Before DENNIS, ELROD, and COSTA, Circuit Judges. GREGG COSTA, Circuit Judge: The prosecution of a medical clinic outside Dallas offers a window into the prescription drug epidemic that is plaguing America. At trial, the parties told a tale of two clinics. The government described a pill mill that prescribed patients more than a million doses of abusable drugs in just two years. The defense described a pain management clinic that helped people who appeared to suffer from chronic pain. A jury agreed with the government’s account and found the clinic’s doctor and office manager guilty of conspiring to distribute controlled substances. We consider a number of challenges to the convictions and sentences. I. Theodore “Tad” Taylor and Chia Jean Lee, a married couple who met while earning their degrees at Yale, ran Taylor Texas Medicine in Richardson, Texas. Taylor was the clinic’s only doctor while Lee, a nurse by training, was the clinic’s office manager. An Eastern District of Texas grand jury indicted the couple for conspiring to distribute controlled substances. The indictment alleged that from 2010 through early 2012, Taylor and Lee conspired to illegally prescribe five controlled substances: oxycodone, amphetamine salts, hydrocodone, alprazolam, and promethazine with codeine. A jury convicted both of them after a seven-day trial. It also made findings about the quantity of drugs the couple distributed, but those quantities did not trigger higher statutory minimum or maximum sentences. See 21 U.S.C. § 841(b)(1)(C). The district court then sentenced Taylor to the 20-year statutory maximum (his Guidelines range would have been higher but for the statutory cap) and Lee to just over 15 years (the bottom of her Guidelines range). Taylor and Lee challenge the sufficiency of the evidence, contend that they were convicted in an improper venue, and argue that three errors infected the trial: premature jury deliberation, unreliable expert testimony, and a deliberate ignorance instruction. They also appeal their sentences. II. We start with the defendants’ claim that there was not enough evidence to convict them. They moved for acquittal at the end of trial, so we review their sufficiency appeal de novo. See United States v. Ollison, 555 F.3d 152, 158 (5th Cir. 2009). That means we do not give deference to the district court’s ruling denying the motion. But, like the district judge, we give great deference to the jury’s factfinding role, viewing the evidence and drawing all inferences in favor of its verdict. United States v. Beacham, 774 F.3d 267, 272 (5th Cir. 2014). Because Taylor was a doctor with prescribing authority, he and Lee could distribute controlled substances as long as they did so for a legitimate medical purpose and within the scope of professional practice. United States v. Norris, 780 F.2d 1207, 1209 (5th Cir. 1986); see also 21 C.F.R. § 1306.04(a). Thus, when a conspirator has prescribing authority, the elements of conspiracy to distribute controlled substances are: “(1) an agreement by two or more persons to unlawfully distribute or dispense a controlled substance outside the scope of professional practice and without a legitimate medical purpose; (2) the defendant’s knowledge of the unlawful purpose of the agreement; and (3) the defendant’s willful participation in the agreement.” United States v. Oti, 872 F.3d 678, 687 (5th Cir. 2017) (footnote omitted). Even by the standards of our adversarial system, the difference in the parties’ portrayals of the clinic is stark. The defendants’ story is that they ran Taylor Texas Medicine as a legitimate pain management operation. Taylor says that he carefully examined patients, refused to prescribe to patients who tested positive for illegal drugs, and attempted conservative treatments before resorting to others prone to abuse. He acknowledges that, in retrospect, he may have made some mistakes. But he contends he acted in good faith and trusted his patients to accurately report their pain. Lee, for her part, asserts that she knew nothing about the prescriptions Taylor wrote. According to her, she was an innocent office manager. The government tells the story of a “pill mill”—a medical practice that serves as a front for dealing prescription drugs. It portrays a clinic packed with drug users and dealers, where one person would often pay for multiple patients’ visits. Also consistent with patients’ trafficking drugs is that, on follow-up visits, many tested negative for the medication Taylor had prescribed them. Others tested positive for illegal drugs like cocaine. Despite the red flags, Taylor kept prescribing these patients drugs. Even when a patient’s wife begged Taylor to stop feeding her husband’s drug addiction, he kept prescribing the husband drugs. And when a pharmacist who filled many of Taylor’s prescriptions told him that some of his patients were also receiving scripts from other doctors, he kept prescribing them drugs too. The pharmacist was so troubled that she contacted the Drug Enforcement Administration for the first time in her career. The government contends that Lee was a key part of the scheme. It says she reviewed failed drug tests, knew some patients had substance abuse problems, and prewrote prescriptions for Taylor to sign. She was also in charge of the clinic’s finances, which improved dramatically as the clinic