Sixth Circuit Clarifies Confrontation Limits for Data-Driven Expert Testimony in Healthcare-Fraud Prosecutions

I. Introduction

In United States v. Kindy Ghussin (Nos. 25-1056/1099/1104/1157/1353, 6th Cir. June 12, 2026), the Sixth Circuit affirmed convictions and sentences arising from a multi-year pharmacy scheme that billed Medicare, Medicaid, and other insurers for prescriptions that were never dispensed (and, in one location, billed for brand drugs while dispensing generics). The defendants—registered pharmacists associated with five pharmacies in Michigan and Ohio—were convicted after trial of conspiracy to commit healthcare fraud and wire fraud, and one defendant was also convicted of a substantive healthcare-fraud count.

The opinion is most significant for its treatment of the Confrontation Clause when the government presents a fraud-loss expert who relies on large datasets and internal validation performed by other analysts. The court also addresses (i) a defendant’s right to present a complete defense, (ii) improper “expert” labeling in front of the jury, (iii) variance in conspiracy proof, (iv) lay interpretation of texts under Rule 701, (v) third-party subpoenas under Rule 17(c), (vi) juror impeachment limits, and (vii) loss, sophisticated-means, and restitution issues at sentencing.

II. Summary of the Opinion

The Sixth Circuit affirmed across the board. On the central constitutional question, it held that the defendants failed to show a Confrontation Clause violation from Qlarant director Johanna Sullivan’s testimony and invoice-review analysis: Sullivan testified to her own analysis and interpretation of billing and wholesaler records, and the defendants identified no testimonial hearsay statements she relayed from absent analysts.

The court further held:

  • Excluding evidence that the pharmacies also dispensed legitimate prescriptions did not violate the right to present a complete defense.
  • Excluding generalized evidence of PBM “bias” and industry distrust was proper as marginally relevant and speculative.
  • Referring to Sullivan as an “expert” in front of the jury was error under circuit precedent, but not reversible under plain-error review.
  • No variance: the evidence supported a single overarching conspiracy.
  • Allowing a cooperating co-conspirator (Abdallah) to interpret texts was permissible lay opinion under Rule 701.
  • Denying a broad Rule 17(c) subpoena for wholesalers’ records was proper under Nixon (lack of specificity and alternative access).
  • A juror’s post-verdict letter could not impeach the verdict.
  • Loss calculations based on the invoice review were reasonable; relevant conduct supported attributing certain pharmacy losses to Ghussin.
  • The sophisticated-means enhancement was properly applied based on falsified/engineered signature practices and audit evasion.
  • Restitution procedures were at most harmlessly flawed, and joint-and-several restitution was within MVRA discretion.

III. Analysis

A. Precedents Cited

1. Confrontation Clause and “surrogate” expert testimony

The court framed Confrontation Clause analysis through the modern “testimonial hearsay” line: Crawford v. Washington (founding-era exceptions; confrontation targets testimonial statements), Melendez-Diaz v. Massachusetts (the government cannot substitute certificates/reports for live testimony), and Bullcoming v. New Mexico (no “surrogate” analyst testifying to another’s test results). It then relied heavily on Smith v. Arizona, emphasizing two limits: the Clause reaches only (i) testimonial statements and (ii) hearsay offered for their truth.

The opinion also drew definitional guidance from Davis v. Washington (testimonial hearsay focus), Ohio v. Clark (primary purpose test), Michigan v. Bryant (consider all circumstances), and the “core class” description recited from Melendez-Diaz v. Massachusetts. For the hearsay component, it cited Anderson v. United States and Rule 801(c).

Critically, the court distinguished Sullivan from the prohibited “substitute” experts in Smith v. Arizona and Bullcoming v. New Mexico by characterizing her testimony as based on her personal review, quality checks, and interpretation of the datasets and outputs—rather than repeating another analyst’s conclusions. It underscored Melendez-Diaz v. Massachusetts footnote 1: confrontation does not require “everyone who laid hands on the evidence,” and gaps in foundational chains often go to weight, not admissibility.

