Read our white paper on AI forgeries in 2026
New AI fake detection model. Try it now
  1. Home
  2. Industries
  3. Healthcare
Home care, pharmacies, health insurers

The healthcare document processing
that reads handwriting and catches the fakes

Healthcare document processing, from reading to checking: prescriptions, care forms and invoices scored in seconds.

  • 94.7% on handwritten prescriptions
  • Fraud score from 0 to 100
  • SOC 2 audited
4.6 / 5 on Trustpilot
Reimbursement claim #2207Example
  • PrescriptionPrescriber, date and dosage read, handwriting includedCompliant
  • Optician invoicePDF re edited after creation, total alteredAlert
  • Insurer statementBase amount differs from the invoiceReview
  • Proof of entitlementName and cover period consistent with the claimCompliant

Illustrative example. Every document gets a fraud score from 0 to 100 and a justification, test by test.

The document checker used by insurers, banks and large companies

4.8 / 5
In short

What is healthcare document processing, from medical records processing to claims?

Healthcare document processing reads prescriptions, medical records, care forms, invoices and claims automatically, the prescriber's handwriting included, and turns them into fields your software can use. The useful version also checks each document: amounts recalculated, file and image examined for edits, the whole claim compared document by document. OCR here means optical character recognition, not the US Office for Civil Rights.

  1. Read the handwriting94.7% accuracy published on handwritten prescriptions, prescriber's corrections included.
  2. RecalculateBase amount, insurer's share, patient's share: an invoice or a benefit statement is recalculated line by line.
  3. Cross check the documentsSame patient, same dates, same amounts across prescription, invoice, statement and proof of entitlement.
  4. Read the fileProducer software, successive edits, modification dates of an invoice or certificate PDF.
  5. Spot duplicatesContent based duplicate detection surfaces a document submitted twice.
  6. Examine the imageEdited or AI generated document photo: K-Detect spots the altered areas.
Documents

The documents checked in a healthcare file

Method

Healthcare fraud detection, document by document

France's public health insurer detected and stopped €723 M of fraud in 2025, and French complementary insurers identified €112 M, with false invoices and statements first and document fraud right behind (ALFA). The NHS Counter Fraud Authority estimates £1.346 bn of NHS funding vulnerable to fraud in its 2025 assessment, and the US Department of Justice charged 324 defendants for $14.6 bn of intended losses in its June 2025 takedown.

Optical, dental, hearing: the invoice that inflates the reimbursement

Optical accounts for 53% of the health fraud identified by French complementary insurers, ahead of dental at 15% and hearing aids at 8% (ALFA, 2026). Equipment is expensive, invoices are easy to falsify, and direct billing pays fast. The techniques recorded range from the fake optician's invoice to the fake insurer statement, through the inflated amount and the purchase that never happened (Magnolia, 2026). On this kind of document, three checks do most of the work: recalculate the invoice, verify that the file really comes from the optician's software, and compare invoice, estimate, prescription and statement.

The fabricated or recycled prescription

In 2025, French pharmacists reported more than 10,000 suspicious prescriptions through the Asafo-Pharma alert system, and 75% of the reports proved founded (Assurance Maladie, 2026). Recycling is the other scheme. In a case cited by the French health insurer, an orthopaedic equipment supplier attached near identical prescriptions to its invoices, same signatures, same stamps, only the dates changed. The first ones were genuine, the following ones fake, for more than €280,000 of loss. A comparison of the documents from the same supplier sees it at once; a handler who opens files one by one, rarely.

The fake sick note and the fake payslip

Fake sick notes, sold mostly on social media, represented €34 M detected and stopped in France in 2025, up 15%. A secure form, mandatory for paper sick notes since 1 September 2025, and online transmission closed part of the door on the public insurer's side. But employers and income protection insurers still receive copies. The same insurer cites fake payslips supplied to collect sick pay, sometimes in the name of fictitious companies. In the UK, the ABI's 2025 figures include a traveller who filed more than £300,000 of bogus medical emergency claims with forged medical paperwork (Insurance Business, October 2026). A copy on an old, unsecured form becomes a signal in itself.

Checks

Healthcare OCR, then checks: where Koncile fits in the flow

No need for two tools, one to read and one to check: the same analysis extracts the data and returns the score.

