What the knowledgebase holds about you
Nothing. Not a thinned-out version, not a pseudonymous one, not a token that could be joined back to you later. There is no patient record in the knowledgebase at all, which is a different and much stronger claim than promising to look after one carefully.
If you create an account to control who can read your EHR records, that account holds your name, your user id and the records you linked to it. You can see all of it under My account. None of it ever reaches the knowledgebase.
Knowledgebase records about you
0
Accounts needed to read
0
Identifiers removed in this demo
112
What gets removed before anything is shared
Counted by the de-identification engine across every case document in this build
| Identifier | Removed |
|---|---|
| Date element | 54 |
| Patient name | 27 |
| Medical record number | 18 |
| Street address | 6 |
| Telephone number | 3 |
| Email address | 3 |
| Account or device identifier | 1 |
| Total | 112 |
These numbers are produced by running the redaction engine right now, in your browser, over the demonstration documents. They are not written into this page.
What is deliberately kept
The treating clinician's name
A clinician is not a patient. Attribution is what lets other clinicians judge whether a contribution is worth trusting, and it is what makes peer review possible at all.
The clinical content
Symptoms, findings, what was prescribed, what happened next. This is the knowledge, and removing it would leave nothing worth sharing.
An age band, not an age
'65 to 74' rather than '68'. Anyone over 89 is grouped into a single band, because a very high age is identifying on its own.
The year, not the date
A date of service becomes a year. Dates are one of the easiest routes back to a person.
Why a group and not a person
Removing a name is not enough. A rare combination of age, sex, location and diagnoses can identify someone even with every obvious identifier stripped out. So before anything is published, the cohort it describes is checked against the rest of the network.
If the group is too small to hide in, it is widened - a broader age band, one fewer condition - until it is not. If it still cannot be widened enough, it is not published at all.
The case references you see, like CASE-7F2A9C, are random. There is no table anywhere that maps one back to a person - not held by us, not held by the contributing organization, not recoverable by anyone. That is deliberate: a lookup table would be the single most valuable thing to steal, so it does not exist.
And in this browser?
This demonstration stores which insights you saved and which topics you clicked, in this browser only, so a page reload does not lose your place. It is never transmitted. Clearing your browser data removes it completely.
Currently held locally: 0 saved insights, 0 session actions, 6 documents marked processed.