The parents of a handicapped child brought this civil action for monetary damages against the government pursuant to the Federal Tort Claims Act (“FTCA”). Because the government admitted liability for medical malpractice prior to trial, the district court held a bench trial on the issue of damages only and awarded the plaintiffs over eight million dollars, most of which was to be placed in, a trust.
This case primarily raises questions about the computation and structuring of damage payments to a handicapped child under the FTCA. Specifically, we address the following issues on appeal: (1) the dis*1502trict court’s power to order the parties to place the damages awarded to the handicapped child in a trust that will fully revert to the government upon the child’s death; (2) the date on which interest on the award should begin to accrue; (3) whether the guardian ad litem’s fees should be considered attorney’s fees deductible from the damages award or costs assessed against the government; and (4) the propriety of various damages awarded. For the reasons stated below, we vacate the district court’s decision and remand for proceedings consistent with this opinion.
I. FACTS
On June 28, 1987, Phillip Lee Hull (“Lee”) was born at the Claremore Indian Hospital in Oklahoma, which is owned and operated by the United States (“the government”). The government admits that its doctors were negligent before and during Lee’s delivery. As a result of the doctors’ malpractice, Lee received the following injuries: severe hypoxic ischemic encephalopathy (brain damage), cerebral palsy, with spastic quadriplegia, and developmental delay. Lee will require constant aid and therapy for the rest of his life.
In December 1988, Phillip Gene Hull and Tanya Lee Hull (“the parents”) brought this civil action on behalf of their infant son and on their own behalf for monetary damages under the Federal Tort Claims Act (“FTCA”), 28 U.S.C. §§ 1346(b), 2671-2680. On behalf of Lee, the parents sought damages for economic losses (including lost wages; lost earning capacity; medical treatment costs; and special housing and transportation costs) and noneco-nomic losses (including loss of enjoyment of life; mental and physical pain, suffering, and disfigurement, past, present, and future; and permanent disability). On their own behalf, the parents sought damages for pain and suffering; loss of Lee’s household services; and loss or impairment of the aid, comfort, society, and companionship of Lee.
Due to the government’s admission of liability and causation, the District Court for the Northern District of Oklahoma held a bench trial for the sole purpose of determining damages. On the government’s motion, the court appointed Judith A. Finn as a guardian ad litem to represent Lee’s interests.
The court found that, with proper care, Lee would have a normal life expectancy of 69.8 years at the time of trial. On this basis, the court awarded Lee $5,527,872.37 for medical services and support, $784,-717.50 for fund management, $1,601,474.00 for lost wages and impairment of earning capacity, and $250,000 for pain and suffering. In addition, the court awarded $150,-000 to Mrs. Hull for pain and suffering and $100,000 to Mr. Hull for pain and suffering. In total, the court awarded $8,414,-063.87.
The court instructed the government to pay the parents directly for their pain and suffering, but ordered that the damages awarded to Lee be placed into a trust fund managed by a third-party trustee. The record indicates that the court wanted to establish a trust that would revert fully to the government upon Lee’s death. Apparently, the court desired such a reversionary provision because it was concerned that Lee’s safety would be in jeopardy if his parents were to receive the unspent portion of this damage award upon Lee’s death. However, the court believed that it lacked the power to impose a full reversionary provision absent the consent of the parents.
Instead, the court imposed a partial re-versionary provision whereby if Lee were institutionalized during his life, any unspent funds remaining from the damage award would be divided at the time of his death as follows: His surviving parents or his estate would receive that percentage of the remaining trust funds obtained by dividing Lee’s age at the time of institutionalization by 69.8, his life expectancy at the time of the trial. The remainder of the trust’s assets would revert to the government.1 Thus, Lee’s parents potentially *1503would receive an economic benefit upon Lee’s death to the extent that they chose to provide home care for him rather than to institutionalize him. The district court concluded that Lee would benefit from home care rather than institutional care, and it apparently felt that this provisión would give the parents an incentive to provide such care.2 The parents agreed to this reverter provision.
The government appeals the form of the trust agreement, the date interest began to accrue, the district court’s ruling that the government must pay the attorney’s fees of the guardian ad litem as costs, and the amount of the various damage awards. The parents cross-appeal on the grounds that the district court failed to compensate them, or inadequately compensated them, for certain damages. The guardian ad li-tem, as cross-appellee, argues that it is in Lee’s best interest for the trust fully to revert to the government upon his death and that the government should pay her fees as costs. We have jurisdiction to hear these appeals under 28 U.S.C. § 1291.
II. DISCUSSION
A. Form of the Trust
The government and the guardian ad li-tem argue that the district court should have established a trust that grants a full reversionary interest to the government upon Lee’s death. In fact, they suggest that Lee’s life may be compromised because the existing trust agreement provides that the parents inherit the entire trust assets if Lee dies while in their care.
