OPINION

An employee of Wells Fargo, whom we will call Jane Jones, was covered under the Wells Fargo & Company Health Plan (the “Plan”), governed by the Employee Retirement Income Security Act of 1974 (“ERISA”). United Behavioral Health (“UBH”) is a third-party claims administrator of the Plan. Jones was admitted to Pacific Shores Hospital (“PSH”) for acute inpatient treatment for severe anorexia nervosa. UBH refused to pay for more than three weeks of inpatient hospital treatment. UBH based its refusal in substantial part on mischaracterizations of Jones’s medical history and condition. PSH continued to provide inpatient treatment to Jones after UBH refused to pay. Jones assigned to PSH her rights to payment under the Plan.

PSH sued the Plan and UBH, seeking payment for the additional days of inpatient treatment. We conclude that UBH abused its discretion in refusing to pay for these days of treatment.

I. Background

The Plan is self-insured by Wells Fargo, which is both the sponsor and administrator of the Plan. Wells Fargo contracts with third-party administrators to review claims made under the Plan. The third-party administrator responsible for reviewing mental health and substance abuse claims, including anorexia nervosa, is OptumHealth Behavioral Solutions, which is a trade name of UBH.

Jones was admitted to PSH on January 25, 2010. Jones’s brother had recently sent an email to Jones’s entire family saying that he felt that he was planning her funeral. Jones’s 17-year-old daughter feared for her mother’s life. UBH’s case management notes listed Jones’s “Reason for admission” as “severe depression, SI [Suicidal Ideation], and anorexia.” Jones’s admitting diagnoses were (1) “Major Depressive Disorder, Recurrent, Severe Without Psychotic Features”; (2) “Anorexia Nervosa”; (3) “pneumonia”; and (4) “Problems with primary support group.” UBH initially authorized four days of inpatient hospital treatment.

UBH case notes for January 27 provide:

(Emphasis added.)

UBH ease notes for January 27 describe Jones’s condition as “Emergent-Life Threatening.” The “Treatment Plan” in the January 27 notes provides:

(Emphasis added.)

(Emphasis added.) Hospital staff were checking on Jones every fifteen minutes as a “suicide precaution! ]” and were supervising her for 2 1/2 hours after every meal. Her laxative taper was “down to 50 tablets of Sena per day from 130 tabs per day.” UBH authorized two more days of inpatient treatment.

(Emphasis added.) UBH authorized four more days of inpatient treatment, through

February 4, and scheduled another review date for February 5.

(Emphasis added.) UBH authorized inpatient treatment for an additional two days, through February 6.

(Emphasis added.)

On February 9, Nurse Wolpert reported to UBH by voicemail:

(Emphasis added.) UBH treatment notes for February 9 recorded: “we’re repeating her labs. 1550 cal, 84 grams of protein. ql5 min checks, meal supervision, and post-meal supervision 2.5 hrs.” (Emphasis added.)

Also on February 9, UBH Regional Director Dr. Zucker conducted a peer-to-peer review, speaking by telephone to Dr. Nomi Fredrick, Jones’s attending physician at PSH. Dr. Zucker wrote a summary of the conversation. (There is no summary of the conversation written by Dr. Fredrick.) Dr. Zucker wrote:

I advised the following:

(Emphasis added.)

On February 10, UBH case notes indicate that Nurse Wolpert

On February 12, Nurse Wolpert reported to UBH:

(Emphasis added.) UBH treatment notes for February 12 recorded: “still keeping cal plan @ 1550, .... d/c [discharge] criteria: prov [provider] wants pt @ 75% of IBW [Ideal Body Weight], around 90-95 lbs.” (Emphasis added.) Up to this date, UBH’s notes had consistently listed Jones’s weight at admission as 88 pounds. On February 12, for the first time, her weight at admission was listed at 81 pounds. UBH authorized inpatient treatment through February 14.

On February 16, Dr. Zucker conducted a second telephone peer-to-peer review with Dr. Fredrick. After his conversation with Dr. Fredrick, he wrote a summary. (Again, there is no summary of the conversation written by Dr. Fredrick.) Dr. Zucker wrote:

(Emphasis added.)

UBH formally notified PSH of its decision not to pay for acute inpatient treatment beyond February 14 in a letter dated February 18. On February 23, Nurse Wolpert requested on Jones’s behalf an “urgent appeal” of Dr. Zucker’s denial of benefits coverage for inpatient hospital treatment after February 14. On either February 23 or 24, the appeal was referred by UBH to Dr. Barbara Center of Prest & Associates. Dr. Center spoke by telephone to Dr. Fredrick on February 24 and on the same day sent written findings to UBH. Dr. Center wrote:

(Emphasis added.)

