an acceptable method for determining the amount of Medicaid overpayment induced by fraud)). Second, in responding to Life Care’s due process argument, the court asserted, “courts that have considered the issue of statistical extrapolation to calculate overpayment have found that it is an acceptable practice which does not violate a defendant’s due process rights.”61×61. Id. at *60 (citing Yorktown Med. Lab., Inc. v. Perales, 948 F.2d 84, 90 (2d Cir. 1991) (holding that the use of statistical sampling to calculate the amount of Medicaid overbilling did not violate due process)). Each time, the court cited cases in which the extrapolation at issue was used to calculate damages or amount of loss. By relying on these arguably inapt precedents while ignoring the body of case law critical of statistical proofs of liability, the Life Care court left its decision open to the same “Trial by Formula” critiques seen in Wal-Mart and analogous cases.

But while the court’s opinion did not significantly engage with the damages/liability divide, the court may yet have reached the right doctrinal result. Paradoxically, despite the fact that the Bigelow line of cases implies that probabilistic evidence is generally insufficient to prove liability, these cases’ reasoning supports the court’s allowing of statistical sampling to establish the extent of liability under the FCA. If the jury agrees with the government that at least some of the sampled claims are false, then the government will have established that Life Care committed fraud and is thus a “wrongdoer.” Bigelow declared that “[t]he most elementary conceptions of justice and public policy require that the wrongdoer shall bear the risk of the uncertainty which his own wrong has created.”62×62. Bigelow v. RKO Radio Pictures, Inc., 327 U.S. 251, 265 (1946). To allow Medicare defrauders to escape liability as a result of the large scope of their fraud would be to allow these “wrongdoer[s] to profit . . . at the expense of [their] victim[s]” due to the effectiveness of their crimes.63×63. Id. at 264. The Life Care court could not — and Bigelow counsels it should not — have countenanced this result.64×64. Although the court did not discuss Bigelow or its reasoning, policy implications played an important part in the court’s ruling. The court recognized that granting the motion “would materially limit the efficacy of the FCA as a tool to combat fraud against the government,” meaning that “potential perpetrators of fraud would be emboldened by the fact that a claim-by-claim review is often impractical.” Life Care, 2014 U.S. Dist. LEXIS 142660, at *62–63.

Whether Wal-Mart’s admonition against “Trial by Formula” will supersede Bigelow’s rationale in future decisions remains to be seen. After all, Life Care allows the government to prove the validity of unlitigated claims through statistics, just as the trial plan invalidated in Wal-Mart would have. But there is reason to believe that Bigelow’s reasoning will survive Wal-Mart. While lower courts had held that similar trial plans would violate the Constitution,65×65. See, e.g., McLaughlin v. Am. Tobacco Co., 522 F.3d 215, 231–32 (2d Cir. 2008) (holding that sampling violated the Due Process Clause), abrogated on other grounds by Bridge v. Phoenix Bond & Indem. Co., 553 U.S. 639 (2008); Cimino v. Raymark Indus., Inc., 151 F.3d 297, 319–20 (5th Cir. 1998) (holding that sampling violated the Seventh Amendment). the Supreme Court decided Wal-Mart on statutory grounds.66×66. See Wal-Mart Stores, Inc. v. Dukes, 131 S. Ct. 2541, 2561 (2011); Joshua A. Rosenthal, Comment, The Case Against Constitutionalized Commonality Standards for Collective Civil Litigation, 32 Yale L. & Pol’y Rev. 309, 310 (2013) (“[T]he Supreme Court has so far limited access to collective litigation on statutory grounds . . . .”). The Wal-Mart decision rested on the Court’s finding that the Ninth Circuit’s suggested trial plan would interpret Rule 23 of the Federal Rules of Civil Procedure (FRCP) in a manner that would modify a substantive right, thereby violating the Rules Enabling Act.67×67. 28 U.S.C. § 2072 (2012); see Wal-Mart, 131 S. Ct. at 2561. Because the question of whether liability under the FCA may be proven by sampling does not involve an interpretation of Rule 23 or any other Rule of the FRCP, the question does not implicate the Rules Enabling Act and therefore Wal-Mart does not control. Absent a finding of a constitutional prohibition against sampling, Bigelow’s policy concerns will continue to advise that courts should allow the use of sampling to prove liability.

While courts have long accepted statistical sampling as an established method of proving the extent of damages, Life Care represents the first time a court has allowed the use of sampling to prove liability under the FCA. Although the court’s analysis at times elided a key distinction between damages and liability, the reasoning of the cases establishing that distinction actually supports the court’s conclusion. Just as Bigelow ensured wrongdoers would not be able to escape liability because of the effectiveness of their wrongdoing, Life Care ensured perpetrators of fraud would not be able to escape liability because of the broad scope of their fraud.