Opinion
CIV 20-264 SCY/JFR
03-10-2022
BRIAN LAX, TRACEY BURON-HAHNLEIN, WERNER HAHNLEIN, and JEREMY HADER, on their own behalf and on behalf of all others similarly situated, Plaintiffs, v. APP OF NEW MEXICO ED, PLLC, f/k/a ALIGNMD OF NEW MEXICO, PLLC, and LOVELACE HEALTH SYSTEM, LLC, Defendants.
MEMORANDUM OPINION AND ORDER GRANTING PLAINTIFFS' AMENDED MOTION TO REMAND
STEVEN C. YARBROUGH, UNITED STATES MAGISTRATE JUDGE
This proposed class action case arises from alleged over-billing practices by Lovelace Health System, LLC (“Lovelace”) and APP of New Mexico ED, PLLC (“APP”). Plaintiffs are former patients who sought treatment at a Lovelace facility and allege they were overbilled by APP, a company that provides emergency room physician and nurse practitioner staffing for Lovelace facilities. Presently before the Court is the parties' dispute over whether the Court should decline jurisdiction under the Class Action Fairness Act's (“CAFA”) local controversy exception. Finding that Plaintiffs have met their burden to establish all four elements of the local controversy exception, the Court grants Plaintiffs' Amended Motion to Remand (Doc. 70).
BACKGROUND
Plaintiffs filed a class action complaint against Lovelace and APP in the New Mexico Second Judicial District Court on February 11, 2020. Doc. 1-1. The Complaint brings claims for violations of the New Mexico Unfair Practices Act, conversion, willful breach of contract, unjust 1 enrichment, and civil conspiracy. Id. Plaintiffs' proposed class includes “all New Mexico residents who, beginning four years prior to the filing of this lawsuit, were billed by APP for amounts greater than the in-network amount permitted by their insurance provider for medical services provided at Lovelace facilities.” Id. ¶ 93. On March 23, 2020, Defendant APP removed the action to federal court, citing diversity jurisdiction under the CAFA. Doc. 1. Defendant Lovelace consented to removal. Doc. 6.
In response, Plaintiffs filed a motion to remand, asserting this case should be remanded to state court because: (1) Defendants have failed to establish that greater than $5 million is in controversy and (2) even assuming more than $5 million is in controversy, the “local controversy exception” mandates that this lawsuit remain in state court. Doc. 25. The Court entered a Memorandum Opinion and Order finding that Defendants had met their burden to show that more than $5 million was in controversy and so remand was not appropriate on that ground. Doc. 39. The Court also found that Plaintiffs failed to meet their burden to show that the local controversy exception applied because they failed to show that greater than two-thirds of the proposed class are citizens of New Mexico. Id. However, the Court took the Motion to Remand under advisement and provided Plaintiffs the opportunity to conduct limited discovery on the question of class citizenship. Id.
Judge Robbenhaar thereafter entered a scheduling order for limited discovery regarding class citizenship. As part of this discovery, each party obtained expert opinions. Doc. 47. Following the limited discovery period, Plaintiffs filed the present Amended Motion to Remand on October 8, 2021 (Doc. 70), and the parties fully briefed it by November 22, 2021 (Docs. 77, 2 78, 82). The Court held a hearing on this Motion on January 27, 2022. Doc. 86. Pursuant to 28 U.S.C. § 636(c) and Fed.R.Civ.P. 73(b), the parties have consented to me serving as the presiding judge and entering final judgment. Docs. 9, 14, 15, 16. Supplementing its previous Memorandum Opinion and Order (Doc. 39), the Court now addresses whether this matter should be remanded under the local controversy exception.
In response to the amended motion to remand, Lovelace asserts that Plaintiffs' proposal for citizenship discovery was too vague. Doc. 78 at 3-4. Judge Robbenhaar already addressed this argument in ruling on Plaintiffs' opposed motion for limited discovery. See Doc. 47.
ANALYSIS
“CAFA was enacted to respond to perceived abusive practices by plaintiffs and their attorneys in litigating major class actions with interstate features in state courts.” Coffey v. Freeport McMoran Copper & Gold, 581 F.3d 1240, 1243 (10th Cir. 2009). Congress recognized, however, that when it comes to “class actions with a truly local focus . . . state courts have a strong interest in adjudicating such disputes.” Id. (quoting S. Rep. No. 109-14, at 39 (2005)). Therefore, “class actions with a truly local focus should not be moved to federal court under this legislation . . . .” Id. (citing S. Rep. No. 109-14 at 39 (2005)).
Under the local controversy exception, codified at 28 U.S.C. § 1332(d)(4)(A), four elements must be met for the court to decline jurisdiction: (1) “greater than two-thirds of the members of all proposed plaintiff classes in the aggregate are citizens of the State in which the action was originally filed, ” here, New Mexico; (2) “at least 1 defendant is a defendant from whom significant relief is sought, ” “whose alleged conduct forms a significant basis for the claims asserted, ” and “who is a citizen of the State in which the action was originally filed;” (3) the “principal injuries resulting from the alleged conduct or any related conduct of each defendant were incurred in the State in which the action was originally filed;” and (4) “during the 3-year period preceding the filing of that class action, no other class action has been filed asserting the same or similar factual allegations against any of the defendants on behalf of the 3 same or other persons.” 28 U.S.C. § 1332(d)(4)(A). Plaintiffs bears the burden to show, by a preponderance of the evidence, that the exception to the CAFA is applicable. Nichols v. Chesapeake Operating, LLC, 718 Fed.Appx. 736, 740 (10th Cir. 2018) (“The preponderance of the evidence standard requires the party with the burden of proof to support its position with the greater weight of the evidence.” (internal quotation marks and citation omitted)). The Court will address each element of the local controversy exception in turn.
1. Citizenship of the Proposed Class
Plaintiffs' proposed class is defined as “all New Mexico residents who, beginning four years prior to the filing of this lawsuit, were billed by APP for amounts greater than the in-network amount permitted by their insurance provider for medical services provided at Lovelace facilities.” Doc. 1-1 ¶ 93 (emphasis added). Residency alone is not equivalent to citizenship. Reece v. AES Corp., 638 Fed.Appx. 755, 772 (10th Cir. 2016); see also Siloam Springs Hotel, L.L.C. v. Century Sur. Co., 781 F.3d 1233, 1238 (10th Cir. 2015) (“An individual's residence is not equivalent to his domicile and it is domicile that is relevant for determining citizenship.”); Nichols, 718 Fed.Appx. at 741 (“We agree with . . . this court's non-precedential decision in Reece, and other circuits that reject the applicability of a rebuttable presumption of citizenship in the context of a CAFA exception invoked based on the mere allegation of residence.”). Instead, “a person is a citizen of a state if the person is domiciled in that state.” Middleton v. Stephenson, 749 F.3d 1197, 1200 (10th Cir. 2014) (internal citation omitted). “And a person acquires domicile in a state when the person resides there and intends to remain there indefinitely.” Id. Said another way, domicile (i.e., citizenship), requires “the demonstration of two factors: residence and the intention to remain.” Preston v. Tenet Healthsystem Mem. Med. Ctr., 485 F.3d 793, 798 (5th Cir. 2007). 4
Because Plaintiffs defined their class using residency, they now face the burden to show that “greater than two-thirds of the members of all proposed plaintiff classes in the aggregate are citizens of [New Mexico].” 28 U.S.C. § 1332(d)(4)(A)(i)(I) (emphasis added). See, e.g., In re Sprint Nextel Corp., 593 F.3d 669, 676 (7th Cir. 2010) (“The plaintiffs might have defined their class as all Kansas citizens who purchased text messaging from Sprint Nextel or an alleged coconspirator. By using that definition, the plaintiffs could have guaranteed that the suit would remain in state court. There would have been no concern that out-of-state businesses, college students, soldiers, and the like comprised greater than one-third of the class, and it doesn't take any evidence to establish that Kansas citizens make up at least two-thirds of the members of a class that is open only to Kansas citizens.”); Nichols, 718 Fed.Appx. at 741 (“The need for this evidence [of citizenship] was [the plaintiff's] own making; he chose to define the class in terms 5 of residence rather than citizenship . . . By defining the class in terms of residence [plaintiff] saddled himself with an evidentiary burden.”); Reece, 638 Fed.Appx. at 774 (“We understand that evidence of class citizenship might be difficult to produce in this case. That difficulty, however, is to a considerable degree a function of the composition of the class designed by plaintiffs. The local controversy exception is designed to ensure that state courts hear cases of a truly local nature.”) (quoting Evans v. Walter Indus., Inc., 449 F.3d 1159, 1166 (8th Cir. 2006))).