The court’s emphasis on “weight vs. admissibility” and the adequacy of cross-examination was reinforced by Delaware v. Fensterer (opportunity for effective cross-examination despite limitations). It also reiterated the general purposes of confrontation through Delaware v. Van Arsdall, Davis v. Alaska, and Pointer v. Texas.

Notably, the panel cited United States v. Contreras (4th Cir.) for the proposition that an expert’s “exposure to testimonial hearsay” does not automatically bar admission of the expert’s “independent assessment,” aligning the Sixth Circuit’s resolution with a broader trend in handling modern data-driven investigations.

2. Right to present a complete defense

The court grounded the defense-right analysis in Holmes v. South Carolina (meaningful opportunity to present a defense), United States v. Reichert (de novo review), United States v. Odeh (right to exculpatory evidence to negate an element), and Taylor v. Illinois (the right is not unfettered). The constraint—trial courts may exclude confusing or prejudicial evidence under “well-established rules”—was supported by United States v. Reynolds and Rule 403, with the arbitrariness/disproportionality test drawn from United States v. Scheffer.

For why “legitimate prescriptions” were not a defense, the court relied on conspiracy/fraud framing decisions including United States v. Dimora, United States v. Ifediba (11th Cir.), United States v. Betro, and United States v. Daulton (unpublished) to reinforce that partial lawful conduct does not negate a scheme charge unless the indictment effectively alleges ceaseless criminality.

The PBM-bias exclusion was supported by relevance/Rule 403 principles and analogized with bias doctrine in Delaware v. Van Arsdall; the court also referenced McKee Foods Corp. v. BFP Inc. as context for claims of PBM anticompetitive conduct, while finding the defense theory too speculative without evidence of data manipulation.

3. “Expert” bolstering in front of the jury and plain error

The panel applied Sixth Circuit prohibitions on tendering/declaring “expert” status before the jury: United States v. Johnson (risk of judicial endorsement), United States v. Maya, and United States v. Campbell (proper process is foundation + testimony, not ceremonial qualification). Because defendants forfeited the challenge, the court applied plain-error review through Greer v. United States and the prejudice inquiry through Molina-Martinez v. United States. It found no reasonable probability of a different outcome, pointing to limiting instructions and substantial independent evidence, analogizing to United States v. Majors (unpublished) and citing United States v. Sibley (unpublished).

4. Variance and single vs. multiple conspiracies

The variance framework came from Stirone v. United States and United States v. Miller (trial must match grand jury charges), with the Sixth Circuit test articulated via United States v. Robinson, United States v. Guerrero, United States v. Siefert, and United States v. Hughes. The court relied on United States v. Beals to reject the idea that divisible ownership interests compel multiple conspiracies.

5. Evidence rulings: lay opinion on texts; subpoenas; juror impeachment

On Rule 701 lay interpretation, the court relied on United States v. Freeman (experiential observations), United States v. Reed (interpretations grounded in participation/personal knowledge), and reviewed under United States v. Fox. It distinguished United States v. Kaplan because Abdallah had personal experience with Abdelrazzaq.

On Rule 17(c), the decision applied United States v. Nixon (relevance, necessity, specificity, good faith), cited United States v. Theunick for review standards, and analogized to United States v. Llanez-Garcia in rejecting a “fishing expedition.”

On juror impeachment, it relied on the longstanding bar in McDonald v. Pless and Sixth Circuit authority United States v. Gonzales.

6. Sentencing and restitution framework

The court applied abuse-of-discretion sentencing review under Gall v. United States, with substantive reasonableness guided by United States v. Rayyan, United States v. Tristan-Madrigal, United States v. Boucher, and deference to weighing factors reaffirmed in United States v. Frei.

