Not legal advice; educational only. Cases are described according to their public posture; civil settlements resolve allegations without an admission of liability unless a court found otherwise, and dismissed cases are described as such.
Most case studies in this section are about a method that worked. This one is about a method that failed twice in federal court and then made history anyway, which makes it the most useful story of all, because it shows exactly where the line sits between a data-mining theory that gets dismissed and one the government is willing to stand behind.
Integra Med Analytics was not a disgruntled employee or a hospital insider. It was a company built for one purpose: run statistical analyses against public Medicare claims data, find providers coding high-value diagnoses far more often than their peers, and file False Claims Act (FCA) suits alleging upcoding. It is the purest test we have of whether an outsider can win on data alone. The answer it produced is a careful “not by itself”, and the nuance is the whole lesson.
The model
Hospitals are paid more by Medicare when a patient’s stay carries certain secondary diagnoses, complications and comorbidities that raise the severity, and the reimbursement, of a case. Integra’s thesis was that some hospitals were coding those high-value secondary diagnoses (conditions like severe malnutrition, encephalopathy, and respiratory failure) at rates well above the national average, and that the gap was evidence of systematic upcoding. The analysis was sophisticated, the datasets were public, and the targets were large hospital systems. On the ACFE (Association of Certified Fraud Examiners) Fraud Tree, this is the false-billing corner of asset misappropriation, exactly the typology a claims-data screen is built to surface.
The two losses
It did not work the way Integra hoped. Its case against Baylor Scott & White was dismissed, and the Fifth Circuit affirmed in 2020. Its case against Providence Health was ordered dismissed by the Ninth Circuit in 2021.
Both courts said a version of the same thing, and it is the single most important sentence a data relator can internalize: a statistical outlier is equally consistent with fraud and with a hospital that is simply better, and earlier, at lawful, well-documented coding. CMS (the Centers for Medicare & Medicaid Services) had actively encouraged hospitals to capture these severity codes accurately. So the high rates Integra flagged had an obvious, lawful explanation, and Integra had not pleaded facts excluding it. (Both rulings, notably, are unpublished and non-precedential, influential, but not binding law.) We unpack that doctrine in Why Data Alone Usually Loses; here, the point is what Integra did next.
The history-making turn
Integra kept going, and it refined its targets. In a case against a group of New York skilled-nursing facilities, alleging schemes to keep patients on higher, more lucrative levels of rehabilitation therapy than they needed, the Justice Department intervened in June 2021, joining a complaint that alleged roughly \$129 million in fraud. It was reported as the first time the government had ever intervened in a case originated purely by a data-analytics firm.
Why that case and not the others? The honest answer, drawn from the public records, is that the nursing-home theory tied the data to a tighter, more particularized story of how the billing was manipulated, in a setting (therapy-minute thresholds) where the lawful explanation is far harder to assert. The government did not bless data-mining in the abstract; it backed a data case that had crossed from anomaly into specific, knowing falsity.
| Integra Med Analytics matter | Public posture |
|---|---|
| v. Baylor Scott & White (5th Cir.) | Dismissed; affirmed 2020 (unpublished) |
| v. Providence Health (9th Cir.) | Ordered dismissed 2021 (unpublished) |
| NY skilled-nursing facilities | DOJ (Department of Justice) intervened 2021; complaint alleged ~\$129M |
The lesson
Read as a single arc, two public dismissals, then the first-ever government intervention in a pure data firm’s case, Integra is the most honest teacher in this section. Rejection, refinement, and then validation is the realistic shape of the work, not a single clean swing. And the difference between the losses and the win is the difference this entire series keeps returning to: a statistic that is “consistent with both” fraud and good medicine is a lead, and the cases that survive are the ones that pair the data with particular facts of a knowing false claim.
That is the bridge between finding the outlier and bringing the case. The screen is where Integra started. The forensic work after the screen is where it eventually won.
What Integra teaches the next data miner
Integra’s arc landed at the exact moment the government decided it wanted more cases like it. When the Justice Department launched its FOCUS initiative (the “Fraud Oversight through Careful Use of Statistics” initiative) in 2026 to recruit data-miner relators, it said it would prioritize data miners who pair analytical rigor with “familiarity with program rules” and “legally sufficient allegations”, the very things Integra’s dismissed cases lacked and its winning case supplied. That is the lesson Integra paid to learn: a statistic becomes a case when it is tied tightly to a documented program rule, in a setting where the lawful explanation cannot survive the facts.
This is also why the ACFE‘s discipline of professional skepticism is a relator’s asset, not a constraint. The Report to the Nations finds that tips, not analytics, catch most occupational fraud, roughly three times as often as the next method, because a person who can explain how and why a scheme works supplies exactly the particularized facts a statistic cannot.
| The data-miner reality, in numbers | Figure |
|---|---|
| Qui tam suits filed in FY2025 (record) | 1,297 |
| FY2025 FCA recoveries (record) | >\$6.8 billion |
| FY2025 recoveries from whistleblower-initiated cases | >\$5.3 billion |
| Integra cases won on statistics alone | 0 |
Sources: DOJ FOCUS initiative (2026); DOJ FY2025 recoveries.
The last row is the one to remember. The firm that pioneered outside data-mining never prevailed on statistics alone, it won when the statistics pointed at a rule, in a setting where “they’re just better at lawful billing” could not survive contact with the facts.
By Noah Green CPA CFE, for Sheepdog Prosperity Partners. Educational only; not legal advice. Court decisions are summarized for a general audience.
Primary sources: Integra v. Baylor Scott & White (5th Cir. 2020) · Integra v. Providence (9th Cir. 2021) · Constantine Cannon, DOJ intervenes in Integra nursing-home case (2021) · 31 U.S.C. § 3730 · ACFE, Fraud Tree