Independent Expert Interpretation of Fraud Analytics Does Not Trigger Confrontation Rights Absent Relayed Testimonial Hearsay

I. Introduction

United States v. Ali Abdelrazzaq (6th Cir. June 12, 2026) arises from a multi-year scheme involving multiple pharmacist-owners and five pharmacies (Harper Drugs; Wayne Campus Pharmacy, LLC; Heartland Pharmacy, LLC; Heartland Pharmacy 2; and Eastside Pharmacy). The government alleged that the pharmacists repeatedly billed Medicare, Medicaid, and other insurers for prescriptions that were not dispensed, sometimes using tactics such as keeping “never dispensed” labels, waiving copays to drive volume, substituting generics while billing for brand-name drugs, and falsifying patient signatures to survive audits.

After a jury trial, defendants Raef Hamaed, Tarek Fakhuri, Kindy Ghussin, and Ali Abdelrazzaq were convicted of conspiracy to commit healthcare fraud and wire fraud; Fakhuri was also convicted of a substantive healthcare fraud count. The appeal presented a cluster of issues, but the opinion’s core contribution is its treatment of a recurring modern evidentiary problem: whether an analytics-based fraud expert violates the Confrontation Clause when team members handled data extraction/verification.

II. Summary of the Opinion

• Convictions affirmed: No reversible Confrontation Clause error in Qlarant expert Johanna Sullivan’s testimony; no violation of the right to present a complete defense; no reversible “expert bolstering” despite the court’s improper references; no variance from the single-conspiracy indictment; no abuse of discretion in admitting a cooperator’s lay interpretations of texts; no Rule 17(c) error; no juror-impeachment relief.
• Sentences affirmed: Loss calculations upheld; sophisticated-means enhancement properly applied to Ghussin; Ghussin’s below-Guidelines term substantively reasonable.
• Restitution affirmed: Even assuming procedural misstep in timing/pronouncement, any error was harmless given opportunities to contest loss and restitution.

III. Analysis

A. Precedents Cited (and How They Shaped the Decision)

1. Confrontation Clause: “testimonial hearsay” limits expert testimony

The court’s Confrontation Clause framework is built on Crawford v. Washington, 541 U.S. 36 (2004), and its successor line emphasizing that the Clause bars admission of testimonial hearsay absent unavailability and prior cross-examination. The court draws the two-part boundary—testimonial + hearsay—through Smith v. Arizona, 602 U.S. 779 (2024), stressing that confrontation is triggered only when an out-of-court statement is both testimonial and offered for its truth.

To define “testimonial,” the panel relies on the “primary purpose” inquiry from Ohio v. Clark, 576 U.S. 237 (2015), and the contextual “relevant circumstances” approach from Michigan v. Bryant, 562 U.S. 344 (2011), while anchoring the “core class” of testimonial materials in Melendez-Diaz v. Massachusetts, 557 U.S. 305 (2009).

On experts as “surrogates,” the opinion treats Bullcoming v. New Mexico, 564 U.S. 647 (2011), and Smith v. Arizona as prohibiting the government from calling a substitute witness to relay another analyst’s conclusions. Critically, however, the panel reads those cases as turning on whether the testifying witness relayed (i.e., functioned as a “mouthpiece” for) an absent declarant’s testimonial assertions, as opposed to offering an independent assessment of data.

The panel also uses Delaware v. Fensterer, 474 U.S. 15 (1985), for the proposition that the Constitution guarantees an opportunity for effective cross-examination—not the most defense-favorable cross-examination imaginable. It invokes Melendez-Diaz v. Massachusetts again (notably n.1) to emphasize that confrontation is not a chain-of-custody requirement demanding every “hand” involved in evidence preparation.

2. Right to present a complete defense and Rule 403 limits

The court applies Holmes v. South Carolina, 547 U.S. 319 (2006), Taylor v. Illinois, 484 U.S. 400 (1988), and United States v. Scheffer, 523 U.S. 303 (1998), as the familiar trio recognizing a meaningful defense right constrained by legitimate evidentiary rules (especially confusion/prejudice under Rule 403). Sixth Circuit applications—United States v. Reichert, 747 F.3d 445 (6th Cir. 2014), United States v. Odeh, 815 F.3d 968 (6th Cir. 2016), and United States v. Reynolds, 86 F.4th 332 (6th Cir. 2023)—frame the defendant’s “demanding” burden to show an exclusion was arbitrary/disproportionate and infringed a weighty interest.

On “legitimate acts” evidence, the panel relies on United States v. Dimora, 750 F.3d 619 (6th Cir. 2014), and treats the Eleventh Circuit’s United States v. Ifediba, 46 F.4th 1225 (11th Cir. 2022), as persuasive: where the theory is not “everything was fraud,” evidence that “some things were legitimate” is typically not exculpatory. The panel also cites United States v. Betro, 115 F.4th 429 (6th Cir. 2024), and United States v. Daulton, 266 F. App’x 381 (6th Cir. 2008), to emphasize that “noncriminal activities” become relevant chiefly when the charge is effectively “ceaseless” criminality.

