A Practical Guide to using AI to transform Data to openEHR

New to openEHR?
, ,
1

Hello openEHR Community!

I am currently in the process of finishing up a thesis about transforming clinical data to openEHR. One of the goals was to find out where and in what form AI (or more specifically, LLMs) can improve the process of doing so.

During my own time of getting into openEHR, I felt that it would definitely be nice to have a small practical guideline that not only has all important links in one place but also gives me a rundown on how to transform the data itself. Because of this, I created this guideline using my own findings: https://drive.google.com/file/d/1WoJqV8duMkV9MbphRCPXFWwKnEpxsjcu/view?usp=sharing (As I am unable to upload the PDF directly)

The guide covers the practical transformation process from start to finish. It also includes recommendations on where LLMs can be useful during these steps.

I would really appreciate some feedback from people with more openEHR experience:

  • Does this workflow reflect how you would approach such a transformation in practice?
  • Is there anything important that is missing or potentially misleading?
  • Are there parts where you would recommend a different approach?

Any feedback would be greatly appreciated!

2

Thanks for mentioning the openEHR-assistant MCP server - I went down probably to same trajectory as you did, It is the result of my own experiences and views combine d to formal specifications in informal guides available in the community.

One thing that I stumble upon in your doc is the workflow image in page 2 is focusing too much on Templates, where as sometimes you need to model aspects that might not covered yet by the archetype sets. SO the process should also involve discovering, using or creating new archetypes .