Stream: analytics on FHIR
Topic: RonFHIR
Moritz Lehne (May 06 2019 at 14:17):
Dear FHIR experts,
I'm relatively new to FHIR and would like to better understand how it can be used for data analytics and statistics. I'm familiar with typical statistics tools like R, and I'm aware that there is the "R on FHIR" package for connecting to FHIR servers from R. However, there seems to be a lot of preprocessing required to transform the hierarchical, tree-like FHIR resources into the rectangular tables (R data frames) necessary for doing statistics in R. Wouldn't it be easier to work directly on the underlying databases (e.g. SQL) instead of doing the analysis on the FHIR resources? Or to put it differently: For what exactly do I need the "R on FHIR" package? Do I need this only if the data itself are stored as FHIR resources (instead of, say, in a relational database)?
Any help is highly appreciated. Thank you very much!
Alexander Zautke (May 06 2019 at 14:29):
CC: @Sander Laverman
Grahame Grieve (May 06 2019 at 15:37):
there's 2 different issues here:
- format
- structure
The structure of the data is tree-like. Because the data in real EHRs is a deep heirarachy, loaded with context and process information. If you get access to the EHR database, it will be dizzyingly deep. Typically, people doing analytics run some kind of ETL process that flattens out the data into some simpler structure that is more suitable for analysis (or for their particular analysis).
FHIR is a format that reflects pre-analysis structure - direct access to the operational systems. Of course it will be easier to access simpler databases prepared by ETL - in which case you wouldn't need FHIR. But there are also uses for accessing the data directly, and that's what RonFHIR is good for
Moritz Lehne (May 07 2019 at 11:57):
Ok, that makes things a bit clearer, thank you very much for the swift reply!
Last updated: Apr 12 2022 at 19:14 UTC
