Hi - I am curious if anyone could recommend a faster way to include all possible medications given for a specific condition other than simply looking up the treatment guidelines and individually identifying each medication’s concept code in the NDFRT domain. For example, if I am trying to verify that a cohort with asthma has received at least one of the standard asthma medications (albuterol, tiotropium, cromolyn sodium, etc) without adding each individual medication from the Drug domain / Chemical Structure concept class ID / NDFRT vocab ID, is there a way to do so? Thank you for your advice!
There are indication drug classes in NDFRT and ETC. For example, all the descendants from NDFRT N0000000498 concept_id 4264780 “Asthma” are drugs used in the treatment of the disease:
select concept.*
from concept_ancestor
join concept on concept_id=descendant_concept_id
where ancestor_concept_id=4264780
and concept_class_id='Ingredient'
If you remove the last clause you will not just get the ingredients, but all the drug products.
I just had a discussion with @mgkahn about NDFRT and ATC. Whenever I try to use NDFRT, it doesn’t seem to have the concept I seek. An example is “depression”. I search the term in Atlas and narrow the results to the NDFRT vocabulary, but I only receive two concepts: Depression, Post Partum and Central Nervous System Depression. What’s the trick to the NDFRT?
sorry to butt in, but it’s really simple 
You should use ‘depressive disorder’, as ‘depression’ is just a symptom or a part of a narrow diagnosis.
http://athena.ohdsi.org/search-terms/terms/4344727
Ah ha! Thanks, @aostropolets! I was so close, yet so far away. Another question, when I use Christian’s query for the depressive disorder concept_id, it returns concepts that aren’t generally considered treatments for depressive disorder. Examples include: psyllium, glucose, etc. How or why do these medications relate to depressive disorder? And when would I use the ATC vocabulary versus the NDFRT?
Friends:
Please don’t use the Indications provided in the vocabularies blindly. There is everything in there from FDA-approved conditions that have an effect on the pathogenetic mechanism and are mentioned in the label, all the way to support medication used under certain circumstances. No idea why those examples are in there, @MPhilofsky, but in a different example anti-emetic medication is used to treat cancer. Why? Because it is used together with chemotherapy. In itself it has no effect on the disease.
So, going back to @Swammer: If you want a list that has a direct and causal therapeutic effect you need to create that yourself. The indications in NDFRT and ETC can be used as a starting point.
Our experience has been similar to @Christian_Reich - not sure how much is iffy indications, and how much is things like excipients or combination ingredients getting pulled in to the subtree by other components of the formulation.
The approach we’ve adopted is to use the ATC or NDF-RT term to assemble a list of "Clinical Drug Form"s (SCDFs – maybe someday SCDGs if OHDSI picks them up, since they’re more efficient for this purpose), manually review that to weed out things that don’t appear to fit, then assemble the final list of drug terms by collecting the children of the SCDFs via concept_ancestor. That keeps the manual review at a high enough level that you’re not wading through hundreds of different drug strengths or brands.
Thank you Christian. I’m curious as to whether the drug definition or the query that I’ve constructed is incorrect. I have a cohort defined on the public ATLAS site that includes all patients with an ED visit who received a dx of Asthma at that visit and received asthma medications at that visit as defined per your suggestion. In a query on our internal databases, the ED visits with Dx of Asthma returns about 90K patients. However, when adding “received asthma medications” to the query, the population drops to ~1,000. Seems like I’m missing something in the query? I’ve exported these queries to the OHDSI/ATLAS as well: [DBMI] ED Visits, [DBMI] ED Visits w/ Asthma Dx, [DBMI] ED Visits w/ Asthma Dx w/ Asthma medications. Any thoughts? Thank you!
This is a classical situation. Based on expectation of clinical practice we create a bunch of plausible criteria, only to find out that they are not actually happening that way in the data. Often, the problems come from low-sensitivity concepts or lousy time stamps defining the index or qualifying criteria (e.g. the asthma concepts are put in there after the fact).
I would do the following: Take your cohort and run a Full Analysis in the Report tab. There, you can find out what Drugs, Conditions and Procedures actually happening during the duration of your cohort. You’ll quickly see what’s going on.
Thank you Christian!