Stream: analytics on FHIR
Topic: Pathling
John Grimes (Jan 10 2020 at 07:48):
Hi all,
Now that I have some documentation and a public demonstration server up and running, I just wanted to share this: https://pathling.app
The demo server is: https://server.pathling.app/fhir
It's an experimental FHIR analytics server implementation that I have been working on, aimed at enabling analytics app development and augmenting health care analytics workflow. It's still a work in progress, but hopefully the documentation helps to articulate the ideas and direction.
I'd love to hear your thoughts.
Jose Costa Teixeira (Jan 10 2020 at 11:12):
- Closed source?
Jose Costa Teixeira (Jan 10 2020 at 11:13):
- Roadmap link gives a 404
Jose Costa Teixeira (Jan 10 2020 at 11:15):
- If i understand, you created a custom operation $aggregate and a graph-like syntax?
Jose Costa Teixeira (Jan 10 2020 at 11:16):
3.1 advantages over graphql?
Jose Costa Teixeira (Jan 10 2020 at 11:17):
- Anyway, this is cool. We are looking at analytics and will play with it.
Jose Costa Teixeira (Jan 10 2020 at 11:17):
4.1 any way to try with our own data?
Jose Costa Teixeira (Jan 10 2020 at 11:17):
thanks for the initiaitve!
Théo Ryffel (Feb 10 2020 at 12:43):
Roadmap is here I think :) https://pathling.app/docs/roadmap/
John Grimes (Feb 14 2020 at 02:27):
Thanks for the feedback @Jose Costa Teixeira, sorry for the delayed reply - I have been away on leave.
RE: open sourcing the code - we are super keen to do this in the near future, but have not quite got there yet.
I've fixed the Roadmap link, thanks.
Yes, $aggregate is a custom operation, as is the upcoming $summarise operation. I did not come up with the syntax, this already existed in the form of FHIRPath, which has been in use within the FHIR specification itself for quite some time.
I have done a lot of analysis around GraphQL - it is technically possible to implement these operations using a GraphQL schema, but in my opinion the resulting solution would be quite messy and not really very usable. GraphQL is primarily concerned with selecting and shaping data, rather than transforming and aggregating it. It is possible to define functions and pass arguments, but there are other complications such as the polymorphic nature of elements and resource references in FHIR.
If you would like to try it out with your own data, just send me a PM and I will find a way to give you access to the Docker image.
Last updated: Apr 12 2022 at 19:14 UTC