On February 24, the same day UBH received Dr. Center’s report, Dr. William Barnard, UBH Assistant Medical Director, denied PSH’s appeal. In a letter addressed to Jones, he wrote:

(Emphasis added.) Dr. Barnard then quoted nearly verbatim the three num bered paragraphs contained in Dr. Center’s report.

Dr. Zucker made a number of obvious mistakes in his summaries of his two peer-to-peer reviews. In his February 9 summary, written after his first review, Dr. Zucker wrote that Jones’s weight at admission had been either 75 or 81 pounds. Both weights are contradicted by information then in the administrative record. Jones’s admission weight was never listed as 75 pounds. From January 25, the date of her admission, until February 12, three days after Dr. Zucker’s summary, Jones’s weight on admission was consistently listed in UBH treatment notes as 88 pounds. On February 12, her admission weight was changed in UBH notes to 81 pounds. Dr. Zucker also wrote in his report that Jones had made “no prior attempts” at suicide. This statement is contradicted in the administrative record. UBH’s January 27 treatment notes state, “In 2007 took whole bottle of laxatives in a suicide attempt.”

Dr. Zucker minimized the risk of suicide. He wrote in his February 9 summary, “[Hojwever, on fur[th]er questioning [Dr. Fredrick] reports that pt has no immediate] plan, has not gathered means, has made no prep, and ther[e] is no 1:1 or even line of site [sic] in the program.” But Dr. Fredrick’s “reports,” “on further questioning” by Dr. Zucker, do not undermine her assessment that Jones was at risk for suicide. Repeated entries in UBH treatment notes indicate that Jones continued to have active suicidal ideation, with plans either to overdose or starve herself to death. Given that Jones was in acute inpatient care, she did not have access to large quantities of Tylenol or laxatives, her planned means of overdosing. So long as Jones remained in acute inpatient care, she would not have been able to “gather[ ] means” or otherwise “prep” for suicide. The lack of line-of-sight supervision did not support Dr. Zucker’s implicit suggestion that the hospital did not itself believe that Jones was at risk for suicide, for UBH treatment records indicate that PSH staff continued to check Jones every fifteen minutes as a precaution against suicide.

In Dr. Zucker’s February 16 summary, written after his second peer-to-peer review, he again made a number of obvious mistakes, despite his self-described review of “all available information.” Dr. Zucker wrote that Jones had a chronically low weight. He then relied on this “chronicity” to refuse further authorization of inpatient treatment. Dr. Zucker wrote that Jones was “now at day 21 with minimal wt gain desp[it]e diet of 2100 cal.” This statement is contradicted in the administrative record. UBH treatment notes nowhere indicate that Jones had been eating 2100 calories per day. UBH notes on January 27 indicate that, at the time of her admission, Jones was eating just 200 calories per day, and that her treatment plan required “go[ing] slow on carbs and fat.” Notes on February 9 indicate that Jones was eating only 1550 calories per day. Three days later, on February 12, the hospital was “still keeping cal plan @ 1550.”

Dr. Zucker wrote in his February 16 summary, “MD states: [patient] has been do[in]g well ... un[ti]l dietician raised cals today.” One may possibly infer from this statement that Jones’s calories were raised “today” to 2100 calories, from 1550 calories on February 12 (though Dr. Zucker does not specify the number of calories by which Jones’s “cals” were raised). But this does not support Dr. Zucker’s rationale for discontinuing coverage after February 14. First, it is unclear whether “today” means February 14, the last date for which UBH authorized payment, or February 16, the date of Dr. Zucker’s conversation with Dr. Fredrick and of his report. Read in context, the later date seems more likely. Even if Dr. Zucker meant February 14 when he wrote “today,” this would mean only that Jones’s caloric intake was raised to 2100 on the last day of her then-three-week stay. On either reading, Dr. Zucker’s statement— that Jones had minimal weight gain after 21 days “desp[it]e diet of 2100 eal[ories]”— makes no sense.

Dr. Zucker wrote further that Jones’s laboratory test results were “normal,” and that she had “finished laxative taper without refeeding [symptoms].” But according to UBH notes, as of February 16, the date of Dr. Zucker’s summary, Jones’s most recent lab results came from tests performed on February 12. Those tests showed elevated levels of amylase and lipase related to Jones’s “acute pancreati-tis” — a condition caused by refeeding. Jones’s amylase levels on February 12 were higher than they had been on February 5, when UBH case management notes already described her labs as “abnormal.”