concentrated its practice on pain management. Because the jury found the defendants guilty, we must honor the government’s telling if it is backed by evidence. It is. The government called seventeen witnesses, including the pharmacist who reported Taylor to the DEA, the patient’s wife who asked Taylor to stop prescribing drugs to her husband, undercover officers who pretended to be patients, an actual patient, medical experts, clinic staff, and case agents. It also introduced documentary evidence like financial records, patient files, and prescription data. Taylor testified too. All this evidence was more than enough for the jury to convict on. What follows is just a sampling. Taylor is not a pain management specialist, yet the clinic shifted its focus to pain patients when he and Lee began having financial difficulties. Eventually 80% of the clinic’s patients were pain patients. The proportion of prescriptions Taylor wrote for the commonly abused drugs hydrocodone and alprazolam grew from about 50% of prescriptions in January 2010 to over 80% by August 2011. Almost all those prescriptions were for the maximum dosage. Cf. United States v. Moore, 423 U.S. 122, 143 (1975) (“[The defendant] did not regulate the dosage at all, prescribing as much and as frequently as the patient demanded.”). He seldom offered patients conservative treatments not prone to abuse. Taylor did little to justify the prescriptions. By 2011, he was seeing 40 to 50 patients a day. The undercover visits confirmed the brevity of the examinations; Taylor spent between two-and-a-half and eleven minutes per visit with the pretend patients. Cf. Oti, 872 F.3d at 688 (describing pill mill where typical patient visits were between four and eight minutes long). One of the medical experts, Graves Owen, estimated that a pain doctor complying with the standard of care might spend 30 to 60 minutes with a new patient and between 10 and 15 minutes for an ordinary follow-up. Cf. id. at 687 (expert testified it would have been “impossible” for a doctor acting within the normal scope of professional practice to see 40 to 50 patients per day). What time Taylor spent with patients often involved only a cursory physical examination. A patient, the undercover officers, and the medical experts all testified that Taylor’s physicals were brief and that he rarely requested imaging to corroborate claims of pain. Sometimes Taylor would enter the examination room with a prefilled prescription form. Agents even found presigned (but otherwise blank) prescription forms when they searched the clinic. For some patients, Taylor wrote prescriptions without any examination at all; they could just stop by the clinic and pick them up. Cf. Moore, 423 U.S. at 142–43; United States v. Evans, 892 F.3d 692, 703–07 (5th Cir. 2018); Oti, 872 F.3d at 688 (all recognizing similar patterns indicative of a pill mill). For at least some of these prescriptions, Taylor had direct knowledge that the patients exhibited obvious drug-seeking behavior. Recall that a pharmacist told Taylor he was prescribing drugs to patients who were getting the same drugs from other doctors. And a patient’s wife called and emailed Taylor asking him not to prescribe to her husband because he had substance abuse problems and was getting prescriptions from other doctors. He ignored their concerns. The undercover operation again corroborated what was happening with clinic patients: Taylor prescribed drugs when the undercovers indicated their pain was fake. One testified that Taylor “coach[ed]” him to come up with an injury to “ legitimize” a prescription. The defendants’ responses to patient drug tests are also telling. A positive test for an illegal drug, such as cocaine, is a warning sign in flashing neon. Less apparent but no less damning is a negative test for a prescribed drug: it is a red flag that the so-called patient is selling medications rather than using them. Yet when many of Taylor’s patients “failed” drug tests— either testing positive for illegal drugs or testing negative for the drugs Taylor had prescribed them—he continued to sign off on scripts. Cf. Moore, 423 U.S. at 143 (recognizing this practice as incriminating evidence in a pill mill case). More than that, the clinic’s irregular pricing structure nakedly compensated Taylor and Lee for assuming the risk of prescribing to these patients with troubling drug tests. It charged a premium to patients who tested positive for illegal substances and gave a discount to patients who tested positive for the drugs they had been prescribed. As the clinic built up its pain management practice, monthly revenues rose fivefold, from just over $20,000 in early 2010 to more than $100,000 by mid-2011. Most of the clinic’s receipts were in cash. Pain patients could not use insurance for their first visit, and they could never use Medicaid. Cf. Oti, 872 F.3d at 684–85 (describing pill mill that accepted only cash, not insurance, Medicaid, or Medicare). Still, patients traveled from all over the Dallas–Fort Worth metroplex to see Taylor. Many patients seemed to know each other, and one man would sometimes pay for several patients’ prescriptions. To make matters worse for Taylor, the jury could have also concluded that he lied to try and hide his guilt. Taylor told DEA agents that he discharged patients who tested positive for illegal drugs (with the exception of marijuana). As we have discussed, the evidence told a much different story. Then, when he took the stand, Taylor repeatedly claimed he could not remember key facts such as whether he continued to prescribe to the patients who were receiving pain medication from other doctors. Patient records show that he did. These statements that the jury could view as coverups are yet more evidence that Taylor knew what he was doing was wrong. Lee fares little better in contesting her guilt. She knew some of the clinic’s patients failed drug tests but facilitated their prescriptions anyway: she sometimes administered drug tests; she saw that one of the undercover’s drug tests came back negative for a drug Taylor had prescribed him; she agreed to let an undercover avoid taking a drug test; and she charged prices that depended on drug test results. The jury could have also determined that Lee, a nurse, knew Taylor saw more patients than he could treat under the proper standard of care. Nevertheless, she continued to help Taylor run the clinic. When the woman who asked Taylor to stop prescribing to her husband emailed, Taylor made sure to copy Lee on the exchange. As the clinic’s business took off, the couple discussed patient volume and pricing. Lee even kept a prescription pad in her office area and sometimes prewrote prescriptions for Taylor to sign. So despite her claim that she just the office manager, the jury could have concluded that she was in on the scheme. All this evidence—and remember, there is more—is easily enough to support the jury’s verdicts. III. But even the guiltiest of defendants cannot be convicted in a venue where their crime did not occur. U.S. CONST. art. III, § 2; id. amend. VI. The defendants contend that is what happened here. Although the trial took place in the Eastern District of Texas (where the defendants lived), the clinic was located in the nearby Northern District. As the defendants emphasize, venue has posed a problem for Eastern District drug prosecutions. We recently vacated a conviction because there was no venue in the Eastern District. See United States v. Niamatali, 712 F. App’x 417 (5th Cir. 2018) (per curiam). In another case, this author raised concerns about prosecuting an Indianapolis drug ring in east Texas merely because a conspirator drove drug proceeds through the district. See United States v. Romans, 823 F.3d 299, 324–26 (5th Cir. 2016) (Costa, J., specially concurring). Yet even though the government prosecuted the Romans defendants hundreds of miles away from where their conspiracy was headquartered, venue was proper. That is because of how broad the venue rule is when it comes to conspiracy cases: “[V]enue is proper in any district where the agreement was formed or an overt act occurred.”[1] Id. at 310 (majority opinion) (citation omitted); accord Hyde v. United States, 225 U.S. 347 (1912); see also 18 U.S.C. § 3237(a) (providing that, for offenses “committed in more than one district,” venue is proper “in any district in which such offense was begun, continued, or completed”). “An overt act is an act performed to effect the object of a conspiracy.” United States v. Kiekow, 872 F.3d 236, 243 (5th Cir. 2017). It does not need to be a criminal act, but “it must be done in furtherance of the object of the conspiracy.” Romans, 823 F.3d at 310 (citation omitted). There is one other feature of our venue review that makes the defendants’ burden a difficult one. Venue is a fact question the jury answers. So similar to our review of the guilty verdicts, we must uphold the jury’s venue finding as long as any rational jury could have concluded that an overt act occurred in the Eastern District. Kiekow, 872 F.3d at 243. The one difference between the jury’s findings on guilt and venue makes our review even more forgiving to the venue determination: the government’s burden on venue is only a preponderance of the evidence. United States v. White, 611 F.2d 531, 534–35 (5th Cir. 1980).[2] The jury could have found two types of overt acts in the Eastern District. First, a rational jury could have concluded that the couple conducted clinic business at their home in the Eastern District. A DEA search of Taylor and Lee’s home in Plano uncovered clinic bookkeeping records dated November 11, 2011, on a CD labeled “Work Files.” Also in the home were bank records and tax documents addressed to Taylor Texas Medicine. Taylor quibbles with this evidence, arguing it was not enough to show that the couple worked at home or that the work related to pain patients. But while the evidence may not have compelled a finding that the couple did clinic business at their Eastern District home, it certainly allowed such a finding. Bookkeeping and tax work for the clinic furthered the couple’s scheme by keeping the pill mill up and running. See Castillo v. Scott, 51 F.3d 1042, 1995 WL 152993, at *2 (5th Cir. 1995) (per curiam) (unpublished) (describing “accounting for the drug transactions” as an “overt act”).[3] Second, Taylor and Lee regularly wrote checks from a bank located in the Eastern District to fund clinic operations. The checks paid clinic staff, a medical billing company, and the clinic’s rent. Those payments perpetuated the scheme just as bookkeeping and doing taxes for the clinic did. In addition to its compliance with the broad, modern interpretation of the venue requirement, this prosecution does not implicate the original vicinage right concern that a defendant might be “dragged to a trial . . . away from his friends, witnesses, and neighbourhood” and “subjected to the verdict of mere strangers.” 