3. “Expert” bolstering in front of the jury (plain error, but no prejudice)

The court treats as error the practice condemned in United States v. Johnson, 488 F.3d 690 (6th Cir. 2007), reinforced by United States v. Maya, 966 F.3d 493 (6th Cir. 2020), and restated in United States v. Campbell, 135 F.4th 376 (6th Cir. 2025): parties should not “tender” an expert before the jury and courts should not announce “expert” status in a way that appears to endorse credibility. But under the plain-error framework of Greer v. United States, 593 U.S. 503 (2021), and prejudice principles in Molina-Martinez v. United States, 578 U.S. 189 (2016), the defendants failed to show a reasonable probability of a different outcome given limiting instructions and other evidence.

4. Variance: single vs. multiple conspiracies

The variance analysis rests on Grand Jury Clause principles from Stirone v. United States, 361 U.S. 212 (1960), and United States v. Miller, 471 U.S. 130 (1985), filtered through Sixth Circuit tests in United States v. Guerrero, 76 F.4th 519 (6th Cir. 2023), United States v. Siefert, 161 F.4th 379 (6th Cir. 2025), United States v. Hughes, 505 F.3d 578 (6th Cir. 2007), and United States v. Beals, 698 F.3d 248 (6th Cir. 2012). The panel applied the “common goal / nature of scheme / overlap” factors and held the evidence supported a single profit-driven conspiracy spanning all pharmacies.

5. Other evidentiary and procedural tools

• Lay opinion on texts: Rule 701 permitted Abdallah’s interpretations based on participation and relationship, consistent with United States v. Freeman, 730 F.3d 590 (6th Cir. 2013), and United States v. Reed, 163 F.4th 338 (6th Cir. 2025). The panel distinguished United States v. Kaplan, 490 F.3d 110 (2d Cir. 2007), because Abdallah had personal experience with Abdelrazzaq.
• Rule 17(c): The subpoena request failed the United States v. Nixon, 418 U.S. 683 (1974), standards (especially specificity and non-fishing), as applied in United States v. Theunick, 651 F.3d 578 (6th Cir. 2011), and United States v. Llanez-Garcia, 735 F.3d 483 (6th Cir. 2013).
• Juror impeachment: Post-verdict confusion about instructions cannot impeach unanimity under McDonald v. Pless, 238 U.S. 264 (1915), and United States v. Gonzales, 227 F.3d 520 (6th Cir. 2000).

6. Sentencing and restitution precedents

On loss, the panel applies the “reasonable estimate” approach (and deferential clear-error review of facts) through United States v. Wala, 166 F.4th 583 (6th Cir. 2026), United States v. Matthews, 155 F.4th 845 (6th Cir. 2025), and United States v. Washington, 715 F.3d 975 (6th Cir. 2013), along with methodology review described in United States v. Betro. Relevant conduct attribution follows United States v. Donadeo, 910 F.3d 886 (6th Cir. 2018), and foreseeability reasoning echoes United States v. Kennedy, 714 F.3d 951 (6th Cir. 2013).

Sophisticated means is assessed under United States v. Simmerman, 850 F.3d 829 (6th Cir. 2017), United States v. Crosgrove, 637 F.3d 646 (6th Cir. 2011), and the total-scheme approach in United States v. Yousef, 170 F.4th 498 (6th Cir. 2026).

Restitution analysis centers on the MVRA and procedural flexibility recognized in United States v. Vandeberg, 201 F.3d 805 (6th Cir. 2000), with amount/precision guidance from United States v. Kilpatrick, 798 F.3d 365 (6th Cir. 2015), and joint-and-several apportionment discretion confirmed by United States v. Hunt, 521 F.3d 636 (6th Cir. 2008).

B. Legal Reasoning (What the Court Actually Did)

1. The opinion’s central rule: independent interpretation of data ≠ surrogate testimony

The defendants’ Confrontation Clause theory depended on characterizing Qlarant’s internal validation of data completeness as testimonial statements by non-testifying analysts, implicitly communicated to the jury through Sullivan. The panel rejected this by demanding what Smith v. Arizona requires: identify the actual out-of-court statement purportedly introduced for its truth. Because Sullivan largely testified to (i) her understanding of Qlarant’s general processes from other cases and (ii) her own analysis and interpretation of admitted records and SAS outputs, the panel found no “testimonial hearsay” was conveyed.

The court treated “data completeness” concerns as classic weight-of-evidence issues. In line with Melendez-Diaz v. Massachusetts (n.1) and Delaware v. Fensterer, it held the defense could cross-examine Sullivan on her lack of personal involvement in early collection steps, and could use the underlying records (which defendants stipulated into evidence) to probe for missing invoices or input errors. The Clause, the panel emphasized, does not compel production of every contributor to a large data pipeline.

Distinguishing Smith v. Arizona and Bullcoming v. New Mexico, the panel stressed that Sullivan “touched everything that went to the jury”: she performed the final quality check, confirmed timeframes, verified data population into SAS, checked typographical corrections, interpreted the results, and helped create trial charts. This level of personal analytical engagement made her a genuine opinion witness rather than a surrogate “reading from records.”

