USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 1 of 15

PUBLISHED

UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT

No. 22-6464

UNITED STATES OF AMERICA,

Petitioner - Appellee,

v.

NATHANIEL WILLIAMS,

Respondent - Appellant.

Appeal from the United States District Court for the Eastern District of North Carolina, at Raleigh. W. Earl Britt, Senior District Judge. (5:21-hc-02244-BR)

Argued: September 14, 2022 Decided: November 17, 2022

Before RICHARDSON and HEYTENS, Circuit Judges, and MOTZ, Senior Circuit Judge.

Vacated and remanded by published opinion. Judge Heytens wrote the opinion, in which Senior Judge Motz joined. Judge Richardson wrote a dissenting opinion.

ARGUED: Jennifer Claire Leisten, OFFICE OF THE FEDERAL PUBLIC DEFENDER, Raleigh, North Carolina, for Appellant. Genna Danelle Petre, OFFICE OF THE UNITED STATES ATTORNEY, Raleigh, North Carolina, for Appellee. ON BRIEF: G. Alan DuBois, Federal Public Defender, OFFICE OF THE FEDERAL PUBLIC DEFENDER, Raleigh, North Carolina, for Appellant. Michael F. Easley, Jr., United States Attorney, OFFICE OF THE UNITED STATES ATTORNEY, Raleigh, North Carolina, for Appellee. USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 2 of 15

TOBY HEYTENS, Circuit Judge:

Federal prisoners on the cusp of being released may be civilly committed if they are

“presently suffering from a mental disease or defect as a result of which [their] release

would create a substantial risk” to the person or property of others.

(d).

Here, the primary question is whether—in making such a risk assessment—a court must

consider any terms of supervision that would govern the prisoner’s conduct post release.

The answer, we hold, is yes. And because the record offers no assurances the district court

appropriately considered the terms of Nathaniel Williams’ supervised release before

ordering him committed, we vacate the court’s order and remand for further proceedings.

I.

A.

Williams has long struggled with mental illness and a proclivity to violent outbursts.

In 2017, Williams assaulted a security guard in Portland, Oregon—a federal crime because

it happened at a Social Security office. See

(a)(1). Williams pleaded guilty

and was sentenced to just over four years in prison, to be followed by three years of

supervised release. The sentencing court imposed 13 standard and 14 special conditions of

supervised release, including that Williams “must take all mental health medications that

are prescribed by [his] treating physician” and “must participate in a mental health

treatment program.” JA 128.

For several years, Williams did not fare well in federal custody. In March 2019, he

was transferred to a medical center in Minnesota after failing to take medication,

experiencing mania and hallucinations, and assaulting his cellmate. Even after the transfer,

2 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 3 of 15

Williams continued to rack up incident reports and failed to reliably take prescribed

medication. After more than a year of new incidents—and Williams’ formal request to

withdraw from voluntary mental health treatment—the government successfully petitioned

to have Williams hospitalized under

, which permits the involuntary

transfer of a federal prisoner who “is presently suffering from a mental disease or defect”

“to a suitable facility for care and treatment.” § 4245(a) & (d).

Having secured the involuntary hospitalization order, Bureau of Prisons officials

concluded Williams met the constitutional and regulatory criteria for involuntary

medication. See Washington v. Harper,

(1990);

(a)(5) &

(b). In June 2021, Williams began receiving monthly injections of Haldol, which

“psychiatrically stabilized” Williams and “remitt[ed]” his symptoms. JA 43, 47. Since

starting Haldol injections, Williams has engaged in no violent behavior.

Three months before Williams’ custodial sentence was set to expire, the Bureau of

Prisons transferred him to a specialized facility in North Carolina to assess whether he

could be safely released. The evaluating psychologist acknowledged that, at the time of

intake, Williams displayed no overt signs of a formal thought or mood disorder and his

medications had largely remitted his past symptoms. The evaluating psychologist, though,

concluded that releasing Williams would create a substantial risk of injury to people or

property because she thought Williams would stop taking his medication, which would, in

turn, render him dangerous to others.

