FHIR to OMOP Desk
Choose which simulated trusts train together and see what each one gains, with no patient record leaving a trust. Then translate a health record into the OMOP research format. All patients are synthetic.
Choose the trusts to see what they gain.
Waking the model
The demo sleeps when nobody is using it. Your request runs as soon as the machine is up.
✓ machine scheduled · lhr
· loading results for 31 groups of trusts
· loading the code map
0.0 s elapsedThe API isn't responding
Retry now, or read the model card while the service recovers.
GET /healthOne trust on its own isn't a federation. Switch on another trust to see what sharing a model adds.
Early rounds score badly because the weights are still being averaged into shape. Stop too soon and federated learning looks worse than it is.
Pick an example or paste a FHIR record, then translate it.
Waking the model
The demo sleeps when nobody is using it. Your record is queued and runs as soon as the machine is up.
✓ machine scheduled · lhr
· loading the code map
0.0 s elapsedThe API isn't responding
Your record is kept. Retry now, or read the model card while the service recovers.
GET /healthThe same model, three ways
One logistic regression trained on one trust's patients, on everyone's pooled data, and federated across trusts. Only where the data sits changes.
Loading the model card…