Finally, Dr. Zucker wrote in his February 16 summary, “MD states [patient] ... not expressing s/i.” We have only Dr. Zucker’s summary of his conversation with Dr. Fredrick to support this statement. We know from UBH treatment notes that on February 12, only two days before the February 14 cut-off date, that Jones “still c/o [complains of] SI w/plan to starve or OD.” There is nothing in UBH’s treatment notes, as distinct from Dr. Zucker’s recounting of his conversation with Dr. Fredrick, to indicate that Jones was no longer experiencing suicidal ideation two days later.

Dr. Center also made a number of obvious mistakes. She wrote, “At the time of this admission, the patient was 5'5" tall and weighed 84 pounds.” There is nothing in the administrative record to support an admission weight of 84 pounds. From January 25 to February 12, UBH treatment notes repeatedly indicate that Jones’s admission weight was 88 pounds. On February 12, the treatment notes indicate that her admission weight was 81 pounds. Dr. Center wrote further, “She is described as taking 75 to 100 laxatives daily.” This is contradicted in the administrative record. UBH treatment notes repeatedly indicate that Jones was taking 180 Sena-S laxatives per day when she was admitted.

Dr. Center wrote further, “The patient was not suicidal, homicidal, or psychotic. The patient reported some vague suicidal thoughts, including thoughts of overdosing on the laxatives that she had been abusing.” This is contradicted in the administrative record. On admission, Jones was actively suicidal, and she continued to have specific (not “vague”) suicidal ideation until at least February 12. UBH treatment notes on January 27 state that Jones had attempted suicide in 2007 by overdosing on a “whole bottle of laxatives.” UBH treatment notes on January 27, January 29, February 3, February 9, and February 12 consistently record Jones’s suicidal ideation and plan to commit suicide through overdose or starvation. Nowhere in UBH treatment notes after February 12 is there any statement that Jones no longer had suicidal ideation.

Dr. Center also wrote, “The patient’s medical history is remarkable for acute pancreatitis which was diagnosed on admission.” This, too, is contradicted in the administrative record. UBH treatment notes are specific in stating that Jones did not have pancreatitis on admission. Rather, as her January 27 treatment plan stated, PSH expected that Jones would later develop pancreatitis as a consequence of her treatment. As stated by Nurse Wol-pert, “She’ll have pancreatitis and anemia from refeeding.” Treatment notes indi cate that Jones was suffering from pan-creatitis on January 29, four days after being admitted to PSH, but not before.

Finally, Dr. Center wrote, “Following admission, the patient had slow weight gain.... By the current last covered date, 2/14/10, the patient [weighed] 83 lbs.... The patient was ... steadily gaining weight.” These statements are contradicted in Dr. Center’s own report, as well as in the administrative record. According to the inaccurate numbers recited by Dr. Center, upon which she ostensibly relied, Jones did not “steadily gain[] weight.” Dr. Center wrote that Jones weighed 84 pounds at admission on January 25, even though there is nothing in UBH notes to indicate that this was Jones’s admission weight. Dr. Center then notes that Jones weighed 83 pounds on February 14. To state the obvious, this is a loss rather than a gain. If Jones’s weight at admission was 88 pounds, as indicated in UBH treatment notes up to February 12, Jones lost five pounds between the date of her admission and February 14. Or if Jones’s admission weight was 81 pounds, as indicated in treatment notes of February 12, she did gain weight; but she did not do so steadily. On this assumption, she started out at 81 pounds on January 25 and dropped to 75.5 pounds on January 29. She then weighed 79 pounds on February 3, and 84 pounds on February 9. UBH treatment notes do not give a weight after February 9, but Dr. Center wrote that Jones weighed 83 pounds on February 14, one pound less than she weighed five days earlier, on February 9.

Dr. Barnard wrote in his February 26 letter to Jones that he had decided on February 24 to deny her appeal “[a]fter fully investigating the substance of the appeal, including all aspects of clinical care involved in this treatment episode.” After referring to his own “full investigation,” Dr. Barnard quoted almost verbatim from Dr. Center’s erroneous report, which Dr. Center had sent to UBH earlier that same day.