2. Defense-evidence exclusions: relevance and speculation

The court affirmed excluding evidence that the pharmacies also filled legitimate prescriptions because the government did not claim “every claim was fraudulent”; therefore, lawful conduct did not negate an element of the charged conspiracy. On PBM “bias,” the court treated industry criticisms as too attenuated and speculative absent evidence of actual data manipulation in this case; admission risked confusing the jury under Rule 403.

3. Improper “expert” endorsement: error, but not outcome-determinative

Although the panel found it plainly improper for the district court to repeatedly call Sullivan an “expert” in the jury’s presence (and for the government to invite it), it held defendants could not show prejudice given instructions that the jury could reject her opinions and the presence of substantial independent evidence (including coconspirator testimony and communications).

4. Variance, lay opinion, Rule 17(c), and juror impeachment

The single-conspiracy proof was supported by shared profit motive, shared methods, and overlapping ownership/personnel, defeating the multiple-conspiracies theory. Abdallah’s interpretations of texts were permissible lay opinions grounded in relationship and participation. The Rule 17(c) subpoena request was overbroad and speculative, amounting to a fishing expedition. And post-verdict juror confusion could not be used to impeach the verdict.

5. Sentencing and restitution: deference to reasonable estimates and flexible MVRA procedure

On loss, the panel upheld reliance on Sullivan/Qlarant invoice review because defendants offered no concrete counter-calculation and did not demonstrate a methodology error that moved the Guidelines needle. For Ghussin, the court treated Wayne Campus losses as foreseeable relevant conduct given ownership stake and profit participation. The sophisticated-means enhancement was supported by falsified patient signatures/records used to defeat audits.

On restitution, the panel acknowledged the district court’s irregular sequence (entering judgment before promised briefing), but held any error harmless under United States v. Vandeberg because Hamaed had meaningful opportunities to contest the loss/restitution figure at sentencing and via post-judgment motion, and still offered only speculation rather than proof.

C. Impact (Why This Opinion Matters)

• Fraud prosecutions increasingly rely on “team-built” analytics. The decision provides a practical Confrontation Clause roadmap: the government may call an expert who personally analyzes and interprets admissible data outputs, even if other staff handled extraction/validation, so long as the expert does not convey absent analysts’ testimonial assertions for their truth.
• Defense strategy implications: The opinion reinforces that “data integrity” attacks often sound in weight (cross-examination and competing proof), not admissibility (constitutional exclusion), especially where defendants stipulate to records and do not identify a specific testimonial statement being laundered through the expert.
• Trial management: Despite reiterating that labeling a witness an “expert” before the jury is improper, the opinion signals that appellants must show concrete prejudice—particularly where limiting instructions and other evidence support the verdict.
• Sentencing and restitution: The court continues to tolerate “reasonable estimate” methodologies for loss in complex billing fraud and confirms that forged pickup documentation and audit-evasion can qualify as “sophisticated means.” On restitution, it highlights MVRA procedural flexibility and the difficulty of reversing restitution absent a developed, evidence-backed alternative calculation.

IV. Complex Concepts Simplified

Confrontation Clause (Sixth Amendment)

The right to cross-examine witnesses against the accused. It mainly blocks the prosecution from introducing testimonial out-of-court statements for their truth unless the declarant is unavailable and the defendant previously cross-examined them.

Testimonial hearsay

A statement made with the primary purpose of serving as a substitute for trial testimony, and offered to prove what it asserts. Both elements—testimonial and hearsay—must be present to trigger the constitutional bar.

“Surrogate analyst” problem

When the government calls someone other than the person who performed the key test/analysis to report results. This is forbidden when the witness merely relays an absent analyst’s testimonial conclusions, but permitted when the witness offers an independent opinion based on admissible data.

Variance (single vs. multiple conspiracies)

A mismatch between the conspiracy charged and what the evidence proves. Courts look for a common goal, the nature of the scheme, and participant overlap to decide whether the proof supports one conspiracy or several.

Rule 17(c) subpoena

A mechanism to obtain specific third-party documents for trial—not a broad discovery tool. The request must be specific, relevant, necessary, and made in good faith (not a fishing expedition).

Loss (Guidelines) vs. Restitution (MVRA)

Guidelines loss can be a reasonable estimate and may involve intended loss; MVRA restitution is based on actual loss to victims. In practice, when the court uses actual loss for Guidelines purposes, it often aligns closely with restitution.

Sophisticated means enhancement

A sentencing increase when the execution or concealment of fraud is especially intricate. Even “simple” fraud can qualify if concealment (e.g., falsified records to defeat audits) is sophisticated.

V. Conclusion

United States v. Ali Abdelrazzaq strengthens Sixth Circuit guidance on modern, data-driven fraud prosecutions: an expert does not violate the Confrontation Clause by offering an independent interpretation of large-scale claims and invoice data, even where other team members performed extraction or validation steps, so long as the expert does not transmit absent analysts’ testimonial assertions. The opinion also reinforces Rule 403 limits on marginal or speculative defense narratives, requires concrete prejudice for “expert” bolstering errors, and affirms deferential review of loss estimates, sophisticated-means findings, and MVRA restitution procedures where defendants fail to provide evidence-based alternative calculations.