MERRITT, Circuit Judge.
In this medical malpractice action brought under the Federal Tort Claims Act for wrongful death arising from surgery in a Veterans Administration hospital, the District Court after a bench trial awarded plaintiff-appellee $606,203.95 in damages against the United States. Plaintiff’s decedent died from an E. Coli infection of the brain (meningitis) after a prolonged period of decline that followed two brain operations in 1980, the first on June 9 to remove a large benign brain tumor and the second on October 15 to treat the infection. The United States argues on appeal that the verdict was based upon insufficient evidence of negligence and proximate causation and upon impermissible inferences drawn by the District Court. The outcome of this appeal turns upon what effect a defendant’s negligent destruction of crucial evidence under its exclusive control should have on the plaintiff’s burden of proof. Two acts by the hospital surgeons in this case create a rebuttable presumption of negligence and proximate causation against the defendant — the negligent destruction of a skull bone flap after the second operation, and the consequent failure at that time to undertake a pathological [1240]*1240examination of this evidence in accordance with customary standards of medical practice. The defendant has not rebutted that presumption, and the judgment of the District Court therefore is affirmed.
I.
Plaintiff’s decedent, Francis Welsh, entered the VA Medical Center in Lexington, Kentucky on May 27, 1980, complaining of depression, loss of appetite and memory, fatigue, and headaches. Tests led to the diagnosis of a brain tumor, and on June 9, 1980, a large benign tumor called a menin-gioma was removed surgically. The surgeons used an artificial material called du-raplast to replace the brain lining that was removed at surgery, but the flap of skull bone that was removed to permit access to the brain was put back in place. It was this piece of the skull that later acquired crucial evidentiary value and was removed and discarded at the second operation.
Six days after surgery, Mr. Welsh began to run a low grade fever; a spinal tap revealed bloody fluid, which is often seen in the postoperative period. Although there was no definite evidence of infection, he was treated with antibiotics for seven days. On the fourth day of antibiotic treatment, he had an epileptic seizure; six days after that, swelling was noted in the area near the skin incision. Three days later, on June 26, his surgeons placed a drain in his lower back to siphon spinal fluid away from the brain; the swelling receded. On June 30 the drain was removed, and on July 9 Mr. Welsh was finally discharged from the hospital.
During the succeeding three months Mr. Welsh visited the VA clinic either two or four times — there is no record of the first and last putative visits, and the parties’ testimony concerning the very existence of these visits is in dispute. The first disputed visit was on July 9, for which there was no doctor’s note. Some independent evidence exists that it took place, however, in the form of a nurse’s note. Plaintiff testified that at the next visit, on July 15, Mr. Welsh complained of headaches, upset stomach, and loss of appetite. The VA medical record for July 18 does not reflect these complaints, noting only that he was “nervous at times then depressed but coping well.” The doctor scheduled him for a return visit three months later. At the next visit, an unscheduled but undisputed appearance on August 22, Mr. Welsh complained of “extreme weakness and nausea,” according to the medical record. Despite this complaint, his temperature was not taken (nor was it taken on any other clinic visit). The resident surgeon who saw him on August 22 noted only a possible problem with “hyper-reactive smell.” Suspecting possible brain seizures as a cause, the resident scheduled an electroencephalogram (EEG), or brain-wave test, to be done two months later in October.
The testimony was conflicting on whether Mr. Welsh came back for another visit to the clinic around September 25 — his family members testified with some specificity that he had, but VA witnesses denied it. The VA records do not reflect such a visit. On the basis of other indications of sloppy recordkeeping by the VA and of credence placed in the testimony of Welsh family members, the District Court found that a visit did occur around September 25 and that — once again — Mr. Welsh’s complaints were not fully investigated.
Finally, on October 10, Mr. Welsh made a scheduled visit to the clinic, again made similar complaints, and again did not have his temperature taken. The VA doctor did order a computed tomography (CT) study of the head, and the EEG ordered in August was performed. The results of both tests were within normal limits. On the way home, Mr. Welsh began to feel dramatically worse, and the family returned him to the VA that night. Examination at this time determined his temperature to be 102 degrees Fahrenheit; later that night, it rose to 104.6 degrees. The next day a needle tap of the area immediately under the incision on his forehead revealed pus, and his doctors instituted a course of an antibiotic treatment. Infectious disease consultants within the VA recommended a still deeper tap to search for a collection of pus near the surface of the brain called a [1241]*1241subdural empyema.” This procedure, however, was not performed until October 15, four days later.