3 JOSEPH STORY, COMMENTARIES ON THE CONSTITUTION OF THE UNITED STATES § 1775 (1833); see also Romans, 823 F.3d at 325 (Costa, J., specially concurring) (discussing the Founders’ desire for a jury to decide local cases). Instead, it is like others we have seen in which criminal conduct takes place in the area north of Dallas that straddles the Northern and Eastern Districts. See United States v. Brown, 898 F.3d 636, 638 (5th Cir. 2018). Although the pill mill was in the Northern District, the defendants lived in the Eastern District and, as we have noted, helped run the business out of that district. The U.S. Attorney for the Eastern District had an understandable interest in squelching a pill mill that district residents operated and patronized. IV. Taylor and Lee next allege several problems with the trial. A. The first one is an allegation of premature jury deliberation. Premature deliberations threaten a defendant’s Sixth Amendment right to trial by an impartial jury. United States v. York, 600 F.3d 347, 356 (5th Cir. 2010). We nevertheless presume a jury was impartial unless the defendant proves otherwise. Id. at 358. And because the district court is best positioned to assess jury misconduct, we review its denial of a motion on that ground for abuse of discretion. Id. at 355. Its “discretion is broadest when the allegation involves internal misconduct such as premature deliberations” as opposed to “external misconduct such as exposure to media publicity.” Id. at 356. As is customary, the district court instructed the jurors at the beginning of the trial not to “discuss the case, even with other jurors, until all the jurors are in the jury room actually deliberating at the end of the case.” See FIFTH CIRCUIT PATTERN JURY INSTRUCTIONS (CRIMINAL CASES) § 1.01 (2019). It also admonished them to “not form any opinion” and to “keep a[n] open mind” until starting deliberations. It even followed the trend of instructing the jury on the offense’s elements before the trial started, which can help orient jurors to the issues and frame the evidence they will soon hear.[4] Despite those efforts, on the third day of trial the jury submitted a note to the court: “We are not clear on exactly what the charges are. Can we get specific clarification, or is that something we hear once all the testimony is complete[?] Are they together only, or separately charged[?]” The court decided to interview each juror individually to see if the jury had begun talking about the case. It asked a series of questions including, “ Have you discussed the merits of the case or reached any decision[?]” All but one juror answered in the negative. Juror 10 acknowledged reaching a decision but denied discussing it with anyone. In addition, several jurors reported that a court security officer said they could discuss the case only if they were all together in the jury room. Others maintained that the security officer had just told them not to discuss the case until deliberations. The defendants asked for a mistrial, arguing that the note showed the jury had begun to discuss the case and that the security officer had seemingly told the jurors that was okay. The district court denied their motions. It noted that all the jurors stated that they had not discussed the merits of the case with anyone. The court nevertheless excused Juror 10 for reaching a decision before the close of evidence. It then reminded the jurors not to discuss the case and to keep an open mind throughout trial. The district court did not abuse its broad discretion. Most significantly, all the jurors said they had not talked about the merits. The chatter among the jury was concerned only with clarifying what the charges were. Furthermore, there is a strong presumption that a jury will follow a court’s “wait to deliberate” instruction. York, 600 F.3d at 358; United States v. Patino- Prado, 533 F.3d 304, 313 (5th Cir. 2008). A security officer’s remark to some jurors that they could discuss the case in the jury room—a remark that does not directly contradict the court’s instruction to avoid deliberating until the close of evidence—does not overcome that presumption, especially when the jurors denied doing any such thing. See York, 600 F.3d at 355–58. Indeed, we have deferred to district courts in this area even when, unlike in this case, there was some evidence indicating that the jury discussed the evidence during the trial. See United States v. Arriola, 49 F.3d 727, 1995 WL 103275, at *5 (5th Cir. 1995) (per curiam) (unpublished) (rejecting premature deliberation appeal when an affidavit from a juror’s son averred that “jurors were discussing the testimony and weight of the evidence”).