United States v. James

• Court: Court of Appeals for the Second Circuit
• Decided: August 29, 2025
• Docket: 24-849
• Status: Published

Opinion

24-849-cr United States v. James

United States Court of Appeals for the Second Circuit August Term, 2024

(Argued: January 28, 2025 Decided: August 29, 2025)

Docket No. 24-849-cr

_____________________________________

UNITED STATES OF AMERICA,

Appellee,

v.

MATHEW JAMES,

Defendant-Appellant. _____________________________________ Before:

SACK, LYNCH, and LOHIER, Circuit Judges.

A jury convicted Mathew James of health care fraud, conspiracy to commit health care fraud, wire fraud, and identity theft for a scheme that involved falsifying insurance claims, manipulating medical records, and impersonating patients when communicating with insurance companies. The United States District Court for the Eastern District of New York (Seybert, J.) sentenced James to 144 months’ imprisonment and ordered him to forfeit $63,382,049.02 and to pay $336,996,416.85 in restitution. Challenging both his conviction and sentence on appeal, James argues that the District Court mishandled the jury’s potential exposure to extra-record material, incorrectly applied sentencing enhancements, improperly lengthened his sentence based on his potential eligibility for earned time credits under the First Step Act and rehabilitation programs, and erred in calculating the forfeiture and restitution amounts. We reject James’s challenge to his conviction, but we conclude that the District Court improperly enhanced James’s sentence under the United States Sentencing Guidelines, including in violation of Tapia v. United States, 564 U.S. 319 (2011), and failed to properly calculate the forfeiture and restitution amounts. Accordingly, we AFFIRM James’s conviction, VACATE his sentence, including the forfeiture and restitution orders, and REMAND for resentencing.

JULIAN S. BROD (Alexandra A.E. Shapiro, C. Eric Hintz, on the brief), Shapiro Arato Bach LLP, New York, NY, for Defendant-Appellant.

ANTOINETTE N. RANGEL, CATHERINE M. MIRABILE, Assistant United States Attorneys (David C. James, Amy Busa, Assistant United States Attorneys, Miriam L. Glaser Dauermann, Trial Attorney, on the brief), for Breon Peace, United States Attorney for the Eastern District of New York, Brooklyn, NY, for Appellee.

LOHIER, Circuit Judge:

Mathew James, the owner of a medical billing business that submitted

insurance claims on behalf of doctors, was tried and convicted on charges of

health care fraud, conspiracy to commit health care fraud, wire fraud, and

aggravated identify theft. The charges stemmed from James’s role in a complex

scheme to defraud insurance companies by “upcoding” medical procedures,

falsifying reports, directing patients to emergency rooms for pre-planned

surgeries, and even impersonating patients in phone calls to the insurance

companies. The United States District Court for the Eastern District of New York

2 (Seybert, J.) sentenced James to 144 months’ imprisonment and ordered him to

forfeit $63,382,049.02 in criminal proceeds and to pay $336,996,416.85 in

restitution.

On appeal, James challenges his conviction and sentence. As for his

conviction, he contends that the District Court mishandled the jury’s potential

exposure to recorded calls that were not introduced into evidence at trial. As for

his custodial sentence, he claims that the District Court improperly enhanced his

sentence under the United States Sentencing Guidelines (based on his leadership

role and abuse of trust in the scheme) and impermissibly considered his potential

eligibility for earned time credits under the First Step Act and prison

rehabilitation programs in imposing a higher sentence, the latter in violation of

Tapia v. United States, 564 U.S. 319 (2011). Finally, James challenges the District

Court’s forfeiture and restitution orders as excessive because they include

revenue derived from his legitimate business activities.

We hold that any error regarding the jury’s exposure to the material not

admitted into evidence was harmless. But we conclude that the District Court

erroneously relied on the potential for earned time credits as a stand-alone factor

and improperly considered James’s eligibility for rehabilitation programs in

3 imposing a term of imprisonment, applied two enhancements under the

Sentencing Guidelines without adequate factual findings, and erred in

calculating the forfeiture and restitution amounts that James must pay. We

therefore affirm James’s conviction but vacate his sentence, including the

forfeiture and restitution orders, and remand for resentencing consistent with

this opinion.

BACKGROUND

I

We recount the trial evidence “in the light most favorable to the

Government.” United States v. Litwok, 678 F.3d 208, 211 (2d Cir. 2012).

James is a former nurse who established what appeared to be a successful

third-party medical billing practice that prepared and submitted medical

insurance claims to insurance companies on behalf of doctors. Doctors retained

James to submit their high-dollar value, out-of-network claims to insurers in

exchange for ten to twelve percent of the insurance reimbursements they

received from those claims.

At all relevant times, James’s business employed and trained two to three

employees, typically young and inexperienced, who reviewed medical reports,

4 identified key medical terms, and inputted billing codes on the medical claim

forms submitted to insurance companies. From 2013 to 2019 his business served

nearly 150 physicians and submitted tens of thousands of insurance claims.

As it turns out, James was masterminding a long-running scheme to

defraud the insurance companies. The Government’s evidence focused on two

ways in which he falsified insurance claim forms as part of his billing practice:

“upcoding” and “unbundling.” “Upcoding” involved billing for costlier medical

procedures than were actually performed. Gov’t App’x 265. “Unbundling”

involved billing “for multiple services that would be included in the primary

procedure.” Gov’t App’x 267.

The upcoding scheme entailed conspiring with doctors to falsify their

reports in order to justify fake codes. To maximize reimbursements, James

directed patients to use emergency rooms for pre-planned surgeries and also

instructed employees to add certain codes to claims forms for more complex

procedures than had actually been performed or for procedures that had not

been performed at all.

As for unbundling, James instructed his employees to improperly use a

coding modifier—Modifier 59—to “unbundle” claims that ordinarily would be

5 billed as a single procedure. In essence, misusing Modifier 59 falsely represented

to insurance companies that component procedures were performed separately.

It also permitted James to bypass the insurance companies’ automated systems

designed to detect and rebundle such codes.

Worse yet, equipped with personally identifiable patient information,

including names, dates of birth, addresses, and insurance details, James went so

far as to impersonate (or instruct his female employees to impersonate) patients

or pose as their family members in phone calls to insurance companies when

claims were denied or reduced.