During the trial below, the district judge apparently shared this concern and he expressed a desire to award Lee’s damages in the form of a fully reversionary trust, whereby any funds remaining in the trust at the time of Hull’s premature death would revert to the government. However, the district judge stated that he was not convinced that he had the power to impose such reversionary conditions on the trust when the parents did not agree to it:
[T]he real issue here is what is [in] the best interest of Phillip Lee Hull. If I had the power I would make every dollar awarded reversionary. I’m not convinced that I have that authority under the law. If I could, I would make every bit other than the attorney’s fee award— that is, all amounts that would go into the trust should in my view be reversion-ary. That is, if anything were to happen to Phillip, if he were to die or if he were to be institutionalized.... I’m not at all convinced that I can enter such an order.
Government’s App. at 177A; see also Findings of Fact and Conclusions of Law at 8, App. of Appellants at 38 (“Although the Court is firmly convinced that a structured settlement would be in the best interests of both parties, under the prevailing law, the Court cannot order such a settlement.”). Accordingly, the judge granted a reversion *1504to the government upon Hull’s death only in the event that Lee was ever institutionalized; if Lee’s parents cared for him at home until his death, then they would be entitled to the entire remainder of the trust’s assets.
The scope of the district court’s authority to specify the form of an award of damages is a question of law, which we review de novo. In re Ruti-Sweetwater, Inc., 836 F.2d 1263, 1266 (10th Cir.1988). For the reasons that follow, we hold that the district court has the inherent power to order the parties to place the money judgment into a fully reversionary trust if such an arrangement is in Lee’s best interest. Accordingly, we remand in order to give the court the opportunity to establish a fully reversionary trust if it determines that such a trust is still appropriate. In making this determination, the court should consider that Lee’s consent to such a trust, through his guardian ad litem, is a strong indication that the trust is in Lee’s best interest.
The relevant jurisdictional statute for suits brought under the FTCA provides that
the district courts ... shall have exclusive jurisdiction of civil actions on claims against the United States, for money damages, accruing on and after January 1, 1945, for injury or loss of property, or personal injury or death caused by the negligent or wrongful act or omission of any employee of the Government while acting within the scope of his office or employment, under circumstances where the United States, if a private person, would be liable to the claimant in accordance with the law of the place where the act or omission occurred.
28 U.S.C. § 1346 (emphasis added).
Some courts have interpreted this provision to require that district courts award only lump sum money judgments in FTCA cases. Frankel v. Heym, 466 F.2d 1226 (3d Cir.1972), appears to be the seminal case on this topic. In Frankel, the court disapproved a proposal whereby a trust was to be established for the benefit of a woman who was permanently disabled by an army vehicle. Under the trust agreement, the government would have been required to supplement the trust from time to time as needed, and any trust funds remaining at the time of the woman’s death were to revert to the government. Id. at 1228. The Third Circuit stated that “in administering the [FTCA,] a district court should not make other than lump-sum money judgments unless and until Congress shall authorize a different type of award.” Id. at 1228-29. The court justified this rule under two rationales: (1) the federal waiver of sovereign immunity under the FTCA incorporates the traditional common law principle that awards in civil suits must take the form of common law money judgments, id., and (2) lump sum money judgments are preferable to judgments that would impose a continuing burden upon the judiciary to supervise the award, id. at 1229.
Other courts have adopted the rule enunciated in Frankel. Some have done so without any further analysis. See, e.g., Andrulonis v. United States, 724 F.Supp. 1421, 1519-20 n. 616 (N.D.N.Y.1989) (“the FTCA contemplates a lump-sum award, and this court lacks the power to enter judgment in this case ‘in terms other than a simple award of money damages,’ and thus cannot mandate the purchase of an annuity or otherwise structure its award”) (citations omitted), affd in part and rev’d in part on other grounds, 924 F.2d 1210 (2d Cir.), vacated sub nom. New York State Dep’t of Health v. Andrulonis, — U.S. -, 112 S.Ct. 39, 116 L.Ed.2d 18, previous op. reinstated after remand, 952 F.2d 652 (2d Cir.1991), cert. denied, — U.S. -, 112 S.Ct. 2992, 120 L.Ed.2d 869 (1992). Others have done so only after a detailed analysis of the rule. See, e.g., Reilly v. United States, 665 F.Supp. 976, 1016-17 (D.R.I.1987) (“in the absence of agreement by the parties to structure the future care award, the law leaves the Court with no alternative but to order payment of a lump sum judgment”), aff'd in part and remanded on other grounds, 863 F.2d 149, 170 (1st Cir.1988) (“Because the FTCA represents a waiver of the government’s sovereign immunity, it is to be construed con*1505servatively. We, therefore, ... hold that, absent unusual circumstances [such as specific statutory authority], the court below was right in insisting that its award of damages be rendered in lump-sum form.”) (footnote and citation omitted).
