A bank can follow a payment into an account.

It may see the money leave again, a large cash withdrawal, or a business account used to pay what looks like a personal expense.

Whether the patient received care is a different question.

That distance between a financial record and a person's treatment runs through a new analysis by the U.S. Financial Crimes Enforcement Network, known as FinCEN. Banks can help identify transactions that deserve investigation. Usually, they cannot establish what happened in a clinic or a patient's home.

A large number with strict limits

Published on 9 September 2026, the analysis covers 5,702 Bank Secrecy Act reports submitted by 471 financial institutions in the year ending 28 February 2026.

Together, the filings describe approximately $17.5 billion in completed or attempted transactions potentially connected to healthcare fraud. That is the value of reported suspicious activity. It does not measure proven theft, confirmed losses or recovered money.

The same money can appear more than once. FinCEN warns that totals may count both incoming and outgoing transfers, movements between accounts, amended filings and repeated reports about the same subject. A filing can also include lawful transactions alongside suspected wrongdoing, and the reporting institution can make mistakes.

The total is therefore a starting point for investigation, rather than a single sum of illicit proceeds.

Reporting was concentrated. Depository institutions submitted about 89 per cent of the filings and accounted for almost 87 per cent of the amounts. The two largest banks by assets in this dataset submitted 29 per cent of the reports; one submitted around 1,100.

Those shares tell us who reported. Customer numbers, monitoring systems and reporting practices may all influence them. They cannot establish the corresponding share of suspected conduct at those banks.

FinCEN reports an average of approximately $3.3 million per filing and a median of about $600,000. Those figures need a qualification: dividing the stated $17.5 billion total by all 5,702 filings gives about $3.07 million, not $3.3 million. The passage on page 4 does not explain the difference. We retain the agency's figure as an attributed figure, rather than treating it as a calculation independently reproduced by this newsroom.

The reported average is well above the median, indicating that larger amounts pull it upwards. The analysis leaves open how much this reflects large suspected schemes, aggregated transactions or repeated counting.

A denominator changes the story

Another concentration has a particular explanation. Fifty-nine Puerto Rican financial cooperatives submitted 1,641 reports, also about 29 per cent of the dataset. Most concerned suspected manipulation of financial information for Medicaid eligibility. Nearly all described customers reportedly reducing their balances before eligibility meetings, obtaining balance letters and putting the money back later that day or within a few days. Around 67 per cent involved less than $15,000 in reported suspicious activity.

This is a pattern in reports about benefit eligibility. It provides no basis for assigning Puerto Rico, or cooperative customers generally, a higher fraud rate.

It also changes how another figure reads. Treasury's announcement puts home-health businesses at 20 per cent of all reports. The underlying analysis says more than 21 per cent, or nearly 32 per cent when the Puerto Rico eligibility reports are excluded.

The home-health cases have not increased in that comparison. Their share becomes larger because the group against which they are counted becomes smaller.

The payment is not the billing event

FinCEN says institutions generally cannot identify the precise conduct behind a suspicious payment, such as billing for care that never happened, billing twice or charging for unnecessary treatment. Detailed accounts tended to appear after charges or indictments became public, or after investigators told an institution about an existing case.

Information therefore moves both ways. A bank may give investigators an early signal; investigators may later help the bank understand what the signal meant.

FinCEN's advisory of 30 March 2026 describes the reporting duty in terms of knowledge or suspicion. It covers transactions believed to involve or conceal illegal funds, evade Bank Secrecy Act rules, lack an apparent lawful purpose or facilitate crime.

That threshold allows an investigation to begin before wrongdoing has been proved. A report must retain that preliminary status when its figures enter public debate.

Institutions must keep the report and supporting documents for five years and supply them when properly requested by FinCEN, law enforcement or a supervisory body. Retention makes later scrutiny possible; the documents still need to be tested.

The counter-question

For whom might this work?

The potential benefit is substantial. Investigators can connect accounts and providers across jurisdictions. Programme administrators can examine payment controls. Banks can learn which combinations of transactions merit attention. Patients and taxpayers may benefit when those signals lead to well-founded investigations while legitimate care continues.

What could it cost?

That takes people as well as data. After a filing, someone must compare the transactions with bills, programme rules and evidence of the services provided. Producing more reports alone cannot complete that work.

Care is needed in the other direction too. A legitimate small provider may resemble a published risk pattern. A beneficiary's ordinary account activity may attract suspicion. Additional scrutiny or delay could affect access to services, but FinCEN's analysis does not measure those possible costs.

Who may be left out?

Patients, workers and legitimate providers are largely outside the financial dataset's view. Differences in monitoring can also make conduct more visible at one institution than another. These limits matter when deciding what the reporting figures can support.

What evidence would change the conclusion?

A fuller assessment would connect reports to unique transactions and subjects, then follow what investigations found. Under appropriate confidentiality safeguards, it would examine substantiated claims, exonerations, confirmed losses, recoveries, false positives and any effects on access to care.

Evidence of earlier intervention and recovered funds, with few unjustified exclusions, would strengthen the case for monitoring. Extensive duplication, weak substantiation or defensive account closures would call for changes.

What remains unknown

The published figures do not yet provide that account. We cannot tell how much of the $17.5 billion is unique money, how many subjects were cleared or convicted, or how much loss was prevented or recovered.

There is a timing limit as well: reports ending on 28 February cannot show the effect of an advisory issued a month later.

To return to the patient, a bank record can help identify a payment worth examining. Establishing whether the care happened requires other records and other people's knowledge.

The $17.5 billion is important because institutions judged the activity worthy of attention. Its significance depends on what careful investigation can establish next.

That work can help distinguish crime, error and legitimate care, and show which suspicions the evidence supports. It is where a financial signal can become useful protection for the people the healthcare system serves.

Sources and limits

• FinCEN, Health Care Fraud: Trends in Bank Secrecy Act Data, September 2026. Used for the dataset, amounts, concentrations, Puerto Rico subset and methodological limits.
• U.S. Treasury release, 9 September 2026. Used for the release date and public framing.
• FinCEN healthcare-fraud advisory, 30 March 2026. Used for the reporting trigger and five-year retention requirement.

Claims of suspicion, allegation, investigation, adjudication, loss and recovery are kept separate. The report’s examples are filer reports and FinCEN analysis, not findings by this newsroom that a named person or business committed fraud.