On October 14, a repeated CT scan revealed the formation of a “crescent” appearance suggestive of subdural empyema; the next day a deep exploration was undertaken. The old bone flap noted above was removed and discarded, rather than sent to the pathology laboratory for examination; a subdural empyema was discovered and drained. The antibiotics thereafter ultimately cured the infection, but by this time Mr. Welsh had sustained permanent brain damage. He remained in a semi-coma from October 1980 until his death more than four years later in May 1985. Except for a six-month stay at home during 1981, this final period of his life was spent in the care of the VA.
In 1982, Mrs. Welsh filed an administrative claim, which was denied by the VA. In February 1983, she filed suit in the United States District Court for the Eastern District of Kentucky pursuant to the Federal Tort Claims Act, 28 U.S.C. § 1346. At trial the theory of plaintiffs expert witness, a Buffalo, New York specialist in infectious disease, was that Mr. Welsh had developed a postoperative infection in June which was incompletely treated with antibiotics. His condition then smoldered through the summer as an undiagnosed infection of the organism E. Coli in the skull bone flap called osteomyelitis. According to plaintiffs expert, the infection then moved into the spinal fluid under the skull creating the localized collection of pus, or subdural empyema. Finally, the subdural empyema burrowed into and through the brain on October 10, rupturing into a brain ventricle (a lake of inner brain fluid), causing infection of the brain lining and fluid, or meningitis.
Defendant’s theory was that there was no infection in June, and that the complaints Mr. Welsh made during the summer clinic visits were unrelated to the disaster that befell him in October. Defendant’s experts theorized that the subdural empye-ma resulted from a gallbladder infection that occurred no more than a few weeks before the catastrophic downturn in Mr. Welsh’s condition that took place on October 10. This theory essentially was based upon a description given a VA doctor on October 14 by Mrs. Welsh of an episode of transient jaundice — possibly a gallbladder attack — that Mr. Welsh had experienced in late September. According to this defense theory, infection in Mr. Welsh’s gallbladder would have escaped into his bloodstream, travelled to the head, seeded into the devitalized bone flap, and eventually spread into the brain area. The VA resident who first propounded the “gall bladder theory” as an explanation for the intervening cause of Mr. Welsh’s empyema contemporaneously described this hypothesis as “certainly grasping at straws,” App. 473, and a review of the record suggests that this observation, made in a moment of candor, is correct. As we shall see momentarily, the District Court accepted a modified version of defendant’s “gall bladder theory” concerning the onset of the infection.
Both plaintiff’s and defendant’s experts agreed that the infection had gone from the skull bone to the brain; the principal difference in their testimony lay in how long the infection had been in the bone. Plaintiff’s expert said since June, or at least three months; defendant’s experts said since late September, or not longer than two weeks.
In its findings of fact and conclusions of law, the District Court largely rejected plaintiff’s theory in favor of defendant’s, but nonetheless found defendant liable. The District Court found that Mr. Welsh’s “gall bladder problems” had their onset on August 25 and that “shortly before” September 15 the E. Coli infection traveled from “the gall bladder, the liver, or a combination” and seeded in his brain. The District Court thus found that there was no infection anywhere in Mr. Welsh’s head before early September. Therefore, the July and August clinic visits, in the District Court’s own view, could not have represented occasions for negligence by failure to diagnose the infection, because the infection had not yet arisen in the gallbladder, much less seeded in the skull.
[1242]*1242The District Court’s theory of the case, rather, turned upon its findings of fact that a clinic visit occurred approximately on September 25, that a VA resident named Dr. Guidry saw Mr. Welsh on that occasion, that Mr. Welsh’s problem was present and discoverable on that occasion, and that failure to diagnose was both negligent and was the proximate cause of Mr. Welsh’s ultimate catastrophic illness and demise. Although there was no record of a September 25 visit and, as noted, the testimony of the various interested witnesses was in conflict over whether it had actually taken place, the District Court resolved these conflicts by not only crediting the testimony of plaintiff’s witnesses on this point, but also by drawing what it termed a series of adverse inferences against the United States based upon:
1. its failure to obtain the testimony of any expert witnesses other than “colleagues of the doctors whose conduct was in issue and present or former employees” of the Lexington VA;
2. its failure to call as a witness Dr. Guidry, the resident who examined Mr. Welsh on October 10 and who the District Court found had examined Mr. Welsh at the postulated September 25 visit;
3. the fact of several missing records and VA doctors’ failure to “follow established procedures,” including the taking of Mr. Welsh’s temperature; and
4. the failure of the surgeons on October 15 to send the bone and tissue removed from the area of Mr. Welsh’s infection to the pathology lab for examination. Because the bone flap was discarded, it was impossible to tell whether the infection of bone was recent, as defendant claimed, or of longer duration, as plaintiff claimed.
