Why MedX trains with synthetic patient data, never real records

MedX Services · 13 October 2026 · 5 min read

Training people to work with patient records creates an obvious tension. The work is about real patients, but training must never expose real patient information. MedX resolves this by designing all of its training around synthetic data: fictional patients, referrals and pathways that behave like the real thing but describe no real person.

Why real data has no place in training

Patient records are among the most sensitive information any organisation holds. Under UK data protection law, information about health is special category data, which requires extra protection and a specific legal condition before it can be used. Health and care confidentiality rules, including the Caldicott Principles, set a clear expectation that patient information is used only where it is genuinely needed.

Training rarely meets that test, because the same learning can be achieved without identifying anyone. Even "anonymised" extracts can carry a risk of re-identification when combined with other information. Synthetic data removes that risk at the source.

What makes synthetic data safe

Fictional data is not automatically safe. A made-up record can still cause harm if it looks like a real one or reaches a real person. These are the standards MedX applies to its training data:

• Identifiers that cannot be confused with NHS numbers. A real NHS number is 10 digits long, and the last digit is a check digit calculated from the other nine. MedX training records use a separate identifier format, prefixed MX-, so no training record can be mistaken for, or accidentally match, a real patient.
• Contact details reserved for fiction. Ofcom reserves specific UK telephone number ranges for use in drama, such as 07700 900000 to 07700 900999, and some internet domains, such as example.com, are reserved for documentation. Using these ranges means a training phone number or email address can never reach a real person.
• Invented names and addresses that are not drawn from real people or real patient lists.
• Realistic patterns, not real records. Scenarios recreate the shapes of real work, such as long waits, duplicate referrals and missing outcomes, without copying any real case.

Realistic enough to learn from

Synthetic data only works for training if it is believable. MedX scenarios are designed from practical experience of pathway management. They include the inconsistencies learners will meet in practice: a referral date that does not match the letter, an appointment outcome left blank, a patient who appears twice. Learners practise finding and resolving these problems with nothing real at stake.

Read more about how we handle personal information on our Data protection and security page.