We agree that courts cannot subject the government to ongoing obligations like the continuing payments proposed in Frankel. However, provided that the government satisfies its obligation up front in one lump sum, nothing in the FTCA prohibits courts from exercising their inherent authority to structure awards or to impose trusts or reverter conditions to ensure that the damage recovery is in the best interest of the victim.3 Indeed, it would be incongruous for the recovery to place the victim’s life in jeopardy. See Reilly, 863 F.2d at 170 (“When a tortfeasor loses at trial, then — absent a statute or the parties’ contrary agreement — it must pay the judgment in one fell swoop. After the wrongdoer and its funds have been parted, the focus shifts: it cannot be doubted that the court has power (1) to ensure that the recovery benefits the victim, and (2) to exercise strict supervision over investment and use of the funds if the victim is a legal incompetent or otherwise in need of protection.”); see also Kosak v. United States, 465 U.S. 848, 868, 104 S.Ct. 1519, 1530, 79 L.Ed.2d 860 (1984) (courts should decide claims under the FTCA “with justice and equity both to the Government and to the claimants”) (quoting Report of the Joint Committee on the Organization of Congress).
Thus, we hold that the district court has the inherent authority to order that Lee’s damages be paid in the form of a fully reversionary trust if it concludes that is in Lee’s best interest, so long as the government’s obligation to Lee ceases when it pays a fixed lump sum to fund that trust. In determining whether such a trust is appropriate on remand, the court should consider what form or structure of damages best serves Lee’s interests from Lee’s perspective only. Thus, the court need not and should not consider whether either the government or the parents assume a risk or whether either might receive a “windfall” upon Lee’s untimely death because of the presence or absence of a reverter, as the court did in Reilly, 665 F.Supp. at 1017-20. The award is to compensate Lee, and Lee only. Thus, the possibility that the parents might succeed to Lee’s award is not relevant. Similarly, once the amount of the award is calculated fully, but not excessively, to absolve the government of its malfeasance, the possibility of a revert-er to the government is not relevant. Both possibilities are only incidental to what must be the only inquiry after the proper amount of the award has been calculated and paid: How should it best be structured to benefit Lee?
The fact that Lee consented to a fully reversionary trust through his legal representative, the guardian ad litem, is a highly relevant consideration because Lee is the real party in interest.4 Nothing in the language of 28 U.S.C. § 1346 or courts’ interpretation thereof prevents the real party in interest from consenting to receive the lump sum judgment in any form, including a reversionary trust. Here, Lee agreed through his legal representative that it would be in his best interest for the government to place his damages, which totalled over eight million dollars, into a fully reversionary trust; in fact, he requested such an arrangement through his legal representative. Because the guardian ad litem — and not the parents — represents Lee, see Garrick v. Weaver, 888 F.2d 687, 692-93 (10th Cir.1989), her consent is the relevant consideration. The parents’ refusal to consent to a fully reversionary *1506trust is irrelevant, particularly in light of their alleged conflict of interest in this case.
In FTCA cases with similar facts, two district courts have imposed trusts granting a reversionary interest to the government with the consent of the real party in interest. See Robak v. United States, 503 F.Supp. 982, 983 (N.D.Ill.1980) (fully rever-sionary trust), affd in -part and rev’d in part on other grounds, 658 F.2d 471 (7th Cir.1981), and Reilly, 665 F.Supp. at 1018-19 (limited reversionary trust).
In Robak, the parents of an eight-year-old rubella-syndrome child brought an action in their own names against the United States under the FTCA and received $900,-000 in damages. They agreed to place the money awarded to them in a reversionary trust for the lifetime future maintenance of the child. 503 F.Supp. at 983. In addition, they agreed that upon the child’s death, all remaining trust proceeds will revert back to the United States in order to prevent any unjust enrichment. Id. The court held without much discussion that such an arrangement was permissible. Id.
In Reilly, the parents of a child who suffered permanent brain damage because of a naval obstetrician’s negligence during delivery brought an action against the United States under the FTCA and received over $11,000,000 in damages on behalf of the child. 665 F.Supp. at 1008. With the agreement of the parents,5 the district court established a trust for the future care award, with a reversion in favor of the government in the event the parents choose to institutionalize her. Id. at 1019. If the parents choose to provide her with home care, however, they will receive the trust proceeds upon her death. Id. at 1018-19. This arrangement is similar to the trust agreement approved by the district court in the instant case.
In addition to the two courts that have actually imposed a reversionary trust, the Seventh Circuit has indicated in dicta that a reversionary trust might be appropriate under certain circumstances, such as “when there is great uncertainty about the victim’s life expectancy, as in [Robak ].” See Nemmers v. United States, 795 F.2d 628, 636 n. 4 (7th Cir.1986). There will almost always be great uncertainty about the life expectancy of a severely injured child.
Ultimately, the district court has the inherent power to ensure that Lee obtains the maximum benefit from his award. Accordingly, we remand to give the district court an opportunity to approve the establishment of a new trust that includes a full reversionary interest to the government if the court still desires to do so. If the court decides a new trust is unnecessary, we ask the court to consider (1) minor changes that might lessen the tension between Lee’s interests and the parents’ interests (for example, the government suggests that the court might provide the parents with' reasonable access to trust funds to pay family expenses); (2) a provision whereby the government will be reimbursed for services that the Indian Health Service renders to Lee in the future; and (3) the appointment of a different guardian ad litem, as requested by the current guardian ad litem.6 The court’s decision as to the structure of the award will be reviewable only for an abuse of discretion.