At trial, patients confirmed that their medical procedures were far less

complex than the corresponding claims reflected. And according to a forensic

accountant who testified for the Government, James received over $63 million in

proceeds from his medical billing business. Based on the ten to twelve percent

commission James generally charged his clients, the Government calculated that

6 the total amount his doctor-clients received from insurance companies over the

course of the scheme stood between approximately $528 and $634 million.

II

A

James was tried on charges of conspiracy to commit health care fraud, 18

U.S.C. § 1349, health care fraud, 18 U.S.C. § 1347, wire fraud, 18 U.S.C. § 1343,

aggravated identify theft, 18 U.S.C. § 1028A, and money laundering conspiracy,

18 U.S.C. § 1956(h). At trial, James’s primary defense was that he acted in good

faith. While admitting to coding errors, he pointed to examples of coding (or

correcting a coding error) to his financial detriment, and one instance in which he

fired an employee who had made too many coding errors. James also asserted

that he relied on the doctors with whom he worked to ensure the accuracy of the

submitted claims, and that doctors—not billers—were ultimately responsible for

billing decisions. The underlying medical records attached to those claims,

James insisted, gave insurance companies the information they needed to assess

and either approve or deny each claim.

It appears to have been undisputed at trial that not all of James’s business

was tainted by the alleged fraud. Indeed, a Government agent readily testified

7 that James’s “misused . . . codes . . . [were only] a subset of [his] total billing.”

App’x 240.

B

Near the start of trial, the Government provided each juror with a physical

binder of transcripts of recorded phone calls to which the defense had not

objected. App’x 393. The binder therefore included transcripts, marked as

exhibits, of calls that were eventually admitted into evidence as well as some

calls that were not admitted but that the Government, prior to trial, had reason to

believe it might introduce as the trial progressed. The Government “ultimately

did not put” two of the recorded calls into evidence, although transcripts of those

calls were included in the binders provided to the jurors during trial. App’x 393,

401. Both of the unadmitted calls involved James and one of his employees,

Jennifer Flanagan, who had impersonated patients on James’s behalf. App’x 393.

The Government apparently removed the transcripts of the two unadmitted calls

from the jury binder at the end of trial. “For the duration of the trial” and prior

8 to its deliberations, however, the jury had access to the transcripts of both

unadmitted calls. App’x 404.

During its deliberations, the jury asked for “2 copies” of the “[b]inder of

telephone transcripts.” App’x 507. In response, the parties provided the jury

with a binder containing only transcripts of the calls that had been admitted.

App’x 392–93. The jury thereafter requested “[a]ll phone transcripts — not just

those presented in the trial.” App’x 508. And in yet another follow-up note, the

jury asked: “Has there been any materials or transcripts removed from the call

transcript binders, such as call ex[]hibits 1, 2 and 3, or any others? Specif[]ically

we are looking for a call between Mat[hew James] and an employee.” App’x 509.

The parties speculated that this last jury note was prompted by “at least

one juror [who] read at least one of [the transcripts of the two unadmitted] calls

and is now asking” why “those two calls are not in the transcript binders which

they now have.” App’x 404. In proposing responses to the note, James asked the

District Court to conduct an inquiry into what material the jurors had reviewed

and determine whether, if any of that material had not been admitted, they could

9 nevertheless disregard it. The District Court rejected James’s proposal and

instead eventually responded by instructing the jury as follows:

I am telling you now that transcripts one and two, which I think you referred to, . . . are not in evidence. If anyone thinks they have seen something, you are mistaken. They are not in evidence. Please continue your deliberations. This is the court’s ruling. You will treat this statement as you would any other item of evidence that I have told you is not in evidence, is not for your consideration.

App’x 409–10.

Less than an hour later, the jury convicted James of health care fraud,

conspiracy to commit health care fraud, wire fraud, and aggravated identity

theft, while acquitting him on the money laundering conspiracy charge. The

District Court later denied James’s motion for a judgment of acquittal. See Fed.

R. Crim. P. 29.

III

James’s sentencing took place over the course of two days, January 23 and

February 2, 2024.

The first day of sentencing was, to say the least, confused and confusing.

The District Court calculated a Guidelines sentencing range of 168 to 210 months.

It found that James’s gains from the fraud totaled $63 million, which represented

James’s total business revenues. A $30 million reduction in the gain calculation

10 might be in order, the District Court acknowledged, because James’s total

business revenues included “funds that were legitimately put together.” App’x

611–12. The court then also applied, over James’s objections, a four-level

enhancement under U.S.S.G. § 3B1.1 for James’s role as an “organizer or leader”

of the fraudulent scheme, and a two-level enhancement under U.S.S.G. § 3B1.3

for abuse of trust. In imposing the leadership enhancement, the District Court

merely stated: “Your objections are noted, as well as the Government’s objections

. . . . I’m not going to review and repeat those. . . . I don’t think there’s anything

else at this point other than to go into 3553(a) factors.” App’x 614–15.

The District Court initially suggested that it would impose “a sentence of

ten years.” App’x 691. But the Government interrupted, advising that James’s

offense of conviction and absence of criminal history rendered him eligible for

First Step Act credits that “could effectively halve the court’s sentence.” App’x

693. In response, the District Court suggested that it would then impose a

longer, twelve-year sentence and later candidly acknowledged that it had

“intend[ed] to give [James] ten years, but that was based on the court’s

misunderstanding of the ability of the defendant to engage in programs and

potentially halve his sentence. That’s why I changed my initial determination.”

11 App’x 702–03. Defense counsel objected, imploring the District Court not to

“tak[e] into account potential programs that [James] may be eligible for,” App’x

692, and insisted that a twelve-year prison term would actually render James

ineligible to participate in certain prison programs.

Leaving the term of imprisonment unresolved, the District Court then

turned to forfeiture and restitution. Although the District Court had earlier

contemplated a $30 million reduction in the gain calculation based on its finding

“that not all of the” submitted insurance claims “were false” and that “some . . .

were legitimate,” App’x 590, it later retreated from that reduction and ordered

James to forfeit approximately $63 million, as further described below.