[5] The defendants have not offered any evidence to get past the presumption of jury impartiality. B. Next up is the expert issue. Taylor and Lee attack the reliability of the government’s two experts for offering opinions about the clinic’s entire practice based on their examination of a small fraction of patient files. See FED. R. EVID. 702. Even though DEA agents seized over 1100 patient files from the clinic, Owen reviewed only 22, and the other expert, Timothy Munzing, reviewed just 7. Taylor and Lee argue that these samples were not representative. We review the district court’s decision to allow the testimony for abuse of discretion. United States v. Kuhrt, 788 F.3d 403, 418 (5th Cir. 2015). We can overturn the ruling only if it was “manifestly erroneous.” Id. (citation omitted). Even then, as is true for other evidentiary issues, the government can salvage the convictions by proving any error was harmless. Id. That is the case here, so we need not decide the merits of the challenge. To begin, the defendants’ claim goes to very little of the witnesses’ testimony. Most of the experts’ statements were limited to conclusions about the patient files they reviewed. The government usually couched its examination in terms of the files the expert looked at. The prosecutor asked questions like, “Was Dr. Taylor engaged in the legitimate professional practice of medicine when treating the patients that you reviewed?” Only a handful of times did an expert’s testimony arguably go beyond the files he reviewed to opining generally about whether he thought Taylor’s practice was illegitimate. That limited testimony was cumulative of mounds of other evidence pointing to the same conclusion. Take, for instance, the testimony and videos from undercover agents showing that Taylor performed cursory examinations; records showing a correlation between patient volume, clinic revenue, and Taylor’s prescribing of commonly abused drugs; and testimony from clinic staff, patients, and a pharmacist showing that Taylor prescribed drugs to patients with drug-seeking behavior. In light of overwhelming evidence that the defendants ran a pill mill, it is hard to believe the trial’s outcome was affected by a few instances of an expert stating that Taylor was a “drug dealer” or that Taylor operated outside the scope of professional practice “in 2010 and 2011.” Cf. Evans, 892 F.3d at 715 (finding lay opinion testimony that the defendant ran a pill mill was harmless because other evidence would lead to the same conclusion). Indeed, the defendants’ argument for why this testimony was harmful focuses not so much on its impact on the guilty verdicts but on the jury’s drug quantity determination. But that finding had no binding effect because the defendants’ statutory punishment range was zero to twenty years regardless of quantity. See 21 U.S.C. § 841(b)(1)(C) (enhancing the statutory sentencing range for cases involving Schedule II drugs only when the defendant has a prior felony drug conviction or when death or serious bodily injury results from the drug use). To convict Taylor and Lee, the government had to show only that the couple conspired to distribute some controlled substances outside the scope of professional practice. See Evans, 892 F.3d at 707 (explaining that a doctor’s abiding by the standard of care for some patients was “irrelevant” to the charged conduct of unlawfully distributing controlled substances to three other patients). Finally, the defense took full advantage of the “traditional and appropriate means of attacking shaky but admissible evidence,” Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 596 (1993), by pointing out that the experts’ testimony was based on a nonrandom sample of the clinic’s patient files. Taylor’s counsel crossed Owen on whether he knew how the government selected the files he reviewed; he said he did not. Another of Taylor’ lawyers questioned a DEA agent on the same topic, and she admitted that the government “got to pick” the files the experts reviewed. These attacks featured in Taylor’s closing argument. The jury heard the defendants’ impeachment evidence and voted to convict anyway. Any error the district court committed by admitting the experts’ testimony was harmless. C. Taylor and Lee’s last claim of trial error is that the district court should not have instructed the jury on deliberate ignorance. This issue gives us déjà vu all over again. See Yogi Berra’s Most Memorable Sayings, MLB.COM (Sept. 23, 2015), https://www.mlb.com/news/yogisms-yogi-berras-best-sayings/c- 151217962. The instruction “should rarely be given,” United States v. Araiza- Jacobo, 917 F.3d 360, 366 (5th Cir. 2019) (citation omitted), but what seems rare is a health care prosecution without the instruction. The deliberate ignorance instruction—also called the willful blindness, conscious avoidance, or ostrich instruction—”inform[s] the jury that it may consider evidence of the defendant’s charade of ignorance as circumstantial proof of guilty knowledge.” United States v. Ricard, 922 F.3d 639, 655 (5th Cir. 2019) (citation omitted). It ensures that a defendant cannot bury his head in the sand to avoid liability. Id. Equating deliberate ignorance with knowledge dates back to nineteenth- century English common law. Ira P. Robbins, The Ostrich Instruction: Deliberate Ignorance as a Criminal Mens Rea, 81 J. CRIM. L. & CRIMINOLOGY 191, 196 (1990) (citing Regina v. Sleep, 169 Eng. Rep. 1296 (Cr. Cas. Res. 1861)).