3 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 4 of 15

B.

Less than two weeks before Williams’ scheduled release, the government asked a

district court in North Carolina (commitment court) to order him civilly committed. The

government’s request automatically stayed Williams’ release, see

(a), and

the commitment court held a hearing at which Williams, the evaluating psychologist, and

a psychiatrist selected by Williams testified. Williams also provided the judgment of the

sentencing court, which included the terms of his supervised release.

During the hearing, several points of common ground emerged. First, Williams

suffers from schizoaffective disorder, bipolar type, which is in partial or full remission.

Second, Williams has a history of seemingly indiscriminate acts of violence when

unmedicated. Third, while on his current medication—which he had been taking for more

than nine months on an involuntary basis at the time of the hearing—Williams poses no

substantial threat to others.

The issue thus came down to whether Williams was likely to continue taking his

medication if released. See JA 115 (commitment court stating, “I don’t think the

Government would dispute” that Williams is “not a danger as long as he complies with his

medication”). Williams acknowledged his past failures to take prescribed medication but

testified he enjoyed the effects of Haldol, wanted to keep taking it, and likely would be

reincarcerated if he failed to do so. By contrast, the evaluating psychologist concluded

Williams was unlikely to keep taking his medication if released, citing his earlier probation

violations, past inability to stick to medications, and lack of insight into the seriousness of

his mental illness.

4 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 5 of 15

After the parties presented their evidence, the commitment court asked whether it

could order Williams’ release under “conditions which would assure his lack of

dangerousness,” JA 116, and recessed the hearing so the parties could formulate positions

on the matter. When the hearing resumed, the parties agreed that a court considering

whether to halt a person’s release under Section 4246(d) may not impose new conditions

without first committing the person to the custody of the Attorney General.

The parties diverged, however, on whether the commitment court should consider

the terms of supervised release imposed by the sentencing court. Williams urged the

commitment court to do so, noting that his terms of supervised release included “numerous

conditions that directly speak to the [commitment court’s] concerns.” JA 136. Williams

particularly emphasized terms requiring him “to sign a waiver that allows the probation

officer to receive all medical information” and thus verify “if he took his [Haldol] shot.”

JA 136–37. By contrast, the government insisted that “[t]he supervised release conditions

should not impact [the commitment court’s] decision.” JA 149. Indeed, the government

asserted the commitment court should not even “consider” those conditions because “we

cannot base our determination on the actions of another court.”

At the end of the hearing, the commitment court orally granted the government’s

request to commit Williams to the custody of the Attorney General. Consistent with the

parties’ agreement, the court determined it lacked authority “to order a conditional release

of a person who has not been committed . . . under the statute.” JA 154. The commitment

court briefly referenced Williams’ arguments about the conditions of his supervised

release:

5 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 6 of 15

The supervised release you face under your criminal judgment is of a limited duration and is subject to the control of some other court that does not have both your best interest and the mental health of you and others like you and the safety of the public, and that may be a little harsh.

Even if the Court does have that thought in mind or that goal in mind, the Court’s authority to impose conditions on you is very limited. The Court is working there in the state of Oregon in the Federal United States District Court for the District of Oregon with a three-year term of supervised release. I don’t know how much of that has already expired. I don’t know any way at all that it can be extended.

JA 157.

Six days later, the commitment court issued a written order memorializing its

decision. The order reiterated Williams’ release would pose a danger only “if he is

unmedicated,” JA 162, and acknowledged the parties’ differing positions about the

relevance of Williams’ supervised release. See JA 161 (noting that Williams “argues that

the conditions of supervised release imposed in his criminal case, to which he would be

subject if he were not committed, mitigate any risk of danger,” whereas “[t]he government

contends that [Williams’] supervised release terms should not impact this court’s

determination”). The order did not comment on the persuasiveness of these arguments, nor

did it explain whether—or how—Williams’ supervised release factored into the

commitment court’s decision. We review the district court’s factual findings for clear error

and its legal analysis de novo. United States v. Bell,

(4th Cir. 2018).

II.

Repeating its position before the commitment court, the government at first insisted

on appeal that a commitment court need not—and perhaps even may not—consider “any

term of supervised release” imposed by a sentencing court. Gov’t Br. 34. Although the

6 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 7 of 15

government walked back its claim somewhat during oral argument, we reject that view in

all its forms.

We start, as always, with the statutory text. Section 4246(d) does not authorize the

involuntary commitment of any person “suffering from a mental disease or defect” or one

whose “mental disease or defect” could pose a danger to others under some circumstances.

(d). Instead, the statute mandates a predictive assessment, providing

commitment may be ordered only if, “after [a] hearing, the court finds by clear and

convincing evidence that” the person’s “release would create a substantial risk of bodily

injury to another person or serious damage to property of another.”

(emphasis added).

As the government elsewhere recognizes, this standard “requires the court to consider all

the evidence presented and determine whether the person’s release would endanger the

public.” Gov’t Br. 49. The statutory text does not permit—much less instruct—a

commitment court to pretend that the person will not be subject to terms of supervision

imposed by the sentencing court and conduct a thought experiment about whether the same

person would likely endanger others if not so constrained. Nor would such an approach

afford appropriate solicitude to the serious liberty interests imposed by continuing to detain

a person whose term of imprisonment has ended. See generally Addington v. Texas,

(1979).

Perhaps for this reason, this Court regularly considers the length and terms of a

person’s supervised release—and the person’s history of compliance with court-imposed

supervision—in reviewing a commitment analysis. See, e.g., United States v. Antone,

(4th Cir. 2014); United States v. Bolander,

(4th Cir.

7 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 8 of 15

2013); United States v. Francis,

(4th Cir. 2012). Doing so makes perfect

sense. Because a basic aim of a commitment analysis is to predict whether a person would

pose a substantial risk to others if released, any restrictions imposed by the sentencing

court—which are themselves designed to protect the public upon the person’s release,

see

, 3553—are inherently relevant.

That is not to say, of course, that anyone who would otherwise be on supervised

release may not be committed under Section 4246(d) or that a commitment court must

invariably recite every condition imposed by the sentencing court and explain why it cannot

protect public safety. In United States v. Bell,

(4th Cir. 2018), this Court

affirmed a commitment order under a different statute—18 U.S.C. § 4248—even though

the person at issue was already subject to a lifetime term of supervised release.

See

. In Bell, however, the commitment court specifically weighed the

defendant’s supervision, but found it insufficient to protect the public given a probation

officer’s testimony that the defendant was “the most impulsive person that she had ever

supervised during her 20-year career” and the defendant’s “non-compliant behavior

escalated throughout the course of each term of supervision.”

.

Nor do we establish any hard floor for a required explanation. Less exhaustive

treatment may be appropriate, for example, when any future supervision would be minimal

or not directed at a person’s primary risk factors, or when the terms of supervised release

are not a crucial component of one party’s arguments about the appropriateness of

commitment. See United States v. Wooden,

, 606–07 (4th Cir. 2018)

8 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 9 of 15

(rejecting notion that “the court must explain in detail why it rejects each and every

individual piece of evidence”).

But here the fact and nature of Williams’ supervised release was no ancillary matter.

Rather, it constituted Williams’ primary argument against commitment—an argument the

government insisted should be ignored and about which the commitment court said

precious little. Indeed, the commitment court said nothing about the terms of Williams’

supervised release, even those directly addressing the central issue in dispute: Would

Williams keep taking his medication following any release? See JA 128 (providing that

Williams “must participate in a mental health treatment program” and “take all mental

health medications that are prescribed by [his] treating physician”). Whether or not the

terms of supervision the sentencing court imposed were ultimately enough to ensure

Williams’ safe release, the commitment court was “required to at least consider the

evidence, and account for it, when concluding otherwise.” United States v. Wooden,

(4th Cir. 2012).

Indeed, the commitment court’s handful of references to the fact of Williams’

supervised release suggest the court gave no weight to its specific terms. The court’s first

nod to supervised release came during the initial hearing, after it requested “research” on

whether commitment under Section 4246 was warranted if Williams’ “unconditional

release” would pose a substantial risk. JA 115. The government’s lawyer sought

clarification, asking whether the court wanted to know if it could impose new terms on

Williams’ release or was instead referring to the “term of supervised release that has been

9 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 10 of 15

imposed by [the sentencing] court.” JA 116. The commitment court responded: “Doesn’t

have anything to do with the term of supervised release.”

The commitment court’s next mention of supervised release came as it was orally

explaining why it had ordered Williams committed. Here too, the court’s remarks do not

show it analyzed the substance of Williams’ conditions of supervision or identified a

permissible reason to minimize the conditions’ weight.

Take the commitment court’s statement that Williams’ “three-year term of

supervised release” “is of a limited duration.” JA 157. For one thing, that reference to the

length of Williams’ supervised release says little about whether the court gave any

consideration to whether its terms would lessen Williams’ dangerousness—at least so long

as the conditions remained in effect.

At any rate, the commitment court admitted knowing little about how long Williams

would spend on supervised release, stating it did not “know how much of [the supervised

release term] has already expired” or whether there was “any way at all that it can be

extended.” JA 157. Before us, the parties hotly dispute several related matters, including

how much supervised release time remains and whether it would have been permissible for

the commitment court to order Williams committed today based on risks that may not

materialize (if at all) for up to three years after this Court’s ruling. It is far from clear

whether these various issues were aired before the commitment court, and the commitment

court did not purport to resolve them. Accordingly, we decline to consider whether the

commitment court’s decision might have survived appellate review had it done so.

10 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 11 of 15

That leaves the commitment court’s observation that Williams’ supervised release

“is subject to the control of some other court” and its brief remarks about the ability of the

sentencing court to regulate Williams’ conduct and protect public safety during his

supervised release. JA 157. We have already explained why any conditions of supervised

release are relevant to whether a person should be committed, and the fact that such

conditions were imposed or would be enforced by “some other court” does not change the

analysis. (Indeed, the sentencing and commitment courts will often differ because the

former will almost always be in the geographic area of the underlying crime, see generally

U.S. Const. amend. VI, while the latter must be where the inmate is detained—including,

as was true here, for psychiatric evaluation or treatment, see

(a)).

What is more, sentencing courts—including the federal district court that sentenced

Williams—must consider public safety and a defendant’s well-being when selecting a term

of incarceration and imposing conditions of supervised release. See

(c),

3553(a). To fulfill that duty, federal law grants sentencing courts broad authority to fashion

appropriately tailored terms of supervision, including the 14 special conditions imposed

here. See

(d); United States v. Magdaleno,

, 1222–23 (9th

Cir. 2022); United States v. Douglas,

(4th Cir. 2017). Accordingly, any

reliance on a belief that the sentencing court’s “authority to impose conditions on

[Williams] is very limited,” JA 157, would constitute legal error.

The commitment court’s written order also fails to illuminate its reasons for

rejecting Williams’ arguments about the terms of his supervised release. To be sure, the

commitment court did not expressly endorse the government’s claim “that [Williams’]

11 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 12 of 15

supervised release terms should not impact th[e] court’s determination,” JA 161, or that

the court should not even “consider” them, JA 149. At the same time, however, the

commitment court’s written order never directly addressed Williams’ assertion “that the

conditions of supervised release imposed in his criminal case . . . mitigate any risk of

danger.” JA 161. So, once again, the record is silent about whether the commitment court

gave any weight to the conditions of supervised release to which Williams would be subject

if it denied the government’s commitment petition and, if so, why it still concluded

Williams should be committed.