The District Court then drew several inferences adverse to the VA, including:
1. The missing records would have shown that the disputed clinic visits actually had taken place;
2. Mr. Welsh’s temperature, if taken, would have been elevated at least on October 10, and possibly on September 25; moreover, further studies would have found the cause of the fever and prevented the catastrophic events that occurred;
3. Dr. Guidry’s testimony would have been adverse to the VA; and
4. The pathology report on the bone flap would have been adverse to the VA.
The District Court concluded that the defendant’s agents had been negligent in failing to discover and treat Mr. Welsh’s infection on the occasion of the clinic visits of September 25 and October 10.
Judge Nelson, in his separate concurring opinion, states that there is sufficient support for the District Court’s findings that negligent acts or omissions in July, August, and September and on October 10, e.g., the failure to take the patient’s temperature, were “substantial factors” in causing the decedent’s subsequent coma and death. There are several problems with this approach.
First, the District Court’s own reasoning would not permit any causative link between negligence that may have occurred during the July or August clinic visits and subsequent damages suffered by Mr. Welsh. Had the District Court adopted plaintiff’s logical and internally consistent theory that the infection smoldered inside Mr. Welsh’s skull from the time of the first operation in June, negligence in July or August would be causally relevant. As noted above, however, the District Court adopted defendant’s “gall bladder theory” of the case but then backdated its onset from late September (the time frame given by Mrs. Welsh) to August 25 (the earliest date of onset acknowledged as possible by defendant’s expert). The postulated August 25 date nevertheless is a full three days after the clinic visit on August 22. This backdating of the gallbladder infection is speculative at best; but even if we adopt it, defendant’s agents cannot be found negligent for failing in July or August to detect an infection that had not yet begun. We should not struggle to uphold a finding that is plainly incorrect.
Second, defendant’s purported negligence in September depends critically upon [1243]*1243a finding that there was in fact a VA clinic visit on September 25, but that records of this visit were either never made or lost. In view of the conflict and vagueness in the testimony on this point, the September 25 “visit” seems a very slender reed upon which to hang liability.
Finally, Judge Nelson would also find liability for failure to discover the infection on October 10. It is unclear, however, that discovery of the infection on October 10 would have been in time to prevent Mr. Welsh’s subsequent rapid deterioration. The testimony of defendant’s witness, Dr. Noble, on this point was in fact equivocal: on the one hand, that “the earlier you got the treatment started the better off you’d have been,” and on the other, unable to say “whether [Mr. Welsh] would have survived or not or had less damage.” App. 182.
In summary, Judge Nelson’s rationale for affirming the District Court would require this Court to approve inconsistent findings by the District Court based upon its acceptance of the “gall bladder theory” that the onset of the infection came in late August. If this theory is accepted, it is doubtful that any negligence on the part of the government substantially contributed to the infection or the patient’s condition, because the infection would have arisen entirely independently of any proven act by the government.
However, there is a much better and more principled way to resolve this case.
II.
The United States argues that there is an insufficient evidentiary basis, including an absence of expert testimony, for two critical findings by the District Court: (1) that the E. Coli infection seeded in Mr. Welsh’s head by September 15, and (2) that the infection could have been diagnosed and treated on September 25 and October 10. This argument must fail.