B. Accrual of Interest
Interest is recoverable against the United States only when expressly provided for by statute. Library of Congress v. Shaw, 478 U.S. 310, 314, 106 S.Ct. 2957, 2961, 92 L.Ed.2d 250 (1986). 31 U.S.C. § 1304(b)(1)(A) permits the recovery of interest on an FTCA judgment for the period *1507during which the judgment is on appeal: “Interest may be paid from the appropriation made by this section on a judgment of a district court, only when the judgment becomes final after review on appeal or petition by the United States Government, and then only from the date of filing of the transcript of the judgment with the Comptroller General through the day before the date of the mandate of affirmance, [emphasis added]” Rule 58 of the Federal Rules of Civil Procedure delineates the requirements for a judgment: “Every judgment shall be set forth on a separate document. A judgment is effective only when so set forth and when [the clerk enters the judgment in the civil docket] as provided in Rule 79(a).”
In its June 5,1991 Order Nunc Pro Tunc, the district court declared that October 19, 1990 — the date on which the court entered its original Findings of Fact and Conclusions of Law, which specified the amount of damages awarded — was the relevant judgment date for purposes of determining when interest on Lee’s award began to accrue. On appeal, the government contends (1) that it was by no means clear that the October 19, 1990 findings and conclusions constituted a “judgment” until the district court entered its Amended Findings of Fact and Conclusions of Law on April 12,1991 and (2) that the June 5,1991 Order Nunc Pro Tunc — in which the district court directed the clerk of the district court to “officially enter on its docket the October 19, 1990 Findings of Fact and Conclusions of Law as a judgment” — was the first document to meet the requirements of Rule 58.
Because this is a purely legal issue, we review the court’s decision de novo. Ruti-Sweetwater, 836 F.2d at 1266. For the following reasons, we conclude that interest on Lee s award began to accrue on June 5, 1991.
First, we agree with the government that it was not clear that the October 19, 1990 findings constituted a “judgment” until the district court entered its Amended Findings of Fact and Conclusions of Law on April 12, 1991, when the district court declared:
The amended judgment rendered herein does not alter the amount delineated by the Court in its original judgment of October 19, 1990. From that date the total amount of damages has been fixed, certain and final, and was accordingly filed of record with the United States Government Accounting Office by Plaintiff’s counsel on October 23, 1990. Therefore, the Court orders that interest should be paid on the total damage award in the amount of 6.26% from October 23, 1990 to date of deposit by Defendant.
Amended Findings of Fact and Conclusions of Law at 24, App. of Appellants at 68. Thus, the parents could not have filed “the transcript of the judgment” required by section 1304 when they filed the court’s findings and conclusions with the Comptroller General on October 23, 1990.
In fact, both parties repeatedly evidenced confusion and requested that the court enter a judgment.7 Indeed, during a hearing on January 17,1991, the district court itself indicated that no judgment had yet been filed and declined to enter judgment because of the possibility of a structured award:
MR. WOLFE [attorney for parents and Lee]: Do we have a final appealable judgment, was that the intent of the Court October the 19th?
THE COURT: ... [T]he judgment has not been entered at this point....
*1508[I]t's my understanding until the judgment itself is entered there is no appeal that can be made....
Government's App. at 178-79; see id. at 190.
Second, October 19,1990 cannot serve as the date of judgment because the district court did not satisfy Rule 58 at that time. The district court’s June 5, 1991 Order Nunc Pro Tunc is the first document that satisfies Rule 58.
This Court has strictly applied the separate document rule. See United States v. City of Kansas City, Kan., 761 F.2d 605, 606-07 (10th Cir.1985); United States v. Clearfield State Bank, 497 F.2d 356, 358-59 (10th Cir.1974) (“Rule 58 applies where it is uncertain whether a final judgment has been entered, as where a trial judge writes an opinion or memorandum providing only the basis for the entry of judgment, but containing apparently directive or dispositive words such as ‘defendant’s motion for summary judgment is granted.’ In such a situation, a judgment must be set out on a document separate from the opinion or memorandum.”).
The parents erroneously construe Bankers Trust Co. v. Mallis, 435 U.S. 381, 98 S.Ct. 1117, 55 L.Ed.2d 357 (1978) (per cu-riam), to preclude the application of Rule 58 for purposes of determining when post-judgment interest begins to accrue. In Mallis, the Supreme Court determined that the Court of Appeals appropriately assumed jurisdiction under 28 U.S.C. § 1291 despite the district court’s failure to satisfy Rule 58 because the parties waived the separate judgment requirement. The parents focus upon the statement in Mallis that “the sole purpose” of the separate document requirement of Rule 58 “was to clarify when the time for appeal ... begins to run.” Id. at 384, 98 S.Ct. at 1120. They ignore, however, the legislative history cited immediately after this statement, which indicates that the Advisory Committee intended for Rule 58 to eliminate uncertainties regarding the commencement of time “for post verdict motions and for the purpose of appeal.” Id. at 385, 98 S.Ct. at 1120. We reject the parents’ narrow construction of Mallis because the opinion as a whole reveals that the drafters of Rule 58 intended to clarify the time that judgment was entered, whether for purposes of calculating the time for appeal or for other purposes where certainty as to the date of judgment is important.
