OHDSI Study #2: Study on Early Treatment Pathways in Chronic Disease

jon_duke · February 2, 2015, 6:35pm

Hi all,

A new proposed OHDSI study protocol has been posted on the Wiki and on Github.

This study is a variant of the prior Treatment Pathways Study, but uses a shorter required period of observation and drug adherence (12 months). The goal is to capture initial treatment practices and also expand the ability for sites with shorter continuity periods to run the study.

Please provide any comments or suggestions over the next 2 weeks. If you would like to join as a contributing investigator, contributing to the analysis and write-up of this work, please let me know. (@rkboyce you have been added based on our previous discussion.)

The commenting period will close on 2/16/2015.

Thanks,

Jon

rkboyce · February 3, 2015, 11:22am

THanks Jon. Will get to work on this this week. -R

jon_duke · February 5, 2015, 3:09pm

A huge thanks to @msuchard for making running OHDSI studies even easier! OHDSI Study #2 is now available directly as an R package. Meaning you no longer need to download it and edit your parameters before running. It is literally one stop shopping!

Full documentation on GitHub, but it is as simple as opening up R and…

library(devtools)
install_github(c("OHDSI/SqlRender","OHDSI/DatabaseConnector","OHDSI/OhdsiStudy2"))
library(OhdsiStudy2)

then execute with your parameters:

execute(dbms = "postgresql",
        user = "joebruin",
        password = "supersecret",
        server = "myserver",
        cdmSchema = "cdm_schema",
        resultsSchema = "results_schema")

It will even e-mail the results files directly to the study coordinator.

email(from = "collaborator@ohdsi.org",
      dataDescription = "CDM4 Simulated Data")

Again, huge thanks to Marc!

jon_duke · February 9, 2015, 6:00pm

Thanks to @Juan_Banda, @Vojtech_Huser, @Patrick_Ryan, we now have Study #2 run on 5 datasets!

We look forward to other submissions. This study produces fairly large CSVs that may not be able to be sent directly from R as described above (due to Gmail size limitations). Can either zip or if preferred I can provide a dropbox location.

Thanks!

Jon

schillil · February 9, 2015, 6:18pm

In order to do the heaving lifting for our distributed partners (in several states) for the 12-mo Pathway Study, we’re planning on extracting the necessary data, and then doing the analysis and Achilles assessment centrally. For the IRB, I’ll need to list all the variables we’re pulling for the patients that meet the cohort inclusion criteria - so I’m wondering if any one else had to create such a list and if they could share it. I have George’s generic IRB template, which is a great start. thanks

Patrick_Ryan · February 9, 2015, 6:53pm

I’m not aware of anyone creating a ‘variable list’, but the protocol
details the specifics of how the analysis is performed. The code only uses
elements from PERSON, OBSERVATION_PERIOD, CONDITION_OCCURRENCE,
CONDITION_ERA, DRUG_EXPOSURE, DRUG_ERA.

jon_duke · February 9, 2015, 6:54pm

No problem Lisa. Patrick’s list is the extent of it. Also if helpful here is a list of all concepts used by the study. If you look at the individual sheets, you will see the concepts included (and their source codes) as well as the concepts excluded (and their source codes).

This may be too much for the IRB but perhaps as an appendix?

rkboyce · February 9, 2015, 9:32pm

Jon - a bit confused — thought the commenting period had to close before the study would be executed?

jon_duke · February 9, 2015, 9:46pm

Rich,

It’s actually helpful to have people test it out during the commenting period, recognizing that they might need to re-run should code changes be made. But as I’ve received no suggested changes or input on the study design, I expect this will be final. Thanks for pointing out.

Jon

rkboyce · February 9, 2015, 10:23pm

I will have my comments to you shortly along with those of two senior colleagues here who specialize in the treatment of depression and who are interested in this study. Testing is important and I am glad that folks are working that out.

rkboyce · February 10, 2015, 11:03am

This is a great idea! Unfortunately, when I try to run the installation process I get the following error pasted below. Is there any reason why R 3.1.0 is required for OhdsiStudy2 as opposed to R 3.0.2? I have previously installed and ran ACHILLES without running into this error. Upgrading is feasible but not desirable since right now I have installed R from the Ubuntu 14.04 package.

ERROR: this R is version 3.0.2, package 'OhdsiStudy2' requires R >=  3.1.0
Error: Command failed (1)

jon_duke · February 10, 2015, 11:57am

Rich,

I know that many of the more sophisticated analyses (e.g., CohortMetthod) will require a higher version of R. @msuchard may be able to comment further. For now, are you able to run from the regular R version of the study?

Jon

rkboyce · February 10, 2015, 12:19pm

I did get it to run using the normal R code but see the issue I found : https://github.com/OHDSI/StudyProtocols/issues/3

Also, I notice that the link to the “Treatment Pathways Discussion Thread” on http://www.ohdsi.org/web/wiki/doku.php?id=research:treatment_pathways_in_chronic_disease_12_mos should probably point to http://forums.ohdsi.org/t/ohdsi-study-2-study-on-early-treatment-pathways-in-chronic-disease/273/3.

msuchard · February 10, 2015, 1:58pm

I downgraded the required R version (to 3.0) in the OhdsiStudy2 package, primarily to see if it then works for Rich. I don’t know if SqlRender, DataConnector or mailR work with R < 3.1. CohortMethod and Cyclops certainly do not, and Martijn and I have no plans for adding R 3.0 support since these packages heavily depend on C++11.

Going forward, I highly recommend upgrading to R 3.1; it is the default distribution on Windows, Mac OS X and most Linux flavors. For Ubuntu, RStudio provides a deb package. (See, e.g., http://www.sysads.co.uk/2014/06/install-r-base-3-1-0-ubuntu-14-04/).

Alternatively, run R in a docker container. In my hands, all of the OHDSI packages appear happy in the container rocker/hadleyverse . Anyone want to build an OHDSI docker container? I’d use it.

rkboyce · February 10, 2015, 4:29pm

Great! I will test with 3.0 just to because its already installed and then work on getting 3.1 installed so that I can run these other tools. I think it would be a good policy moving forward to state the exact version requirements in the “Requirements” section of a protocol page. I have added R >= 3.1 to this protocol.

jon_duke · February 10, 2015, 4:37pm

Thank you Rich, I have made this change.

Vojtech_Huser · February 11, 2015, 1:20pm

I like the fact that we have the study #2 and I provided data using IMEDS datesets (CCAE, GE).
In order to contribute to publications writing / ideas - having access to the visualisation framework (what is the update there, URL, feedback) or some comments on how to work with the sequence table would be great.

jon_duke · February 11, 2015, 2:17pm

Thanks @Vojtech_Huser. @lee_evans is setting up a schema on the OHDSI server so I can upload all the results to allow people to access and explore. There are quite a fe ways to slice the data, and more importantly, prune and boil down to a meaningful story. I will let the forums know as soon as this is ready.

Re: the visualization (Sunburst), I’ll defer to @Patrick_Ryan and @Frank as to any updates. I know it’s been tricky to port externally. But perhaps a publicly accessible URL may be possible?

Frank · February 11, 2015, 4:00pm

Porting it isn’t necessarily all that tricky its just been a moving target with features changing. Apologies though, I’ve been distracted by a few other projects. I’m on a deadline to have it available by the end of the month.