It is true that in the ordinary case, it would be plaintiff’s burden to establish these two critical propositions of fact leading to liability. Liability under the Federal Tort Claims Act is determined according to the substantive law of the state in which the alleged negligence occurred. Rayonier, Inc. v. United States, 352 U.S. 315, 318-20, 77 S.Ct. 374, 376-77, 1 L.Ed.2d 354 (1957). In Kentucky medical malpractice cases, a plaintiff ordinarily must prove negligence and proximate cause by expert medical testimony. Jarboe v. Harting, 397 S.W.2d 775 (Ky.1965); Turner v. Reynolds, 559 S.W.2d 740, 741 (Ky.App.1977). Kentucky adheres to the substantial-factor test for causation. Deutsch v. Shein, 597 S.W. 2d 141, 144 (Ky.1980); Morris v. Hoffman, 551 S.W.2d 8, 11 (Ky.App.1977). This also is a case in which state law determines “the effect of a presumption respecting a fact which is an element of a claim ... as to which state law supplies the rule of decision.” Fed.R.Evid. 302. Thus Kentucky law governs the application of rules that would create presumptions or shift the parties’ burdens of proof.
Plaintiff did not establish these two factual elements by direct proof; rather, the District Court inferred them. Because these determinations by the District Court involve the application of rules of law concerning presumptions in order to justify those inferences, our review is necessarily de novo.
At the time of the operation at which the YA surgeons searched for and identified the subdural empyema, and during which they elected to discard the skull flap, it was well known that focal osteomyelitis was a condition commonly associated with sub-dural empyema. See J. Greenlee, Subdural Empyema, in G. Mandell, J. Douglas & J. Bennett, eds., Principles & Practice of Infectious Disease 592 (2d ed. 1984) (citing four pre-1980 publications). Also at this time, the Lexington VA hospital was governed by the standards of the Joint Commission on Accreditation of Hospitals. The Commission’s 1980 Manual provides:
Specimens removed during a surgical procedure shall ordinarily be sent to the pathologist for evaluation....
Every gross specimen sent to the laboratory shall be examined by a pathologist. ...
[1244]*1244Exceptions to sending specimens removed during a surgical procedure to the laboratory should be made only when the quality of care has not been compromised by the exception, when another suitable means of verification of the removal has been routinely employed, and when there is an authenticated operative or other official report that documents the removal. The limited categories of specimens that may be exempted from the requirement to be examined by a pathologist include, but are not necessarily limited to, the following:
Specimens that by their nature or condition do not permit fruitful examination, such as a cataract, orthopedic appliance, foreign body, or portion of rib removed only to enhance operative exposure;
[six other obviously inapplicable categories] ....
When exemptions are not authorized because of federal or state regulations, the requirements of a training program, or the medical staff bylaws, rules and regulations, the most strict requirements shall apply.
Accreditation Manual for Hospitals 130-31 (1980 ed.) (emphasis added).
The VA physicians attempted to assert at trial that the discarded skull flap was akin to “a portion of a rib removed only to enhance operative exposure,” relying on testimony that discarding the specimen did not compromise Mr. Welsh’s case and that it was common practice to discard skull flaps removed to enhance operative exposure. Although it appears true that Mr. Welsh’s care was not compromised by the discarded bone, the remainder of defendant’s justification for this action is merit-less.
To qualify for the exemption claimed, the specimen must “by [its] nature or condition ... not permit fruitful examination.” Even this exemption applies only when authorized by external or internal regulations, training requirements, or bylaws. In the absence of such authorization, “the most strict requirements” must apply.
Given the common association between osteomyelitis and subdural empyema, and therefore the likelihood that the skull flap was infected, it can hardly be contended that its examination would have been per se “unfruitful.” Moreover, defendant made no showing that any regulations, requirements, or bylaws had authorized the creation of such an exemption. The discarding of the skull flap was flatly contrary to the requirements of the Joint Commission Manual. Although this violation does not constitute negligence per se, it is strong evidence that defendant’s agents were at fault in discarding the specimen; this conclusion was buttressed by additional expert testimony of other physicians at trial. The District Court was therefore surely correct in its finding that the VA surgeons’ act of discarding the skull flap was, if not intentional, at least seriously negligent.
The government argues that negligent destruction of the skull flap does not lead to a conclusion that the medical care of Mr. Welsh was negligent. It argues, in other words, that negligence in this respeet did not cause any injury. It may be true that destruction of the bone caused no medical injury to Mr. Welsh. The destruction did, however, foreseeably prejudice his legal rights. The reason the destruction of the bone was negligent was because the bone would otherwise have been the subject of a thorough pathological examination. This examination would have shown how old the infection was and hence would have shown whether the hospital surgeons were negligent in their diagnosis and treatment.
