OHDSI Home | Forums | Wiki | Github

HIV research in the CDM

(Kristin Kostka, MPH) #1


@msuchard @schuemie and I are looking at a PLE question for HIV patients. We’re experiencing some challenges with the creating an appropriate concept set for HIV coding.

Is there anyone in the OHDSI community doing HIV research? Anyone have HIV cohort definitions they could share? We could really use help!

(Julianna Kohler) #2

@krfeeney, I assume you said this to get me involved? :smile: I don’t have an HIV cohort definition offhand but could come up with one fairly easily and would love to get involved.

(Kristin Kostka, MPH) #3

@julie_kohler, we always welcome your expertise! Please share with the class if you come up with any good concept sets. :wink:

I just remembered… @Vojtech_Huser and @Sigfried_Gold… aren’t you guys working on HIV characterization?

(Sigfried Gold) #4

Yes! End of the semester course work is interrupting my progress for a couple weeks, but we’re eager to hear what you all are doing and hope to have something to show in a month or two.

(Julianna Kohler) #5

Happy to share–can you give me an example of the kinds of cohorts you’re trying to find? I’m the slacker that’s been sitting in the back of the class and is a little behind. :smiley:

(Kristin Kostka, MPH) #6

Here’s one cohort… but we think this may not be the right order of logic and/or the coding is too restrictive. :slight_smile:

Initial Event Cohort

People having any of the following:

  • a drug exposure of Atorvastatin2
    • for the first time in the person’s history

with continuous observation of at least 180 days prior and 0 days after event index date, and limit initial events to: earliest event per person.

Inclusion Rules

Inclusion Criteria #1: NoPriorStatin

Having all of the following criteria:

  • exactly 0 occurrences of a drug exposure of AnyStatins1where event starts between all days Before and 1 days Before index start date

Inclusion Criteria #2: ProteaseInhibitor

Having all of the following criteria:

  • at least 1 occurrences of a drug era of ProteaseInhibitor4where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Limit qualifying cohort to: earliest event per person.

End Date Strategy

Custom Drug Era Exit Criteria

This strategy creates a drug era from the codes found in the specified concept set. If the index event is found within an era, the cohort end date will use the era’s end date. Otherwise, it will use the observation period end date that contains the index event.

Use the era end date of Atorvastatin2

  • allowing 7 days between exposures
  • adding 0 days after exposure end

Cohort Collapse Strategy:

Collapse cohort by era with a gap size of 0 days.

Appendix 1: Concept Set Definitions

  1. AnyStatins


Concept Id Concept Name Domain Vocabulary Excluded Descendants Mapped
21601855 HMG CoA reductase inhibitors Drug ATC NO YES NO
  1. Atorvastatin


Concept Id Concept Name Domain Vocabulary Excluded Descendants Mapped
1545958 atorvastatin Drug RxNorm NO YES NO
  1. HIVpositive


Concept Id Concept Name Domain Vocabulary Excluded Descendants Mapped
439727 Human immunodeficiency virus infection Condition SNOMED NO YES NO
4013106 HIV positive Condition SNOMED NO YES NO
  1. ProteaseInhibitor


Concept Id Concept Name Domain Vocabulary Excluded Descendants Mapped
21603148 Protease inhibitors Drug ATC NO YES NO

(Vojtech Huser) #7

Our team is working on HIV research. Details can be found here https://github.com/lhncbc/CDE/tree/master/hiv

Our R code (where Sigfried is trying to learn to use the FeatureExtraction package) is here https://github.com/Sigfried/HivDescriptive/blob/master/R/main.R#L231

We would love to join forces with like-minded OMOP (and Sentinel or VDW or PCORNet or |put-your-favorite-model-here|) researchers

(Julianna Kohler) #8

One possible thought–if you’re searching for Protease Inhibitors using ATCs, many patients take fixed dose combinations, which are harder to code for ATCs (since the PI will only be part of the therapy). Have you checked to see if you get patients when you look just for that code?