[6] The Supreme Court first approved the concept at the end of that century. Id. at 197–98 (citing Spurr v. United States, 174 U.S. 728 (1899)). But many attribute the modern rise of deliberate ignorance instructions to the Model Penal Code’s defining knowledge to include a situation in which “a person is aware of a high probability of [a fact's] existence, unless he actually believes that it does not exist.” United States v. Alston-Graves, 435 F.3d 331, 339 (D.C. Cir. 2006) (quoting MODEL PENAL CODE § 2.02(7)); see also Robbins, supra, at 200–01 (citing Leary v. United States, 395 U.S. 6 (1969),[7] and Turner v. United States, 396 U.S. 398 (1970), as drug prosecutions in which the Supreme Court relied on the Model Penal Code’s definition). The increasing use of ostrich instructions has prompted fears that “the jury might convict for negligence or stupidity.” Ricard, 922 F.3d at 655 (citation omitted). We are not alone in our concern with their overuse. Similar to our admonitions, other courts use the words “rarely,” “sparingly,” and “caution” when discussing the instruction. Alston-Graves, 435 F.3d at 340–41 (quoting cases from the First, Fourth, Fifth, Ninth, Tenth, and Eleventh Circuits giving these and similar warnings). We are also not alone in observing that these calls for restraint often go unheeded. See id. at 337 (“Why in the face of this mountain of evidence the prosecution sought, and the district court gave over a defense objection, a willful blindness instruction is difficult to fathom.”). The limitations we have emphasized are that a deliberate ignorance instruction should be given only “when a defendant claims a lack of guilty knowledge and the proof at trial supports an inference of deliberate ignorance.” Ricard, 922 F.3d at 655–56 (citation omitted). To allow that inference, there must be evidence showing: “(1) subjective awareness of a high probability of the existence of illegal conduct, and (2) purposeful contrivance to avoid learning of the illegal conduct.” Id. at 656 (citation omitted). We review a district court’s jury instructions for abuse of discretion if the defendants preserved their challenge and for plain error if they did not. United States v. Gibson, 875 F.3d 179, 195 (5th Cir. 2017). The parties dispute which standard applies, with the disagreement centering on the specificity of the defendants’ objections at the charge conference. We can sidestep this issue. Even though this case did not warrant a deliberate ignorance instruction, any error was, again, harmless. See United States v. Roussel, 705 F.3d 184, 190 (5th Cir. 2013) (deciding that the parties’ disagreement over the standard of review was “moot” because instruction error was harmless). Our concern about the instruction is not with the first requirement. The couple claimed a lack of guilty knowledge at trial. Only Lee argues otherwise. She says that she had to testify to claim lack of knowledge. But a defendant does not need to take the stand to claim lack of guilty knowledge. A defense attorney can raise the issue in an opening statement, through any witness’s testimony, or during closing argument. See United States v. Wofford, 560 F.3d 341, 353 (5th Cir. 2009); United States v. Wisenbaker, 14 F.3d 1022, 1027 (5th Cir. 1994). That is what Lee’s counsel did during opening, cross-examination of Taylor, and closing. The best example is this statement from the closing: “[T]here has been no evidence in this case, none, of her knowledge that these prescriptions were being illegally written, or that they were going out to patients without medical need.” With knowledge in dispute, we look to whether the evidence supported an inference of deliberate ignorance. For the first step of that analysis, the evidence demonstrated that both defendants were subjectively aware of a high probability that some illegal conduct was occurring at the clinic. This first prerequisite “often overlaps with an inquiry into a defendant’s actual knowledge,” Araiza-Jacobo, 917 F.3d at 366, as it does here. Because we have already explained that there was plenty of evidence indicating each defendant knew Taylor was improperly prescribing drugs to patients, we will not repeat ourselves. The second step is where we have doubts. For evidence that Taylor and Lee contrived to preserve deniability, the government points to testimony that their roles were separate. Taylor testified that the defendants had distinct jobs each one stuck to: he was the doctor who saw patients while Lee was the office manager who dealt with finances. He said the couple generally did not discuss patients or finances. Indeed, he claimed he did not even know how much the clinic charged patients. The separate roles translated into independent physical spaces. Taylor kept to his exam room and thus knew little about what went on in the reception area. The front office was Lee’s domain. Clinic staff testified that she was “very present” there. She managed the clinic’s employees, and sometimes performed patient intake. But she was not in the examination room when patients met with Taylor. This evidence does not show the purposeful contrivance required to raise an inference of deliberate ignorance. It falls short of the conduct that has satisfied this prerequisite in other health care prosecutions. See Gibson, 875 F.3d at 188, 196 (defendant responded to coconspirator’s proposal, “I don’t want to know what you-all are doing or how you-all doing [sic] it”); United States v. Brown, 871 F.3d 352, 356 (5th Cir. 2017) (defendant did nothing after consultant pointed out several problems with the billing practices of her medical equipment company); United States