One can conceive of cases in which mere negligent destruction of bone or tissue might violate medical standards but have no significant legal effect — for example, where a discarded bone sample discloses a fracture the existence of which is provable by other evidence such as X-ray evidence, or when an operation and the related pathological examination is aimed at diagnosis and treatment of Condition A, but later litigation is concerned with whether defendants were negligent in failing to diagnose and treat Condition B. In Mr. Welsh’s case, however, the medical purpose [1245]*1245and the evidentiary purpose were identical. The outcome of the pathological examination would have proven or disproven the crucial fact at issue — the duration and cause of the infection.
III.
In the absence of pathological examination of the skull flap, plaintiff was not able to carry her burden of proof in the ordinary sense on the proximate cause issue. The District Court assisted plaintiff by drawing the several inferences which the United States argues are impermissible.
On this point, the District Court’s drawing of inferences adverse to the defendant based on defendant’s failure to produce evidence that was within its control finds general support in Kentucky case law. See, e.g. Electronic Sales Engineers, Inc. v. Urban Renewal & Community Development Agency, 477 S.W.2d 814, 816 (Ky.1972) (in dispute over whether a lease was signed, failure of plaintiff to produce its employee who had custody of lease justifies presumption her testimony would be adverse); Welch v. L.R. Cooke Chevrolet Co., 314 Ky. 634, 236 S.W.2d 690, 691-92 (1950) (presumption adverse to bailee when it failed to produce ex-employee who was only witness to fire); Rice v. Rice, 243 Ky. 837, 50 S.W.2d 26, 30 (1932) (inference created by a letter confirmed by failure to call its author); cf. Whitcomb v. Whitcomb, 267 S.W.2d 400, 402 (Ky.1954) (adverse presumption unwarranted when uncalled witness was expert equally available to both parties).1
Although the District Court’s general approach conformed with analogous Kentucky law, the United States argues that there is no specific authority for basing the inferences on the discarded skull flap. The parties have cited, and the Court’s own research has found, no Kentucky case dealing with pre-litigation disposal of ultimately critical evidence by a negligent defendant. However, in this Tort Claims Act case our task, as in diversity, is to make our best prediction, even in the absence of direct state court precedent, of what the Kentucky Supreme Court would do if it were confronted with this question. In that inquiry we may rely upon analogous cases and relevant dicta in the decisional law of the State’s highest court, opinions of the State’s intermediate appellate courts to the extent that they are persuasive indicia of State Supreme Court direction, and persuasive opinions from other jurisdictions, including the “majority rule.” See Erie R.R. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938); Bailey v. V & O Press Co., 770 F.2d 601, 604 (6th Cir.1985); 19 C. Wright, A. Miller & E. Cooper, Federal Practice & Procedure § 4507.
The general rule is that the burden of proof lies where the pleadings place it, unless reasons of probability, policy, or fairness dictate otherwise. See Selma, R. & D.R.R. v. United States, 139 U.S. 560, 567-68, 11 S.Ct. 638, 640-41, 35 L.Ed. 266 (1891); see generally E. Cleary, Presuming & Pleading: An Essay on Juristic Immaturity, 12 Stan.L.Kev. 5 (1959); F. James & G. Hazard, Civil Procedure § 7.8 (2d ed. 1977).
When, however, the customary approach would result in placing the burden upon a party who is not in a bétter position to produce the required proof, the courts have not hesitated to allocate the burden to the opposing party. See, e.g., United States v. New York, N.H. & H.R.R., 355 U.S. 253, [1246]*1246256 n. 5, 78 S.Ct. 212, 214 n. 5, 2 L.Ed.2d 247 (1957); Pace v. Hymas, 111 Idaho 581, 726 P.2d 693, 697-98 (1986); see generally Cleary, ed., McCormick on Evidence § 337 (3d ed. 1984).
That an adverse presumption may arise from the fact of missing evidence is a generally accepted principle of law that finds its roots in the 18th century case of the chimney sweeper’s boy who found a jewel ring, took it to a jeweler for appraisal, got back the ring minus the jewel, and brought an action in trover. See Armory v. Delamirie, 1 Strange 505, 93 Eng.Rep. 664 (1722); see generally Stier, Revisiting the Missing Witness Inference, 44 Md.L.Rev. at 142 & n. 22.
