Opinion
3:12-cv-00295-LRH-VPC
09-01-2015
UNITED STATES OF AMERICA ex rel. CECILIA GUARDIOLA, Plaintiff, v. RENOWN HEALTH, et al., Defendants.
ORDER
This order concerns the "data universe" for statistical sampling of claims in this case. At the June 8, 2015 case management conference (#127), the parties apprised the court of their dispute regarding the data universes for several claims. To resolve the matter, the court ordered relator to file a brief on the issue, to which defendants would respond and relator would reply. Relator filed her opening "memorandum" (#135) on July 27, 2015. Defendants filed a "response" (#138) on August 7, and relator replied (#142) on August 17. Relator's brief indicates that only one universe remains a point of contention. Having considered the parties' positions and other relevant papers, this order follows.
I. BACKGROUND
This is a qui tam action under the False Claims Act, 31 U.S.C. §§ 3729-33. Relator alleges that, while working at Renown from June 2006 through June 2014, she discovered the existence of a scheme whereby Renown knowingly submitted false inpatient reimbursement claims to Medicare that should have been billed on a less expensive, outpatient basis. Of relevance to this order is relator's claim that Renown failed to address unreliable billing software in use at Renown, the Siemens billing system, by which Renown submitted fraudulent claims to Medicare.
To streamline this case, the parties will utilize statistical sampling. In June 2014, the parties filed briefs on the viability of statistical sampling. Over two days in August 2014, the court held an evidentiary hearing at which statistical experts for both parties testified about statistical sampling. Relator's expert, Dr. Constantin Yiannoutsos, explained that relator desired to extrapolate from a universe of claims data a statistically valid estimate of how many claims were improperly submitted for reimbursement to Medicare. (See #88 at 17-19.) By using statistical sampling, the parties are relieved from examining every single claim on an individual basis, a time-consuming and costly exercise. The adequacy and completeness of the underlying data universe necessarily determines the validity of his sampling plan. (See generally #75.) Following the hearing, the parties filed supplemental briefs.
In a November 5, 2014 order (#98), the court held that statistical sampling of claims was appropriate and would make this litigation more efficient and less costly. The court ordered that Renown produce certain data, including data for inpatient "zero-day stays." (Id. at 6.) The court adopted relator's proposed definition of a zero-day stay as one that was "less than 24 hours." (Id. at 3.) The court's order expressly observed that, as "[t]his case is in the discovery phase, . . . a ruling on admissibility or use of the statistical sampling will no doubt be debated later at the pre-trial stage of this case." (Id.)
The instant dispute concerns the meaning of a zero-day stay and finalization of relator's proposed sampling plan. In various papers and at the case management conferences, defendants repeatedly contended that, rather than expeditiously turning her attention to the task, relator has laboriously reviewed the produced data and compared it to "inadequate and over-inclusive data" from Noridian, the contracted Medicare claims processor in Nevada. (#125 at 16.) Because relator was surprised that there were fewer claims than she expected—which Renown contends lowers her chance of recovery and corresponding settlement pressure—Renown argues that she seeks at this late hour to acquire more data and alter the definition of a zero-day stay to include said data. (#138 at 5-6.)
Relator counters that Renown's characterization is disingenuous because its production revealed a serious discrepancy in a portion of the data that complicates her statistical sampling. During the time period relevant to this case, Renown utilized two electronic billing systems: Siemens, and later Epic. (#135 at 6.) Siemens documented a patient's "admit" time and discharge time. Based upon those times, relator initially determined which claims constituted zero-day stays, based upon the adopted definition. However, relator later learned that the Siemens "admit" time was actually the patient's registration time, i.e., the moment in which arrangements for the inpatient stay began. This is distinct from the patient's admission time, i.e., the moment at which the patient presents and begins receiving inpatient medical care. (Id. at 7.) As a result, an inpatient stay that Siemens documented as twenty-six hours in length, and thus would fall outside of a zero-day stay, may actually have lasted twenty-three hours, when measured from the actual admission time through discharge, and thus is a zero-day stay.
Because the Epic software accurately documents both the registration time and the admission time, Dr. Yiannoutsos determined through the Epic data that the median time difference between registration and admission is two hours for Emergency Room ("ER") admissions, i.e., admissions that follow initial presentation in the ER, and 4.5 hours for non-ER admissions. (Id. at 7-9.) Thus, relator has informed Renown that she will include within her data universe for zero-day stays non-ER visits up to 28.5 hours in length, and ER visits up to twenty-six hours in length (hereinafter, the "time-adjusted claims"). She believes that the zero-day stay definition, as set in the court's November order, is not so narrow that it turns on how Siemens documented the visit length. Instead, her position is that the zero-day stay definition measures the visit from the time care began—that is, admission—through discharge.