As for restitution, the Government initially proposed that James pay

$342,822,962.93, representing the total amount paid out by insurance companies

based on billed claims in which James either impersonated patients or used

Modifier 59. In support, the Government supplied two spreadsheets that simply

added up the amounts paid out on all claims involving either impersonation or

the use of Modifier 59. James objected, insisting that the Government’s

methodology failed to account for properly-billed claims (such as legitimate uses

of Modifier 59), the actual impact of the alleged fraud on the payouts (where, for

12 example, the fraud only partially increased the amount of a payout), and

payments that doctors had returned to the victim insurance companies. James

also contended that the Government had double-counted some claims involving

both impersonation and Modifier 59.

Given the remaining uncertainty as to James’s eligibility for First Step Act

credits, as well as the appropriate forfeiture and restitution figures, the District

Court adjourned sentencing and solicited further briefing from the parties.

On the second day of sentencing (February 2), James pushed for a ten-year

sentence on the ground that the District Court could not rely on the possibility of

credits or rehabilitation to impose a longer sentence. The Government countered

with a proposed twenty-five-year sentence, doubling down on its view that the

District Court should consider the probability that James’s sentence would

effectively be cut in half through a combination of good conduct time, earned

time credit under the First Step Act, and reduced time from successful

completion of the Residential Drug Abuse Program (“RDAP”). During the

hearing, the District Court confirmed that it would impose a twelve-year (rather

than ten-year) sentence but denied that the longer sentence “entirely” reflected

James’s eligibility for credits under the First Step Act. App’x 719. The longer

13 sentence, the court said, better reflected “the enormity and the difficulty . . . this

fraud imposes not only on the insurance companies, but on self-funded plans,”

as well as James’s “horrible” and “abusive” impersonation calls. App’x 719, 734.

The District Court also explained that it had considered “a variety of reductions”

available to James in prison, App’x 734, including benefits from spending more

“time on RDAP in addition to the year that comes off. And there is also

placement and a home detention and a residential center. I mean, there are a

variety of ways that you may come up with half the sentence.” App’x 724.

The District Court then turned to forfeiture. Pointing to evidence that he

personally earned roughly only ten percent of the proceeds from fraudulent

billings, James maintained that the Government’s restitution calculation of

approximately $336 million yielded a forfeiture amount of approximately $34

million. The District Court responded: “There’s gain and there’s loss, okay. . . .

The loss was much higher, and that’s why I’m doing what I’m doing here.”

App’x 752. In imposing its $63 million forfeiture order, the court found that the

“gain to [James] was a minimum of [$]63 million,” not $34 million, App’x 752,

because “the Government put in the work in their submission . . . to demonstrate

that this fraud was much higher than they’re asking for,” App’x 747.

14 James fared no better with the District Court’s resolution of the disputed

restitution amount. Between the first and second days of sentencing, the

Government had reduced its restitution calculation by about $6 million—to

$336,996,416.85—to correct the double-counting James had previously identified.

At the second hearing, the Government described the $336 million figure as “a

conservative figure” that “doesn’t encompass the entirety of the fraud that the

defendant committed.” App’x 752. Over James’s objection, the District Court

simply adopted the Government's position without further explanation.

This appeal followed.

DISCUSSION

Challenging his conviction on appeal, James argues that the jury’s

exposure to transcripts of unadmitted phone calls was highly damaging and that

the District Court’s refusal to inquire about the effects of the extra-record

information was error. “It is well-settled that any extra-record information of

which a juror becomes aware is presumed prejudicial.” United States v. Greer, 285

F.3d 158, 173 (2d Cir. 2002). To overcome this presumption, the Government

must show that “the information is harmless.” Id. With that in mind, we review

15 the District Court’s handling of jury exposure to extra-record material for abuse

of discretion, see United States v. Farhane, 634 F.3d 127, 168 (2d Cir. 2011),

understanding that “the trial judge’s conclusions regarding the effect of the

extra-record evidence on the jury are entitled to substantial weight,” United States

v. Nkansah, 699 F.3d 743, 751 (2d Cir. 2012) (quotation marks omitted), abrogated

on other grounds by United States v. Bouchard, 828 F.3d 116 (2d Cir. 2016).

Recall that early in James’s trial, the Government provided jurors with

binders containing transcripts of recorded phone calls that it might offer into

evidence. During deliberations, the jury sent a note asking: “Has there been any

materials or transcripts removed from the call transcript binders, such as call

ex[]hibits 1, 2 and 3, or any others? Specif[]ically, we are looking for a call

between Mat[hew James] and an employee.” App’x 509. The note apparently

referred to exhibits that one or more jurors may have seen in the binders during

the trial before the non-admitted exhibits were removed.

Although the jury note referred to three exhibits, on appeal the parties

focus, as we do, on Government Exhibits 1 and 2 (“GX1” and “GX2”). 1 We

1 We are persuaded that the jury was not exposed to Government Exhibit 3 (“GX3”). The Government represented that GX3 was never in the jury’s binders because it was subject to objection, unlike GX1 and GX2. The jury’s note specifically requested “a call

16 conclude that any exposure to these transcripts was harmless. The “touchstone”

of our analysis is “not the mere fact of infiltration of some molecules of extra-

record matter,” but rather “the nature of what has been infiltrated and the

probability of prejudice.” United States ex rel. Owen v. McMann, 435 F.2d 813, 818

(2d Cir. 1970) (Friendly, J.). Here, the contents of GX1 and GX2 largely

duplicated evidence that was properly admitted, and portions of the calls align

with rather than contradict James’s defense.

GX1 and GX2 reflect calls recorded by one of James’s employees, Jennifer

Flanagan, concerning her discussions with the FBI following a search of James’s

office. In the calls, James repeatedly told Flanagan to “t[ell] the truth” and

insisted that she had “nothing to worry about” because he ran a “legitimate

business.” App’x 426–27, 429, 431–32, 435, 442. While James admitted that “one

thing [he] did wrong was impersonating patients,” App’x 441, that admission

merely duplicated other trial evidence, including and especially James’s

stipulation at trial that he had impersonated patients, see Gov’t App’x 568–72.

And if there were any remaining doubt, at trial James’s employees (including

between Mat and an employee,” a description that fits GX1 and GX2 but not GX3. And defense counsel accepted the Government’s explanation that GX3 was never in the binder. We therefore have no reason to doubt that the jury was not exposed to GX3. 17 Flanagan) and patients themselves confirmed that James routinely impersonated

patients and their relatives. That fact was thus never really disputed, and the

jury heard essentially the same undisputed fact through multiple sources.