On February 25, one month after her admission to PSH and eleven days after UBH ceased paying for her treatment, Jones was discharged. After her discharge, Jones assigned to PSH her right to payment under the Plan. PSH brought suit under 29 U.S.C. § 1132 (a)(1)(B), alleging that UBH and the Plan had wrongfully denied benefits to Jones. The district court held that, despite numerous errors in Dr. Zucker’s and Dr. Center’s reports, the administrative record provided a reasonable basis for determining that acute inpatient care was not necessary after February 14. The court concluded that it was “not left with a definite and firm conviction that UBH’s benefits determination was in error,” and therefore could not disturb that decision. This appeal followed.

II. Standard of Appellate Review

We review de novo the district court’s choice and application of the standard of review of an ERISA plan administrator’s decision. Abatie v. Alta Health & Life Ins. Co., 458 F.3d 955, 962 (9th Cir. 2006) (en banc). We review for clear error the district court’s underlying findings of disputed fact. Id.

III. Discussion

A. Standard of Review of UBH’s Benefits Denial

We begin by addressing the standard under which we should review the denial of benefits by UBH. “The essential first step of the analysis ... is to examine whether the terms of the ERISA plan unambiguously grant discretion to the administrator.” Id. at 963 . When the terms of the plan do not grant discretion to determine eligibility for benefits or to construe the terms of the plan, we review de novo the administrator’s denial of coverage. Id. When a plan unambiguously confers such discretion, we review a denial of benefits for abuse of discretion. Id. If there are procedural irregularities or if an administrator operates under a conflict of interest, we consider the irregularities or conflict as a factor in determining whether there has been an abuse of discretion. Id. at 965, 972 .

The district court concluded that the Plan unambiguously granted discretion to the administrator. It then reviewed UBH’s denial of benefits for abuse of discretion. PSH does not challenge the district court’s determination that the Plan explicitly grants discretion to Wells Fargo, and derivatively to its third-party administrator UBH. However, PSH makes three arguments in favor of less deferential review of UBH’s denial of benefits.

First, PSH contends that there were procedural irregularities in UBH’s benefits denial such that we should review the denial de novo. Even when a plan confers discretion on an administrator, if that administrator engages in “wholesale and flagrant violations of the procedural requirements of ERISA,” its decision is subject to de novo review. Id. at 971 . However, most procedural errors are not sufficiently severe to transform the abuse-of-discretion standard into a de novo standard. Anderson v. Suburban Teamsters of N. III. Pension Fund Bd. of Trustees, 588 F.3d 641, 647 (9th Cir.2009). Instead, we weigh any procedural errors as a factor in determining whether UBH abused its discretion. Id.; Abatie, 458 F.3d at 972 .

It is painfully apparent that UBH did not follow procedures appropriate to Jones’s case. UBH treatment notes describe her case as requiring evaluation by UBH Regional Director Dr. Zucker due to its “medical and psychiatric complexity.” Yet the treatment notes in UBH’s administrative record, upon which UBH ostensibly made its decision in this “complex” case, are based entirely on telephone conversations and voicemail messages. No PSH hospital records were ever put into the administrative record. No UBH doctor or other claims administrator ever examined Jones. The choice to conduct only a paper review “raise[s] questions about the thoroughness and accuracy of the benefits determination.” Montour v. Hartford Life & Accident Ins. Co., 588 F.3d 623, 634 (9th Cir.2009) (alteration in original) (quoting Bennett v. Kemper Nat'l Servs., Inc., 514 F.3d 547, 554 (6th Cir.2008) (internal quotation marks omitted)).

Moreover, Dr. Zucker’s “peer-to-peer” evaluations and coverage decisions, made after telephone conversations with Dr. Fredrick, were based on obvious factual errors that could easily have been corrected if only the UBH administrative record, let alone PSH hospital records, had been consulted. Similarly, Dr. Center’s “independent” evaluation and coverage decision were based on obvious factual errors that could easily have been corrected if the UBH administrative record had been consulted. See Saffon v. Wells Fargo & Co. Long Term Disability Plan, 522 F.3d 863, 873 (9th Cir.2008) (explaining that an administrator’s decision was entitled to less deference because the administrator “took various of [claimant’s] doctors’ statements out of context or otherwise distorted them in an apparent effort to support a denial of benefits”).