The venerable principle of Armory v. Delamirie remains good law. In transporting its wisdom to modern cases, the critical question for the courts has been not whether some kind of adverse consequence should flow from the fact of destruction of evidence, but rather how best to integrate the teaching of Armory into a coherent scheme of 20th century evidentiary principles that includes inferences, presumptions, and shifting burdens of production and persuasion. Compare Nation-Wide Check Corp. v. Forest Hills Distributors, Inc., 692 F.2d 214, 216-20 (1st Cir.1982) (adverse inference from document destruction sufficient to shift burden of tracing proceeds of money order sales) with Stanojev v. Ebasco Services Inc., 643 F.2d 914, 923-24 & n. 7 (2d Cir.1981) (adverse inference from non-production of personnel records not sufficient to cure plaintiff's failure to make out prima facie age discrimination case).
As the Nation-Wide Check court explained, the policy rationales for this type of adverse inference are both evidentiary and deterrent. The evidentiary rationale springs from the common sense notion that a party with notice of an item’s possible relevance to litigation who proceeds nonetheless to destroy it is more likely to have been threatened by the evidence than a party in the same position who does not destroy it:
The fact of destruction satisfies the minimum requirement of relevance [under Fed.R.Evid. 401]: it has some tendency, however small, to make the existence of a fact at issue more probable than it otherwise would be_ Precisely how the document might have aided the party’s adversary, and what evidentiary shortfalls its destruction may be taken to redeem, will depend on the particular facts of each case.
692 F.2d at 218.
The second rationale acts to deter parties from pretrial spoliation of evidence and “serves as a penalty, placing the risk of an erroneous judgment on the party that wrongfully created the risk.” Id.
Destruction of potentially relevant evidence obviously occurs along a continuum of fault — ranging from innocence through the degrees of negligence to intentionality. The resulting penalties vary correspondingly. Some jurisdictions have created causes of action against intentional spoliators. See Williams v. California, 34 Cal.3d 18, 664 P.2d 137, 192 Cal.Rptr. 233 (1983); Smith v. Superior Court, 151 Cal.App.3d 491, 198 Cal.Rptr. 829 (1984); Bondu v. Gurvich, 473 So.2d 1307 (Fla.Dist.Ct.App.1984); see generally Comment, Spoliation: Civil Liability for Destruction of Evidence, 20 U.Rich.L.Rev. 191 (1986).
Some courts assign no adverse evidentia-ry consequences to destruction of evidence that is unintentional or satisfactorily explained. See, e.g., INA Aviation Corp. v. United States, 468 F.Supp. 695, 700 (E.D.N.Y.), aff'd mem., 610 F.2d 806 (2d Cir.1979). Others apply a panoply of sanctions against spoliators, including those who pri- or to litigation or to discovery requests discard evidence they know or should know will be relevant or discoverable. See, e.g., National Ass’n of Radiation Survivors v. Turnage, 115 F.R.D. 543, 556-58 (N.D.Cal.1987) (VA destruction of radiation records warrants Fed.R.Civ.P. 11 sanctions); Struthers Patent Corp. v. Nestle Co., 558 F.Supp. 747, 763-66 (D.N.J.1981) (dicta).
In the medical malpractice context, at least two state appellate courts have endorsed the creation of a rebuttable presumption that would shift the burden of persuasion to a health care provider who [1247]*1247negligently alters or loses records of medical care. See Public Health Trust v. Valcin, 507 So.2d 596, 599-601 (Fla.1987) (negligent loss shifts burden to create jury question on negligence), modifying 473 So. 2d 1297 (Fla.Dist.Ct.App.1984); Thor v. Boska, 38 Cal.App.3d 558, 569 n. 8, 113 Cal.Rptr. 296, 303 n. 8 (1974) (dicta); see also Barker v. Bledsoe, 85 F.R.D. 545 (W.D.Okla.1979) (negligent destruction of autopsy materials by expert hired in advance of litigation warrants more than re-buttable presumption).