Considering “the nature of what has been infiltrated and the probability of

prejudice,” McMann, 435 F.2d at 818, we conclude that any exposure to GX1 and

GX2 was harmless and did not deprive James of a fair trial. The District Court

therefore did not abuse its “wide discretion in deciding how to pursue an inquiry

into the effects of extra-record information.” Nkansah, 699 F.3d at 751 (quotation

marks omitted); see United States v. Weiss, 752 F.2d 777, 783 n.2 (2d Cir. 1985).

We now turn to the more troublesome issue of James’s sentence. The

District Court added two years to his term of imprisonment after learning about

his potential eligibility for early release under the First Step Act and through

RDAP. That was error.

We first address the First Step Act, under which eligible prisoners can

reduce their sentence by earning time credits for “successful participation in

evidence-based recidivism reduction programming or productive activities.” 18

18 U.S.C. § 3632(d)(4)(A)(i)–(ii); see also Giovinco v. Pullen, 118 F.4th 527, 528 (2d Cir.

2024), cert. denied sub nom. Giovinco v. Flowers, 145 S. Ct. 1947 (2025). An

individual’s “[s]uccessful participation” in these programs “requires a

determination by Bureau [of Prisons (“BOP”)] staff that an eligible inmate has

participated in” the relevant programming “and has complied with [its]

requirements.” 28 C.F.R. § 523.41(c)(2). An individual “will generally not be

considered to be ‘successfully participating’ in” qualifying programs in certain

“situations,” including if they have been placed in “a Special Housing Unit” or

sent “for extended medical placement in a hospital or outside institution.” Id.

§ 523.41(c)(4)(i)–(ii).

At James’s first sentencing hearing, after the District Court announced a

twelve-year sentence, it explained that it had “intend[ed] to give [James] ten

years, but that was based on the court’s misunderstanding of the ability of the

defendant to engage in programs” available under the First Step Act “and

potentially halve his sentence. That’s why I changed my initial determination.”

App’x 702–03. The District Court tried to walk back this remark at James’s

second sentencing hearing by explaining that its decision to impose the same

twelve-year sentence it had originally proposed was “not based solely on the

19 defendant’s being eligible for the First Step Act.” App’x 734. At the very least,

though, the record reflects that the District Court considered the potential for

earned time credits under the First Step Act as a stand-alone reason, apart from

the factors included in 18 U.S.C. § 3553(a), to lengthen James’s sentence. To the

extent the District Court did so, it erred.

This issue appears to be one of first impression among the Courts of

Appeals, and we base our conclusion on the plain text of 18 U.S.C. § 3553(a) and

the purpose of the First Step Act. 2 Section 3553(a) directs a district court to

consider certain factors “in determining the particular sentence to be imposed.”

(emphasis added). As we have explained in another context, “[i]t does not ask

the district court to determine whether any term of imprisonment that the

defendant may ultimately serve fulfills those purposes.” United States v. Kimbell,

2 A recent decision of the Tenth Circuit considered a district court’s comments about earned time credit at sentencing but did not address the specific question at issue here. See United States v. Nickols, No. 24-5056, 2025 WL 1625407, at *3 (10th Cir. June 9, 2025). There the circuit affirmed a life sentence where the district court, in denying the defendant’s motion for a downward variance from the Guidelines recommendation of life, mistakenly commented that if it granted the motion the defendant would be eligible for earned time credits despite having been convicted of an offense that disqualified him from receiving such credits. Id. at *3. This was found to be harmless error, however, as the “comment did not affect [d]efendant’s advisory sentence under the Sentencing Guidelines nor the court’s reasoning in imposing a life sentence.” Id. at *1. 20 No. 21-288, 2021 WL 5441249, at *3 (2d Cir. Nov. 22, 2021) (summary order). A

district court bypasses that statutory imperative when it extends a sentence to

account for how long it anticipates a term of imprisonment will actually last.

Our reading of § 3553(a) comports with the purpose of these credits under

the First Step Act. Relying on eligibility for earned time credits as a stand-alone

factor at sentencing disrupts the balancing Congress dictated in § 3553(a) by

“expand[ing] relevant sentencing considerations beyond those enumerated.”

United States v. Park, 758 F.3d 193, 198 (2d Cir. 2014). In Park, we found a

sentence to be procedurally unreasonable where the district court based its

decision to “impos[e] a term of probation rather than incarceration” solely on the

“cost of incarceration to the government,” because cost “is not a relevant

sentencing factor under” § 3553(a). Id. The same is true with the potential for

earned time credits. In providing for time credits, Congress clearly aimed to

offer “an incentive for prisoners to attend the recidivism reduction programs

Congress was devising, and the obvious incentive was that the time credits

would reduce a prisoner’s incarceration time.” Guerriero v. Miami RRM, 2024 WL

2017730, at *3 (11th Cir. May 7, 2024). Permitting a district court to increase a

defendant’s sentence at the front end to account for potential reductions under

21 the First Step Act would contravene Congress’s intent to allow a prisoner to earn

a reduction in the time to be served under the sentence that the District Court

already found proper based on the § 3553(a) factors.

Our holding also accounts for the fact that, under the First Step Act, a

prisoner is not certain to receive earned time credit. The decision to award

earned time credit is discretionary and, as noted above, “[s]uccessful

completion” of qualifying programs is “determine[ed] by [BOP] staff.” 28 C.F.R.

§ 523.41(c)(2); see also Noe v. Ciolli, 2025 WL 1662659, at *2 (10th Cir. June 12, 2025)

(rejecting argument that plaintiff had a “liberty interest in his right to earn

credits” under the First Step Act, because “[e]ven if he were to participate in

those programs, . . . prison authorities would need to assess [defendant’s

performance]” before awarding credits) (citing 28 C.F.R. § 523.41(c)(2)); Sedlacek

v. Rardin, No. 24-1254, 2025 WL 948485, at *1 (6th Cir. Jan. 21, 2025) (recognizing

earned time credits as “conditional”). Credits cannot “generally be” awarded

when a prisoner faces certain circumstances, such as placement in an outside

“hospital” or a “mental health/psychiatric hold[].” 28 C.F.R. § 523.41(c)(4)(ii),

(iv). Prisoners may find themselves in one of these situations for reasons that are

unexpected and outside their control. A district court’s decision to extend a

22 sentence based on potential earned time credits may therefore rest on an

expectation of results that never come to pass.