On loss, it relied on United States v. Betro (actual vs intended loss), United States v. Siefert, United States v. Wala (reasonable estimate; heavy burden), United States v. Matthews (clear error), and United States v. Washington (absence of contrary estimates). Relevant-conduct attribution was grounded in United States v. Donadeo and foreseeability in United States v. Kennedy.

On sophisticated means, the court relied on United States v. Simmerman (totality of conduct), United States v. Yousef (concealment sophistication can suffice), and United States v. Crosgrove (falsified records as indicative).

On restitution, the court cited statutory limits and review standards from United States v. Fike and United States v. Clay, the right to be present from United States v. Hayden, MVRA procedural flexibility from United States v. Vandeberg, overlap between guideline loss and restitution when both are actual loss from United States v. Simpson, precision limits from United States v. Kilpatrick, and joint-and-several discretion from United States v. Hunt.

B. Legal Reasoning

1. The core new clarification: Confrontation is not triggered by internal data-validation steps absent relayed testimonial statements

The opinion’s principal doctrinal move is to force a disciplined question: what “statement” was introduced for its truth? Borrowing directly from Smith v. Arizona, the court held the defendants failed at step one because they did not identify any specific out-of-court testimonial assertions conveyed by Sullivan. General testimony about how Qlarant “typically functioned” was treated as testimony from Sullivan’s personal knowledge of institutional practice—not the admission of a testimonial statement made in this case by a non-testifying analyst.

The court also treated the defense claim (“someone else verified completeness”) as an inferential leap: even if Sullivan said data “had already been verified,” she did not testify that the data were complete because an absent declarant told her so, nor did she repeat an absent declarant’s conclusion.

Finally, the court anchored the admissibility/weight boundary: challenges to completeness or integrity of underlying business records ordinarily affect weight, not constitutional admissibility—particularly where defendants stipulated to admissibility of the underlying records and had access to probe the analysis through cross-examination of the testifying analyst. This is consistent with Melendez-Diaz v. Massachusetts (chain-of-custody gaps go to weight) and Delaware v. Fensterer (constitution guarantees the opportunity for effective cross-examination, not the best possible one).

2. A practical boundary for modern fraud trials

The court’s reasoning reflects the reality that large-scale claims-and-invoice analyses are team-based and software-mediated. Without this boundary, any complex healthcare-fraud case using claim datasets could become constitutionally untriable unless the government called every data handler. The Sixth Circuit avoided that outcome by insisting on the testimonial-hearsay predicate: confrontation attaches to testimonial assertions; it does not attach to the mere fact that an expert relied on a workflow in which others performed clerical extraction, formatting, or preliminary validation.

3. Defense-right limits: “Some lawful conduct” does not negate a fraudulent scheme

In rejecting the “legitimate prescriptions” evidence, the court took the government’s theory as the reference point (per United States v. Dimora) and held that a partial lawful operation does not negate a fraud scheme, aligning with United States v. Ifediba and United States v. Betro. The court characterized the excluded evidence as confusing under Rule 403 because it risked leading jurors to treat “not 100% fraud” as a legal defense to conspiracy.

The PBM-bias theory failed for lack of a concrete nexus to the evidence: absent proof PBMs manipulated the defendants’ own claims data, the theory invited speculative leaps and risked sidetracking the trial into industry-policy disputes.

4. Error acknowledged but contained: “expert” endorsement and the prejudice requirement

The court candidly recognized the district court and government violated Sixth Circuit guidance by qualifying and labeling Sullivan as an “expert” before the jury (per United States v. Johnson, United States v. Maya, and United States v. Campbell). But under Greer v. United States and Molina-Martinez v. United States, the court required a showing of outcome prejudice. It found the limiting instructions and non-Sullivan evidence (cooperator testimony, written communications) broke any causal link.

5. Conspiracy proof: operational interdependence over formal corporate separateness

The variance analysis emphasized functional conspiracy indicators: common profit goal, shared tactics (labels, reversals avoided, forged signatures), overlapping participants and ownership, and profit-sharing checks. The court treated separate pharmacy entities as administrative containers, not separate conspiratorial universes—consistent with United States v. Hughes and United States v. Beals.

