OHDSI Home | Forums | Wiki | Github

Extension of Treatment Pathways study in Europe

Hi All,

Within the European Medical Information Framework (EMIF) project we are considering to repeat the treatment pathways study as published in PNAS in the European databases that have been or are being mapped to the OMOP CDM. We are also considering to extend this to other disease areas that are of interest within this project.

I have clearance from our IPCI database governance board to run the PNAS study and like to do some tests before getting other databases involved.

What I cannot find on the wiki is the code used to produced the output as presented in the paper, like the graphs. Where can i find these post-processing steps in the community? I think we should share the full pipeline all the way to the final reported results.

@hripcsa, @Frank, @Patrick_Ryan can you help me with this?

If any of you like to be involved in this extended treatment pathways study do please contact me. We could add other disease area’s etc.

It would also be very nice to think about an online dashboard that can show these results and would allow other databases in the future to add their data on the fly (which i think should be the standard approach for all OHDSI studies!).

Peter

I assume you mean the sunburst plots. Patrick’s team did those using d3.js library. Patrick gave an example with code here: https://bl.ocks.org/kerryrodden/7090426 https://bl.ocks.org/kerryrodden/7090426. The analysis that they applied to compute the treatment pathway results is here: https://github.com/OHDSI/StudyProtocols/tree/master/Study%201%20-%20Treatment%20Pathways https://github.com/OHDSI/StudyProtocols/tree/master/Study%201%20-%20Treatment%20Pathways . They then transformed the final result set into a JSON object that would work with the d3 sunburst.

Definitely happy to help. I expect Patrick’s team may have more to offer.

George

Great idea peter. I know there are other data partners who werent able to
participate the first time around thatd like to. We are keen to run
treatment pathways for the ims disease analyzer france, germany, and
australia datasets once the cdms are completed.

This will also be a good forcing function to make sure we clean up the
postprocessing code. We can get the sql script we used to compile the
results and produce the sunbursts on github. Perhaps more exciting is that
we can post a public website with a repository of all the results from any
sites that opt in. The full results from our original pnas paper are
already publicly available, so that could be our first starting point.

Patrick, has your group been able to make any headway here? I didn’t see any updates in the protocol’s github, so I wanted to double check. I’m visiting with a group who found them really neat, and I’d love to be able to have their team executing the pipeline as an example!

Hey Robert, @Rijnbeek is leading this effort out of Erasmus MC. I will
reach out to him, but would suggest you post directly too.

Hi Robert,

Thanks for reaching out.

Your timing is great! This study will be discussed in two weeks in Barcelona in the EMIF consortium meeting.

I am currently testing the automatic creation of the graphs as a webpage together with Chris from JnJ and we I am adding this to
the study package so you can visualize the graphs locally as well in a Shiny app.

I will keep you in the loop, I expect to give an update after the Barcelona meeting and we will formally kick-off the study.

Peter

Great, thank you both for the helpful information. I’ll keep my eyes out for the update!

@Rijnbeek.
Hello Peter,
I am interested in making similar sunburst graphs that are depicted in the PNAS Treatment Pathway paper and was wondering if you were able to implement the Shiny app you mentioned in your earlier post?

Thank you for your time,
Jonathan Dekermanjian.

Hi Jonathan,

In EMIF we have finished a treatment pathway paper (submitted) on the treatment patters before and after dementia diagnosis on 6 European databases but using an alternative non-cdm approach (but including interactive sunburst plots.)

Running this study against the CDM has been much slower than i hoped for unfortunately in EMIF. This is mainly due to ongoing slow mapping efforts that needed to be finalised first to get all data sources onboard. I therefore postponed the effort to build the shiny app from the already available javascript interactive visualisations.

Running other treatment pathway studies using the CDM (and possibly repeating the PNAS paper) are still on my agenda though in the context of OHDSI Europe. This is also part of the upcoming project European Health Data and Evidence Network (EHDEN), including dashboard development. However, I also know of others that are implementing treatment pathway tools (ultimately in ATLAS is suppose), so this may even be done before EHDEN kicks off (October most probably). I think multiple people are implementing the same tool and we need to coordinate this a bit to avoid duplicate efforts.

I suggest to make a new post in the forum to understand better who is working on adding this to the toolset already etc.

I know of the work done by John Duke that is waiting to be implemented in the WebAPI and Atlas I think:

Peter

1 Like

Another question regarding the treatment pathways work has arisen as Jonathan Dekermanjian, a data science summer student at Colorado, plows thru the treatment pathways code. We are scratching our heads about what to do with combination (multi-ingredient) medications and distinct medications that are started on exactly the same day. We understand that only the first exposure to an ingredient is illustrated in the sunburst plot. But what should be done when that first exposure includes more than one ingredient? Was there a previous thread that I could not find that discussed how the community resolved this?

2 Likes
t