Home / Data types / Healthcare
De-identified healthcare data for AI training
Datadhara sources clinical notes, prescriptions, lab reports and discharge summaries directly from clinics, hospitals and diagnostic labs. Patient identifiers are removed before data leaves the source, and every record is licensed for AI training under a written agreement.
What we can source
| Record type | Format | Typical labels |
|---|---|---|
| Outpatient notes | Free text, scanned forms | Specialty, diagnosis codes, medications |
| Prescriptions | Handwritten and printed scans | Drug, dose, frequency, duration |
| Lab reports | PDF + structured values | Test, value, unit, reference range, flag |
| Discharge summaries | Text, PDF | Sections, diagnoses, procedures |
We collect to your spec, so volumes, languages and labels are agreed per project. Tell us what you need.
Example record
{
"test": "HbA1c",
"value": 6.1, "unit": "%",
"ref_range": "4.0–5.6",
"flag": "high",
"patient_id": "[REDACTED]",
"consent_ref": "src-0311 / v1"
}
Values are illustrative. Real deliveries follow the schema agreed in your spec.
How personal information is removed
Patient names, record numbers, phone numbers, addresses and ID numbers are removed. Exact dates are shifted or coarsened, and rare conditions in small clinics are reviewed so they can't single a patient out. We use the HIPAA Safe Harbor identifier list as a baseline checklist. See the full methodology.
What teams use it for
- Clinical NLP and summarisation
- Handwriting and form understanding for prescriptions
- Lab-result extraction and structuring
- Evaluation sets for medical assistants
How licensing works
Every source signs a written agreement that covers AI training. You start with a small pilot and a dataset card, and if the data works for you we agree volume, timeline and license terms (exclusive or non-exclusive) in writing.
Questions
Is the healthcare data de-identified?
Yes. Patient identifiers are removed before data leaves the clinic, and the method is documented in the dataset card.
Can I get data for a specific specialty?
Usually, yes. Tell us the specialty, document types and volume and we'll confirm what we can source before scoping a pilot.
Tell us what your models are missing
A few lines is enough: the domain, the format, and roughly how much.
Request data →