II. DISCUSSION
Before turning to the underlying dispute, the court observes that this issue presents in an odd procedural posture. Relator's brief is styled as only a memorandum, and it states that she moves for nothing at all. As stated, she believes that the court's prior order embraces her expert's adjustment in methodology; thus, her position is that this issue is a sly attempt by Renown to achieve a premature evidentiary ruling through the discovery process.
In contrast, Renown argues that relator is attempting to revise the zero-day stay definition without first carrying the high burden for reconsideration. Unsurprisingly, Renown's "response" to relator's memorandum argues that reconsideration is unwarranted, largely on the basis of a challenge to Dr. Yiannoutsos's conclusion regarding the time adjustments. The brief discusses and disputes at some length his underlying factual premises and conclusions. Further, Renown argues that not only that inclusion of the time-adjusted claims in the sampling plan is unwarranted by the facts or law regarding inpatient admissions, but also will render relator's sampling plan unreliable and confusing. Citing the court's power to control discovery, Renown suggests that the court should embrace prevent the inclusion of these claims in the sampling plan.
This said, the briefs leave unaddressed a few pertinent questions. Relator does not state that Renown has denied any particular data requests that are necessary to her plan to include the time-adjusted claims; thus, it is unclear whether Renown need be ordered to produce the relevant data. Like relator, Renown does not move for any particular order, aside from urging the court to apply the reconsideration standard. Yet its "power-to-control-discovery" argument hints that it desires a ruling that would disallow the inclusion of these claims—a possible basis for Renown's refusal to produce such data. If this inference is correct, Renown may seek sub silentio a protective order that would obviate its obligation to produce time-adjusted data. Notwithstanding these uncertainties, the court proceeds to rule on the matter.
A. The time-adjusted data is discoverable. An order excluding the time-adjusted claims from the sampling plan would require an evidentiary ruling that is properly reserved for the District Court. Therefore, the court shall not exclude those claims from relator's data universe.
In its duty to aid the District Court in fair and speedy resolution of this case, this court supervises only discovery. The court begins with a pertinent inquiry: is the time-adjusted data discoverable?
"Parties may obtain discovery regarding any nonprivileged matter that is relevant to any party's claim or defense . . . . Relevant information need not be admissible at the trial if the discovery appears reasonably calculated to lead to the discovery of admissible evidence." Fed. R. Civ. P. 26(b). When weighing whether discovery is warranted, "[t]he question of relevancy should be construed liberally and with common sense and discovery should be allowed unless the information sought has no conceivable bearing on the case." Soto v. City of Concord, 162 F.R.D. 603, 610 (N.D. Cal. 1995). The court has broad discretion in determining whether evidence is relevant for discovery purposes, see Surfvivor Media, Inc. v. Survivor Productions, 406 F.3d 625, 635 (9th Cir. 2005), and it has similarly broad discretion to manage the discovery process, Hunt v. County of Orange, 672 F.3d 606, 616 (9th Cir. 2012).
The time-adjusted data is discoverable, for it is indisputably relevant. Evidence is relevant when "it has any tendency to make a fact more or less probable than it would be without the evidence" and "the fact is of consequence in determining the action." Fed. R. Evid. 401. Relator has adequately explained the basis for her belief that the time-adjusted claims properly fall within the data universe for zero-day stays, based upon the guidelines for an inpatient stay and the problem with the Siemens' "admit time." Within these claims, there may be additional instances of fraud that may be extrapolated to the entire universe, which similarly includes these time-adjusted claims. Ultimately, inclusion may render Dr. Yiannoutsos's statistical sampling—and therefore, his expert testimony—more reliable and accurate at trial. Notably, one of relator's claims in this case concerns the use of the Siemens software. Hence, it is apparent that the data is relevant and discoverable. Fed. R. Civ. P. 26(b). Even if the data (or testimony) is not ultimately admissible, it cannot be said at this stage that discovery of the time-adjusted data is (or was) not reasonably calculated to lead to admissible evidence.
Here, relator cites the Medicare regulations and Noridian's guidance. (See #142 at 7-8; #142-3.)