The Government contends that it was appropriate for the District Court to

consider the potential for earned time credits because James’s actual served

sentence was relevant to its determination of how to achieve the goals of

sentencing under § 3553(a). Appellee’s Br. 46. But the District Court does not

appear to have meaningfully tied its consideration of earned time credit to the

§ 3553(a) factors. The Fourth Circuit’s decision in United States v. Fowler, 948 F.3d

663 (4th Cir. 2020), which involved the similar context of good-time credits, is

instructive insofar as it suggests when it might be permissible to consider credits

at sentencing. In Fowler, the court affirmed a sentence where the district court

“did not consider good-time credits as a stand-alone factor during sentencing,

nor did [it] utilize them as any sort of vehicle for an enhanced sentence term.”

948 F.3d at 669; see also United States v. Roberts, 919 F.3d 980, 992 (6th Cir. 2019)

(suggesting that a court may not “consider good-time credit as a stand-alone

factor in fashioning the length of a sentence”). Instead, the circuit explained, the

district court had “only considered the potential impact of good-time credits in

relation to Fowler’s age at release,” which was relevant to particular § 3553(a)

23 factors. Fowler, 948 F.3d at 669–70. The Fourth Circuit observed that the district

court validly considered “age at release” because it “correlated with recidivism

rates, a necessary consideration for a judge charged with setting a sentence that

protects the public and deters additional criminal acts.” Id. (citing 18 U.S.C.

§ 3553(a)(2)(B)-(C)). The court in Fowler also understood that the district court

“wished to encourage Fowler’s rehabilitation by avoiding a life-equivalent

sentence.” Id. (citing 18 U.S.C. § 3553(a)(2)(D)). That the District Court failed to

make any comparable findings in the present case, by contrast, strongly suggests

that it used the First Step Act as a “sort of vehicle for an enhanced sentence

term.” Id. at 669.

We therefore conclude that the District Court erred in relying on the

potential availability of earned time credits under the First Step Act as a stand-

alone factor to lengthen James’s sentence.

RDAP, which typically runs for nine months, “is an intensive drug

treatment program” run by the BOP “for federal inmates with documented

substance abuse problems.” See Reeb v. Thomas, 636 F.3d 1224, 1225 (9th Cir.

2011). The BOP may reduce the sentence of inmates who successfully complete

24 the program by up to a year. See id. at 1226; 18 U.S.C. § 3621(e)(2)(B). The BOP

has sole discretion to determine an inmate’s eligibility for RDAP, 18 U.S.C.

§ 3621(e)(5)(B), as well as to grant or deny sentence reductions upon successful

completion of the program, id. § 3621(e)(2)(B); see also Reeb, 636 F.3d at 1226.

18 U.S.C. § 3582(a) directs courts to “recogniz[e] that imprisonment is not

an appropriate means of promoting correction and rehabilitation” when

determining whether to impose imprisonment and, if so, for how long. Tapia,

564 U.S. at 327 (quotation marks omitted). In Tapia v. United States, the Supreme

Court held that while sentencing courts may discuss rehabilitation and the

benefits of specific treatment programs, they “may not impose or lengthen a

prison sentence to enable an offender to complete a treatment program or

otherwise to promote rehabilitation.” Id. at 335; see id. at 321; United States v.

Gilliard, 671 F.3d 255, 258 (2d Cir. 2012). “Congress did not intend that courts

consider offenders’ rehabilitative needs when imposing prison sentences.” Tapia,

564 U.S at 331.

As the record reflects (and the Government concedes), at James’s second

sentencing hearing, the District Court considered the effects of RDAP in

imposing James’s sentence. As our discussion of the First Step Act indicates, we

25 cannot confidently tell how significant a role RDAP or any other factor newly

articulated at the second sentencing hearing played in justifying the same

sentence the District Court had already announced. But we are persuaded the

District Court erred by considering RDAP at all. True, this appeal presents a

somewhat novel application of Tapia. Instead of lengthening James’s sentence to

ensure his rehabilitation, the District Court appears to have imposed a longer

sentence to counteract the potential reduction in prison time that might result

from his efforts to rehabilitate. But this twist does not warrant a different

outcome.

Among other errors, the District Court assumed that James would qualify

for, participate in, and successfully complete RDAP. But the court could not

ensure James’s participation, since the BOP alone determines eligibility for and

admission to the program. See Tapia, 564 U.S. at 330–32; 18 U.S.C. § 3621(e)(5)(B).

In addition, the District Court’s approach in effect penalized James for disclosing

his substance abuse problems prior to sentencing. If approved, that approach

would discourage defendants from disclosing their substance abuse problems in

26 the first place, for fear that doing so might lengthen their sentence. That result is,

of course, the opposite of what Congress intended.

The District Court’s reasoning contravenes congressional policy judgments

more generally. Congress determined that certain offenders should be eligible

for reduced sentences after completing a substance abuse treatment program.

See 18 U.S.C. § 3621(e)(2)(B). Congress also directed courts to “recogniz[e] that

imprisonment is not an appropriate means of promoting correction and

rehabilitation.” See id. § 3582(a); Tapia, 564 U.S. at 326–32. It follows that issues

relating to the defendant’s rehabilitation should not affect the term of

imprisonment for any reason. By increasing James’s sentence to counteract the

benefits of participating in RDAP, the District Court effectively also countered a

clear congressional mandate.

The Government responds by pointing us to 18 U.S.C. § 3661, which

provides that “[n]o limitation shall be placed on the information concerning the

background, character, and conduct of a person convicted of an offense which a

court of the United States may receive and consider for the purpose of imposing

an appropriate sentence.” Section 3661, the Government argues, permitted the

District Court to consider the possibility that James would obtain credit from

27 participating in RDAP as a factor in “imposing an appropriate sentence.” We

disagree. The broad sentencing discretion that § 3661 affords district courts does

not displace § 3582(a)’s specific statutory bar against using rehabilitation to

increase a term of imprisonment even in part. A defendant’s potential future

participation in a rehabilitation program is in any event not meaningfully related

to their “background, character, and conduct” for purposes of determining an

appropriate sentence. See United States v. Cozad, 21 F.4th 1259, 1263 (10th Cir.