At oral arguments held September 3, 2020, Plaintiffs offered an explanation as to why they did not define the class as citizens of New Mexico. Recording at 43:00. The Court summarized this explanation in a previous Opinion:
Although Plaintiffs acknowledged at oral argument that they could have easily met the first requirement of the local controversy by simply defining the class in terms of citizens rather than in terms of residents, they asserted that this would have later created complications for them at the class certification stage. In other words, Plaintiffs argue a de facto requirement of the local controversy exception to define the class in terms of citizens rather than residents would create an unfair dilemma for Plaintiffs. As everyone agrees, Plaintiffs' argument begins, an element in the test for citizenship is the intent to remain in the state. Given this, had Plaintiffs defined the class as citizens of New Mexico, Defendants at the class certification stage would argue that Plaintiffs must complete the Herculean task of demonstrating that each member of the putative class subjectively intends to remain in New Mexico. See also Doc. 35 at 6 (Plaintiffs' Reply argument that “[s]trictly requiring proof of domicile for all class members, including proof of ‘intent to remain,' would inject complex individualized inquiries into every class action where the CAFA's ‘local controversy' exception is at issue.”). The Court does not now opine on whether, had Plaintiffs defined the class as citizens rather than residents, Plaintiffs would be required to complete such a Herculean task.Doc. 39 at 21.
After allowing Plaintiffs to conduct limited discovery related to the citizenship of the proposed class members, Plaintiffs produced an expert report and filed their amended motion to remand, asserting that, in their sampling, more than two-thirds of the class members were determined to be citizens of New Mexico. Doc. 70 at 15. Plaintiffs reached this conclusion “through a sampling procedure under the direction of Professor James H. Degnan.” Doc. 70 at 6. APP provided Plaintiffs with a spreadsheet of 45, 128 lines of information, including all people who received services from APP between February 11, 2016 and February 11, 2020. Doc. 70-1 ¶ 7 (expert report of Prof. Degnan); Doc. 73-1 at 7 (Defendants' expert report). Plaintiffs' counsel removed from that list transactions in which a customer did not have a New Mexico address, did not receive a bill from APP, or received the first bill from APP after a payment had been received by APP, leaving 29, 351 class member records. Id. ¶¶ 8, 9. Prof. Degnan then conducted random sampling:
Plaintiffs' expert produced both a report and, in response to Defendants' expert report, a supplemental report. As explained in the Order filed contemporaneously, the Court strikes Plaintiffs' supplemental report. Doc. 87. Accordingly, in considering the citizenship question, the Court only takes into account Plaintiffs' original expert report.
In the Expert Report, Prof. Degnan explains that he created a random sample of 100 class members from the class list provided by Defendants. Plaintiffs then surveyed the 100 class members. The survey showed that of the 100 class members in the random sample, 52 affirmed their citizenship, 1 stated he was not a citizen, and the remaining 47 either would not respond or could not be6
contacted. Plaintiffs then “skip traced” class members to determine if additional information could be obtained about the 47 non-respondents. Skip tracing yielded information in three categories: current residential address, property ownership on February 11, 2020, and vehicle registration on February 11, 2020. Of the 47 class members who either would not respond or could not be contacted in the survey, 27 identified New Mexico data in 2 of the 3 skip trace categories (continued residence plus vehicle registration or property ownership) and did not identify non-New Mexico data in any category. In sum, through the survey or the skip trace data, 79 of the 100 class members in the random sample were identified as citizens of New Mexico as of the date of filing the complaint (February 11, 2020).
Prof. Degnan's conclusion is that “there is high confidence (95%) that the true proportion of New Mexico citizens from the population of 29, 351 individuals sampled is larger than 2/3, even if the none of the 21 remaining class members in the sample were citizens of New Mexico.” Two points are worth emphasizing. First, Prof. Degnan was able to find proof of citizenship to a much higher degree of confidence (95%) than the preponderance standard requires. Second, Prof. Degnan reached this conclusion despite the highly conservative assumption that people not contacted in the survey and not located through skip tracing are non-citizens.Doc. 70 at 6.
Defendants produced their own expert report (from Prof. Erik Erhardt) criticizing the scientific reliability of Prof. Degnan's report. For example, APP argues that “Prof. Degnan's statistical analysis is not scientifically valid, lacks evidentiary reliability, and Plaintiffs cannot meet their burden of proof.” Doc. 77 at 4. Lovelace likewise argues that Plaintiffs are unable to produce reliable evidence to establish that greater than two-thirds of the proposed class are citizens of New Mexico. Doc. 78 at 14.
Before turning to the competing expert reports, the Court reviews whether Plaintiffs used a valid sampling process.
a. Plaintiffs' Sample
As a starting point, the Court must consider whether sampling itself (i.e., not asking every member of the class, but asking a random sample and extrapolating from that sample) is an acceptable process to determine if more than two-thirds the proposed class are citizens of New 7 Mexico. In Sprint Nextel, the Seventh Circuit suggested such sampling would be an appropriate method to determine citizenship of the proposed class:
Given that there are probably hundreds of thousands of putative class members, if not more, it would be infeasible to document each class member's citizenship individually, but the district court could have relied on evidence going to the citizenship of a representative sample. This evidence might have included affidavits or survey responses in which putative class members reveal whether they intend to remain in Kansas indefinitely, or, if they are businesses, their citizenship under the relevant test. Given those results and the size of the sample and the estimated size of the proposed class, the district court could then have used statistical principles to reach a conclusion as to the likelihood that two-thirds or more of the proposed class members are citizens of Kansas.593 F.3d. at 675-76 (internal citations omitted).
