OHDSI Home | Forums | Wiki | Github

Mapping Adverse Event data to OMOP and to FHIR

I am new to OMOP. I am working on mapping adverse event clinical data to OMOP CDM and then from OMOP to FHIR. Are there any working groups where I could get involved to get some help around mapping AE to OMOP?


Hi, @priagopal. It’s great that you decided to join to OMOP. May be useful to start from the basic courses in EHDEN Academy, for example, OMOP CDM and Standardised Vocabularies course (https://academy.ehden.eu/course/view.php?id=4). This course as many others is free of charge and very useful for future work with OMOP data.
Also feel free to ask questions of interest on this forum in the appropriate sections.

Hi Mikita, Thanks for getting back to me. I watched the tutorial video in OHDSI website. I am particularly interested in AE data mapping. Based on the information I could see in the website, looks like AE clinical trial data is mapped to Observation and condition_occurrence. I heard in the training that Observation is used for any data that cannot go into any other table. AE is pretty important. Just wondering whether we are going to have an Events table in the future?


Depending on the nature of adverse events Observation or Condition_occurence tables may be used for such purposes. Currently a separate Event table isn’t planned in the future.

Hi @priagopal

You may be interested in joining the FHIR and OMOP Working Group. The HL7 and OHDSI communities have formed the WG and four (4) subgroups to advance efforts on FHIR-OMOP use cases/projects. Please refer to the OMOP on FHIR Confluence page for more info.

For subgroup updates, resources, meetings times, etc. refer to the direct links below:

Upcoming WG Meeting: There is also an upcoming all-hands meeting (8th Dec @ 10 ET). Subgroups will provide updates on what they have been working and plans for 2022. You can register here

Hello @priagopal and welcome to OHDSI!

Are your AE data coded with a common vocabulary such as ICD or SNOMED? If yes, then you just follow the standard process as outlined in the Book of OHDSI Chapter 6, section 3. It explains how data are mapped to standard concept_ids. And the domain_id for a standard concept_id tells you where the data live in the CDM. However, I don’t think there is a standard way to identify an adverse clinical event from any other clinical event. You should ask on the Vocabulary forum, they might be able to help.

I host the Electronic Health Record working group every other Monday. We’re not solely based on EHR data, we help support all OHDSI friends on their journey. Here’s a link to our ongoing forum post. Our December 20th meeting will be open for Q&A. Join us then to introduce yourself and your explain your use case. We’ll be able to give you guidance.

Hi Melanie,

Thanks so much for the insights. I am looking to convert clinical trial AE data to OMOP. The verbatim terms are usually free text and coded to Meddra in clinical studies.


The Book of OHDSI also covers the Usagi tool. Usagi will help you map the free text terms to standard concept_ids which are used in the CDM.

Priya, for mapping AE concepts, all the ideas apply as for other concepts. So the material Melanie kindly pointed out above is definitely worth examining. As for AE qualifiers like seriousness, severity, causality, etc., there is no straightforward way to store such information in CDM, but the OHDSI community has developed few conventions for this. They based on utilizing observation_event_id and obs_event_field_concept_id / modifier_of_event_id and modifier_of_field_concept_id fields of the observation and measurement tables respectively. Here I should note that these fields became standard only in OMOP CDM v5.4. More on this topic you can see here in the Clinical Trials Workgroup conventions.

1 Like

It is important to understand the overarching use case for the OMOP CDM, @priagopal. It is not some generic vessel to store data captured somewhere in healthcare. It is there to support large-scale analytics of observational data, and that from the perspective of the patient.

With respect to your use case: AEs from the patient perspective are just Conditions (symptoms, signs, diagnoses) or other Observations, which indicate something (bad) happened affecting health. There is no need for other “events”. The fact that those are due to a drug, or some other intervention, or some other condition, or completely spontaneous, is not input to the research, but output (causal effect estimation). Ordinary observational data rarely come with those stated causes. So, using the modifier of event mechanism is possible, but not part of the general standard and therefore not utilized by any method. However, the Clinical Trial WG is at liberty to establish conventions for their purposes as it sees fit, as @Philip_Solovyev indicated.

Severe AE as well as Serious AE are also artifacts of clinical trials and missing from ordinary observational data. Most Condition concepts speak for themselves if they are severe or if they have serious consequences: E.g., suffering from a Disseminated intravascular coagulation cannot possibly be a good thing, and so nobody expects that fact being explicitly stated.

Bottom line: Primary analysis of clinical trial data is not the stated application of the OMOP CDM, but could be done using extra conventions folks in the Clinical Trial WG are coming up with. However, secondary use of such data (e.g. finding patients meeting certain inclusion criteria) that does not expect any special cross-linking between the data is well supported.