Stream: research
Topic: Short interview for Internship
Inèz Sikkel (Jan 29 2019 at 13:19):
Hi Everyone! My name is Inèz Sikkel and I'm a Medical Informatics student. For my internship I'm doing a small study on how HL7 FHIR can be used to semantically map clinical (research) data. I would like to Interview experts who have already tried to map data with FHIR. I am specifically interested in how FHIR was implemented (what servers, resources were used). Where the mapping was used for and if you succeeded.
The questions are the following:
1. How did you use FHIR to map clinical data?
2. In what setting did you use FHIR? Data extraction, consents, recruitment, data standardization, device integration
3. How was FHIR implemented? Has your organization a FHIR server integrated?
4. What FHIR resources did you use?
5. Is the FHIR mapping still used?
6. What would you recommend for a new organization to use FHIR?
It shouldn't take more than 15-20 minutes. Is anyone available for a phone call/zoom call this week?
Lloyd McKenzie (Apr 17 2019 at 19:05):
@Nagesh Bashyam @Geoff Low
Geoff Low (Apr 23 2019 at 09:05):
Hey, sorry for missing this.
The primary use case is to access the subject data for incorporation in Case Report Forms (CRFs) - the primary unit of work for clinical trial data capture. The CRFs are (unfortunately) electronic representations of paper forms that would be entered by the Investigator for a study participant to record the information required per protocol. The broad requirement is to gather resources for a Patient (subsetted to cover the information required for a study) and push this data into the CRFs. You should look at the ResearchStudy and ResearchSubject resources.
In terms of the datan, on the clinical side we have our own set of standards for representation of common concepts. These are defined by CDISC (http://www.cdisc.org) and are driven by the collection and analysis use cases. The data are partitioned more discreetly than FHIR resources into domains based on core clinical concepts (and I use this term loosely) - domains are broadly broken down into observation classes (Events, Findings, Interventions and Special Purpose).
Events are based around patient state changes (such as Medical History/Adverse Events/Subject Disposition)
Findings are based around observations on a patient (such as Lab results, Vital Signs, ECG)
Interventions are based around changes applied to a patients (such as Concomitant Medications, Investigational Product Administration, Study Procedures)
Special Purpose are based around other things (such as Demographics)
As you can see, they cut across many contexts of FHIR resources (ie Findings are broadly ~ Observations, Events ~ Condition, Interventions ~ MedicationStatement/Administration, etc). In addition, as research concepts they can be much more prescriptive about the types of data collected.
The mapping for the moment is based on taking a CDISC domain/variable set and then aligning these with FHIR resources/attributes. There are some issues around coding systems in use (research is generally restricted to WHODrug/MEDDRA/CDISC defined NCI subsets) which are not broadly used outside the research domain (the NCI-t values are often annotated so there is some crosswalk there).
In terms of presentations that I'm aware of:
- papers on this have been presented at PhUSE conferences by Research on FHIR group
- There have been a couple of papers presented at CDISC conferences (the EHR to CDASH project is one to lookup)
- I presented something at the Boston DevDay last year, and Transcelerate presented something at the Amsterdam DevDay (and I believe they will also be in Seattle)
There's a bunch of information there. Let me know if you have other queries
Last updated: Apr 12 2022 at 19:14 UTC