2022) (observing that “a district court does not enjoy boundless discretion with

respect to the facts it relies on at sentencing” and that “a factor may be

impermissible because its consideration is prohibited by statute,” and citing

§ 3582 and its prohibition against “considering the defendant’s need for

rehabilitation”); 18 U.S.C. § 3661 (affording sentencing courts broad discretion to

consider “information concerning the background, character, and conduct” of a

convicted defendant (emphasis added)).

C

To summarize, we hold that district courts may not rely on the potential

for earned time credit under the First Step Act as a stand-alone factor to enhance

a defendant’s sentence. We also hold that a district court cannot lengthen a

28 defendant’s sentence based on the BOP’s possible reduction of the sentence for

participating in a rehabilitation program. We therefore vacate James’s sentence

and remand for resentencing consistent with this opinion.

James separately urges that the District Court erred in applying a four-

level Guidelines enhancement for his role as an “organizer or leader” under

U.S.S.G. § 3B1.1(a) and a two-level enhancement for abuse of a position of trust

under U.S.S.G. § 3B1.3. We review the application of the Guidelines de novo and

the factual determinations underlying Guidelines calculations for clear error.

United States v. Cramer, 777 F.3d 597, 601 (2d Cir. 2015). On this record, we agree

that the District Court should also reconsider both of these enhancements on

remand for resentencing.

“In enhancing a defendant’s sentence based on his role in the offense, a

district court must make specific factual findings as to that role. . . . Although

this requirement of making specific factual findings may interfere with the

smooth operation of the sentencing hearing, we require specific factual findings

to permit meaningful appellate review.” United States v. Carter, 489 F.3d 528, 538

29 (2d Cir. 2007) (quotation marks omitted); see 18 U.S.C. § 3553(a)(4), (c). The

district court may satisfy its “obligation to make the requisite specific factual

findings” by explicitly adopting in open court “the factual findings set forth in

the presentence report [(‘PSR’)].” Carter, 489 F.3d at 539.

Unfortunately, the District Court applied the four-level enhancement for

James’s leadership role without making any findings in open court about either

James’s actual role in the offense or the number of participants involved in the

criminal activity. 3 And while the District Court later did explicitly adopt the

findings in James’s PSR, it again failed to do so in open court. The District Court

opted instead to refer to the PSR in a sealed statement of reasons issued nearly

two months after sentencing. As we held in Carter, adopting a PSR “only in a

written judgment” that “was not available to the public” constitutes error. 489

3 Section 3B1.1(a) provides for a four-level enhancement only “[i]f the defendant was an organizer or leader of a criminal activity that involved five or more participants or was otherwise extensive.” U.S.S.G. § 3B1.1(a). Application Note 4 to § 3B1.1 lists the “[f]actors the court should consider” in evaluating a defendant's role in the offense, including “the exercise of decision-making authority, the nature of participation in the commission of the offense, the recruitment of accomplices, the claimed right to a larger share of the fruits of the crime, the degree of participation in planning or organizing the offense, the nature and scope of the illegal activity, and the degree of control and authority exercised over others.” U.S.S.G. § 3B1.1 cmt. n.4; see Carter, 489 F.3d at 539. A “participant,” meanwhile, is “a person who is criminally responsible for the commission of the offense, but need not have been convicted.” U.S.S.G. § 3B1.1 cmt. n.1.

30 F.3d at 539–40. This is because a central purpose of § 3553(c) is “to enable the

public to learn why [a] defendant received a particular sentence.” Id. at 539

(quotation marks omitted). For these reasons, we conclude that the District

Court failed to satisfy the open court requirement of § 3553(c).

We also note that the PSR offers nothing more than conclusory

descriptions of James’s role and fails to specify the five “criminally responsible”

participants. See United States v. Ware, 577 F.3d 442, 453–54 (2d Cir. 2009). The

PSR thus “provide[s] an insufficient basis for [James] or this Court to determine

why the [D]istrict [C]ourt did what it did.” Carter, 489 F.3d at 540 (quotation

marks omitted).

As noted, James also challenges the District Court’s application of the

abuse-of-trust Guideline, § 3B1.3, which provides for a two-level enhancement

“[i]f the defendant abused a position of public or private trust . . . in a manner

that significantly facilitated” the offense. U.S.S.G. § 3B1.3. 4 The enhancement

applies “only where the defendant has abused discretionary authority entrusted

4 U.S.S.G. § 3B1.3 also applies where a defendant makes “use[] [of] a special skill.” But the Government does not contend that the enhancement here was or could be based on that provision. 31 to [him] by the victim,” United States v. Jolly, 102 F.3d 46, 48 (2d Cir. 1996); see

United States v. Broderson, 67 F.3d 452, 456 (2d Cir. 1995), or “abused a position of

fiduciary or quasi-fiduciary status,” United States v. Huggins, 844 F.3d 118, 124 (2d

Cir. 2016). Examples of positions of trust include “an attorney serving as a

guardian, a bank executive’s fraudulent loan scheme, or the criminal sexual

abuse of a patient by a physician under the guise of an examination.” U.S.S.G.

§ 3B1.3 cmt. n.1.

The District Court determined that the abuse-of-trust enhancement

applied to James’s offense because he “had individuals’ personal information in

terms of their health, their status, all sorts of things.” App’x 619. This apparently

referred to individual patient information. But James had no direct

relationship—contractual, fiduciary, or quasi-fiduciary—with the victimized

patients, let alone the insurance companies in this case.

Of course, that is not the end of our analysis. The commentary to § 3B1.3

suggests that the enhancement may apply to a defendant who misuses personal

information even in the absence of a relationship of trust. Application Note 2(B)

states that “[n]otwithstanding . . . any other provision of this [G]uideline, an

adjustment under this [G]uideline shall apply to . . . [a] defendant who exceeds

32 or abuses the authority of his or her position in order to . . . use without

authority[] any means of identification.”

At least one circuit has relied on this comment to apply the enhancement

in a similar circumstance. In United States v. Abdelshafi, the Fourth Circuit held

that the enhancement was properly applied to a defendant who fraudulently

billed for “transportation services” that he provided to Medicaid patients as a

“third-party vendor” for an HMO that had contracted with Virginia’s Medicaid

agency. 592 F.3d 602, 605, 611 (4th Cir. 2010). The court described “Abdelshafi’s

arguments in regard to his lack of discretion and [lack of] fiduciary relationship

with Medicaid [as] beside the point in the face of the plain reading of application

note 2.” Id. at 611. That was the case “regardless of whether [the court]

consider[ed] the victim in this case to be Medicaid, [the HMO], the patients

whose identifying information Abdelshafi misused, or all of them.” Id.