v. Delgado, 668 F.3d 219, 228 (5th Cir. 2012) (defendant sought advice from Medicare and Medicaid experts on only select issues and gave newsletters to coconspirators that omitted pages on the regulations their enterprise violated); cf. Wofford, 560 F.3d at 354 (embezzlement case in which defendant failed to investigate employees’ concerns after they asked him, “Don’t you think this might be illegal?”). It is a stretch to say that the defendants’ roles were so separate that one could plausibly infer—over the competing, commonsense explanation that the two merely had different jobs—that each was consciously avoiding the other’s domain to maintain deniability. Indeed, the government repeatedly pointed out that the defendants worked together frequently at the clinic, citing evidence like Lee’s habit of prewriting prescriptions for Taylor to sign. That narrative supported its main theory that Taylor and Lee both knew what they were doing. Cf. Oti, 872 F.3d at 698; Kuhrt, 788 F.3d at 417 (both recognizing that a deliberate ignorance instruction was improper when the government premised its case on actual knowledge). It is troubling that an instruction that should be given rarely has become commonplace. With someone’s liberty on the line, there must be a compelling justification for an instruction that runs the risk of “confus[ing] the jury” and convicting a defendant who merely “should have been aware” of criminal conduct. United States v. Cartwright, 6 F.3d 294, 301 (5th Cir. 1993). Prosecutors and district courts should carefully scrutinize the facts before deciding they warrant the instruction. The key is whether there is evidence showing the defendant took proactive steps to ensure his ignorance. See United States v. Mendoza-Medina, 346 F.3d 121, 133 (5th Cir. 2003) (“The sine qua non of deliberate ignorance ‘is the conscious action of the defendant—the defendant consciously attempted to escape confirmation of conditions or events he strongly suspected to exist.’” (citation omitted)). The deliberate ignorance instruction cannot be a “backup or supplement” when the case genuinely “hinges on a defendant’s actual knowledge.” Kuhrt, 788 F.3d at 417. Nevertheless, we end up where we often do when the district court gives the instruction in a case with strong evidence of actual knowledge. The error in giving the instruction was harmless because “there is substantial evidence of actual knowledge.” Id.; see also Alston-Graves, 435 F.3d at 342. As we have recounted, that evidence comes from a variety of sources: undercover agents, witnesses who raised concerns about patients’ drug abuse, the defendants’ practice of ignoring highly suspicious drug tests, the clinic’s odd pricing structure, and Taylor’s cursory medical examinations. In fact, it is hard to find another Fifth Circuit pill mill case with such overwhelming evidence of guilt. V. Taylor and Lee challenge their sentences too. They contend that the district court miscalculated the drug quantity involved in the conspiracy and also translated that amount into the wrong offense level. Taylor also argues that the district court erred by applying a firearm enhancement to his Guidelines range. A. The jury attributed the following drug quantities to the conspiracy: 114g of oxycodone, 580g of amphetamine salts, 3896g of hydrocodone, 542g of alprazolam, and 279g of promethazine with codeine. Each defendant’s revised Presentence Report (PSR) used those amounts. And the district court adopted the PSRs’ findings. Both defendants objected, so this court reviews the district court’s fact findings for clear error. Oti, 872 F.3d at 699.[8] A district court may estimate drug quantity. United States v. Betancourt, 422 F.3d 240, 246 (5th Cir. 2005). It can base its findings on “any information that has ‘sufficient indicia of reliability to support its probable accuracy,’ including a probation officer’s testimony, a policeman’s approximation of unrecovered drugs, and even hearsay.” Id. at 247 (citation omitted). Here, the district court estimated the drug quantities based on testimony from a member of the DEA task force. The officer took the number of prescriptions Taylor wrote for each charged controlled substance from January 2010 through September 2011 and then discounted those figures by 25% to account for prescriptions that may have been legitimate. The results were doubly conservative, he testified. For one thing, the base number did not include prescriptions written during the last five months of the conspiracy. For another, after reviewing all the clinic’s patient files, he said there was no evidence that even 25% of the clinic’s prescriptions were legitimate. The district court’s crediting this testimony is enough to support its drug quantity findings. Contrary to Taylor’s assertion, a medical expert was not required to determine how many prescriptions the clinic had dispensed outside the scope of professional practice. Oti, 872 F.3d at 700. There was no clear error. B. The next issue takes us deep into the weeds of the Sentencing Guidelines, but the answer ends up being simple. Taylor’s and Lee’s revised PSRs used the 2011 Sentencing Guidelines to turn the above listed drug quantities into a marijuana equivalency of about 14,312kg (the Guidelines used to convert each type of drug, based on its perceived harmfulness, into an amount of marijuana).