In a case decided prior to Mallis, we held that Rule 58 precluded us from awarding interest prior to the entry of formal judgment. Millers’ Nat’l Ins. Co. v. Wichita Flour Mills Co., 257 F.2d 93, 104 (10th Cir.1958). In Millers’, we applied Rule 58 and declined to accrue interest on the date of a verdict when there was a two-year delay in the entry of formal judgment because the statute authorizing interest in that case, 28 U.S.C. § 1961, authorized interest only from the date of entry of the judgment. Id. The interest statute involved here, 31 U.S.C. § 1304, similarly requires a judgment before the interest can begin to accrue. We hold that Millers’ remains good law even after Mallis.
A number of circuits similarly have applied Rule 58 subsequent to Mallis to determine the date of judgment for purposes of calculating accrual of interest. See, e.g., In re Red Carpet Corp., 902 F.2d 883, 893-94 (11th Cir.1990); Marshall v. Perez-Ar-zuaga, 866 F.2d 521, 523 (1st Cir.1989). Other circuits, however, have declined to apply Rule 58 to determine when interest begins to accrue on a judgment, not on the basis of Mallis, but because the purpose of post-judgment interest statutes is to protect a successful party from losing the value of his or her award during the time that the award is contested. See, e.g., Poleto v. Consolidated Rail Corp., 826 F.2d 1270, 1280-81 (3d Cir.1987); Merrill Lynch, Pierce, Fenner & Smith, Inc. v. Knudsen, 749 F.2d 496, 497-98 (8th Cir. 1984); Turner v. Japan Lines, Ltd., 702 F.2d 752, 756-57 (9th Cir.1983). Accordingly, where the district court did not enter formal judgment promptly, these courts have held that interest should begin to accrue on the date of the verdict or award. Thus, it appears that these courts have adopted an equitable definition of “judg*1509ment” for purposes of awarding post-judgment interest.
The precedent in our Circuit precludes us from adopting such an approach. See Millers’, 257 F.2d at 104. Furthermore, in this case we are dealing not with the general interest statute, 28 U.S.C. § 1961, but rather with the specific statute that allows interest to be recovered against the government, 31 U.S.C. § 1304(b)(1)(A). Because waiver of sovereign immunity is always construed narrowly, Shaw, 478 U.S. at 318, 106 S.Ct. at 2963, we must give the term “judgment” the narrow definition supplied in Rule 58.
The parents next contend that the parties have waived the separate document requirement if Rule 58 is applicable. We disagree. This requirement is deemed waived if
[1] [T]he District Court clearly evidenced its intent that the opinion and order from which an appeal was taken would represent the final decision in the case. [2] A judgment of dismissal was recorded in the clerk’s docket. And [3] [the party against whom the appeal was taken] did not object to the taking of the appeal in the absence of a separate judgment.
Mallis, 435 U.S. at 387-88, 98 S.Ct. at 1121-22 (bracketed numbers added). These prerequisites are not satisfied in the instant case; in fact, the record contains affirmative evidence to the contrary.8
The district court’s June 5, 1991 Order Nunc Pro Tunc is the first document that meets the requirements of Rule 58. The government concedes in its brief that this order satisfies the requirements of 31 U.S.C. § 1304(b)(1)(A) even though it was never filed with the Comptroller General; the parents did file the original and the amended Findings of Fact and Conclusions of Law, both of which were discussed in the Order Nunc Pro Tunc. See Brief for Appellant, Cross Appellee United States of America at 22-23.
Accordingly, we reverse the district court’s determination that interest began accruing on October 23, 1990 and remand for the court to recalculate the accrued interest from June 5, 1991.
C. Fees of the Guardian Ad Litem
The district court ordered the government to pay for the costs of the guardian ad litem pursuant to Federal Rule of Civil Procedure 54(d), which provides that “costs against the United States, its officers, and agencies shall be imposed only to the extent permitted by law.” See Amended Findings of Fact and Conclusions of Law at 25, App. of Appellants at 69. The government argues on appeal that the fees of the guardian ad litem should be deducted from the damages awarded to Lee pursuant to 28 U.S.C. § 2678 (“Attorney fees”) because the guardian ad litem acted as an attorney. We remand for the district court to separate that portion of the guardian ad litem’s fee that should be considered attorney’s fees from that part that should be considered costs consistently with the following discussion.
“Except to the extent it has waived its immunity, the Government is immune from claims for attorney’s fees.” Ruckelshaus v. Sierra Club, 463 U.S. 680, 685, 103 S.Ct. 3274, 3278, 77 L.Ed.2d 938 (1983). “[W]e must construe waivers strictly in favor of the sovereign and not enlarge the waiver ‘beyond what the language requires.’ ” Shaw, 478 U.S. at 318, 106 S.Ct. at 2963 (citations omitted). 28 U.S.C. § 2678 pro*1510vides the sole authorization for attorney’s fees in FTCA actions. This section permits attorney's fees of up to 25 percent of the judgment to be taken from the judgment itself.