Prescription processing: keep the prescriber's correction

Koncile reads the prescriber and their registration number, the medicines, dosages and directions, by hand included, with 94.7% accuracy published on handwritten prescriptions. The lesson learnt at HealthNet Homecare, a UK home care provider: when the doctor crosses out a printed value and writes the right one next to it, the correction is what must be kept. The completeness check follows, signature, date, dosage, then the data goes to your dispensing or practice software through the API. An incomplete prescription comes out as an exception with the missing field named, instead of being retyped by hand.

Intelligent document processing in healthcare claims: the whole file, not the document alone

On an optical claim, the insurer receives the prescription, the estimate, the invoice, the statement and sometimes a proof of entitlement. Koncile reads each document and checks that they tell the same story: same patient, dates in the right order, amounts that follow from estimate to invoice to statement. Each document gets a fraud score from 0 to 100 and the reason test by test. Clean files go through, the others reach the handler with the gap already pointed out, for example a base amount on the statement that does not match the invoice.

Business rules written in plain language

Every organisation has its rules, and they change. Rather than coding them, you write them in plain language inside the document's template: the invoice date must follow the prescription date, the patient on the invoice is the patient on the prescription, the dosage must be consistent with the prescribed medicine. Koncile applies these checks to every document and adds them to the report. You can start with optical invoices, where fraud concentrates, then extend to the rest of the flow.

Case studies

Health data: the framework to respect, and what we do not do

A healthcare document is sensitive data, and public controls are tightening. Three points matter.

GDPR: sensitive data by nature

Data concerning health belongs to the special categories of GDPR (article 9), with reinforced safeguards. A checking tool must therefore limit what it keeps. At Koncile, documents are hosted and analyzed securely, encrypted, deleted after analysis and never used to train a model. Koncile is audited SOC 2 Type II. The same regulation frames automated decisions (article 22): a fraud score remains decision support, and refusing a reimbursement is the job of a handler who has read the anomalies. For US or UK specific requirements, ask us for the security documentation before you commit.

Public locks change what you need to check

In France, the public insurer is closing doors one by one: the secure sick note form is mandatory since 1 September 2025 and unsecured paper notes are rejected; almost one GP prescription in two is now electronic, and some sensitive medicines require a secure prescription since 1 March 2025; in hearing aids, billing with the health card went from 45% to over 95% of cases. For your checks, a document that bypasses these locks, a paper copy or an old form, deserves a second look. The pattern is the same wherever electronic prescribing grows: the paper copy becomes the exception, and the exception is where to look.

What Koncile does not check

Medical relevance. Koncile verifies that a prescription is complete, consistent and that its file has not been edited, not that the treatment was right. It makes no diagnosis, does not know the patient's cover and cannot see who hands over a health card. A high score is a signal to verify, never an accusation. We publish each detection rate with what we have not measured yet, and we have no healthcare specific fraud rate to publish today.

Data

What the 2025 fraud reports taught us

Reports published in 2025 and 2026 in France, the UK and the US say where fraud concentrates, and therefore which documents to check first.

€723 Mof fraud detected and stopped by France's public health insurer in 2025, up 15%
£1.346 bnof NHS funding vulnerable to fraud, bribery and corruption (NHS Counter Fraud Authority, 2025)
$14.6 bnof intended losses in the US health care fraud takedown of June 2025, 324 defendants
75%of the 10,000 suspicious prescriptions reported by French pharmacists in 2025 were founded

Insured individuals represent 53% of the fraud cases detected by the French public insurer, but only 16% of the amounts. Health professionals, health centres included, weigh close to three quarters of the loss for 28% of cases, with €138 M for health centres alone. The large amounts go through invoices, prescriptions and statements issued by professionals: documents that can be recalculated and compared. In the US, 96 of the 324 defendants of the 2025 takedown were licensed medical professionals. Across the 1.6 million documents Koncile processed, all sectors combined, 5% showed a manipulation or a transformed file; we have no healthcare specific rate yet, and we say so.