Because the evidence is relevant, this court must permit its discovery. To the extent that questions of evidence arise in discovery, the court's only role is to verify and grant discovery requests thatare "reasonably calculated" to produced admissible evidence, see Fed. R. Civ. P. 26(b), even if the parties disagree about ultimate admissibility. It is true that district courts serve as gatekeepers for expert testimony, as they alone must ensure that "any and all scientific testimony or evidence admitted is not only relevant, but reliable." Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 589 (1993). Yet "[w]hether Daubert's specific factors are, or are not, reasonable measures of reliability in a particular case is a matter that the law grants the trial judge broad latitude to determine." Kumho Tire Co. v. Carmichael, 526 U.S. 137, 153 (1999) (emphasis added). Implicit within this authority is an important limitation: so long as requested discovery is relevant under discovery's broad standard, this court—supervising discovery as it does—is not to employ the discovery rules to prematurely decide evidentiary issues that are reserved for the District Court or the jury.
For this reason, district courts within the Ninth Circuit have declined to prematurely adjudicate evidentiary issues related to expert testimony during the discovery process. "Questions concerning the accuracy and reliability of a witness' factual basis, data and methods goes to the weight and credibility of the witness' testimony and are questions of fact properly reserved for the jury." Urrutia v. Ekdahl, No. cv-031990-PHX-PGR, 2005 WL 5976564, at *4 (D. Ariz. July 15, 2005). "Unless beyond the realms of reason and reliability, [issues that speak to an expert's methodology] are fully within the province of experts to debate and a jury to resolve." Apple iPod iTunes Antitrust Litig., No. 05-CV-0037-YGR, 2014 WL 4809288, at *6 (N.D. Cal. Sept. 26, 2014).
Renown's argument, that the court must forbid inclusion of the time-adjusted data on discovery grounds, is entirely unpersuasive. In fact, its brief belies its position that this is a discovery issue at all. Renown vehemently disputes that inclusion of the time-adjusted data will produce a reliable sampling plan. Supported by its outside medical billing consultant, Christine Melnykovych, Renown vigorously contests the individual cases upon which Dr. Yiannoutsos relied when calculating the median time differences by delving into the particular facts of those cases. (See #138 at 14-16, 18-19; #138-10 at 4-5; #140.) This position is plainly a challenge to the reliability of Dr. Yiannoutsos's methods. Necessarily encompassed within the ultimate determination as to the reliability of his methods, and therefore the admissibility of his testimony, may be inquiries into the facts of those cases, the proper definition of inpatient stays, and other related issues. None of those matters is properly before this court. Renown "retain[s] the right, and will have the opportunity, to question or attack the reliability of [Dr. Yiannoutsos and his methods] and any opinions and conclusions he bases thereon[]" through motions in limine or at trial. Arredondo v. Delano Farms Co, No. 1:09-CV-01247-MJS, 2014 WL 5106401, at *10 (E.D. Cal. Oct. 10, 2014).
In sum, the court finds that the data is relevant under the broad discovery standard. The court declines Renown's invitation to prevent relator from including the time-adjusted claims in her sampling plan because to do so during the discovery process would improperly and prematurely decide an evidentiary issue. Rule 26(b) compels the court to allow discovery and obligates Renown to produce the time-adjusted data if it has not already done so. Moreover, to the extent Renown's brief can be construed as a motion for a protective order, it is denied in absence of a demonstration of good cause. See Foltz v. State Farm Mut. Auto. Ins. Co., 331 F.3d 1122, 1130 (9th Cir. 2003) ("A party asserting good cause bears the burden, for each particular document it seeks to protect, of showing that specific prejudice or harm will result if no protective order is granted. "). B. The zero-day stay definition encompasses the time-adjusted claims. To the extent reconsideration is necessary, the court finds that it is appropriate.
The court next considers a second question: does the November order regarding statistical sampling (#98) permit the inclusion of the time-adjusted data in the sampling plan? Here, even if the data is discoverable, Renown seemingly suggests that relator has not been granted permission to include it under the terms of the November order. In reviewing the August 2014 hearing transcript and other papers, the court is persuaded that the meaning of a "zero-day stay," one of "less than 24 hours," contemplates a stay from the time of admission through discharge. At heart, the issue is not about individual number of hours in the stay, but instead how a "stay" is to be measured.