Whether payments to the guardian ad litem should be taxed as costs pursuant to Rule 54(d) or deducted as attorney’s fees from the judgment pursuant to 28 U.S.C. § 2678 depends upon the role -that the guardian ad litem plays. To the extent the guardian ad litem acts as an officer of the court, looking after the interests of the minor, compensation is taxable as costs. To the extent the guardian ad litem performs legal services as an attorney, compensation should be deducted as attorney’s fees. See duPont v. Southern Nat’l Bank, 771 F.2d 874, 882-83 (5th Cir.1985) (distinguishing between guardian ad litem and attorney ad litem and holding that only expenses of guardian ad litem are taxable as costs), cert. denied, 475 U.S. 1085, 106 S.Ct. 1467, 89 L.Ed.2d 723 (1986); Franz v. Buder, 38 F.2d 605, 607 (8th Cir.1930) (“Where the services as guardian ad litem and as an attorney are rendered by the same person, a separate allowance should be made,for each service; the amount allowed for services as guardian ad litem being taxable as costs and the amount allowed for services as counsel fees assessable against the estate of the incompetent.”); cf. Schneider v. Lockheed Aircraft Corp., 658 F.2d 835, 854-55 (D.C.Cir.1981) (legal research and analysis expenses of guardian ad litem taxed as costs when performed on behalf of court rather than in course of representing infant).
In this case, the parents argue that the fees for the guardian ad litem’s services can in no way be considered attorney’s fees for purposes of section 2678 because the court appointed her over the parents’ objections arid she often argued against the parents’ interests. We reject this argument because, as established supra, Lee is the primary real party in interest in this case, and it appears that the guardian ad litem performed some legal services on his behalf.
The government contends that the guardian ad litem denominated many of her claims for compensation as attorney’s fees. Specifically, the government asserts, the “Application of the Guardian Ad Litem For Attorney Fees and Expenses” includes charges for various legal services, such as “trial preparation, preparation for and attendance at depositions, preparation of motions, and trial of the lawsuit.” See Reply Br. for Appellant, Cross Appellee United States at 12. Unfortunately, we cannot evaluate the government’s assertion because it failed to include this application as part of the appellate record. Accordingly, we must remand for the district court to examine that application and to determine what portion of the guardian ad litem’s compensation was properly taxed as costs and what portion should have been deducted from the damages award as attorney’s fees.
D. Specific Items of Damages
1. Adequacy of Damages -Awarded For Home Health Aide Services
The- parents contend that the district court awarded an insufficient amount for the cost of Lee’s home health aide services. During the trial, the parties presented conflicting evidence as to what services would be required. We cannot say that the court’s decision to adopt the. recommendation of the government’s expert was clearly erroneous.
2. Government’s Claim That The District Court Did Not Properly Adjust The Damages Awards to Present Value
The government contends that the district court (1) improperly failed to adjust portions of the award for future damages to present value and (2) improperly adjusted the award for lost future earnings by using a negative discount rate.
In calculating damages, the trier of fact should discount awards for future damages to present.value to account for the effects of investment, but the effects of inflation should also be considered, and, with regard to wages, other factors may *1511need to be considered that may reasonably lead to a projection of increased wages over time. See Hoskie v. United States, 666 F.2d 1353, 1355-56 & n. 2 (10th Cir. 1981); Steckler v. United States, 549 F.2d 1372, 1378 (10th Cir.1977). See generally Jones & Laughlin Steel Corp. v. Pfeifer, 462 U.S. 523, 532-47, 103 S.Ct. 2541, 2548-56, 76 L.Ed.2d 768 (1983) (analyzing methods of computing damage awards and selecting discount rate in the context of the Longshoremen’s and Harbor Workers’ Compensation Act).
Some courts have sought to avoid the complexity of such computations by assuming that investment interest and inflation rates will normally have a fairly fixed relationship to each other so that a net discount rate may be used that merely reflects the rate by which investment interest normally will exceed inflation. This is also referred to as the real rate of return that an investor receives. Although some courts have assumed parity between investment interest and inflation, e.g., Barnes v. United States, 685 F.2d 66, 70 (3d Cir.1982) (applying zero discount rate as part of “total offset” method, whereby investment interest rates and inflation rates are presumed to be equal and to cancel one another out), most courts assume that the net discount rate will be in the range of one to three percent. E.g., Doca v. Marina Mer-cante Nicaraguense, S.A., 634 F.2d 30, 39-40 (2d Cir.1980) (fixing discount rate at two percent unless litigants offer evidence to the contrary), cert. denied, 451 U.S. 971, 101 S.Ct. 2049, 68 L.Ed.2d 351 (1981); Trevino v. United States, 804 F.2d 1512, 1519 (9th Cir.1986) (extending to FTCA case Supreme Court’s statement in Pfeifer that a trial court should not be reversed “if it adopts a rate between 1 and 3% and explains its choice,” 462 U.S. at 548-49, 103 S.Ct. at 2556), cert. denied, 484 U.S. 816, 108 S.Ct. 70, 98 L.Ed.2d 34 (1987). Whether a net discount rate is used or actual investment interest and inflation figures are used, the basis of the computation must be explained and supported by competent evidence.