Second, PSH argues that materials outside the administrative record—specifically, Jones’s hospital records—-should be considered by the court in any review of UBH’s benefits denial. PSH requested that the district court expand the record beyond the administrative record compiled by UBH and consider these records. UBH vigorously opposed this request. The district court declined to consider any documents beyond the administrative record. When reviewing for abuse of discretion a plan administrator’s benefits determination, our review is typically limited to the contents of the administrative record. See Abatie, 458 F.3d at 969-70 . However, when procedural irregularities are apparent in an administrator’s determination, we may consider extrinsic evidence to determine the effects of the irregularity. “[T]he court may, in essence, recreate what the administrative record would have been had the procedure been correct.” Id. at 973 .

There was good reason for the district court to consider hospital records, in addition to the administrative record compiled by UBH, in a case involving a confessedly high degree of “medical and psychiatric complexity.” All the information in UBH’s administrative record concerning Jones’s medical condition is based on telephone conversations and voice mail messages, with the predictable result that the administrative record contains conflicting (and necessarily incorrect) information about some of the most important issues in the case, such as Jones’s weight at various times during her treatment. Where the administrator makes a coverage determination based solely on an administrative record such as this one — and where actual medical records would be helpful to determining the accuracy of the medical facts upon which the administrator makes its coverage determination — expansion of the record in the district court is appropriate.

Third, PSH contends that UBH, even though a third-party administrator, was operating under a conflict of interest, and that we should consider that conflict as a factor in determining whether there was an abuse of discretion. PSH points to UBH’s self-interest in continuing its contractual relationship with Wells Fargo, and to Wells Fargo’s self-interest, as a direct funder of the Plan, in minimizing benefit payments authorized under the Plan by UBH.

However, we need not reach these contentions. Even conducting an abuse of discretion review uninfluenced by any procedural irregularity or conflict of interest — and considering only the record that UBH had before it when it made its benefits determination — we hold that UBH improperly denied benefits to Jones.

In reviewing for abuse of discretion, we consider all of the relevant circumstances in evaluating the decision of the plan administrator. As we wrote in our en banc decision in Abatie , “A straightforward abuse of discretion analysis allows a court to tailor its review to all the circumstances before it.” 458 F.3d at 968 (emphasis added). The Supreme Court in Glenn, decided two years after Abatie , made clear that abuse of discretion review, whether or not including conflict of interest as a factor, entails a review of all the circumstances. The Court cautioned against talismans or formulas that would “falsifly] the actual process of judging,” 554 U.S. at 119, 128 S.Ct. 2343 (alteration in original) (internal quotation marks omitted), and endorsed a process in which reviewing courts consider all relevant factors, of which, depending on the circumstances, conflict of interest may be one, id. at 117, 128 S.Ct. 2343 (“[W]hen judges review the lawfulness of benefit denials, they will often take account of several different considerations of which a conflict of interest is one.”).

We wrote twenty-three years ago in Horan v. Kaiser Steel Retirement Plan, 947 F.2d 1412 (9th Cir. 1991), that we will uphold a plan administrator’s decision if it is grounded in “any reasonable basis.” Id. at 1417 (internal quotation marks omitted); see also Sznewajs v. U.S. Bancorp Amended & Restated Supplemental Benefits Plan, 572 F.3d 727, 734-35 (9th Cir.

2009). This language in Horan could be read to mean that we should make an “any reasonable basis” determination without looking at all the circumstances of the case. To take a simple example, factors favoring discharge from the hospital might provide reasonable bases if considered in isolation. A patient might be eating well, have proper blood sugar levels, have no infections, and have a supportive family. Those factors, considered in isolation, would support discharge. But if the reason for the patient’s hospitalization is severe congestive heart failure, those factors would not be reasonable bases to support discharge. In the wake of Glenn, we have recognized that this unrealistic reading of the any-reasonable-basis test is not “good law when ... an administrator operates under a structural conflict of interest.” Salomaa v. Honda Long Term Disability Plan, 642 F.3d 666, 674 (9th Cir.2011). It is also not “good law” even when an administrator is not operating under a conflict of interest and we are performing a “straightforward abuse of discretion analysis.” See Abatie, 458 F.3d at 968 ; cf. Conkriglit v. Frommert, 559 U.S. 506, 521 , 130 S.Ct. 1640 , 176 L.Ed.2d 469 (2010) (“Applying a deferential standard of review does not mean that the plan administrator will prevail on the merits. It means only that the plan administrator’s interpretation will not be disturbed if reasonable.” (internal quotation marks omitted)). In all abuse-of-diseretion review, whether or not an administrator’s conflict of interest is a factor, a reviewing court should consider “all the circumstances before it,” Abatie, 458 F.3d at 968 , in assessing a denial of benefits under an ERISA plan.