In this case, we need not decide whether that commentary permits a

finding of “abuse of trust.” Even if it did, its application would constitute

impermissible double counting, which “occurs when one part of the [G]uidelines

is applied to increase a defendant’s sentence to reflect the kind of harm that has

already been fully accounted for by another part of the [G]uidelines.” United

33 States v. Watkins, 667 F.3d 254, 261 (2d Cir. 2012) (quotation marks omitted).

Here, the District Court already imposed a mandatory two-year sentence for

James’s aggravated identity theft convictions. See 18 U.S.C. § 1028A. “To avoid

enhancing the defendant’s sentence twice for the same offense conduct—once

under the guideline for the predicate offense and again under § 1028A—the

Sentencing Commission has directed judges not to apply any specific offense

characteristic for the transfer, possession, or use of a ‘means of identification.’”

United States v. Xiao Yong Zheng, 762 F.3d 605, 606 (7th Cir. 2014) (citing U.S.S.G.

§ 2B1.6 cmt. n.2). For James’s sentence, the only permissible application of the

abuse-of-trust enhancement would punish the possession and use of

identification information. Applying the abuse-of-trust enhancement based on

James’s misuse of patient identification information would increase his sentence

for “the kind of harm that has already been fully accounted for by another part of

the [G]uidelines,” constituting impermissible double-counting. Watkins, 667 F.3d

at 261.

Our conclusion does not foreclose the possibility, although we see no

evidence, and the Government points to none, that James’s status as a third-party

medical biller spawned a “fiduciary relationship with [his clients’] patients” that

34 would independently justify application of the enhancement. United States v.

Ntshona, 156 F.3d 318, 321 (2d Cir. 1998). Were that the case, the enhancement

would not punish James’s misappropriation of patients’ personal information

but rather his abuse of their trust. See United States v. Thorn, 317 F.3d 107, 122 (2d

Cir. 2003) (vacating application of § 3B1.3 based on “use of a special skill” but

remanding to consider whether defendant was “in a position with discretionary

authority” such that the enhancement should apply).

After increasing James’s offense level by six levels due to the leadership

role and abuse-of-trust enhancements, the District Court calculated a Guidelines

range of 168 to 210 months. We have recognized “the powerful anchoring effect

of a miscalculated Guidelines range on a district court’s thinking about the

appropriate sentence.” United States v. Seabrook, 968 F.3d 224, 234 (2d Cir. 2020)

(cleaned up). Because it “is certainly not clear from this record” that the District

Court’s calculation of the Guidelines “had no influence” on the below-Guidelines

144-month sentence it ultimately imposed, id. (quotation marks omitted), we

35 cannot say with confidence that the District Court would have imposed the same

sentence absent these enhancements.

Accordingly, on remand for resentencing, the District Court should

recalculate James’s Guidelines range. If the District Court concludes that the

leadership enhancement or abuse-of-trust enhancement should be imposed, it

must make specific factual findings in open court to support each enhancement

consistent with this opinion. Among other things, if it determines that an

enhancement under § 3B1.1(a) or (b) is warranted, the District Court should

specifically identify the five or more participants involved in the offense. 5

IV

We next turn to forfeiture. On appeal from a forfeiture order, we review

“the district court’s legal conclusions de novo and its factual findings for clear

error.” United States v. Treacy, 639 F.3d 32, 47 (2d Cir. 2011). “We must

determine whether the trial court’s method of calculation was legally acceptable,

but we will not disturb a district court’s reasonable estimate of the amount, given

5The parties have not briefed on appeal whether an enhancement could be justified without a finding of five or more participants under the “or otherwise extensive” clause of U.S.S.G. § 3B1.1. See U.S.S.G. § 3B1.1 cmt. n. 3. We therefore express no view on whether that portion of the Guidelines may be relevant here. 36 the available information.” United States v. Walters, 910 F.3d 11, 31 (2d Cir. 2018)

(cleaned up).

On de novo review, we start with the basic principle that “forfeiture is gain

based,” United States v. Torres, 703 F.3d 194, 203 (2d Cir. 2012) (quotation marks

omitted), and that, as relevant here, only “gross proceeds traceable to the

commission of” a federal health care offense may be forfeited, 18 U.S.C.

§ 982(a)(7). “Congress’s intent [is] that forfeiture proceedings be used to recover

all of the [defendant’s] ill-gotten gains but not to seize legitimately acquired

property.” United States v. Daugerdas, 892 F.3d 545, 548 (2d Cir. 2018) (quotation

marks omitted) (emphasis added).

In its pre-sentencing submission, the Government argued that James

should forfeit $63 million—a figure that it acknowledged represented James’s

total business revenues. James asserted that the forfeiture amount should be

limited to the portion of his revenues actually attributable to his fraud and

submitted an expert report concluding that just under twenty percent of his

revenue—or about $12 million—represented proceeds from fraud. With

virtually no explanation, the District Court eventually determined that the “gain

37 to [James] was a minimum of [$]63 million” and accordingly ordered forfeiture

in that amount. App’x 752.

So far as we can tell from the record, the District Court’s forfeiture

calculation reflected James’s legitimately earned revenue as well as gross

proceeds traceable to the fraud. The District Court erred in ordering James to

forfeit a much higher amount that included legitimate revenue not traceable to

his fraud. Of course, courts “may use general points of reference as a starting

point for a forfeiture calculation and make reasonable extrapolations.” United

States v. Roberts, 660 F.3d 149, 166 (2d Cir. 2011) (quotation marks omitted). But

the District Court provided little to no other permissible explanation to justify its

$63 million forfeiture figure, even though it had initially found that “not all of

the claims [submitted by James] were false,” App’x 590, and the Government had

conceded that $63 million represented James’s total business revenues. Nor did

the District Court attempt either to calculate the percentage of submitted claims

tainted by fraud or to determine whether partially fraudulent claims were subject

to forfeiture. This lack of explanation makes it difficult for our Court to

38 “determine whether the trial court’s method of calculation was legally

acceptable.” Walters, 910 F.3d at 31.

The Government responds that all proceeds of James’s medical-billing

business are forfeitable because the business was “permeated with fraud.”

United States v. Warshak, 631 F.3d 266, 332 (6th Cir. 2010) (quotation marks

omitted); see United States v. Bikundi, 926 F.3d 761, 793 (D.C. Cir. 2019). But the

District Court did not find that James’s business was permeated with fraud. To

the contrary, evidence at trial showed that James earned revenue derived from

completely legitimate claims backed by accurate coding. See App’x 254–56, 788–

92. The FBI investigating agent testified, for example, that the improper codes

yielded merely a “subset” of James’s total billing, App’x 240, while a defense

expert attributed only twenty percent or so of James’s revenues associated with

the patients featured at trial to improper coding.

Any forfeiture order must be tailored to “ill-gotten gains.” Daugerdas, 892

F.3d at 548 (quotation marks omitted). Without an adequate justification for the

forfeiture amount and in light of the evidence justifying a much lower amount of

gain traceable to James’s fraud, we have little choice but to vacate the forfeiture

order. We remand with instructions for the District Court to recalculate the

39 amount of forfeiture, explain its methodology, make specific findings justifying

the final amount, and exclude from its calculation all revenue derived from

James’s legitimate claims. 6

V

Finally, James challenges the District Court's restitution order of

$336,996,416.85. The Government proposed that amount to account for all

insurance claims paid out in cases involving impersonation or James’s use of

Modifier 59. Because the District Court employed a flawed methodology to

calculate restitution, we vacate the restitution order and remand for a

recalculation.

The Mandatory Victims Restitution Act (MVRA) provides for mandatory

restitution to victims of conspiracy to commit health care fraud, health care

fraud, and wire fraud. See 18 U.S.C. § 3663A(c). “We review an MVRA order of

6 We do not mean to imply that the Government must conduct a complete audit of every claim James submitted so long as it employs some “reasonable means” in making the determination. It may be appropriate for the Government to develop a statistical sampling method that allows it to review a representative subset. Cf. Yorktown Med. Lab’y, Inc. v. Perales, 948 F.2d 84, 89–90 (2d Cir. 1991) (rejecting due process challenge to use of statistical sampling to calculate Medicaid overpayments). Any method the Government develops, however, must account for claims that consist of both legitimate and fraudulent billing. 40 restitution deferentially, and we will reverse only for abuse of discretion.”

United States v. Yalincak, 30 F.4th 115, 121 (2d Cir. 2022) (quotation marks

omitted). This deferential review acknowledges that “ordering restitution

requires a delicate balancing of diverse, sometimes incomparable, factors, some

of which not only lack certainty but may indeed be based on mere probabilities,

expectations, guesswork, even a ‘hunch.’” United States v. Rossi, 592 F.3d 372, 376

(2d Cir. 2010) (cleaned up).

Under the MVRA, “restitution may be awarded only in the amount of

losses directly and proximately caused by the defendant’s conduct.” United

States v. Gushlak, 728 F.3d 184, 194–95 (2d Cir. 2013) (citing 18 U.S.C.

§ 3663A(a)(2)). While a restitution amount need not be “mathematically precise”

and “a reasonable approximation will suffice, especially in cases in which an

exact dollar amount is inherently incalculable,” id. at 195–96 (quotation marks

omitted), the approximation must be “supported by a sound methodology,” id.

at 196. A “failure at least to approximate the amount of the loss caused by the

fraud without even considering other [legitimate] factors” that may be relevant

41 to calculating restitution “requires a remand to redetermine the amount of the

loss.” United States v. Rutkoske, 506 F.3d 170, 180 (2d Cir. 2007).

The District Court’s restitution order does not reflect a “reasonable

approximation of losses” directly and proximately caused by James’s fraud.

Gushlak, 728 F.3d at 196. The District Court’s methodology for calculating

restitution failed to separate legitimate from fraudulent revenue. It instead

included all payments associated with all impersonated patients, regardless of

when the impersonation occurred. 7 It likewise assumed that all payments

associated with Modifier 59 were directly or proximately caused by fraud, even

though some legitimate claims also used Modifier 59, including claims as to

7 For example, James impersonated one patient in connection with a claim submitted to United Healthcare in September 2018, but every United Healthcare payment that the Government includes in its spreadsheet supporting the District Court’s restitution amount for that patient predates the impersonation by nearly a year.

42 which Modifier 59 did not materially affect the insurance company’s payment

decision. 8

The Government urges us to ignore these methodological flaws. Its

restitution calculations (and the District Court’s) were conservative, it claims,

insofar as they excluded some high-dollar payments attributable to James’s

fraud, as well as all claims under $10,000. But we are not persuaded that the

payments included in the Government’s restitution calculations and

subsequently adopted by the District Court were directly and proximately

caused by fraudulent conduct. While these calculations need not be

“mathematically precise,” a reasonably approximated restitution amount must

still be “supported by a sound methodology.” Gushlak, 728 F.3d at 195–96

(quotation marks omitted). The Government’s basic methodology appears

marred by serious flaws that undermine our confidence in the causal connection

between James’s conduct and the District Court’s final restitution amount.

Accordingly, we vacate the restitution order and remand with instructions

to the District Court to redetermine a restitution amount that, consistent with this

8 Although James routinely instructed staff to use this modifier improperly, some uses were legitimate, and certain insurers sometimes ignored the modifier entirely for payment purposes. 43 opinion, reflects a “reasonable approximation” of losses “directly and

proximately caused” by James’s conduct. See id. at 195–96 (quotation marks

omitted). On remand, insofar as it relies solely on the impersonation of a patient,

the District Court should exclude payments made by insurance companies before

the impersonations associated with that patient occurred; distinguish between

legitimate and fraudulent uses of Modifier 59; and account for claims as to which

insurers may have ignored the use of Modifier 59 in their payment decisions.

VI

In addition to his claims of error, James also asks that we reassign this case

to a different judge on remand. That is not at all warranted where, as here, we

do not doubt “the judge’s fairness or the appearance of the judge’s fairness.”

United States v. Elfgeeh, 515 F.3d 100, 137 (2d Cir. 2008) (quotation marks omitted).

James’s request for reassignment is therefore denied.

CONCLUSION

For the foregoing reasons, we AFFIRM James’s judgment of conviction,

VACATE the District Court’s sentence, including the forfeiture and restitution

orders, and REMAND for resentencing consistent with this opinion.