(Andrew Williams) #9

I’m a Co-I on study that meets your 2+ source definition: COMpAAAS Tripartite. The three sources are the Antiretroviral Therapy Cohort Collaboration (ART-CC) which itself is composed of many - mostly European - sources, Kaiser Permanente, and the Veterans Healthcare System (VA). Interoperability hasn’t been top of mind despite my advocating for its value. That might change as progress is made toward shared analyses. Each part has been concentrating on analyses using it’s own source so far. Let me know if you want to me to investigate who you might talk to about working on CDEs that meet study needs.

(Andrew Williams) #10

Do you have any experience comparing cohorts defined with good PI codes vs those that skip PI codes altogether? In other words a comparison that tests the assumption that nearly everyone with an HIV diagnosis is on a PI.

(Julianna Kohler) #11

Interesting question–I had assumed PI was of interest for interactions between statins and PI, specifically. But if PI is intended to be a proxy for HIV infection, it’s probably not a very good proxy–I think efavirenz remains pretty popular, and there are lots of other drug classes that are used as the backbone of ART.

(Andrew Williams) #12

Your assumption about the interest in the interaction was correct. That answers the question and - betrays my ignorance about the variations of ART.

The person who question motivated all this is a colleague with drug development background. She would have known. I just wondered whether we could safely avoid the trickiness of representing PI properly. Thanks.

(Julianna Kohler) #13

Of course! And worth noting–I’m not sure if the fixed dose combos are the issue–just that it’s a possibility.

(Nanguneri Nirmala) #14

Sorry for the late reply, I Just joined the forum.
Yes, we are looking for HIV patients on PI + statins to study the interaction between the two. So the question @Andrew was asking is pertinent to our analysis.

(Vojtech Huser) #15

I have question to a drug expert (Anna O.?) (and fellow HIV experts) @aostropolets

SNOMED CT links SNOMED CT ingredients to SNOMED CT drug classes. I like that A LOT!
I have to do more hoops accross terminologies (and joins prone to mistakes) to do that on the standard drug terms side.

This standard RxNorm term here tenofovir disoproxil http://athena.ohdsi.org/search-terms/terms/1710281

is nicely linked with a snomed term

And that term has this relationship to a class

and this class nicely leads me to other ingredients (in SNOMED CT; unfortunately)

My HIV (and drug) question is: How can I define a cohort of patients on reverse transcriptase inhibitor ? (this NON STANDARD term http://athena.ohdsi.org/search-terms/terms/4159314 )

I realize I am walking the hierarchy in the non standard terms - but is the solution to use ATC on the standard term side? And if yes, how exactly (in Atlas)

I also like this Retroviral protease inhibitor class here http://athena.ohdsi.org/search-terms/terms/4029212

This ATC class would be a possible ATC parent

(so how do I do the correct jumps (or more ideally, would not have to and Atlas will help me not to make a mistake)

I don’t need a full solution to a thorny problem. Just some/any advice would be great.
I am basically defining my cohort here http://www.ohdsi.org/web/atlas/#/cohortdefinition/1770613 and want to enter something super smart here (note that my CDM does not have drug_eras generated and I must use drug_exposure:

(Anna Ostropolets) #16

Reverse transcriptase inhibitors belong to J05AF class in ATC, which you can use to get all RxNorm drugs that go under this ATC class.
I added J05AF and J05AE to your concept sets in your cohort, so you can use them.
So it seems like you can use ATC for these purposes, although I’m not sure that this is exactly what you need… Is it?

(Vojtech Huser) #17

Thanks Anna. For our Rx-driven HIV cohort, what you replied is very helpful.

Another update: Seems like the LOINC code most important for our Lab-driven HIV cohort would be http://athena.ohdsi.org/search-terms/terms/3010747

(also listed in this CDC list https://www.cdc.gov/hiv/guidelines/reporting.html (go to file LOINC map xlsx)

All HIV cohorts are in this package. (which we hope to soon test at some pilot sites) https://github.com/Sigfried/HivDescriptive/blob/master/inst/settings/CohortsToCreate.csv