After a bench trial in an action brought under the Federal Tort Claims Act (FTCA), see 28 U.S.C. § 1346, §§ 2671-2680, the trial court1 concluded that a nurse at a Veterans Administration (VA) hospital killed veteran Elzie Havrum. The court ordered the United States to pay $450,000 in damages to Mr. Havrum's wife, Helen Havrum, because of the hospital's negligent failure to protect her husband from the nurse. On appeal, the government challenges the sufficiency of the evidence. Ms. Havrum cross-appeals, maintaining that the trial court should have awarded attorney's fees to her. We affirm.

The alleged acts of negligence occurred in Missouri, and thus the elements of Ms. Havrum's claim are determined by Missouri law. See 28 U.S.C. § 1346(b)(1). To prevail, she was required to show that the VA hospital had a duty to protect Mr. Havrum from injury and that its failure to perform that duty caused his death. See Behrenhausen v. All About Travel, Inc., 967 S.W.2d 213, 216-17 (Mo. Ct. App. 1998). The government does not challenge the trial court's conclusion that the hospital breached its duty to protect Mr. Havrum from the nurse, who presented a danger to patients. The government contends, however, that Ms. Havrum failed to establish causation because the evidence did not support the court's finding that the nurse killed Mr. Havrum. We review the court's factual findings for clear error. See Cook v. Nebraska Public Power District, 171 F.3d 626, 630 (8th Cir. 1999).

The trial court found that Richard Williams, the nurse in question, gave Mr. Havrum a lethal dose of codeine and, alternatively, that even disregarding the evidence of codeine poisoning, the circumstantial evidence indicated that Mr. Williams killed Mr. Havrum. After carefully reviewing the record, we conclude that the circumstantial evidence sufficiently supports a finding that Mr. Williams did indeed kill Mr. Havrum; thus we need not address the admissibility of expert testimony that Mr. Havrum received a codeine overdose.

In finding that Mr. Williams killed Mr. Havrum, the trial court relied, in part, on a study by the hospital's epidemiologist, Dr. Gordon Christensen, who investigated a suspected link between Mr. Williams and an increase in deaths on the ward where Mr. Williams customarily worked. The study concluded that patients who were under Mr. Williams's care were almost ten times more likely to die as other patients were: Although statistically one would expect five deaths to occur during Mr. Williams's shifts during the relevant period, the death toll was actually forty-eight. In addition, Mr. Williams was associated with many unexpected deaths that occurred in private rooms. Dr. Christensen also testified that he had never seen anything so unusual as the number of patients who died on the relevant ward from May through July, 1992, between 1:00 a.m. and 3:00 a.m. (a period when fewer deaths generally occur); Mr. Williams was present for eleven of the thirteen deaths in that interim, although he worked on only one-third of the shifts.

Dr. Christensen concluded that there was only one chance in a million that the pattern of deaths on the ward was random, and that there was a compelling correlation between the deaths and Mr. Williams for which Dr. Christensen could find no benign explanation. Although we agree with the government, as we must, that this statistical evidence alone does not establish that Mr. Williams caused Mr. Havrum's death, that evidence is nevertheless probative. It was, moreover, only one aspect of the circumstantial evidence upon which the trial court relied in finding causation.

With regard to Mr. Havrum specifically, the court noted that his death was among those that Dr. Christensen found highly unusual. Mr. Havrum died on the ward in question at 1:15 a.m. in a private room with Mr. Williams present. Mr. Havrum was not expected to die, and the government offered no evidence that he faced death as part of some short-term natural progression. Although Mr. Havrum suffered from a serious illness, the admitting physician did not place him in intensive care and did not believe that his death was imminent. Mr. Havrum actually reported feeling better while he was in the hospital, but sixteen hours after his admission he was pronounced dead.

