OHDSI Home | Forums | Wiki | Github

ATLAS use for Pharmacovigilance

(Pantelis Natsiavas) #1

We are investigating of using/adapting ATLAS User Interface to support pharmacovigilance (PV) activities, i.e. potential signal identification, signal strengthening or even exploratory analysis of clinical data. The overall idea is that we could have hospital data (e.g. EHR data) in OMOP-CDM and use/modify ATLAS to explore these data for PV. I am not a statistics expert and therefore I struggle to follow all the details and the statistical measures provided by ATLAS. I wonder if my rationale has a substantial flaw and I would really like to hear views on that.

I think that the following ATLAS views might be useful in terms of PV:
-Incidence rate analysis
We could consider the target patient population (e.g. the ones receiving a drug) as the “target cohort” and the part of this population where the Adverse Drug Reaction (ADR) under investigation actually occurred, as the “outcome cohort”
The “outcome cohort” could be analyzed in terms of various features (e.g. gender, age groups etc.).
-Cohort pathways
It could potentially be used to highlight sequential events for the “outcome cohort”.

Having these said, I have the following questions:
Q1: Do you see any substantial flaws in the above rationale?
Q2: Could you please suggest any papers demonstrating the use of the above ATLAS parts in terms of PV?
Q3: I am struggling to understand the “drug group era long term” feature used in the “Characterizations” view. Could you please provide some reference on that?

(Chris Knoll) #2

No problems with what you described. We’ve used characterization and IR analysis to inform regulatory reports to the FDA.

I think others in the community may be able to provide references to papers, but my focus has been to take output from the tools for regulatory communication, not papers.

Sure: the ‘long term’ component of this means that it is looking 365d prior (and including) the cohort entry. The drug era part means it’s reading from the DRUG_ERA table, which normalizes the drug exposure to the ingredient class.,

The group means that it is going to attempt to roll-up to ATC classes using CONCEPT_ANCESTOR. Meaning: if one person has DrugA, another person has DrugB, and both of these roll up to ATC1, the report will show a count for DrugA from the one person, a count for DrugB from the other person, and 2 counts of ATC1 because each person contributes a count to that group.

(Pantelis Natsiavas) #3

Thank you Chris for the input. I really appreciate it.