In revisiting this issue, the Seventh Circuit again suggested a similar process:
Counsel for the proposed class assumed that there were only two options: determine the citizenship of every policyholder (expensive) or rely on assumptions (cheap). But there's at least one more option: take a random sample of policyholders (100, say), ascertain the citizenship of each of these on the date the case was removed, and extrapolate to the class as a whole. If the sample yields a lopsided result (say, 90% Illinois citizens or only 50% Illinois citizens) then the outcome is clear without the need for more evidence. (The more lopsided the result, the smaller the sample needed to achieve statistical significance.) If the result is close to the statutory two-thirds line, then do more sampling and hire a statistician to ensure that the larger sample produces a reliable resultMyrick v. WellPoint, Inc., 764 F.3d 662, 665 (7th Cir. 2014). Other courts have also accepted such random sampling to determine the citizenship requirement. See, e.g., Vodenichar v. Halcon Energy Properties, Inc., No. 13CV0360, 2013 WL 1386954, at *4 (W.D. Pa. Apr. 4, 2013) (accepting a sample size of 72 putative class members questioned out of 1800); Keltner v. SunCoke Energy, Inc., No. 13-cv-01374, 2015 WL 3400234, at *7 (S.D. Ill. May 26, 2015) (accepting a sample of 181 putative class members); McAllister v. St. Louis Rams, LLC, No. 16-cv-172, 2018 WL 827968, at *3 (E.D. Mo. Feb. 12, 2018) (reviewing a random sample of the proposed class to determine citizenship); Nichols, 718 Fed.Appx. at 738, 741 (considering plaintiff's sample of 100 potential class members out of 28, 929, but finding that the plaintiff 8 failed to meet his burden to prove that at least two-thirds of the proposed class were Oklahoma citizens because of the sample itself (not the sample size) was flawed).
Here, the proposed class includes potentially thousands of members and it would be nearly impossible for Plaintiffs to contact all those people to determine their citizenship. See Preston, 485 F.3d at 816 (“From a practical standpoint, class action lawsuits may become totally unworkable in a diversity case if the citizenship of all members of the class, many of them unknown, had to be considered.” (internal quotation marks and citation omitted)). Plaintiffs therefore used a sample of 100 proposed class members to conduct a statistical analysis and extrapolate information. Such sampling is an appropriate method to determine the citizenship of the entire proposed class.
For a sample to be valid, however, the sample must be representative of the population. See Hostetler v. Johnson Controls, Inc., No. 3:15-CV-226 JD, 2016 WL 3662263, at *10 (N.D. Ind. July 11, 2016) (“The first step in conducting a reliable survey is defining the universe of people to be surveyed.”). APP provided Plaintiffs with a spreadsheet of 45, 128 people who received services from APP between February 11, 2016 and February 11, 2020 and who were in-network with Lovelace and out of network with APP. Doc. 70-1 ¶ 7 (Plaintiffs' expert report); Doc. 73-1 at 7 (Defendants' expert report). Plaintiffs' counsel, in consultation with APP (see Doc. 82-4), removed from that list transactions in which a customer did not have a New Mexico address, did not receive a bill from APP, or received the first bill from APP after a payment had been received by APP, leaving 29, 351 class member records. Doc. 70-1 ¶¶ 8, 9. 9
APP says that the spreadsheet it produced to Plaintiffs contained “68, 490 lines of information” (Doc. 77 at 7) but Defendants' expert clarifies that the spreadsheet contained 45, 128 “unique name and date-of-birth combinations.” Doc. 73 at 7.
APP and Lovelace take issue with the universe of people from whom Plaintiffs drew their sample (i.e., the 29, 351 people) because that group “excluded 2, 034 persons who did not have a New Mexico address.” Doc. 77 at 7; Doc. 78 at 16. Defendants' expert, Prof. Erik Erhardt, asserts that “[t]he filtering out of non-New Mexico addresses represents potential selection bias which undermines the validity of the sampling methodology because the sample may not represent the class population.” Doc. 73-1 at 7. If the relevant question were what percentage of those treated (irrespective of residency) were New Mexico citizens, the Court would agree with Defendants.
But Plaintiffs defined their class as “all New Mexico residents who, beginning four years prior to the filing of this lawsuit, were billed by APP for amounts greater than the in-network amount permitted by their insurance provider for medical services provided at Lovelace facilities.” Doc. 1-1 ¶ 93 (emphasis added). So it follows that, when determining if greater than two-thirds of the potential class are citizens of New Mexico, Plaintiffs would start with a universe of people who are in the class, i.e., those who have New Mexico addresses and thus reside in New Mexico. In other words, by removing from the sampling universe patients without New Mexico addresses (and who thus do not reside in New Mexico), Plaintiffs were narrowing the universe to those people who fit their proposed class definition. Such narrowing is appropriate.
The Court next considers the sample Plaintiffs drew from the universe of people in the proposed class. See Hostetler, 2016 WL 3662263, at *11 (“The second step in conducting a reliable survey is to select an appropriate sample of the total population from which to take the survey.”). After receiving the list of 29, 351 potential class members, Plaintiffs' expert, Prof. Degnan, pulled a random sample of 100 names and provided that sample to Plaintiffs' counsel. 10 Doc. 70-1 ¶ 9. Prof. Degnan explains that he “randomly sampled the spreadsheet by generating 100 random numbers with replacement from 1 to 29, 351 using the statistical software language called R to create a random sample of 100 class members.” Id. The Court finds that Plaintiffs chose a truly random sample of the proposed class to survey. This random sample is different than a “sample of convenience” that is less likely to mirror the population from which the sample is drawn. For example, the plaintiffs in Hostetler had a list of 2, 130 individuals who fit the potential class. 2016 WL 3662263, at *12. In drawing a sample from this population, the plaintiffs eliminated individuals from the sample that could not be reached by telephone. Id. The court found that, by doing so, the plaintiffs created a “‘sample of convenience,' meaning the sample was chosen based on which individuals were easiest to survey, not whether they would fairly represent the composition of the class.” Id. Such a sample “creates a potential for systemic bias that would prevent the results of the survey from being reliably extrapolated from the sample to the class as a whole.” Id. The Court see no similar problem here as Plaintiffs' expert chose a random sample of 100 people from the potential class to survey without excluding any particular group from the sampling.
Once Plaintiffs had their sample group of 100 potential class members, Plaintiffs' counsel's staff attempted to contact all 100 by telephone to conduct a survey. Doc. 70 at 6. Of that group, 52 people affirmed their citizenship, 1 stated he was not a citizen, and the remaining 47 either did not respond or could not be contacted. Id. Defendants argue that having Plaintiffs' counsel's staff conduct the telephone survey tainted the survey. Doc. 77 at 8; Doc. 78 at 17. Specifically, they argue that the interviews were not impartial, there is no evidence the interviewers were properly trained, and that there is no evidence that the survey forms were checked for accuracy. Doc. 77 at 8; Doc. 78 at 17. The Court rejects these arguments because the 11 interviewers, Plaintiffs' counsel's staff, used a specific script to conduct the survey, one that left little, if any, room to indicate a bias and one that would not require any specialized training. See Doc. 82-3 (script). The script was short, included an introduction paragraph stating that the caller is from Plaintiffs' counsel's law firm, and contained only three questions. Id. Additionally, defense counsel was involved in designing the script. See Doc. 47 at 24 (discovery order requiring Plaintiffs to provide to Defendants a script of any questions they intend to ask class members and setting some guidelines for the survey); Doc. 82-2 (email between the parties revising the script). Defendants point to nothing specific that raises concerns about the impartiality or accuracy of the telephone survey results. Instead, the Court trusts Plaintiffs' representation that 52 people affirmed their citizenship and 1 stated he was not a citizen. Plaintiffs' counsel is, after all, bound by Federal Rule of Civil Procedure 11, regarding representations to the Court, and the Court has no reason to doubt the survey results.