The trial court, in deciding that Mr. Williams killed Mr. Havrum, also referred to suspicious inconsistencies and alterations in the medical records. According to the records, Mr. Williams was the last nurse to see Mr. Havrum alive. A physician called to the bedside by Mr. Williams stated in her progress note that Mr. Havrum had no pulse, respiration, heart rate, or blood pressure and that she pronounced him dead at 1:15 a.m. Although the physician's note further states that Mr. Havrum was "found unresponsive by nursing," Mr. Williams's medical note states that he found Mr. Havrum in "severe respiratory distress," a description that Mr. Williams frequently used in his medical notes on a patient's death. "Respiratory distress," according to several nurses, indicates that the patient is struggling to breathe and is inconsistent with being "unresponsive."

The trial court also remarked that Mr. Williams first wrote a medical note indicating that he found Mr. Havrum in severe respiratory distress at about 1:15 a.m., the same time that the doctor pronounced him dead. The time in the note was then changed to 1:10 a.m., a line was drawn through the note, and the note was marked "error R.W." Mr. Williams then wrote another medical note; this time he stated that he found Mr. Havrum in severe respiratory distress at about 1:10 a.m. and that the doctor arrived at about 1:15 a.m., just as Mr. Havrum stopped breathing. Although the government suggests possible innocent explanations for the changed entries and omissions, the trial court, which noted that Mr. Williams had been fired by another hospital for inserting a false entry into a patient's chart, was free to draw its own less innocent inferences from Mr. Havrum's hospital records. See Newhouse v. McCormick and Company, Inc., 110 F.3d 635, 643 (8th Cir. 1997). And one permissible inference is that Mr. Williams's evident uncertainty about what to say and to note in the records indicates that there was in fact nothing particularly wrong with Mr. Havrum and that Mr. Williams took his life.

With regard to the government's complaint that the trial court improperly relied on a physician's hearsay statement to find causation, we conclude that the court relied on other evidence that was admissible and sufficient to support its findings, and the court's opinion clearly indicates that it would have reached the same conclusion without the challenged evidence. See Estate of Largent v. United States, 910 F.2d 497, 500 (8th Cir. 1990). Thus the error, if any, was harmless. See Fed. R. Civ. P. 61. The government also contends that the trial court erred in finding causation based on circumstantial evidence because Ms. Havrum's pathologist did not find that a homicide had occurred until he received a report indicating that Mr. Havrum had died of codeine poisoning. Even before he received the report, however, the pathologist specifically stated that Mr. Havrum's death could not be classified as natural. Although, excluding codeine poisoning, the evidence does not establish a specific cause of Mr. Havrum's death, the government's own pathologist testified about several ways to induce a death that cannot be detected post-mortem. We fail to see how the opinion of Ms. Havrum's expert precluded the trial court as the fact-finder from considering all of the circumstantial evidence, drawing reasonable inferences from that evidence, and concluding that, more likely than not, Mr. Williams killed Mr. Havrum. We therefore conclude that the trial court's findings are not clearly erroneous.

In her cross-appeal, Ms. Havrum contends that prior to being sued the Department of Veterans Affairs attempted to hide from the public information indicating that its nurse was endangering patients, and that this misconduct entitled her to an award of attorney's fees under 28 U.S.C.§ 2412(b) of the Equal Access to Justice Act, see 5 U.S.C. § 504, historical and statutory notes, short title, 1980 acts (West 1996). We have held that where the government has acted in bad faith it may be required to pay attorney's fees under § 2412(b), see McLarty v. United States, 6 F.3d 545, 549 (8th Cir.1993), and we assume, without deciding, that such awards are available in FTCA actions, see Lucarelli v. United States, 943 F. Supp. 157, 158 (D. P.R. 1996). Assuming also that pre-litigation misconduct could support an attorney's fees award, we think that the district court was correct in refusing to award such fees because the government's bad-faith conduct did not prove to be a measurable impediment to the progress of Ms. Havrum's lawsuit. Fees for bad-faith activity on the government's part are available only in exceptional circumstances, see Brown v. Sullivan, 916 F.2d 492, 495 (9th Cir. 1990), and we think that cases in which the activity complained of had no substantial effect on the plaintiff's litigation efforts necessarily fall outside that category.

Accordingly, we affirm the judgment of the trial court.