When Dr. Yiannoutsos testified as to the boundaries of the data universe that he and relator envisioned, he defined a zero-day stay as "a hospital stay of less than 24 hours," which he described with reference to the "time of admission." (#88 at 28:17-24; see also id. at 40:2-23 (describing the zero-day stays in terms of provision of care, rather than registration time).) Defense counsel also spoke of the zero-day claims in reference to time of admission, rather than time of registration. (Id. at 182:9-13.) On the second day of the hearing, relator's counsel asked relator for her exact definition of a zero-day stay, and she stated: "[a] patient is admitted or discharged on the same day . . . or is in the hospital less than 24 hours." (#88 at 291:11-16.) When the court adopted relator's proposed definition (#98 at 3; see also #88 Exh. 2), it adopted the definition as she and Dr. Yiannoutsos explained it—namely, a stay of less than twenty-four hours measured from admission to discharge.
The court observes that relator's statement is somewhat ambiguous, referencing as it does 24 hours "in the hospital," which could be read as encompassing registration or other waiting time before care begins. However, the other zero-day stay discussions at the hearing resolve the ambiguity, for they consistently contemplated the stay as measured from the time of care. The court declines to read relator's words as in conflict with those of her expert. --------
Consequently, the November order permits inclusion of the time-adjusted claims. Relator's explanation of the time-adjusted claim evinces that the adjustment aims to respect this definition while accounting for the Siemens software's complication of the data by its misleading "admit" time. Renown's suggestion that relator is attempting to dramatically expand the data universe, in light of the facts and foregoing discussion, is entirely meritless. Instead, relator is simply making a minor adjustment to her sampling plan in response to newly discovered facts about the Siemens data. Renown invites the court to burden relator with the responsibility for her lack of knowledge about the data one year ago. The court declines that invitation. There is no legitimate dispute that relator's adjustments aim to capture stays of less than twenty-four hours. Therefore, her time adjustments fall within scope of the November order's definition.
Because relator is acting within the terms of the prior order, she need not move for reconsideration. Had she done so, the court would have granted the motion it in light of the circumstances. "'A motion for reconsideration is an 'extraordinary remedy, to be used sparingly in the interests of finality and conservation of judicial resources.'" Peterson v. Miranda, 57 F. Supp. 3d 1271, 1275 (D. Nev. 2014) (quoting Kona Enters., Inc. v. Estate of Bishop, 229 F.3d 877, 890 (9th Cir. 2000)). The court may reconsider a prior order where the court is presented with newly discovered evidence, an intervening change of controlling law, manifest injustice, or where the prior order was clearly erroneous. Fed. R. Civ. P. 59(e); Peterson, 57 F. Supp. 3d at 1275. "Further, a district court 'possesses the inherent procedural power to reconsider, rescind, or modify an interlocutory order' for sufficient cause." Peterson, 57 F. Supp. 3d at 1275 (quoting City of Los Angeles, Harbor div. v. Santa Monica Baykeeper, 254 F.3d 882, 887 (9th Cir. 2001)).
The circumstances satisfy this standard. First, newly-discovered evidence arose. What relator once reasonably understood as the admission times—the Siemens "admit" times—is actually the time of registration. The fact has obvious impact upon the zero-day stay definition when read narrowly. Second, it would be manifestly unjust to shackle relator to a narrow reading of order given the factual developments. To do so would require her to have knowledge about the data before it had been produced. Finally, the court's inherent powers permit it to modify the prior order for "sufficient cause," and there is little doubt that these circumstances are sufficient.
Accordingly, the court advises the parties that the definition of a "zero-day stay" for the purpose of relator's sampling may include stays of less than twenty-four hours, measured from the time inpatient care began through the time of discharge. The definition provided in docket 98 shall be construed by the parties to include the time-adjusted claims. Again, the court acknowledges that Renown disagrees with the propriety of relator's time-adjustments, and it may raise and argue that point as one of evidence at the appropriate moment.
III. CONCLUSION
For the foregoing reasons, relator may include the Siemens time-adjusted claims in her data universe. If it has not already done so, Renown SHALL EXPEDITIOUSLY PRODUCE data consistent with relator's proposal to include the time-adjusted claims. Relator SHALL EXPEDITIOUSLY FINALIZE her statistical sampling plan upon Renown's production of any outstanding data.
The parties are directed to MEET AND CONFER and determine a plan by which they will comply with the requirements of this order. By Friday, September 11, the parties SHALL FILE a joint status report regarding production of the time-adjusted data, if necessary, and the date by which relator's sampling plan will be finalized.
IT IS SO ORDERED.
DATED: September 1, 2015.
/s/ _________
UNITED STATES MAGISTRATE JUDGE