In the facts of this case, potential evidence was discarded by parties who were at least negligent and possibly grossly negligent 2 in doing so. Judge Patel, in dealing with the destroyed YA radiation documents, correctly said that “where one party wrongfully denies another the evidence necessary to establish a fact in dispute, the court should draw the strongest allowable inferences in favor of the aggrieved party.” National Ass’n of Radiation Survivors, 115 F.R.D. at 557. The strength of the inference allowable obviously will vary according to the facts and evidentiary posture of a given case. Whether the defendant’s actions may result or must result in an inference that the missing evidence would be unfavorable to the spoliator, or result merely in a burden-shifting presumption, will depend upon a case by case analysis. See Public Health Trust v. Valcin, 507 So.2d at 600-01 (burden shift); Thor v. Boska, 38 Cal.App.3d at 569 n. 8, 113 Cal.Rptr. at 303 n. 8 (dicta re burden shift); Kane v. Northwest Special Recreation Ass’n, 108 Ill.Dec. 96, 155 Ill.App.3d 624, 508 N.E.2d 257, 261-62 (1987) (when plaintiffs mother washed away possible scientific evidence of mentally retarded daughter’s rape in civil suit against institution, proper to instruct jury on permissible adverse factual inference).
IY.
In this case, the District Court drew adverse factual inferences from the VA’s failure to call Dr. Guidry, its former resident, as a witness and from the discarding of the skull flap. Although, as discussed supra note 1, Kentucky law permits adverse inferences from a “missing witness,” an inference that osteomyelitis was of long standing in the skull flap cannot follow logically from Dr. Guidry’s failure to testify. Dr. Guidry was the resident who attended Mr. Welsh at the October 10 clinic visit and, according to the Welsh family, at the disputed September visit. Permissible adverse inferences about his testimony might include one that the September visit occurred or one that the evaluation performed at that visit was inadequate. But no direct inference may be drawn from any testimony Dr. Guidry might have given as to the duration of osteomyelitis in the skull flap: conclusive proof of duration could only come from pathological examination, and Dr. Guidry was not present at surgery when the specimen was discarded.
The District Court inferred from the VA surgeons’ failure to send the skull flap to pathology that “upon surgically entering the decedent’s skull on October 15 ... the doctors found an infection of long standing that had been walled off in some manner so that it did not manifest itself until October 10.” App. 92 (emphasis added). This inference was not mandatory, but it was permissible. In addition, the negligent destruction of evidence foreseeably pertinent to litigation and the consequent failure to perform pathological examination in accordance with the standards of ordinary medical practice give rise in the circum[1248]*1248stances of this case to a rebuttable presumption that the missing specimen would establish that the defendant was negligent in failing to discover the underlying disease process and that this negligence was the proximate cause of the decedent’s demise.
When, as here, a plaintiff is unable to prove an essential element of her case due to the negligent loss or destruction of evidence by an opposing party, and the proof would otherwise be sufficient to survive a directed verdict, it is proper for the trial court to create a rebuttable presumption that establishes the missing elements of the plaintiffs case that could only have been proved by the availability of the missing evidence. The burden thus shifts to the defendant-spoliator to rebut the presumption and disprove the inferred element of plaintiffs prima facie case.
The United States argues that, because Kentucky follows the so-called “bursting bubble” or “vanishing” presumption favored by Professor Thayer that shifts only the burden of production, its courts would never shift the burden of persuasion. This is incorrect. It is true that Kentucky generally employs the Thayerian approach. See, e.g., Lee v. Tucker, 365 S.W.2d 849, 851 (Ky.1963); see generally McCormick on Evidence §§ 342, 344; Laughlin, In Support of the Thayer Theory of Presumptions, 52 Mich.L.Rev. 195 (1953). This approach, however, is not inflexible. In fact, the Kentucky approach to inferences, presumptions, and burdens of proof pays maximum heed to the strength of the underlying proof and the facts of a particular case. See Commonwealth Life Ins. Co. v. Hall, 517 S.W.2d 488, 492 & n. 4 (Ky.1974); Lee v. Tucker, 365 S.W.2d at 851.
Kentucky law does declare that “negligence is never presumed.” Creech Coal Co. v. Louisville & N. R.R., 308 Ky. 414, 214 S.W.2d 381, 384 (1948). This dictum invariably is modified, however, by language asserting that negligence must be established by direct and positive proof thereof, or by the proof of facts or circumstances from which negligence reasonably can be inferred. See, e.g., Whitt v. Cato’s Administrator, 306 S.W.2d 260, 261 (Ky.1957) (citing Creech Coal); Caney Creek Coal Co. v. Ellis, 437 S.W.2d 745, 748 (Ky.1969).