* * *

Refusing to release a person who has served a criminal sentence based on a

prediction that it would be too dangerous to do so is serious business. Before prolonging a

person’s loss of liberty under Section 4246, courts must consider all evidence bearing on

the issue, including whether and how well other preexisting restrictions on the person’s

conduct (including an unserved term of supervised release) could reduce the risks posed

by the person’s release. Because the record leaves us unable to tell whether the district

court appropriately considered the terms of Williams’ supervised release, we vacate the

commitment order and remand for further proceedings.

SO ORDERED

12 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 13 of 15

RICHARDSON, Circuit Judge, dissenting:

The majority and I largely agree. Williams suffers from bipolar type schizoaffective

disorder. When he is taking his medicine, he’s fine. But when he’s off his medicine, he’s

dangerous. As a federal inmate, Williams sometimes took medication and other times

would not. When the district court held its commitment hearing, Williams had been taking

his medication for about a year. But, during that year, he had no choice in the matter. And

he had recently asserted that he did not need medication all the time, only as needed. So

the district court had to make a tough predictive judgment: Would Williams keep taking

his medicine if released from federal custody? Based on the evidence in front of it, the

court predicted that he would not. It thus held that Williams was highly likely to pose a

substantial risk to others and civilly committed Williams under § 4246. So far, so good.

But the majority says that—even assuming the district court had sufficient

evidence to support its answer—the court didn’t show enough of its work to get credit.

That is where we part ways.

I wholeheartedly agree that the district court needed to consider Williams’s

preexisting supervised release when predicting the risk he posed. But, from my reading of

the record, the district court did just that. During a hearing, Williams’s attorney argued at

length that his supervised release conditions were enough to mitigate his danger. The

district court rejected that argument, reasoning that the supervised release terms were “of

a limited duration.” J.A. 157. The district court also expressed concern that—even while

they were in place—Williams’s existing supervision might not adequately mitigate his risk.

See J.A. 157 (noting Williams’s release conditions were “subject to the control of some

13 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 14 of 15

other court”). 1 The court continued: “I don’t know how much of that has already expired.

I don’t know any way at all that it can be extended.” J.A. 157.

To my eyes, the district court’s discussion revealed a concern that, once Williams’s

supervised release conditions ended—whenever precisely that would be 2—he would once

again stop taking his medication, as he had many times in the past. And, just like those

past instances, once he became unmedicated, he would become violent and dangerous. So

it’s technically true to say that the district court “said nothing about the [specific] terms of

Williams’ supervised release,” or whether those terms would render Williams safe “at least

so long as the conditions remained in effect.” Majority at 9, 10. But the district court did

not need to discuss those questions because its decision rested on a far more fundamental

problem: Whatever the terms were, they were ending soon. And so whatever protection

they provided would soon end as well.

The district court’s finding that Williams was highly likely to pose a substantial risk

to others is entitled to very deferential review. The majority repeats the refrain that a

district court need not “explain in detail why it rejects each and every individual piece of

evidence.” Majority at 9 (quoting United States v. Wooden,

(4th Cir.

1 Indeed, the record reflects Williams had skirted court-ordered supervision before. So it makes sense that the district court was worried that he might do so again. But even if this time might be different, Williams’s supervision is still of limited duration. Whatever protection it provides will end in just a few years. 2 The majority notes that the parties “hotly dispute . . . how much supervised release time remains.” Majority at 10. And, in the hearing, the district court acknowledged that it did not know how much of the three-year term was left. But the point is that the term would eventually expire: not next decade, but soon.

14 USCA4 Appeal: 22-6464 Doc: 38 Filed: 11/17/2022 Pg: 15 of 15

2018)). Yet that promise rings hollow here. The district court did enough given the

inherent challenges with predicting the future risk someone will pose. I respectfully

dissent.

15