De-identified Clinical Notes as AI training data
Clinical notes record how doctors reason: what the patient reported, what was found, what was suspected and what was done next. That narrative, written in real shorthand, is something medical AI developers find hard to source responsibly. Momentra only considers clinical notes that have been properly de-identified; identifiable patient records are never part of a partnership.
Can De-identified Clinical Notes be used as AI training data?
De-identified clinical notes can be used as AI training data; identifiable patient records are never part of a partnership. Progress notes in SOAP format, discharge summaries and radiology or pathology report text are the most valuable content. De-identification, such as HIPAA Safe Harbor or Expert Determination, is agreed and applied before transfer, and identifiers inside free text are removed.
No files needed upfront. We sign an NDA before reviewing any sample.
What can be included
You choose the scope. These are the parts of De-identified Clinical Notes that are typically most useful.
- ✓Progress notes in SOAP or similar formats
- ✓Discharge summaries and transfer notes
- ✓Radiology and pathology report text
- ✓Operative and procedure notes
- ✓Referral and consultation letters
- ✓Diagnosis codes (ICD-10, SNOMED CT) and medication lists linked to the notes they came from
- ✓Nursing notes and care plans
Why AI labs value it
Clinical reasoning in context
The assessment and plan section shows how a clinician moved from findings to a decision. That reasoning is what clinical AI tools need to learn and be tested on.
Narrative linked to codes
When free text is linked to the diagnosis and procedure codes assigned for that encounter, the codes act as labels. This supports coding assistance and summarisation models.
Real abbreviations and style
Notes are full of local abbreviations, mixed languages and incomplete sentences. Models trained only on textbooks struggle with that, which is why real de-identified notes matter.
How it is scoped, exported and anonymized
De-identify before anything is transferred
Only de-identified notes are in scope. Under HIPAA this can mean the Safe Harbor method, removing the 18 listed identifiers, or an Expert Determination. In India, identifiers such as UHIDs, ABHA numbers and Aadhaar numbers are removed too. The method is agreed in writing before transfer.
Handle free text carefully
Names, dates, places and family details appear inside sentences, not just in header fields. Purpose-built tools detect these, dates are shifted consistently, and samples are reviewed for anything left behind.
Follow your own governance
Health data is sensitive under India's DPDP Act 2023 and special category data under GDPR. Hospitals often route requests like this through an ethics committee or data governance board, and rare conditions or small patient groups may need to be excluded to avoid re-identification.
How a partnership begins
Share Basic Information
Share only your contact, company, industry, and partnership interests—no raw data required.
Initial Value Assessment
We discuss the data type, completeness, scarcity, and the tasks it could support.
Define the NDA and Boundaries
Before reviewing any sample, we define confidentiality, permitted use, retention, and security requirements.
Receive an Offer and Begin the Partnership
We provide a purchase proposal and contract, then support de-identification and secure data transfer.
Frequently asked questions
- Can we send identifiable notes and have you de-identify them?
- No. Identifiable patient records are not part of a partnership. De-identification is planned with you and carried out before transfer, using an agreed method.
- Do we need patient consent?
- It depends on the jurisdiction, how the data was collected and the de-identification standard used. For example, HIPAA does not treat properly de-identified data as protected health information. Your legal and ethics advisers should confirm this; nothing here is legal advice.
- We are a small clinic. Is our data worth assessing?
- Possibly, especially in specialties or languages that are under-represented. Smaller datasets need extra care, because fewer records can make re-identification easier.
- Are structured EHR fields included as well?
- Codes, medications and lab values linked to notes can add value. They are de-identified to the same standard as the text.
Related
Let's talk about your De-identified Clinical Notes.
Share only the essential details. No files are required at this stage.
Contact the Momentra team