Moreover, Kentucky departs from its strict approach to proof of negligence in the area of res ipsa loquitur. Kentucky’s form of res ipsa is rather strong; it not only allows a factfinder to infer negligence but also, if the inference is forceful enough, it can create a rebuttable presumption of negligence, possibly resulting in a directed verdict. Sadr v. Hager Beauty School, 723 S.W.2d 886, 887 (Ky.App.1987); Bowers v. Schenley Distillers, Inc., 469 S.W.2d 565, 568-69 (Ky.1971); Bell & Koch, Inc. v. Stanley, 375 S.W.2d 696, 697 (Ky.1964).
The Florida Supreme Court expressly drew upon an analogy to the policies underlying the res ipsa doctrine in formulating its approach to negligent spoliation. See Public Health Trust v. Valcin, 507 So.2d at 600 (citing Marrero v. Goldsmith, 486 So.2d 530 (Fla.1986)). There is reason to conclude that the Kentucky Supreme Court also would find this analogy apt. In Embs v. Pepsi-Cola Bottling Co., 528 S.W.2d 703 (Ky.1975), the State’s highest court said:
In the administration of the law we must be satisfied with proof which leads to a conclusion with reasonable probability where absolute logical certainty is not possible. We may be constrained to act upon indecisive evidence where complete proof is impossible. Then the logical, probative force of the evidence produced is measured in part by the test of whether it is the best evidence available.
528 S.W.2d at 706 (emphasis added).3 See also Bell v. Koch, 375 S.W.2d at 697 [1249]*1249(whether res ipsa “inference rises to the dignity of a presumption depends upon its strength (or persuasive power), which will vary with the circumstances of each individual case”) (emphasis added).
In summary, the approach of the Kentucky courts to the missing-witness problem, their approach to the res ipsa doctrine, and their express reliance upon the approach of Dean Prosser to problems of inference and presumption,4 support the conclusion that it is proper under Kentucky law to apply a rebuttable presumption of causation in this case.
Creating this rebuttable presumption occupies a middle ground — it neither simply condones the defendant’s negligent spoliation of evidence at the plaintiff’s expense nor imposes an unduly harsh and absolute liability upon a merely negligent party. See Valcin v. Public Health Trust, 473 So.2d 1297, 1306-07 (Fla.Dist.Ct.App.1985), modified on other grounds, 507 So.2d 596 (Fla.1987). Instead, this approach merely selects which of two parties — the innocent or the negligent — will bear the onus of proving a fact whose existence or nonexistence was placed in greater doubt by the negligent party. The choice is obvious, and in accord with Kentucky and federal evi-dentiary principles. See Embs v. Pepsi-Cola Bottling Co., 528 S.W.2d at 706; Fed. R.Evid. 302; see also Electronic Sales Engineers, Inc., 477 S.W.2d at 816; Welch v. L.R. Cooke, 314 Ky. 634, 236 S.W.2d at 691-92.
Applying this presumption to the facts of this case, it is clear that the United States failed to show that Mr. Welsh’s subdural empyema was not of long standing. Nor did the United States show that the subdural empyema was not caused by spread from an adjacent area of osteomyelitis present long enough before the October 10. visit that the VA doctors should have discovered it. To carry this burden in the absence of the discarded bone flap, the VA was required to persuade the District Court either that there was no osteomyelitis in the bone flap at all, or that the osteomyeli-tis developed so near in time to the October 10 visit that VA doctors could not have discovered it in the exercise of ordinary care. This burden the VA obviously failed to carry.
This rationale for finding for the plaintiff is different from the series of inferences used by the District Court. But a court of appeals may affirm a judgment of a district court on a ground other than that relied upon by the district court. See, e.g., Helvering v. Gowran, 302 U.S. 238, 245, 58 S.Ct. 154, 157, 82 L.Ed. 224 (1937); Krumholz v. Goff, 315 F.2d 575, 582 (6th Cir.1963); see also Herm v. Stafford, 663 F.2d 669 (6th Cir.1981).
Defendant’s other arguments on appeal, including its argument that the District Court erred in denying its Rule 60 motion to produce its missing physician-witness and other evidence, are without merit. One trial is enough. The judgment of the District Court is affirmed.