B. UBH’s Benefits Denial

“A plan administrator abuses its discretion if it renders a decision without any explanation, construes provisions of the plan in a way that conflicts with the plain language of the plan, or fails to develop facts necessary to its determination.” Anderson, 588 F.3d at 649 . “[T]he test for abuse of discretion in a factual determination (as opposed to legal error) is whether 'we are left with a definite and firm conviction that a mistake has been committed.’ ” Salomaa, 642 F.3d at 676 (quoting United States v. Hinkson, 585 F.3d 1247, 1262 (9th Cir.2009) (en banc)). “[A]n administrator ... abuses its discretion if it relies on clearly erroneous findings of fact in making benefit determinations.” Taft v. Equitable Life Assurance Soc’y, 9 F.3d 1469 , 1473 (9th Cir.1994).

The Plan documents provide two sets of guidelines relevant to Jones’s eligibility for coverage of acute inpatient treatment at PSH. First, UBH level-of-care guidelines provide that acute inpatient care is warranted when any one of six criteria are met:

In its initial authorization of inpatient hospital treatment for Jones, UBH found that at least the fifth of these criteria, a “serious medical condition [that] requires 24-hour management,” was satisfied.

“Continued Service” guidelines under the Plan are used to determine whether previously authorized care should be continued at its current level “as a member’s severity of illness changes.” In order to maintain a current level of care, each of ten criteria must be met.

In denying coverage for Jones’s inpatient treatment beyond February 14, UBH found that criteria 1, 2, and 9 were no longer met.

Because coverage for acute inpatient treatment is indicated whenever any one of the six criteria for that level of care is met, the continued presence of any one of the six level-Of-care criteria necessarily satisfies all ten of the continued service criteria. This may be seen by examining the level-of-eare guidelines for residential treatment — the level of care immediately below acute inpatient treatment, which is the care Jones was receiving at PSH. These guidelines provide that residential care is appropriate only if, among other things, “[t]he member is not at immediate risk of serious harm to self or others,” and “[t]he member ... does not require 24-hour nursing care and monitoring.” Similarly, the level of care guidelines for “Partial Hospital/Day Treatment,” the level of care below residential care — and the level that Dr. Zucker concluded was appropriate for Jones after February 14 — also provide that partial hospital treatment is appropriate only if “[t]he member is not at immediate risk of serious harm to self or others.”

UBH’s decision that Jones no longer qualified under UBH guidelines for acute inpatient care as of February 14 therefore necessarily rested on determinations that Jones no longer presented a “[s]erious and imminent risk of harm to [herjself,” and that her condition no longer warranted 24-hour monitoring and care. These determinations in turn rested on Dr. Zucker’s and Dr. Center’s findings of fact, which were endorsed by Dr. Barnard. As discussed above, Dr. Zucker and Dr. Center made several critical factual errors, upon which they based their conclusion that Jones could safely be discharged.

UBH owed a fiduciary duty to Jones under ERISA. The Supreme Court has described that duty as follows:

Pegram v. Herdrich, 530 U.S. 211, 223-24 , 120 S.Ct. 2143 , 147 L.Ed.2d 164 (2000). Fiduciaries must discharge their duties “with the care, skill, prudence, and diligence under the circumstances then prevailing that a prudent man acting in a like capacity and familiar with such matters would use in the conduct of an enterprise of a like character and with like aims.” Id. at 224 n. 6, 120 S.Ct. 2143 (quoting 29 U.S.C. § 1104 (a)(1)(B)).

UBH fell far short of fulfilling its fiduciary duty to Jones. Dr. Zucker, UBH’s primary decisionmaker, made a number of critical factual errors. Dr. Center, as an ostensibly independent evaluator, made additional critical factual errors. Dr. Barnard, UBH’s final decisionmaker, stated that he arrived at his decision to deny benefits “after fully investigating the substance of the appeal.” He then rubber-stamped Dr. Center’s conclusions. There was a striking lack of care by Drs. Zucker, Center, and Barnard, resulting in the obvious errors we have described. What is worse, the errors are not randomly distributed. All of the errors support denial of payment; none supports payment. The unhappy fact is that UBH acted as a fiduciary in name only, abusing the discretion with which it had been entrusted.

Conclusion

Reviewing for abuse of discretion, we conclude that UBH improperly denied benefits under the Plan in violation of its fiduciary duty under ERISA.

REVERSED.