The government asserts that the district court failed to discount portions of Lee’s future damages awards to present value. We agree that nothing on the face of the court’s Findings of Fact and Conclusions of Law reveals that the court discounted the awards as required to reflect present worth. The parents argue that the court obviously used the offset method of computing present value and assumed a zero net discount rate; therefore, no adjustment was necessary. Because the court’s choice of a discount rate has a significant effect upon the amount of damages awarded, we decline to make such an assumption, especially given the court’s use of a non-offsetting method in computing lost future earnings and given the fact that a zero discount rate ordinarily will not reflect economic reality. Accordingly, we remand for the district court to make appropriate findings of fact explaining its method of calculating present value for all awards for future damages except lost future earnings.
On remand, the district court should bear in mind that Steckler precludes a mechanical application of the offset approach. See 549 F.2d at 1378. In Steckler, we adopted the rationale of United States v. English, 521 F.2d 63 (9th Cir.1975), wherein the Ninth Circuit rejected the offset doctrine in an action under the FTCA. That court stated: “[T]he court may not assume that the discount rate and the inflation rate will net to zero. The lower court must first estimate future income and expenses, taking into account estimated changes in the purchasing power of the dollar, and then discount this future net income stream to its present value.” Id. at 75. Thus, the offset approach is appropriate in this Circuit only if the court uses competent evidence to calculate the interest rate and the inflation rate and they happen to be equal and therefore to cancel each other out.
We affirm, however, the present value calculation of Lee’s lost future earnings. Although the district court did not make specific findings of investment interest rates, inflation rates, and predicted growth in wages, the plaintiffs’ expert did *1512address these factors in his testimony of the present value of Lee’s lost future earnings. Because the district court adopted the figure proposed by this expert, we may conclude that the district court also adopted the expert’s methodology.9 Specifically, the plaintiff’s expert used a discount rate of 1.64 percent, which took into account both inflation and the investment interest rate. However, he also applied a 3.29 percent “growth rate” to Lee’s projected earnings, which represents Lee’s estimated annual increase in earning power— beginning at age 20 — without considering inflation. The expert then concluded that Lee’s earning power, in terms of present value, would increase 1.65 percent (3.29— 1.64) each year.
We review the district court’s choice of a discount rate under a clearly erroneous standard. See Hoskie, 666 F.2d at 1356. Although some courts have generally disallowed the use of a growth rate factor in the estimation of lost future income, see, e.g., McCarthy v. United States, 870 F.2d 1499, 1503 (9th Cir.1989); Cunningham v. Quaker Oats Co., 107 F.R.D. 66, 79 (W.D.N.Y.1985), we cannot say that the district court’s decision to rely upon expert testimony utilizing this methodology constitutes clear error. The Supreme Court recognized this approach in Pfeifer, 462 U.S. at 548, 103 S.Ct. at 2556, and the parents’ expert provided the court with a detailed basis for his calculation. Accordingly, we affirm the $1,601,474.00 award for lost wages and impairment of earning capacity.
3. Fund Management Award.
The government also asserts that the district court’s award of $784,717.50 for fund management — i.e., remuneration for the trustee — is excessive. “The standard for determining whether an award of damages is excessive is whether the award shocks the judicial conscience.” Miller v. United States ex rel. Dep’t of the Army, 901 F.2d 894, 897 (10th Cir.1990). Because the trustee has a significant amount of responsibility to ensure Lee’s continuing welfare, we cannot say that this award is so excessive as to shock the judicial conscience.
4. Miscellaneous Other Damage Items As To Which There Are Inadequate Findings
Finally, the parents argue that the district court erred in failing to award damages for specialized adaptive housing; personal services rendered by Mrs. Hull; housekeeping services required by Lee; lost household services; physical disability and disfigurement; past medical expenses; psychological counseling; and dental expenses. They allege that the existence of these damages was established at trial, leaving only the amount of damages to be determined by the court.
“It is a fundamental legal principle that the determination of the quantum of damages in civil cases is a fact-finder’s function.” Bennett v. Longacre, 774 F.2d 1024, 1028 (10th Cir.1985). Thus, the fact that the parents presented some evidence regarding the above damages during the trial is not determinative. The court was entitled to discount this testimony based upon its observation of the witnesses and its assessment of their credibility.
However, the district court’s power to determine damages is not without constraints. Federal Rule of Civil Procedure 52(a) requires the district court to make a finding of fact to justify its decision to award, or not to award, damages. See Nemmers, 795 F.2d at 634-35 (where parents brought medical malpractice action against the government to recover damages arising from the delivery of their child, district court’s statement that “[t]he Court declines to award compensation for ‘quality of life’ considerations” was deemed to be an insufficient finding of fact under Rule 52(a)).