Defendants also argue that Plaintiffs' survey results are not objective and reliable because Plaintiffs do not account for the high rate of non-responses. Doc. 77 at 9; Doc. 78 at 18. True, of the 100 people in the random sample group, a high number, 47, did not respond or could not be contacted for the telephone survey. However, instead of removing those 47 people from the sample (as the plaintiff in Hostetler did en route to creating a “sample of convenience”), Plaintiffs used other information to attempt to determine their citizenship. See In re Text Messaging Antitrust Litig., Nos. 08 C 7082, 09 C 2192, 2011 WL 305385, at *3 (N.D. Ill. Jan. 21, 2011) (rejecting the argument that any sample bias existed where responses were only obtained from fifteen percent of the sample population because “voter registration information, driver's license registration . . . and Internet search information supports plaintiffs' contention that survey non-respondents had similar rates of Kansas citizenship as respondents”). 12
Specifically, anticipating a high number of non-responses to their survey, before conducting the survey, Plaintiffs' counsel retained American Legal Claims Services to obtain data on all 100 people in the sample group in the following categories: current residential address, property ownership on February 11, 2020, and vehicle registration on February 11, 2020. Doc. 70-1 ¶ 15. This data corroborated the results of the telephone survey of the 52 people who affirmed New Mexico citizenship, as no non-New Mexico data (i.e., data from another state) in any category was associated with these 52 individuals. Id. ¶ 16.
Doc. 70-1 does not indicate when Plaintiffs chose the indicia of citizenship criteria on which they relied. At oral argument, however, Plaintiffs clarified that they did so prior to conducting the survey.
Regarding the 47 people who did not respond to the survey, Plaintiffs used the results of their investigation in the above three categories (a process Plaintiffs refer to as “skip-tracing”) to attempt to determine their citizenship. See Doc. 70-1 ¶ 16. If a person had New Mexico data in 2 of the 3 categories (continued residence plus vehicle registration or property ownership) and did not identify data from another state (i.e., non-New Mexico data) in any category, Plaintiffs categorized that person as a New Mexico citizen. Id. ¶ 16. Through this process, Plaintiffs categorized 27 of the non-responsive people as New Mexico citizens. Plaintiffs categorized the remaining 20 people as non-New Mexico citizens either because (1) the information discovered indicated the person was a non-New Mexico citizen or (2) not enough information was discovered to decide and so the person was assumed to be non-New Mexican. Id. 13
Plaintiffs' expert report states only that the skip trace revealed data in “2 of the 3 categories, ” which could be read to mean any 2 of the 3 categories were met. Doc. 70-1 ¶ 16. However, in their briefs, both parties clarify that the “2 of 3” test meant a finding of continuous residence plus at least one of the two other categories (vehicle registration or property ownership). Doc. 70 at 6; Doc. 77 at 11.
Plaintiffs' decision to categorize as non-citizens all persons about whom insufficient citizenship data was found (either through the telephone survey or skip tracing), is important. See Doc. 70-1 ¶¶ 16-17. In Hostetler, which Defendants cite to support their argument, the court reasoned that “a reliable survey requires an analysis of whether the results were impacted by non-response bias, which occurs when the nonresponses to a survey are not random, but correlate to a trait the survey is intended to measure.” 2016 WL 3662263, at *12. Plaintiffs here introduced a non-response bias into their survey: by assuming all people about whom they had insufficient information were non-New Mexico citizens, Plaintiffs may have undercounted the true number of individuals in the survey who are New Mexico citizens. This bias, however, is not prejudicial to Defendants because it can tilt only in their favor. Thus, the only way Defendants can be prejudiced by a non-response bias is if the “2 of 3” test that led Plaintiffs to categorize 27 of the non-responders as New Mexico citizens is unreliable.
The next question for the Court to consider is, therefore, whether the skip tracing and the “2 of 3” test used by Plaintiffs is sufficient to meet their burden to establish citizenship of the non-responders in the sample. As discussed above, citizenship is determined by domicile and “a person acquires domicile in a state when the person resides there and intends to remain there indefinitely.” Middleton, 749 F.3d at 1200. In considering domicile for diversity jurisdiction, the Tenth Circuit has directed district courts to use an “all-things-considered” approach, considering indicia of citizenship such as “the party's current residence; voter registration and voting practices; situs of personal and real property; location of brokerage and bank accounts; membership in unions, fraternal organizations, churches, clubs, and other associations; place of employment or business; driver's license and automobile registration; payment of taxes; as well as several other aspects of human life and activity.” Middleton, 749 F.3d at 1201 (internal 14 quotation marks and citation omitted); see also Jeter v. Wild W. Gas, LLC, No. 12-CV-411-TCK-PJC, 2014 WL 3890308, at *6 (N.D. Okla. Aug. 7, 2014) (applying Middleton when determining citizenship under the CAFA exceptions).
Here, the Court finds that Plaintiffs' “2 of 3” test sufficient to establish citizenship for purposes of this sample. First, as discussed above, Plaintiffs' sample included only people who resided in New Mexico on February 11, 2020. To establish intent to remain, Plaintiffs used the current residential address (meaning continued residency) and either property ownership or vehicle registration on February 11, 2020. Further, any person who showed non-New Mexico data (i.e., data from another state) in any of the three categories was considered a non-citizen. Defendants argue that property ownership and vehicle registration are poor proxies for inferring citizenship. Doc. 77 at 11; Doc. 78 at 19. The Court agrees that any one factor alone (continued residency, property ownership, etc.) may not be enough to assume citizenship. See Reece, 638 Fed.Appx. at 769 (“[M]ere property ownership in a state does not necessarily equate to citizenship in that state; a person may own property in a state without either being a state resident or intending to remain there.”). But, in this case, a person's prior residence in New Mexico (when the complaint was filed on February 11, 2021), their continued current residence in New Mexico, property ownership or vehicle registration in New Mexico, and lack of attachment to another state through property ownership or vehicle registration, is sufficient indicia of citizenship for purposes of the local controversy exception. Although other indicia of citizenship would certainly strengthen the showing of citizenship, Plaintiffs must only establish citizenship by a preponderance of the evidence, which they have done.
In sum, the Court finds that Plaintiffs properly used a random sample of proposed class members to gather information from a telephone survey or skip tracing to determine the 15 citizenship of the sample group. The Court must next consider Plaintiffs' expert's analysis of that information.
b. Expert Reports
Based on the telephone survey results and the skip-trace data, Plaintiffs identified 79 of the 100 potential class members in the sample group as citizens of New Mexico on February 11, 2020 (when Plaintiffs filed the complaint). Doc. 70-1 ¶ 17. Plaintiffs provided this information to their expert, Prof. Degnan, who concluded:
A 95% confidence interval for this data is from 69.7% to 86.5%, meaning that there is high confidence (95%) that the true proportion of New Mexico citizens from the population of 29, 351 individuals sampled is larger than 2/3, even if none of the 21 remaining class members in the sample were citizens of New Mexico. Statistically, there is very strong evidence that the true proportion of New Mexico citizens from this population is larger than 2/3.Id. ¶ 18.
As an initial matter, APP argues that because Prof. Degnan was not involved in the data selection and collection (rather, Plaintiffs' counsel provided him the results of the telephone survey and skip trace data), his analysis is not reliable. Doc. 77 at 10 n.11. To support this point, APP cites Hostetler, in which the court questioned a statistical expert who simply plugged numbers into a formula to calculate a confidence interval. 2016 WL 3662263, at *13. In that case, however, the court found errors with the representativeness of the sampled class and the expert did not investigate those errors. Id. at *10, *12. Such analysis is not applicable here because, as discussed above, the Court finds no error with Plaintiffs' sample group or the resulting data given to Prof. Degnan.
In response to Prof. Degnan's analysis, Defendants did not conduct an independent investigation of citizenship or introduce any evidence to show that any of the 79 people in the sample group are not citizens of New Mexico (nor are they required to produce such evidence, as 16 Plaintiffs carry the burden of proof). Defendants' argument instead is that Plaintiffs have not met their burden and, in support, submit testimony from their expert, Prof. Erhardt, who asserts that the statistical procedures used by Plaintiffs' expert, Prof. Degnan, are unreliable.
In their briefs, Defendants cite Federal Rule of Evidence 702 for admissibility of expert testimony. Rather than moving to exclude Prof. Degnan's expert report under Rule 702, however, they use their own expert to criticize Prof. Degnan's conclusions.
First, Defendants complain that Prof. Degnan used confidence interval testing instead of p-value testing (also referred to as hypothesis testing). To support their position that confidence interval testing is unreliable, APP cites the Reference Manual on Scientific Evidence (Third, 2011) for the proposition that a “confidence interval is unsuitable for making probability statements about population characteristics.” Doc. 77 at 13 (citing Reference Manual at 247). This is true. But this statement in no way indicates that confidence interval testing is inferior to p-value testing. This statement simply recognizes that the characteristics of a population are already defined, meaning they do not change. Because the characteristics of a population are already set at any given time, they cannot be affected by probability or chance. Thus, talking about population characteristics in terms of probability is inaccurate. Those characteristics either do or do not exist in the population-probability has nothing to do with it. In contrast, whether a random sample of the population accurately captures the characteristics of a population is subject to chance/probability. See Reference Manual at 149 (explaining that “[p]robability statements 17 apply to the behavior of samples” and that, in contrast, “[a]ccording to the frequentists theory of statistics, probability statements cannot be made about population characteristics”).
In his deposition, Prof. Degnan stated that he did p-value testing on the original sample, but that he did not include it in his report because he thought confidence interval testing would be easier to understand. Doc. 77-1 at 35: 11-25 (Prof. Degnan deposition). The Court will not consider the p-value testing not included in Prof. Degnan's original report. Prof. Degnan also conducted p-value testing in his supplemental report, but the Court strikes that report. See Doc. 87.
By definition, when a statistician uses a 95% confidence interval, that statistician estimates that a sample to be drawn from the population will fail to capture the mean population 1 out of 20 times. For the statistician who draws such a non-representative sample, however, the likelihood that the confidence interval of this non-representative sample will fail to capture the population mean is 100%, as the population mean (a constant at any given time) simply falls outside the 95% confidence interval of the sample.
Thus, while it is accurate to say that a “confidence interval is unsuitable for making probability statements about population characteristics” it would be just as accurate to say that “a p-value test is unsuitable for making probability statements about population characteristics.” See Id. at 252 (“It is easy to mistake the p-value for the probability of the null hypothesis given the data . . . probabilities govern the samples, not the models and hypotheses.”). That is because, regardless of the test being employed, it is unsuitable to make probability statements about population characteristics. It is suitable, however, to make statements about how confident we are that a sample interval captured a population's characteristic, such as its mean.
Further, nothing in the Reference Manual supports the notion that p-value testing is valid, whereas confidence interval testing is not. To the contrary, the Reference Manual supports the use of confidence interval testing, particularly when it comes to analyzing survey responses from a random sample. See Id. at 371 (stating “one of the advantages probability sampling has over other types of sampling is that the researcher can calculate a confidence interval that describes explicitly how reliable the sample estimate of the population is likely to be”).
The Reference Manual also indicates that confidence intervals are capable of providing information simple significance testing cannot provide. Id. at 579-80 (“Calculations of a confidence interval permit a more refined assessment of appropriate inferences about the association found in an epidemiologic study, ” and “[t]he advantage of a confidence interval is that it displays more information than significance testing.”). Thus, one advantage of a 18 confidence interval is that it can provide a range of possible effect sizes compatible with the data. A confidence interval can provide information about how close or far a number is from a cut-off point, whereas p-values simply indicate statistical significance based on whether a value is to the left or right of a cut-off point. Thus, in addition to determining whether a result is statistically significant, a confidence interval can provide information about exactly where a value falls on a bell curve.
Further, Defendants fail to cite case law that supports the notion that p-value testing is a valid methodology whereas confidence interval testing is not. The case Defendants cited at oral argument, Frappied v. Affinity Gaming Black Hawk, did consider p-value test results to evaluate a disparate treatment claim under Title VII. 966 F.3d 1038, 1052 (10th Cir. 2020). Frappied, however, did not discuss the validity of confidence interval testing or opine about whether confidence interval testing or p-value testing is superior. That the Tenth Circuit in Frappied accepted an expert's use of p-value testing does not mean that it rejected other types of analysis, such as confidence interval testing.
Indeed, p-value tests and confidence interval tests share the same underlying statistical foundation. For both tests, a researcher draws a sample from a population. Reference Manual at 241. Assuming a normal distribution, this sample creates a range of values that resemble a bell curve. See id. at 277-79, 290. Because of random error, this sample is unlikely to exactly match the population. Id. at 241. Thus, a statistician looks at factors such as the size of the sample and variance within the sample to create a confidence interval. Id. A confidence interval is “[a]n estimate, expressed as a range, for a parameter. For estimates such as averages or rates computed from large samples, a 95% confidence interval is the range from about two standard errors below to two standard errors above the estimate. Intervals obtained this way cover the true value about 19 95% of the time, and 95% is the confidence level or the confidence coefficient.” Id. at 285. Thus, in samples such as the one used in the present case, a statistician can take the mean of the sample and mark about two standard deviations on the bell curve on each side of the mean to obtain a confidence interval of 95%.
By definition, the true mean of the population is likely to fall within this confidence interval 95% of the time. Id. at 381 (“Traditionally, scientists adopt the 95% level of confidence, which means that if 100 samples of the same size were drawn, the confidence interval expected for at least 95 of the samples would be expected to include the true population value.”). Thus, one might say that a number that falls outside this confidence interval is statistically significant in that the true sample mean is expected to fall within either of the far ends of this bell curve (past the cut-off points) only 5% of the time. Placing cut-off points (measured by standard error) even farther away from the sample mean provides an even greater level of confidence that numbers falling on the far extremes of the bell curve, past the cut-off points, are statistically significant. See id. at 245-46.
Like confidence interval tests, p-value tests assume that the population and population sample (if correctly drawn) have a normal distribution (i.e., form a bell curve). See id. at 249-57, 277-79. In a p-value test, however, a null hypotheses is assumed (for example, the null hypothesis for a drug company trying to determine if a newly developed drug has a significant effect would be that the drug being studied had no effect). See id. at 241. Nonetheless as above, a cut-off is chosen to mark a point beyond which a number (such as the sample mean) will be considered statistically significant. Thus, although there are differences in the two tests, both tests are founded on the same statistical principles and can be used to assess the likelihood that a sample captures the true characteristics of a population. 20
Finally, although Defendant Lovelace's expert used p-value testing, it does not appear that he opined that Plaintiff's expert erred in choosing to use confidence interval testing. And, even if that were the import of his testimony, for the reasons stated above, the Court would find that Prof. Degnan's use of confidence interval testing employed a valid statistical methodology. Having determined that Plaintiffs chose a valid test for their statistical analysis, the Court turns to Defendants' arguments that bear on how Plaintiffs performed their statistical analysis.
Prof. Erhardt does take issue with Prof. Degnan's use of a two-sided rather than one-sided confidence interval test. Doc. 73-1 at 31 (noting his p-value testing “differs from the less powerful two-sided confidence interval approach that Prof. Degnan used . . .”). To the Court's knowledge, however, Prof. Erhardt did not opine that the use of a two-sided rather than onesided confidence interval test would affect the validity of Prof. Degnan's ultimate opinion and Defendants do not make such an argument.
Prof. Erhardt asserts that Prof. Degnan engaged in HARKing (hypothesizing after the results are known) and data dredging. Prof. Erhardt explains that because the telephone survey results alone would not permit a conclusion that more than two-thirds of the class are New Mexico citizens, Plaintiffs knew they would need to “create a means of demonstrating that 75 individuals (or 77 using Prof. Degnan's two-sided confidence interval testing) were New Mexico citizens based on the sample size (100) and confidence interval level (95%).” Doc. 73 at 4 (Prof. Erhardt report). To do this, Plaintiffs conducted skip tracing as discussed above. Originally, when briefing their plan for limited jurisdictional discovery, Plaintiffs proposed using drivers licenses, vehicle registration, and voter registration to determine citizenship. Doc. 44 at 11. However, Plaintiffs later changed the skip trace criteria to current residential address, property ownership, and vehicle registration. Doc. 70-1 ¶ 15. Prof. Erhardt asserts that, by changing the criteria, Prof. Degnan engaged in HARKing in order to receive results that confirm the conclusion Plaintiffs sought. Doc. 73 at 5. This improper change of plan for sampling and 21 testing, Prof. Erhardt asserts, “invalidates Prof. Degnan's confirmatory hypothesis testing conclusion.” Doc. 73 at 6.
In other words, Prof. Erhardt criticizes Prof. Degnan for not precisely following his initial proposal that Plaintiffs provided to the Court when explaining how they would conduct limited jurisdictional discovery. However, that initial proposal was an example of how sampling could take place if the Court allowed such limited discovery and was provided before Plaintiffs received a class member list. See Doc. 44 at 11 ¶ 11 (indicating that Plaintiffs' counsel was investigating whether vehicle registration and voter registration could be obtained). It does not appear that Plaintiffs intended this example of how sampling could take place to be a final determination about which criteria they would ultimately use. Indeed, Prof. Degnan confirmed that “[o]nce a [class member] list is provided, I will be able to advise Plaintiffs immediately as to how to proceed with a sampling procedure.” Id. at 11 ¶ 13. In the present motion, Plaintiffs clarify that, after they received the class list, they learned they were not able to obtain vehicle registration and voter registration from the State. Doc. 82-5. Plaintiffs therefore sought skip tracing data on other information (current residential address, property ownership, and vehicle registration) that courts have found to be reliable indicia of citizenship and provided that information to Prof. Degnan to analyze.
Prof. Degnan, therefore, did not receive data based on his original proposal (telephone survey plus drivers licenses, vehicle registration, and voter registration), run an analysis that reached a result he (or Plaintiffs) did not want, and then change the criteria (to current residential address, property ownership, and vehicle registration) in order to reach a different result. Said another way, Prof. Degnan did not “chang[e] aspects of the analysis with knowledge of the data” as Prof. Erhardt suggests. Doc. 73 at 6. Instead, Plaintiffs' counsel provided Prof. Degnan with a 22 different set of criteria (data) before he began his analysis. Although what data was available from the State ultimately dictated what data Plaintiffs chose to use, what is important in determining whether Plaintiffs made a hypothesis after the results were known (engaged in HARKing), is whether Plaintiffs knew the results before they selected which data to use. The facts demonstrate that Plaintiffs selected the criteria (data) to be used and hypothesized about the significance of that data (by developing their “2 of 3 test”) before they conducted their survey. As a result, in choosing, before conducting their survey, not to use drivers licenses and voter registration data as part of their citizenship criteria, Plaintiffs did not engage in HARKing.
At oral arguments, Plaintiffs did inform the Court that, in addition to the skip trace criteria they used to determine citizenship (current residence, vehicle registration, and property ownership), they also gathered employment data. However, Plaintiffs chose not to use this data because they could not draw inferences from it, one way or another, due to various problems such as old information and that some data lacked corresponding addresses or provided an address of a corporate office instead of a local office. In other words, the data could not be used to reliably determine whether a person was employed in New Mexico. This meant that one of the indicia of citizenship Plaintiffs had planned on using-employment in New Mexico-turned out not to be available.
That Plaintiffs developed a test for citizenship using certain criteria and then, after performing their survey, omitted one of those criteria does create a concern. The Court finds, however, that this concern is not fatal to Plaintiffs' statistical analysis. First, although employment data can be a useful tool in determining a person's citizenship, data about a person's employment is not a necessary component in the citizenship analysis. Second, Plaintiffs provided the employment data (and all skip trace data) to Defendants and neither Defendants nor 23 Defendants' expert, Prof. Erhardt, make any arguments specific to Plaintiffs' failure to use the employment data. They only criticize Prof. Degnan for failing to follow his original plan of using drivers licenses, vehicle registration, and voter registration data, as the three criteria to determine citizenship.
Third, and most importantly, Plaintiffs did not run an analysis using the employment data, decide they did not like the outcome, and then exclude that data from their analysis to reach a different outcome. Instead, they never analyzed the employment data because the data was insufficient to make inferences about the employee's relation to New Mexico. Thus, at the time they chose not to use employment data as a criterion, they did not know if that data would have helped them or hurt them. In other words, they did not change their hypothesis about what criteria would indicate New Mexican citizenship before the results of their survey were known. The concern about omitting employment data as one of the criteria for determining citizenship, therefore, cannot be accurately characterized as a HARKing concern. Instead, the concern is whether, after dropping one of the criteria determined to be relevant to citizenship, the remaining criteria can provide sufficiently reliable indicia of citizenship. As to this question, the Court has already concluded that the criteria Plaintiffs used (continued New Mexico residency plus either New Mexico vehicle registration or property ownership, plus no out-of-state positive result in any of these criteria) was sufficient to determine citizenship for purposes of jurisdiction in this class-action.
Finally, Defendants argue that Plaintiffs improperly apply the preponderance of the evidence standard to the interpretation of statistical evidence. Doc. 77 at 18. In essence, Plaintiffs assert that their expert's use of a 95% confidence interval far exceeds what is necessary for the Court to determine citizenship by a preponderance of the evidence, which only requires a finding 24 of citizenship with greater than 50% confidence. See Doc. 70 at 5-7, Doc. 82 at 9-11. In support of this argument, Plaintiffs cite the Seventh Circuit's decision in Sprint Nextel. Doc. 82 at 10. There, the court reasoned that “[s]tatisticians and scientists usually want at least 95 percent certainty, but any number greater than 50 percent would have allowed the district court to conclude that the plaintiffs had established the citizenship requirement by a preponderance of the evidence.” In re Sprint Nextel Corp., 593 F.3d 669, 676 (7th Cir. 2010). This language appears to mix the preponderance of the evidence standard with the standard for scientific reliability. A scientific analysis must be reliable under scientific principles for the Court to accept. Plaintiffs have presented no evidence that statisticians would find statistical significance based on a confidence interval of 51% and the Court would not be inclined to accept an opinion that employed a confidence interval rejected by the scientific community.
As the Reference Manual notes, it is error to equate the level of the confidence interval chosen with the legal burden of proof. Reference Manual at 577 n.1. The Court need not further delve into the many reasons this is true, however, because Plaintiffs' expert in this case applied a confidence interval of 95%, which is an interval commonly used by statisticians. See Doc. 72 at 4 (Defendants' expert, Prof. Erhardt, discussing proper application of a 95% confidence interval level); Reference Manual at 381 (“Traditionally, scientists adopt the 95% level of confidence, which means that if 100 samples of the same size were drawn, the confidence interval expected for at least 95 of the samples would be expected to include the true population value.”). Therefore, the confidence interval Prof. Degnan used in no way undermines his conclusions. Based on the statistical evidence Plaintiffs have presented, the Court finds that Plaintiffs have met their burden to show, by a preponderance of the evidence, that more than two-thirds of the proposed class members are citizens of New Mexico. 25
2. Significant Local Defendant
To meet the second element of the local controversy exception, Plaintiffs must show that “at least 1 defendant is a defendant from whom significant relief is sought, ” “whose alleged conduct forms a significant basis for the claims asserted, ” and “who is a citizen of the State in which the action was originally filed.” 28 U.S.C. § 1332(d)(4)(A)(i)(II). Said another way, “for the local controversy exception to apply there must be ‘at least one real defendant . . . whose alleged conduct is central to the class's claims and from whom the class seeks significant relief.'” Woods v. Standard Ins. Co., 771 F.3d 1257, 1266 (10th Cir. 2014) (citing S. Rep. No. 109-14 at 28). The Tenth Circuit has “interpret[ed] the significant local defendant requirement strictly so that plaintiffs and their attorneys may not defeat CAFA jurisdiction by routinely naming at least one state citizen as a defendant, irrespective of whether that defendant is actually a primary focus of the litigation.” Id. at 1265.
In this case, Defendant APP is undisputedly not a citizen of New Mexico and therefore does not meet this element. Lovelace, however, undisputedly is a citizen of New Mexico. The parties therefore focus their arguments on whether Defendant Lovelace is a defendant from whom significant relief is sought and whose alleged conduct forms a significant basis of the claims asserted. The Court will address each question in turn.
a. Significant Basis
The significant basis element requires that the local defendant's conduct “forms a significant basis for the claims asserted by the class.” Woods, 771 F.3d at 1266. Thus, to begin, the Court examines the claims in Plaintiffs' complaint. The complaint alleges that Lovelace had a business relationship with APP for APP to provide emergency room physicians and nurse practitioners at Lovelace facilities. Doc. 1-1 ¶ 13. The named Plaintiffs and proposed class 26 members sought treatment at Lovelace health facilities, facilities that were in-network with their insurance plans, but were treated in part by APP employees. Doc. 1 ¶¶ 16, 27, 32, 44, 49, 61, 66, 77, 83. Following Plaintiffs' visits, APP billed Plaintiffs for the out-of-network amount, even though it was only entitled to the in-network amount. Id. ¶¶ 17, 37, 54, 71, 88. Although Plaintiffs allege that APP (not Lovelace) overbilled them, they also allege that Lovelace is part of a joint enterprise with APP, including having a profit-sharing agreement where Lovelace benefits from APP's overbilling practices. Id. ¶¶ 21, 22, 23. They allege that Lovelace hides APP's billing practices by holding itself out as in-network for many insurance providers, not advertising to the public that many of its services are provided by APP, and not disclosing to patients, including failing to disclose to any of the named Plaintiffs, that they are being treated by APP employees or that they would receive a separate bill. Id. ¶¶ 16, 18, 19, 33, 50, 67, 84.
Reviewing these claims, although Plaintiffs allege that APP was the party responsible for the actual overbilling, they also allege Lovelace was a significant player in an alleged overbilling scheme. They allege that Lovelace interacted with all of the proposed class members by advertising itself as in-network, failing to disclose its relationship with APP and APP's billing practices, treating the Plaintiffs, and then benefitting from APP's overbilling. Indeed, they allege that both Lovelace and APP are agents for the other and ratified each other's misconduct. Id. ¶¶ 24-25.
For comparison, this differs greatly from other instances in which a local defendant's conduct did not form a significant basis of the case. The plaintiffs in Woods alleged they paid for insurance coverage through payroll deductions, but then did not receive the benefits of this insurance. 771 F.3d at 1260. They filed their proposed class action against an insurance company, the state agency that contracted the insurance company, and an insurance agent who 27 managed the accounts. Id. The Tenth Circuit considered whether the conduct of the insurance agent, the only local defendant, formed a significant basis of the claims asserted by the plaintiffs. Id. at 1265. It found, when comparing the agent's conduct with that of the other named defendants, that the agent was not “a real target of the litigation, rather than an isolated role player in the alleged scheme implemented by” the insurance company. Id. at 1266. Likewise, the Senate Judiciary Committee report explains that in cases of consumer fraud against an insurance company, a local insurance agent
presumably would not be a person whose alleged conduct forms a significant basis for the claims asserted. At most, that agent would have been an isolated role player in the alleged scheme implemented by the insurance company. In this instance, the real target in this action (both in terms of relief and alleged conduct) is the insurance company, and if that company is not local, this criterion [for the local controversy exception] would not be met.Id. (citing S. Rep. 109-14, at 38) (emphasis in original).
Unlike an insurance agent who has an isolated role, Plaintiffs here allege that Lovelace is a real target of the litigation. Plaintiffs have provided information about Lovelace's conduct relative to APP's conduct. See Fulgenzi v. Smith, No. 1:12-CV-1261 RB/RHS, 2014 WL 11497836, at *4 (D.N.M. Nov. 13, 2014) (“Where the complaint contains no information about the local defendant's conduct relative to the other defendants' conduct, the plaintiff fails to meet [the significant basis] requirement.”). They allege that Lovelace induced patients to seek treatment at Lovelace facilities by advertising it as in-network and hiding Lovelace's relationship with APP such that APP could then overbill patients to both defendants' benefit.
Defendants argue that Lovelace's conduct does not form a significant basis of the case because Plaintiffs allege that APP overbilled, not Lovelace. Doc. 77 at 20; Doc. 78 at 23 (incorporating Doc. 30 at 15). Indeed, Defendant APP asserts that “without the overbilling allegedly committed by APP, Plaintiffs would not have a claim against Lovelace, period.” Doc. 28 77 at 20. But the inverse is also true. APP would not be in a position to overbill had Lovelace not advertised itself as in-network, treated the patients, contracted with APP, and failed to disclose to patients its relationship with APP. So, although Plaintiffs do not allege Lovelace overbilled, they allege that Lovelace's conduct formed a significant basis of the overbilling scheme.
In arguing that Lovelace's conduct does not form a significant basis of the case, APP cites the deposition of its 30(b)(6) representative Elizabeth Maxey. Doc. 77 at 21-22. It points to three pieces of Ms. Maxey's testimony. First, she verified the list of individuals who received services from APP at Lovelace facilities in New Mexico; second, she explained that APP does not overbill or balance bill; and third, she confirmed that Lovelace plays no part in APP's billing. Id.
