[Call for Collaborators]: Tool for Extracting Treatment Pathways in Patient Populations Standardized in the OMOP CDM

Hi folks :wave:

This is a call for collaborators for a project I am starting! I’d love collaborators on this work if you are interested in the idea; I’ll be working on it at any rate! Details as follows:

Title: Tool for Extracting Treatment Pathways in Patient Populations Standardized in the OMOP CDM

Goal: Understand the sequencing of treatments in one or more patient cohorts of interest

Methods: To summarize the events following a cohort start date of one or more target cohorts, we create a set of cohorts to identify the clinical events of interest for the target population called event cohort. Once we have our treatment pathway designed, we can execute this design against one or more OMOP CDM databases.

Target Output Example:

An example output visualization could look something like this:

Where the legend section displays summary statistics and who was ascribed to what clinical event (in this case, medical prescription).

The sunburst plot represents the various event pathways taken by persons over time. The center of the plot represents the cohort entry and the first color-coded ring shows the proportion of persons in each event cohort. Then, the first ring in the sunburst plot shows the proportion of persons that initiated a type of first-line therapy defined by the event cohorts. The second set of rings represents the 2nd event cohort for persons. In certain event sequences, a person may never have a 2nd event cohort observed in the data and that proportion is represented by the grey portion of the ring.

Sources and Further Reading:

  1. Chapter 11 Characterization | The Book of OHDSI
  2. https://www.pnas.org/doi/10.1073/pnas.1510502113

The plan would be that this tool would live within the JuliaHealth ecosystem, I’d definitely try targeting a few venues to present this work at, and I would use it for a few studies I am planning (happy to collaborate on that as well). I was thinking that to start with, we could focus on line regimens, think about a data asset we could produce, and some visualizations within Makie. Dream would be to link this up with Genie for really slick dashboarding.

Let me know your thoughts!

~ tcp :deciduous_tree:

P.S. CC’ing some folks who may be interested: @dilumaluthge @Fareeda @Jay-sanjay @cce @arinbasu1 @Phil_Vernes

P.P.S. @Pere , this is what has been driving my interest in “pluggable” modules I can make for Genie dashboards! :smiley: