GOULD, Circuit Judge

Hilary Potashner, Federal Public Defender; Brianna Fuller Mircheff (argued), Deputy Federal Public Defender, Los Angeles, CA, for Defendant–Appellant.

Eileen M. Decker, United States Attorney; Patricia A. Donahue, Assistant United States Attorney Chief, National Security Division; Melissa Mills (argued) and Sarah J. Heidel, Assistant United States Attorneys, Los Angeles, CA, for Plaintiff–Appellee.

Appeal from the United States District Court for the Central District of California, Beverly Reid O'Connell, District Judge, Presiding. D.C. No. 2:13–cr–00676–BRO–1.

Before: RONALD M. GOULD and MARSHA S. BERZON, Circuit Judges, and GEORGE CARAM STEEH III, Senior District Judge.

OPINION

GOULD, Circuit Judge:

Nna Alpha Onuoha appeals the district court's order authorizing the Bureau of Prisons (BOP) to forcibly medicate him to restore his competency to stand trial. Onuoha was charged under 18 U.S.C. §§ 844(e) and 1038(a)(1) for allegedly making phone calls to authorities at the Los Angeles International Airport (LAX) instructing them to evacuate the airport. He was found unfit to stand trial, and the district court ordered him to be forcibly medicated pursuant to Sell v. United States, 539 U.S. 166, 123 S.Ct. 2174, 156 L.Ed.2d 197 (2003). We have jurisdiction over interlocutory appeals of Sell orders under the collateral order doctrine. Sell, 539 U.S. at 176, 123 S.Ct. 2174. We hold that the district court clearly erred in finding that the proposed course of treatment was in Onuoha's best medical interests. We vacate the order and remand for further proceedings consistent with this opinion.

I

Onuoha served in the National Guard from 2004 to 2012, including a stint with a peacekeeping force in Kosovo. After returning from Kosovo, Onuoha worked as a Transportation Security Administration (TSA) screener at LAX from 2006 to September 2013. Except for the charges in this case, he has no criminal history.

In the summer of 2013, Onuoha was suspended from his job with TSA for comments made to a female passenger. On September 10, 2013, Onuoha went to LAX on his day off and passed through security screening at several terminals. He then went to TSA headquarters at LAX and resigned from his job. Hours later, he returned to TSA headquarters and left an envelope for a former supervisor involved with his suspension. The government alleges that Onuoha then called a TSA checkpoint and said that LAX should be evacuated. During the phone call, Onuoha mentioned the package he left for his former supervisor, indicated that it should be read immediately, and said that he would be watching to see if LAX was evacuated. Onuoha then called the LAX Police Department and his TSA supervisor, telling them to evacuate the airport because he was going to “deliver a message” to America and the world. The recipients of these calls believed that Onuoha was threatening to set off bombs or open fire at the airport. The envelope Onuoha left for his supervisor was discovered to contain religious writings, and did not include any explosives. Authorities decided not to evacuate the airport, but they did evacuate TSA headquarters.

Law enforcement officials went to Onuoha's apartment to apprehend him. They found that he had cleared out all of his belongings and left only a large note reading “09/11/2013 THERE WILL BE FIRE! FEAR! FEAR! FEAR!” This message led police to believe that Onuoha was an active shooter seeking to evacuate the airport so that he could target and kill people as they fled. Information about Onuoha's military background and potential access to firearms fed these concerns. It was later discovered that Onuoha had posted to his personal website an open letter “To LAX Passengers” with religious comments. This letter stated that “the news media have probably come to the conclusion that I'm a terrorist,” but also stated “I did not call for any threat.”

Later that same day, Onuoha called LAX police to say that he heard law enforcement was looking for him. He told police he was at a church in Riverside, California, and described the car he was driving. He also told police that he did not intend to make a bomb threat, only to “deliver” a message. Onuoha waited at the church until law enforcement arrived and arrested him. When he was interviewed by police, he reiterated that he did not intend to make a threat, stating that “[k]illing was not on my mind.”

On September 11, 2013, the day after Onuoha was arrested, the government filed a complaint against Onuoha and requested pre-trial detention. Onuoha was later indicted on three counts in violation of 18 U.S.C. § 1038(a)(1) (false information and hoaxes) and three counts in violation of 18 U.S.C. § 844(e) (making telephonic threats). At the detention hearing, the government moved for a competency evaluation, which Onuoha's defense counsel opposed. The motion was denied, and defense counsel indicated that Onuoha intended to proceed to trial. In February 2014, the defense gave notice that it would raise a diminished-capacity defense and submitted a report that Onuoha suffered from paranoid schizophrenia. The government again filed a motion for a competency evaluation, which this time was granted by the district court.

The evaluation was performed by Bureau of Prisons (BOP) medical personnel. The evaluation revealed that Onuoha believed that he received revelations from God and had a message to preach, and that these beliefs rose to the level of delusions. The evaluation concluded that Onuoha was not competent to stand trial. The district court found Onuoha incompetent to stand trial and committed him to BOP custody to determine whether he could be restored to competency.

BOP psychologist Dr. Angela Alden–Weaver and BOP psychiatrist Dr. Robert Lucking evaluated Onuoha for several months. They submitted their evaluation to the district court in November 2014. They agreed with the finding that Onuoha was incompetent to stand trial and diagnosed him with schizophrenia. They also found that Onuoha was not a danger to himself or others. They further determined that anti-psychotic medication would likely restore Onuoha to competency, and recommended a course of long-acting Haldol (haloperidol decanoate), including specific dosages and a timetable. The recommended treatment included an initial test dose of 10 milligrams of short-acting Haldol, followed by 24 hours of observation for adverse side effects. The treatment plan then recommended three 150–milligram doses of long-acting Haldol at two-week intervals to obtain a therapeutic blood level. After gaining this blood level, the treatment plan recommended 150 to 200 milligrams of Haldol every four weeks. Dr. Lucking predicted that this treatment would take around four months to restore Onuoha to competency.

The government filed a motion for an order to involuntarily medicate Onuoha with the goal of restoring him to competency, relying on Sell v. United States, 539 U.S. 166, 123 S.Ct. 2174, 156 L.Ed.2d 197 (2003). Onuoha's attorneys opposed the motion. The district court held several hearings that included taking testimony from government witnesses Dr. Lucking and Dr. Bryan Herbel, a second BOP psychiatrist. After the hearings, the district court granted the government's motion and ordered Onuoha to be involuntarily medicated in accordance with Dr. Lucking's recommendations as articulated in his and Dr. Weaver's evaluation. Onuoha filed a timely interlocutory appeal, and the district court stayed its order pending our decision.

II

In Sell v. United States, the Supreme Court recognized that the government may involuntarily medicate a defendant charged with a serious crime to restore that defendant to competency to stand trial. 539 U.S. at 179, 123 S.Ct. 2174. The Supreme Court held that a court may not grant a Sell motion unless the government proves four factors:

(1) “that important governmental interests are at stake” in prosecuting the defendant for the charged offense; (2) “that involuntary medication will significantly further those concomitant state interests,” i.e., it is substantially likely to restore the defendant to competency and substantially unlikely to cause side effects that would impair significantly his ability to assist in his defense at trial; (3) “that involuntary medication is necessary to further those interests,” i.e., there are no less intrusive treatments that are likely to achieve substantially the same results; and (4) “that administration of the drugs is medically appropriate, i.e., in the patient's best medical interest in light of his medical condition.”

United States v. Ruiz–Gaxiola, 623 F.3d 684, 687–88 (9th Cir.2010) (quoting Sell, 539 U.S. at 180–81, 123 S.Ct. 2174 ) (emphasis in Sell ). Each of these factors must be proven by clear and convincing evidence. Id. at 692. Orders based on Sell authorizing involuntary medication are “disfavored.” United States v. Rivera

Guerrero, 426 F.3d 1130, 1137 (9th Cir.2005).

The district court found that all four Sell factors were satisfied. On this appeal Onuoha challenges only the district court's conclusions on the first and fourth factors, and so we limit our discussion, first, to whether important government interests are at stake in prosecuting Onuoha and, second, to whether administration of the prescribed drugs is medically appropriate, i.e., in the patient's best medical interests in light of his medical condition. We conclude that the first factor is met but that the fourth factor is not: there is an important governmental interest in prosecuting Onuoha, but the proposed treatment is not in his best medical interests.

A

Under Sell we first address whether important governmental interests support prosecuting Onuoha. We review this factor de novo. Ruiz–Gaxiola, 623 F.3d at 693. The Sell Court recognized that “[t]he Government's interest in bringing to trial an individual accused of a serious crime is important,” but it also noted that “[s]pecial circumstances may lessen the importance of that interest.” Sell, 539 U.S. at 180, 123 S.Ct. 2174. The Court mentioned several examples of “special circumstances” that diminish the government's interest in prosecution, including the potential for civil commitment, the length of time needed to restore a defendant to competency, the effect of the potential delay on the government's interest in timely prosecution, the length of time the defendant has already been confined, and constitutional requirements of a fair trial. Id.

Our analysis of the first Sell factor proceeds as a two-step inquiry. In our first step, we consider whether the alleged crime is sufficiently “serious” to establish an important governmental interest. See United States v. Gillenwater, 749 F.3d 1094, 1101 (9th Cir.2014) ; Ruiz–Gaxiola, 623 F.3d at 693. If an important governmental interest is established, we evaluate in the second step of this analysis whether any “special circumstances” lessen that interest. Gillenwater, 749 F.3d at 1101 ; Ruiz–Gaxiola, 623 F.3d at 693–94. This second step requires measuring any mitigating circumstances against the established government interest. See, e.g., United States v. Brooks, 750 F.3d 1090, 1097 (9th Cir.2014) (explaining that courts must consider whether a potential sentence is outweighed by the likelihood of civil commitment and the length of time a defendant has already served).

Onuoha argues that the district court erroneously treated the first Sell factor as a totality-of-the-circumstances test. We agree that a totality test is inappropriate in the context of the first Sell factor. A relatively weak governmental interest could not properly prevail in scenarios without mitigating circumstances, because the Sell Court held that “important governmental interests” must be implicated to justify forcible medication. Sell, 539 U.S. at 180, 123 S.Ct. 2174 (emphasis in the original). Our two-step approach helps to ensure that the interests at stake are important. If the government cannot demonstrate at the outset that its interest in prosecution meets a significant threshold, the inquiry ends there.

1

We next address the facts here. We must consider whether Onuoha's charged crimes are sufficiently “serious” to indicate an important governmental interest. We have previously held that the U.S. Sentencing Guidelines range is “the appropriate starting point” because it is the “best available predictor of the length of a defendant's incarceration.” United States v. Hernandez–Vasquez, 513 F.3d 908, 919 (9th Cir.2007). Both parties agree that the Sentencing Guidelines range for Onuoha's alleged crimes is 27 to 33 months. This range is lower than any range we have previously held to be indicative of a “serious” crime under the first Sell factor. See, e.g., Gillenwater, 749 F.3d at 1101 (range of 33 to 41 months) ; Ruiz–Gaxiola, 623 F.3d at 694 (range of 100 to 125 months) ; Hernandez–Vasquez, 513 F.3d at 911–12 (range of 92 to 115 months).

But the Guidelines range is only the starting point in determining whether the government has an important interest in prosecution. Brooks, 750 F.3d at 1097. In Sell, the Supreme Court stated that courts also “must consider the facts of the individual case in evaluating the Government's interest in prosecution.” 539 U.S. at 180, 123 S.Ct. 2174. Although our analysis begins with the Guidelines range, it is not “the only factor that should be considered” because it does “not reflect the full universe of relevant circumstances.” Hernandez–Vasquez, 513 F.3d at 919.

In addition to the Guidelines range, we have previously considered the specific facts of the alleged crime as well as the defendant's criminal history. In Gillenwater, for example, we determined that the defendant's threats to choke, rape, and kill government officials and employees was sufficiently serious criminal conduct to satisfy the first Sell factor despite the low Guidelines range of 33 to 41 months. 749 F.3d at 1101. And in both Ruiz–Gaxiola and Hernandez–Vasquez, we considered the defendants' extensive criminal history in concluding that the crimes at issue were sufficiently serious. Ruiz–Gaxiola, 623 F.3d at 694 ; Hernandez–Vasquez, 513 F.3d at 919.

Onuoha has no criminal history and his Guidelines range is low. But even so, when we look at the substance of Onuoha's conduct, the stress he placed upon the airport's security systems, and the nature of the crimes charged, we conclude that Onuoha's criminal conduct is without doubt sufficiently serious to support a strong governmental interest. It is not just that he is the subject of prosecution; as the Fourth Circuit notes, this is a “truism [ ] applicable to any case where the government seeks forcible medication: without a prosecution, there would be no case.” United States v. White, 620 F.3d 401, 413 n. 9 (4th Cir.2010) (emphasis in the original). Rather, Onuoha's alleged conduct threatened “the basic human need for security” to such an extent that it weighs heavily in favor of an interest in prosecution. Sell, 539 U.S. at 180, 123 S.Ct. 2174. Onuoha is accused of making phone calls to LAX officials on the eve of the anniversary of the September 11th attacks, urging evacuation of the airport. These phone calls were reasonably perceived as terrorism threats, and they considerably disrupted airport activities and diverted law enforcement resources. The government did not merely have an interest in incarcerating Onuoha for a time for this conduct. It had an interest in gaining a trial conviction to show others that such conduct will result predictably in conviction and a serious penalty of incarceration.

Onuoha argues that his alleged criminal conduct is not sufficiently serious because his statements were “cryptic” and not specifically violent in nature. He has also continually maintained that he did not intend to threaten anyone, only to “deliver” a message. We conclude that these arguments are unavailing. Onuoha knew or reasonably should have known that the recipients of his phone calls would assume he was threatening terrorism. Terrorism, whether real or perceived, threatens our need for security. We agree with the district court's assessment that “[t]hreats of terrorism, whether genuine or fraudulent, are of grave severity, particularly when they involve a highly populated public venue such as an airport.”

The district court also considered Onuoha's potential for future violence as strengthening the need for prosecution. This consideration was wholly unnecessary to justify involuntary medication for the purpose of permitting trial and conviction. And reliance on Onuoha's dangerousness was potentially an error. Whether a defendant should be involuntarily medicated because they pose a danger to themselves or others is governed by a separate test, articulated in Washington v. Harper, 494 U.S. 210, 227, 110 S.Ct. 1028, 108 L.Ed.2d 178 (1990). Courts should “remain mindful of the Supreme Court's distinction between the purposes and requirements of involuntary medication to restore competency and involuntary medication to reduce dangerousness. It should take care to separate the Sell inquiry from the Harper dangerousness inquiry and not allow the inquiries to collapse into each other.” Hernandez–Vasquez, 513 F.3d at 919. Additionally, the record contains no firm evidence that Onuoha is an actual danger to himself or others, or that he will become a danger in the future. Drs. Lucking and Weaver specifically noted in their evaluation that Onuoha did not pose a threat to himself or others and would not qualify for involuntary medication under the Harper analysis.

2

Because the alleged crime is sufficiently serious to support a governmental interest in prosecution, we proceed to the second step of the first Sell factor and consider any “[s]pecial circumstances [that] may lessen the importance of that interest.” Sell, 539 U.S. at 180, 123 S.Ct. 2174. Onuoha argues that the time he has already spent in custody constitutes a “special circumstance” and diminishes the government's interest in incapacitating him. Sell suggests that length of time a defendant has already spent in confinement is a mitigating factor, although it “does not totally undermine” the need for prosecution. Id. Onuoha has been incarcerated since September 2013 and has already served more time than the minimum Guidelines range of 27 months. Possibly, if Onuoha is ultimately restored to competency and convicted, he may conceivably be sentenced to time served. However, a sentence might also include a period of supervised release, which “would help ensure that [Onuoha] does not return to making threats when released into the public.” Gillenwater, 749 F.3d at 1102. Additionally, there is an important distinction between incarceration itself, and the significance for society of gaining a criminal conviction for a defendant's violation of the law. A conviction and resulting sentence serves more purposes than the incapacitation, specific deterrence, and rehabilitation of an individual; general deterrence of the serious crime at issue here is also an important consideration. See, e.g., Furman v. Georgia, 408 U.S. 238, 343, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972) (per curiam) (Marshall, J., concurring) (“Our jurisprudence has always accepted deterrence in general, deterrence of individual recidivism, isolation of dangerous persons, and rehabilitation as proper goals of punishment.”); United States v. Barker, 771 F.2d 1362, 1368 (9th Cir.1985) (“[P]erhaps paramount among the purposes of punishment is the desire to deter similar misconduct by others.”); 18 U.S.C. § 3553(a)(2)(B) (courts should consider “adequate deterrence to criminal conduct” in selecting a sentence). Here, the government had a valid interest in prosecuting Onuoha for generating public fear over terrorism. That interest of the government cannot be served by mere detention; instead, general deterrence for the benefit of society is served when a person is convicted of a serious crime, thus deterring others from making the same mistake. We conclude that in this case, the particular circumstance of Onuoha's detention does not displace the governmental interest in prosecution.

There are no other circumstances that diminish the governmental interest in prosecution. Nothing in the record indicates that Onuoha is a candidate for civil commitment, and Onuoha has not argued that any delay resulting from the restoration process will interfere with the government's interest in timely prosecution or his constitutional rights to a fair trial. See Sell, 539 U.S. at 180, 123 S.Ct. 2174. We agree with the district court's finding that the alleged crimes are sufficiently serious to support an important governmental interest and that special circumstances do not diminish the importance of that interest. The first Sell factor is satisfied.

B

To satisfy the fourth Sell factor, we must conclude that the proposed treatment plan is “medically appropriate, i.e., in the patient's best medical interest in light of his medical condition.” Sell, 539 U.S. at 181, 123 S.Ct. 2174 (emphasis in the original). Whether the course of treatment recommend by the BOP is in Onuoha's best medical interests is a question of fact reviewed for clear error. Hernandez–Vasquez, 513 F.3d at 916–17. This is a deferential standard. See Easley v. Cromartie, 532 U.S. 234, 242, 121 S.Ct. 1452, 149 L.Ed.2d 430 (2001). We may not reverse a factual finding without a “definite and firm conviction that a mistake has been committed.” Id. (quoting United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948) ).

The district court adopted Dr. Lucking's recommended treatment as articulated in his and Dr. Weaver's evaluation. The proposed treatment includes an initial test dose of 10 milligrams of short-acting Haldol in the first 24 hours, followed by three 150–milligram doses of the long-acting version of Haldol at two-week intervals until a therapeutic blood level is obtained. Onuoha raises several objections to this treatment, arguing that the district court clearly erred in concluding that Dr. Lucking's recommendation is in his best medical interest. He primarily argues that the course of treatment increases the risk of side effects; the dosage is significantly higher than is generally recommended; and the use of long-acting Haldol does not conform to the community standard of care.

We first address Onuoha's concerns about the “extrapyramidal” (neurological) side effects of Haldol, including dystonia (muscle contractions that cause abnormal twisting postures), akathisia (the urge to move continuously), and pseudoparkinsonism (drug-induced Parkinson's disease ). Onuoha also points to other, similarly serious, side effects, some of which increase the risk of death. The district court heard testimony from Dr. Lucking that side effects at the recommended dosage of Haldol were infrequent and easily treatable with anticholinergic medication. Based on this testimony, the district court made “the factual finding that the treatment plan proposed by the government is unlikely to cause [Onuoha] significant side effects” and that involuntary medication was in Onuoha's best medical interest given the “significant delusions that have impacted his life.” In Gillenwater, which involved similar testimony from the same Dr. Lucking regarding haloperidol deconoate (long-acting Haldol ), we held that the district court did not clearly err in determining that medication was in the defendant's “best medical interest when the potential harms and benefits of the treatment are viewed against the seriousness of his condition.” 749 F.3d at 1105. However, we did not consider the recommended dosage in Gillenwater. Here, the district court heard testimony from Dr. Lucking that side effects are more likely to occur at higher doses. As Onuoha argued during the Sell hearing and maintains on appeal, Dr. Lucking's recommended dose is much higher than the BOP's own internal recommendations. The district court did not consider this contention in its written analysis.

Under the recommended treatment, Dr. Lucking proposed injecting Onuoha with 10 milligrams of short-acting Haldol to observe adverse side effects for 24 hours before administering 150 milligrams every two weeks for the first three doses. This dosage equates to 300 milligrams of long-acting Haldol in the first month of treatment. The BOP recommendations list the starting dose of short-acting Haldol at two to five milligrams per day, and the starting dose of long-acting Haldol at 25 to 50 milligrams every two weeks. The manufacturer of Haldol and the Physicians' Desk Reference (PDR) similarly recommend a short-acting Haldol test dose of two to five milligrams and state that the initial injections of long-acting Haldol should not exceed 100 milligrams. Dr. Lucking's recommended test dose is two to five times the BOP's starting recommendation, and his recommended starting dose for long-acting Haldol is three to six times the BOP's starting recommendation. Also, the BOP's recommended dose—after a starting dose—for long-acting Haldol is 50 to 200 milligrams every two to four weeks, and its maximum recommended dose is 300 milligrams every three to four weeks. Dr. Lucking's plan skips the starting dose and goes straight to the maximum dose of 300 milligrams per month.

Dr. Lucking testified that he regularly administers these starting doses “so that treatment moves on in a more rapid manner and [the recipient] can be restored in a more timely manner.” But restoring competency quickly is not a controlling concern under the fourth Sell factor—only best medical interests are considered. Dr. Lucking did not set forth any explanation why a dose above what is generally recommended is in Onuoha's best medical interests. Dr. Lucking suggests only that it would let Onuoha reach a therapeutic blood level faster.

The government contends that the BOP's internal standards are just “recommendations” that are not binding on a prescribing doctor. This observation is insufficient affirmatively to demonstrate that a high dosage is in Onuoha's best medical interest, which the government must prove by clear and convincing evidence. Ruiz–Gaxiola, 623 F.3d at 692. The government also argues that the recommendations were not written with restoration in mind. But the pertinent consideration under the fourth Sell factor is not restoration, but best medical interest.

Also, the district court appears to have miscalculated the amount of long-acting Haldol that Onuoha would receive in the first month. The district court incorrectly stated that Onuoha “would be administered doses of 150 milligrams on a monthly basis,” which the district court described as “on the lower end of typical doses of this medication.” In fact, under Dr. Lucking's recommended treatment, Onuoha would receive 300 milligrams in the first month, followed by 150 milligrams in the subsequent months. As previously noted, 300 milligrams as a starting dose is three to six times higher than the BOP's starting recommendation, and is the maximum recommended by the BOP as a non-starting dose. In light of the recommendations of the BOP and other medical sources, 300 milligrams cannot accurately be described as a low starting dose, or even a low dose. Because the district court miscalculated the dosage and failed to take into account the BOP dosage recommendations, it clearly erred in concluding that the proposed treatment was in Onuoha's best medical interest.

Onuoha also argues that the long-acting form of Haldol recommended will not allow doctors to monitor side effects and adjust his dosage as they would on short-acting Haldol. The district court did not consider this point in its written analysis, although it was raised and discussed at the Sell hearing. At the hearing, Onuoha pointed out that the PDR recommends that physicians stabilize patients on short-acting drugs before injecting them with long-acting Haldol. The manufacturer of Haldol also recommends that patients should only be treated with long-acting Haldol if they are stable and able to tolerate the short-acting version of Haldol. Dr. Lucking recommended against the use of short-acting Haldol because it would require daily injections that would be “traumatic” for Onuoha and would put the treating staff at risk, although he later acknowledged that Onuoha would likely acquiesce to injections with “a minimum of resistance.” Another government witness, Dr. Bryon Herbel, testified that long-acting Haldol is used in federal prisons because prison doctors are “trying to balance managing ... the side effects with the safety of repeated use of force.... So you look at the—the risk of, say, getting injured in a repeated forced cell extraction, that's not a minimal risk.”

The district court appears to have accepted these expert witnesses' explanation that short-acting Haldol was not appropriate for Onuoha. In response to Onuoha's arguments at the Sell hearing, the court noted, “[Dr. Herbel] said in the custodial setting, that the PDR, Physicians' Desk Reference, is set up for volunteers out of custody and that—the custodial situation is not set up to engage in the short-term.” However, penological interests do not control under the fourth Sell factor, which considers only “the patient's best medical interest in light of his medical condition.” Sell, 539 U.S. at 181, 123 S.Ct. 2174. It may be significant that the Sell Court used the word “patient” in its explanation of this factor, as opposed to the word “defendant”—a choice that “serves to emphasize that, in analyzing this factor, courts must consider the long-term medical interests of the individual rather than the short-term institutional interests of the justice system.” Ruiz–Gaxiola, 623 F.3d at 703. The record clearly indicates that stabilization on a short-acting anti-psychotic before the introduction of long-acting Haldol is the community standard of care. We agree with Onuoha that “best medical interests are best medical interests, whether that individual is in custody or in the community.”

We acknowledge that courts must rely on the testimony of medical experts in evaluating the constitutionality of involuntary medication. But a physician's word is not absolute, not even the word of a reputable and experienced doctor. Although Dr. Lucking has administered involuntary medication hundreds of times, his recommendations are still subject to Sell 's rigorous analysis. See United States v. Watson, 793 F.3d 416, 424–27 (4th Cir.2015) (holding that Dr. Lucking's proposed treatment did not satisfy the second Sell factor); United States v. Grigsby, 712 F.3d 964, 975–76 (6th Cir.2013) (holding that Dr. Lucking's proposed treatment did not satisfy the Sell analysis). On remand, the district court should evaluate Dr. Lucking's proposed treatment plan against the recommendations of other medical sources in the record, as well as consider any other pertinent evidence.

III

Involuntary medication orders are disfavored in light of the significant liberty interest at stake. Rivera–Guerrero, 426 F.3d at 1137. The government must demonstrate by clear and convincing evidence that all four of the Sell factors are satisfied. Ruiz–Gaxiola, 623 F.3d at 692. Here, we conclude the fourth factor is lacking, and the district court clearly erred in finding that the proposed treatment was in Onuoha's best medical interest. The record demonstrates that the proposed treatment includes dosages higher than are generally recommended and that the use of a long-acting medication does not conform to the standard of care. Although we recognize that the district court took pains to be careful and fair-minded about its decision, we have the firm conviction that the factual finding that the medication is in Onuoha's best medical interest is error on the current record. Although Dr. Lucking testified that the medication and dosage was appropriate, we conclude that the district court could not credit his testimony on that point without exploring and answering the questions posed by contradictory evidence in the record. We vacate the district court's order and remand on an open record for all four Sell factors for proceedings consistent with this opinion.

VACATED and REMANDED.