*1513With the exception of the claim for psychological services, see Findings of Fact and Conclusions of Law at 6, App. of Appellants at 36, the district court did not mention these damage claims in its original or amended Findings of Fact and Conclusions of Law. Accordingly, although we can find support in the record or at law for the district court’s failure to award certain of these damage claims and the court may have concluded that other of these damage claims were subsumed in the large damage awards that were given, we remand for the court to enter appropriate findings of fact or conclusions of law to support its decision denying damage claims for specialized adaptive housing; personal services rendered by Mrs. Hull; housekeeping services required by Lee; lost household services; physical disability and disfigurement; past medical expenses; and dental expenses.
III. CONCLUSION
The decision of the district court is VACATED and REMANDED for proceedings consistent with this opinion.
. A later provision in the Amended Findings of Fact and Conclusions of Law appears to be contradictory, as it suggests that the reverter occurs at the time Lee is institutionalized rather *1503than upon his death. See Amended Findings of Fact and Conclusions of Law at 23-24, App. of Appellants at 67-68 (“The Trust will not be re-versionary in the event of Lee Hull’s early and untimely death, but is directed by the Court to revert to the United States should Lee ever be placed in an institutional environment. This reversion is to be calculated by subtracting from the total Trust Estate remaining at the time of institutionalization, the cost of Lee’s institutional care from that year until he reaches the age of 69.8 years. For example, should Lee be institutionalized at the age of 20, the reversion would be calculated by subtracting from the total Trust Estate the cost of institutional care for 49.8 years.”); see also id. at 25, App. of Appellants at 69 ("The Trust will not be rever-sionary in the event of Lee Hull’s early and untimely death, but is directed by the Court to revert to the United States via the formula indicated should Lee ever be placed in an institutional environment.’’). Such a provision makes no sense because it would potentially deprive Lee of a portion of his damages for pain and suffering, lost economic opportunities, and the like during his lifetime merely because his parents chose to institutionalize him. Upon remand, this provision should be clarified. The proposed trust document drafted by the parents states more consistently that the trust terminates upon Lee’s death and then a portion of the remainder reverts to the government if Lee has been institutionalized. See Government’s App. at 21-22.
. However, this provision would not provide an incentive against Lee’s untimely death while in the care of his parents. In that event, the entire remaining amount of the damage award would go to the parents and there would be no reverter to the government.
. This approach is consistent with the rationales set forth in Frankel in that (1) the government is paying a lump sum and (2) a properly structured trust managed by an independent trustee does not place a continuing burden on the judiciary.
. The government need not consent to the rever-sionary interest. Sovereign immunity mandates that Congress must agree if novel types of awards are to be permitted against the government, but sovereign immunity does not preclude the government from receiving money back after it has satisfied its lump sum obligation.
. The court’s consideration of the parents’ consent in Reilly is not inconsistent with our position that the consent of the parents in the instant case is irrelevant. Reilly differs from this case in two important respects. First, the child in Reilly was represented by her parents, not by a guardian ad litem. Id. at 978. Second, the possibility of a conflict between the parents' interest and the child’s interest was not an issue in Reilly. Id. at 1019.
. On remand, the court may of course reconsid- . er any arguments regarding the provisions of the trust agreement. Indeed, we note that the district court has retained the power to clarify the trust provisions if requested to do so by the trustee or the parents.
. For example, as of November 6, 1990, the government expressed in its “Response to Plaintiffs’ Motion to Alter or Amend the Court’s Findings of Fact and Conclusions of Law and Request to Enter a Final Judgment Providing for a Structured Award Reversionary to the Government” its belief that the court had not yet entered a final judgment. See Government’s App. at 6, 8. In addition, during both the January 17, 1991 hearing and their February 1, 1991 “Motion for Court to Enter Separate Judgments for Plaintiffs and Motion for Court to Bifurcate Trust Proceeding(s) and Brief in Support Thereof,” the parents acknowledged the absence of a judgment and urged the court to enter a judgment so that interest could commence. See id. at 12, 14-15, 179, 183, 187, 189.
. It is not enough, as the parents contend, that the district court retroactively satisfied the first two prerequisites through its June 5, 1990 Order Nunc Pro Tunc. The court’s intent at the time it issued the original Findings of Fact and Conclusions of Law is controlling, and, as discussed supra, the court itself stated on January 17, 1990 that no judgment had yet been entered. If the Mallis waiver requirements could be satisfied retroactively as the parents suggest, a court could theoretically prevent an appeal by declaring that it intended to enter judgment long before it actually did so such that the period in which to appeal would have already expired.
Furthermore, the waiver rule set forth in Mal-lis appears to be geared toward situations in which an oversight results in a deviation from Rule 58 and compliance with the requirements of Rule 58 is never really at issue. In this case, the deviation from Rule 58 was clearly not an oversight and the date of the entry of judgment was an ongoing issue in the litigation.
. By contrast, the district court did not adopt this expert's recommended present value calculations for the other future damages awards. Therefore, we may not assume that the court adopted the expert’s methodology for those awards.