OHDSI Home | Forums | Wiki | Github

Number of Biologic medicine scripts needed for a feasibilty Study

Researchers
#1

Hi All,

I am planning an OHDSI study on the safety and effectiveness of biologic medicines. “Biologics” encompass the immune-based therapies such as the monoclonal antibodies and fusion proteins (eg ipilimumab, denosumab and etanercept)

I have received reviewers responses to our funding application to do this work and we have been asked to provide estimates of the numbers of biologic medicines used across the OHDSI network. At this stage I only need to provide an estimate of the overall number of biologic scripts/dispensings to demonstrate the potential feasibility of this study. These numbers are only to be included in the response to reviewers and will not be used in a publication. I apologise about the tight timeline but my response is due Monday 10th July.

If you can participate I would be very grateful. The code below can be run against your data to generate the count of biologics across all products.

Kind regards, Nicole

SELECT COUNT(*) AS prescription_count
FROM drug_era
WHERE drug_concept_id IN (735843, 909959, 912263, 936429, 937368, 1110942, 1151789, 1186087, 1312706, 1314273, 1315411, 1387104, 1397141, 19041065, 19047423, 19080458, 19080982, 19100985, 40161532, 40167582, 40171288, 40222444, 40238188, 40241969, 40244266, 42801287, 44507676, 45774639, 45775965, 45892628, 45892883);

#2

Hi, @nicolepratt,

Have you considered defining a cohort on http://www.ohdsi.org/web/atlas and then asking people to import it into their own OHDSI environment to generate a cohort and give you those statistics? With a cohort, you can also run other characterization queries and you can get some additional information about recent medical history (to understand if you may need to ruleout people from the study).

-Chris

#3

@nicolepratt:

Ouch. You need to ask for all those ingredients and their descendants. Like this:

select count(8) as prescription_count from drug_era
where drug_concept_id in (
  select descendant_concept_id from concept_ancestor where ancestor_concept_id in (735843, 909959, 912263, 936429, 937368, 1110942, 1151789, 1186087, 1312706, 1314273, 1315411, 1387104, 1397141, 19041065, 19047423, 19080458, 19080982, 19100985, 40161532, 40167582, 40171288, 40222444, 40238188, 40241969, 40244266, 42801287, 44507676, 45774639, 45775965, 45892628, 45892883)
);
#4

C’mon @Christian_Reich, you know better than that! The query is run off
the DRUG_ERA table, which only contains ingredients:)

@Chris_Knoll had a good suggestion…for those who do have ATLAS installed
locally and want to get @nicolepratt’s feasibiility count without running
SQL, here’s a cohort definition that’ll get you want she need…

http://www.ohdsi.org/web/atlas/#/cohortdefinition/913204

#5

Thanks all! I appreciate that there may be many ways to do this! Thanks @Chris_Knoll for the suggestion and @Patrick_Ryan for the cohort definition #awesomecommunity

#6

Seems, @Christian_Reich is right: Drug_era contains drugs, but Dose_era contains ingredients.

1 Like
#7

If only there was some documentation on the drug_era table :wink:

A Drug Era is defined as a span of time when the Person is assumed to be exposed to a particular active ingredient…The drug_concept_id field only contains Concepts that have the concept_class ‘Ingredient’.

#8

really.
Thanks, @schuemie for correcting me:)

#9

@nicolepratt When do you need this information by? We can run this query off a few of our datasets.

#10

Replying for @nicolepratt since she’s on the other side of the world and
timezones won’t cooperate. Per her original post, she’s looking for
results by Monday, July 10. If the Columbia or Janssen experience is any
indicator, the time it took you to write that post should exceed the time
it takes the run the query, either via SQL or ATLAS :slight_smile:

#11

@nicolepratt: Using your query I’ve generated counts for six of our internal databases. I’ve sent the results to your e-mail.

#12

This is a good opportunity for me to ask about best practices for resolving drug eras: we’ve run into serious confounding at the ingredient level in a couple projects where a given active ingredient is available both systemically and non-systemically (e.g. prednisone may be taken systemically as a whole-body immunomodulator, such as for asthma, or may be used topically as a local immunomodulator, such as for poison ivy).

My interim approach is to create new drug_era records based on SCDF rather than IN, thinking they’ll coexist queitly alongside the IN records. The risk here is that queries that don’t restrict by specific concept IDs (or take @Christian_Reich’s approach and roll up descendants) will end up double-counting. The alternative is to create a separate table, notionally drug_form_era, but that won’t be seen by tools like Atlas or Achilles. How are other folks dealing with this issue?

#13

Yeah. It would help to first read and then type, not the other way around. But maybe I just wanted to test the community whether they know their stuff. :smile:

Well, both queries should really return the same result.

The general response would be “Create cohorts”. One for each drug form of interest. It’s not as clean and efficient as the DRUG_ERA table, but hey, who cares. It’s just a database.

@bailey: Care to toss out your study to the community?

#14

Per the convention, DRUG_ERA should be at the ingredient level. But if
for any particular analytical use case, having the data rolled up at the
ingredient level isn’t what you want, then you can always create custom
eras directly off the DRUG_EXPOSURE table. In fact, ATLAS would do that
for you, on a per-cohort basis. Here’s the example to address
@nicolepratt’s request using this approach (note, here you have to follow
@Christian_Reich’s recommendation to use the CONCEPT_ANCESTOR table
expansion): http://www.ohdsi.org/web/atlas/#/cohortdefinition/913208

#15

@Patrick_Ryan @Christian_Reich Thanks. We’ve taken that approach on a single-study basis, but do find that it’s a common enough question that a drug_era-style rollup is useful. We may be an outlier in that respect, though; the meds we tend to study are probably different from the typical over-50 use case. As long as you don’t have a general sense that OHDSI studies will be querying drug_era with anything other than a list of ingredient concept_ids, I’m less worried about having both in the drug_era table.

@Christian_Reich I’m hopeful that some of the PEDSnet studies can get to OHDSI, once we get the bugs hammered out of the analyses.

#16

@nicolepratt Trying to find your email to send it to you. I executed it on 7 data assets we have. There are counts from both your query and the Atlas one.

#17

As a general practice, an OHDSI study does not have to use the DRUG_ERA
table in an analysis and should not be used if the question of interest
requires studying an exposure that is not at the ingredient level (e.g.
restricted to specific strengths, formulations, brand names, etc.).
DRUG_ERA is a derived construct that can be useful to use if the analysis
is at the ingredient level (as @nicolepratt’s request here).

#18

my email is nicole.pratt@unisa.edu.au Thanks so much for participating!

t