Even accepting this extrinsic evidence, the Court finds that Lovelace's conduct forms a significant basis of Plaintiffs' class claims. First, that Ms. Maxey verified the list of individuals who received services from APP at Lovelace facilities does not bear on Plaintiffs' allegations of Lovelace's wrongdoing. Second, Plaintiffs' lawsuit is premised on its allegations that APP overbilled and/or balance billed. This factual dispute cannot be resolved at this early stage of the litigation based on Ms. Maxey's assertion that Defendant APP did not do the wrong Plaintiffs allege. Third, as set forth above, that Lovelace played no part in APP's billing does not mean that Lovelace had no part in the scheme Plaintiffs describe in the complaint. Plaintiffs' allegations against Lovelace are not that it overbilled patients but that it played a significant role in the 29 overbilling scheme by contracting with APP and failing to disclose its relationship with APP, which allowed APP to overbill to the benefits of both defendants.
In evaluating significant basis and significant relief, it is unsettled whether the Court should consider just Plaintiffs' allegations in the complaint or if it should also consider extrinsic evidence. See Woods, 771 F.3d. 1267 n.6 (collecting cases and declining to decide whether courts should consider extrinsic evidence).
b. Significant Relief
To satisfy the significant relief element, the local defendant must be “a primary focus of the plaintiffs' claims - not just a peripheral defendant.” Woods, 771 F.3d at 1266. In other words, the local defendant must be “a target from whom significant relief is sought by the class (as opposed to just a subset of the class membership).” Id. “The Tenth Circuit has held that the ‘significant relief' requirement of the exception is satisfied where the complaint ‘claims that every potential plaintiff is entitled to recover from [the local defendant] and the proposed class seeks to recover damages from all defendants jointly and severally.'” Fulgenzi, 2014 WL 11497836, at *3 (citing Coffey v. Freeport McMoran Copper & Gold, 581 F.3d 1240, 1244 (10th Cir. 2009)). This element requires only a local “defendant from whom significant relief is sought [and] does not mean a defendant from whom significant relief may be obtained, ” such that a court is not required to “assess[] the financial viability of a defendant as part of this preliminary consideration.” Coffey, 581 F.3d at 1245.
Here, Plaintiffs' complaint makes clear that Lovelace is a primary target of litigation and a defendant from whom the entire class of plaintiffs seeks relief. For example, the complaint alleges that “Lovelace and APP are part of a joint enterprise for the purpose of providing emergency room services for profit, including, upon information and belief, a profit sharing arrangement” and that “Lovelace and APP authorized, participated in, and ratified each other's misconduct.” Doc. 1-1 ¶¶ 22, 25. They allege Defendants are jointly and severally liable for all damages and that “each are directly liable; are liable for aiding, abetting, participating in, and ratifying the other parties' conduct; are vicariously liable for each other's misconduct; and are 30 liable as members of a joint enterprise.” Id. ¶¶ 104, 110, 117, 121. Plaintiffs bring five claims for relief, each against both Lovelace and APP for their joint overbilling scheme. Id. at 11-15. These allegations go beyond mere conclusory allegation of joint and several liability. See Woods, 771 F.3d at 1269 (holding that a statement of joint and several liability is not enough to alter the actual significance of an otherwise insignificant defendant).
In its response to the original motion to remand (which it incorporates into its response to the amended motion to remand), Lovelace attacks the viability of each claim brought against it to argue that Plaintiffs do not seek significant relief from it. Doc. 30 at 16-20. This argument is presently misplaced as the Court is not considering the merits of the claims, only whether those claims seek significant relief from Lovelace. Further, the claims do not appear frivolous on their face or brought just for jurisdictional purposes.
Turning again to the example of a local insurance agent in an insurance fraud case, the Senate Judiciary Committee explained that
in a consumer fraud case alleging that an insurance company incorporated and based in another state misrepresented its policies, a local agent of the company named as a defendant presumably would not fit this criteria. He or she probably would have had contact with only some of the purported class members and thus would not be a person from whom significant relief would be sought by the plaintiff class viewed as a whole.Woods, 771 F.3d at 1266 (citing S. Rep. 109-14, at 38) (emphasis in original). Here, Lovelace interacted with the entire class because the entire proposed class sought treatment at a Lovelace facility, and the entire proposed class seeks relief from Lovelace.
In sum, the Court finds that Lovelace, the local defendant, is a defendant from whom significant relief is sought and whose alleged conduct forms a significant basis of the claims asserted. 31
3. Principal Injuries
Element three of the local controversy exception requires that the “principal injuries resulting from the alleged conduct or any related conduct of each defendant were incurred in the State in which the action was originally filed, ” i.e., in New Mexico. 28 U.S.C. § 1332(d)(4)(A). In their complaint, Plaintiffs allege that they were overbilled after seeking treatment at Lovelace facilities in New Mexico. Doc. 1-1 ¶¶ 27, 34, 44, 51, 61, 68, 77, 85. Indeed, the proposed class includes only those people who are New Mexico residents and who sought treatment at Lovelace facilities. Id. ¶ 93. Lovelace Health Systems, for its part, “owns and operates 7 hospitals, 26 clinics, and 5 emergency rooms around the state of New Mexico.” Id. ¶ 11. It is clear from these allegations that the principal injuries alleged by Plaintiffs resulted from conduct that occurred in New Mexico. Further, Defendants do not dispute that this element is met.
4. No Other Class Actions
Finally, to fall under the local controversy exception, “during the 3-year period preceding the filing of that class action, no other class action has been filed asserting the same or similar factual allegations against any of the defendants on behalf of the same or other persons.” 28 U.S.C. § 1332(d)(4)(A). Plaintiffs attach to their amended motion to remand a declaration by their counsel confirming that he has “found no class actions asserting the same or similar allegations filed against Defendants in the three years preceding the filing of this class action.” Doc. 70-3 ¶ 3. Defendants do not dispute that this element is met and the Court finds that Plaintiffs have met their burden for this element.
CONCLUSION
In sum “[t]his case presents a classic example of what Congress intended to cover when it created this exception. It is a truly local controversy-a controversy that uniquely affects a 32 particular locality to the exclusion of all others.” Coffey, 581 F.3d at 1243 (internal quotation marks and citation omitted). Plaintiffs have met their burden to show by a preponderance of the evidence that more than two-thirds of the proposed class are citizens of New Mexico who were affected by a purely local incident-overbilling after being treated at a New Mexico Lovelace hospital. The Court therefore GRANTS Plaintiffs' Amended Motion for Remand (Doc. 70). This matter is remanded to the Second Judicial District Court.
Lastly, Plaintiffs request, in addition to remand, that they be awarded reasonable attorney's fees and costs incurred as a result of removal. Doc. 70 at 16. Indeed, 28 U.S.C. § 1447(c) provides that “[a]n order remanding the case may require payment of just costs and any actual expenses, including attorney fees, incurred as a result of the removal.” “Absent unusual circumstances, courts may award attorney's fees under § 1447(c) only where the removing party lacked an objectively reasonable basis for seeking removal.” Martin v. Franklin Cap. Corp., 546 U.S. 132, 136 (2005). The Court finds that the removing party (APP) had an objectively reasonable basis for seeking removal, as demonstrated in the Court's first Memorandum Opinion and Order regarding the amount in controversy under the CAFA (Doc. 39) and the over-year long process required to determine if an exception to the CAFA applies. Accordingly, the Court denies Plaintiffs' request for expenses. 33