Source: Assurance Maladie, press kit of 16 April 2026; ALFA, 2025 activity report (July 2026); NHS Counter Fraud Authority, Strategic Intelligence Assessment 2025; HHS OIG, 2025 National Health Care Fraud Takedown; Koncile data 2026

Clients

They read and check their healthcare documents with Koncile

See all our customers

HealthNet Homecare

Home care in the UK, prescriptions read with the prescriber's corrections. Sean Hickey: "The success rate is close to 100%, even on an exam code that is not present on the document."

Your flow, in a demo

Health insurer, pharmacy, laboratory or home care provider: bring ten real documents, handwritten ones included. We show the extraction and the fraud score on your own files.

Frequently asked questions

Short answers, with figures when we have them.

Talk to us
  1. Two things share the acronym. OCR, optical character recognition, is the technology that turns a scanned or photographed document into text and fields. In the US, OCR is also the Office for Civil Rights of the Department of Health and Human Services, which enforces HIPAA. This page is about the first one: reading prescriptions, care forms, invoices and claims, then checking them.

  2. Yes, with a model trained on prescriptions, which reads the prescriber, the registration number, the medicines, dosages and directions. Koncile publishes 94.7% accuracy on handwritten prescriptions. What makes the difference: keeping the correction written by the doctor, not the printed value they crossed out, and returning an incomplete prescription as an exception rather than a guess.

  3. Yes, and they report. In France, pharmacists flagged more than 10,000 suspicious prescriptions in 2025 through the Asafo-Pharma system, and 75% proved founded. A pharmacist checks the mandatory mentions, the prescriber's number, the consistency between medicine, dosage and directions, and the date. What the counter cannot see, the producer software of a PDF and its editing traces, a document check adds.

  4. Look at the medium first: in France, some medicines require a secure prescription form since March 2025. Then check the mandatory mentions, the prescriber's registration number, the consistency between medicine, dosage and directions, and the date. On a PDF, the software that produced it and the modification traces complete the examination. Compare with the other prescriptions from the same prescriber or supplier: recycled ones share signatures and stamps.

  5. Check the form version, the dates, the prescriber and the file. In France, a paper sick note must use the secure form since 1 September 2025, so a copy on an old form received by an employer or an insurer is suspect in itself. A certificate edited in a PDF editor keeps the trace of it. Koncile returns the producer software, the modification dates and the edited areas, with a score from 0 to 100.

  6. In France. Documents are encrypted, analysed on secure servers, deleted after analysis unless you keep them in your history, and never shared or used to train a model. Koncile is audited SOC 2 Type II and complies with GDPR, including article 9 on health data.

  7. Medical relevance. Koncile checks that a prescription is complete and consistent and that its file has not been edited, not that the treatment was the right one. It makes no diagnosis, does not know the patient's cover and cannot see who presents a health card. A high score stays a signal to verify, never an accusation.

  8. 94.7% on handwritten prescriptions, the figure we publish. HealthNet Homecare, a UK home care provider, reports a success rate close to 100% on its prescriptions, including on an exam code that is not present on the document. On fraud, we have no healthcare specific detection rate yet: across 1.6 million documents of all types, 5% showed a manipulation or a transformed file. We publish each rate with what we have not measured.

Further reading

  • Assurance Maladie, press kit on the 2025 anti fraud results, 16 April 2026 · assurance-maladie.ameli.fr
  • ameli.fr, €723 million of fraud detected and stopped in 2025, April 2026 · ameli.fr
  • Magnolia, optical as the first health fraud for complementary insurers according to ALFA, 22 July 2026 · magnolia.fr
  • MoneyVox, top 3 frauds by line according to ALFA, 16 July 2026 · moneyvox.fr
  • NHS Counter Fraud Authority, Strategic Intelligence Assessment 2025, 25 September 2025 · cfa.nhs.uk
  • HHS Office of Inspector General, 2025 National Health Care Fraud Takedown · oig.hhs.gov
  • Insurance Business UK, ABI 2025 fraud figures, 6 October 2026 · insurancebusinessmag.com
  • GDPR, Regulation (EU) 2016/679, articles 9 and 22 · eur-lex.europa.eu

Try Koncile's fraud detection on your own documents

Upload your bank statements, pay stubs or IDs: Koncile extracts the data and flags fakes in seconds.

Talk to an expert