6. Sentencing: “reasonable estimate” loss and foreseeability-based relevant conduct

The Sixth Circuit endorsed invoice-review loss estimates as “available information” supporting a “reasonable estimate” under United States v. Wala and United States v. Siefert, stressing the defendant’s “heavy burden” to show computations fall outside permissible ranges. It also affirmed attribution of losses from a pharmacy Ghussin claimed he did not operate, because ownership and profit participation made the fraud reasonably foreseeable under United States v. Donadeo and United States v. Kennedy.

7. Restitution: MVRA procedural flexibility and harmless-error approach

While acknowledging procedural irregularity (judgment entered before full restitution briefing), the court relied on United States v. Vandeberg to hold the MVRA grants discretion in procedure and that any error was harmless because the core dispute (actual loss) was aired at sentencing and again in a motion to correct.

C. Impact

  • Confrontation Clause in data-heavy fraud cases: The decision signals that defendants must identify specific testimonial hearsay statements being conveyed—not merely point to team-based workflows. Experts who independently review and interpret records and software outputs can testify even if they did not personally collect or initially validate every data input.
  • Trial management of “legitimate conduct” evidence: The opinion reinforces that “we also did lawful work” is generally not relevant to negate a scheme-fraud conspiracy when the government does not claim 100% illegality, and trial courts may exclude it under Rule 403 without constitutional violation.
  • PBM bias theories: Broad “industry distrust” narratives will likely be excluded unless tethered to case-specific proof of data manipulation or investigative misconduct.
  • Expert-labeling practice in the Sixth Circuit: The court again admonishes against formal “expert qualification” ceremonies in front of jurors, but also shows that reversal will usually require a concrete prejudice showing.
  • Sentencing loss and restitution: Defendants challenging invoice-review loss must bring concrete counter-evidence. Speculation about “missing records” is unlikely to move either guideline loss or MVRA restitution.

IV. Complex Concepts Simplified

  • Confrontation Clause: You generally have the right to cross-examine the person who made a testimonial statement used against you. But that right is triggered by the admission of testimonial hearsay—not by the mere fact that an expert used data compiled by others.
  • Testimonial vs. non-testimonial: “Testimonial” usually means created primarily for use in prosecution (like affidavits or formal lab reports). Ordinary business records (e.g., invoices, claims data) are typically not testimonial.
  • Hearsay: An out-of-court statement offered to prove what it asserts. If it is not offered for truth (or not a “statement” at all), Confrontation concerns may not apply.
  • Rule 403 (confusing evidence): Even relevant evidence can be excluded if it would confuse jurors or distract from the real issues.
  • Variance in conspiracy cases: A “variance” occurs if the indictment alleges one conspiracy but the proof shows only separate conspiracies. Courts look for a common goal, shared methods, and overlapping players.
  • Loss amount (Guidelines): The court makes a “reasonable estimate,” not a perfect one, based on available information.
  • Restitution (MVRA): Restitution is limited to victims’ actual losses and can be joint-and-several among conspirators. MVRA procedures are flexible, and some procedural missteps can be harmless if the defendant had a meaningful chance to dispute the amount.

V. Conclusion

United States v. Kindy Ghussin establishes a practical and legally disciplined rule for modern healthcare-fraud prosecutions: the Confrontation Clause is not violated by an expert’s testimony that rests on independent analysis of underlying records and software-generated comparisons, even when other team members performed data extraction and preliminary validation, so long as the expert does not relay specific testimonial hearsay statements for their truth. The decision also reinforces trial courts’ authority to exclude “legitimate conduct” and broad policy-bias evidence as confusing or speculative, reiterates limits on juror impeachment, and confirms that invoice-review methodologies can support both guideline loss and MVRA restitution absent concrete rebuttal proof.