[9] That figure put their base offense levels at 36. See U.S.S.G. § 2D1.1(c)(2) (2011). Lee argues a marijuana equivalent weight of 14,312kg should result in a base offense level of 34, not 36. Taylor asserts that his marijuana equivalency calculation ignored caps in the Guidelines for controlled substances listed on Schedules III, IV, and V. a Because neither defendant made those objections before the district court, we review for plain error. See United States v. Le, 512 F.3d 128, 135 (5th Cir. 2007). Taylor and Lee go wrong in mismatching parts of two different Sentencing Guidelines Manuals. Lee takes the marijuana equivalent calculated under the 2011 Guidelines and plugs it into the 2018 Guidelines, which set an offense level of 34 (two points lower than the 2011 manual) for a converted drug weight of at least 10,000kg and less than 30,000kg. U.S.S.G. § 2D1.1(c)(3) (2018). Taylor applies the 2018 Guidelines but treats hydrocodone as a Schedule III drug subject to a weight cap. Some forms of hydrocodone were on Schedule III in 2011, 21 C.F.R. § 1308.13(e)(1)(iii)–(iv) (2011), but the drug in all its forms was on Schedule II by 2018, 21 C.F.R. § 1308.12(b)(1)(vi) (2018), and the 2018 Guidelines treated it exclusively as such, U.S.S.G. § 2D1.1 cmt. n.8(D) (2018). This no doubt sounds complicated, but the bottom line is that defendants cannot have it both ways. Either the 2011 Guidelines apply or the 2018 Guidelines do. See U.S.S.G. § 1B1.11(b)(2) (“The Guidelines Manual in effect on a particular date shall be applied in its entirety.”). Parties cannot pick sections from different Guidelines that help them and meld them into a new Guidelines Manual applicable to just their case. Each Guidelines Manual the defendants invoke hurts them in some respects and helps them in others. For example, compared to the 2011 Guidelines, the 2018 Guidelines set higher converted drug quantity thresholds for every offense level but also provide a much harsher ratio for calculating hydrocodone’s converted drug weight. Compare U.S.S.G. § 2D1.1(c)(2) & cmt. n.10(D) (2011), with U.S.S.G. § 2D1.1(c)(2) & cmt. n.8(D) (2018). When it comes to the defendants’ base offense levels, the countervailing elements are a wash. After careful review, it turns out that 36 is the appropriate base offense level under either Guidelines Manual. That means either there was no error at all—because the offense level was correctly calculated—or any error in using the wrong manual had no effect on the defendants’ substantial rights because the offense level would be the same under either version. Whichever plain-error box one uses, this claim fails. C. The final issue involves a two-level enhancement Taylor received for possessing a firearm in connection with his drug trafficking. See U.S.S.G. § 2D1.1(b)(1) (2011). The enhancement was based on a handgun found in his desk drawer at the clinic. Taylor objected to the enhancement, arguing that the firearm was not present where the offense occurred—it was in his office while he wrote prescriptions in an exam room. We review the district court’s contrary finding for clear error. United States v. King, 773 F.3d 48, 52 (5th Cir. 2014). “The enhancement should be applied if the weapon was present, unless it is clearly improbable that the weapon was connected with the offense.” U.S.S.G. § 2D1.1 cmt. n.3(A) (2011). The government has the burden to prove by a preponderance of the evidence that the weapon was “present” by showing “that a temporal and spatial relation existed between the weapon, the drug trafficking activity, and the defendant.” King, 773 F.3d at 53 (citation omitted). If the government meets its burden, the defendant can avoid the enhancement by showing that a nexus between the weapon and the offense is “clearly improbable.” Id. (citation omitted). The district court did not clearly err in finding that the firearm enhancement applied. There was a temporal and spatial relationship between the gun, the drug trafficking activity, and Taylor: he kept the gun in his office while, in nearby rooms, he gave out illegal prescriptions and patients handed over cash. Although the enhancement requires the weapon to be “found in the same location where drugs . . . are stored or where part of the transaction occurred,” id. (citation omitted), that does not mean Taylor had to have the gun on his person or in the room when he prescribed to patients. See United States v. Gunn, 215 F. App’x 785, 793 (11th Cir. 2007) (per curiam) (rejecting a “room- by-room” approach to the firearm enhancement); see also United States v. McKeever, 906 F.2d 129, 134 (5th Cir. 1990) (affirming enhancement when guns were “scattered throughout the house” but not in the drug lab area, and observing that “the judge was entitled to conclude that the whole premises became the situs of the offense”). It was also not clearly improbable that the gun was connected to the offense. Clinic employees and patients reported feeling unsafe in the waiting room because some patients were unruly. The problem became so pronounced that the clinic hired a security guard. And, as mentioned, most of the clinic’s receipts were in cash. From those facts, the district court was allowed to infer that Taylor kept the gun in his office to keep order among drug-using clientele and to protect his scheme’s proceeds. * * * The judgments are AFFIRMED.
