ComparatorSelectionExplorer

This package is designed to aid the selection of comparators in active comparator, new user cohort designs by computing a similarity metric across all possible pairs of drug comparators.

This package utilises code found within the OHDSI HADES Library and is designed to work on observational databases conforming to the OMOP CDM.

A shiny application is provided for the exploration of results.

Setup

Follow the HADES setup guide to ensure your user space is correctly configured. Following this, use remotes to install from github:

remotes::install_github("OHDSI/ComparatorSelectionExplorer")

Using this package

To run the full analysis:

library(ComparatorSelectionExplorer)
connectionDetails <- createConnectionDetails(
 ...
)


executinSettings <- createExecutionSettings(connectionDetails = connectionDetails,
                                            databaseName = 'my_cdm',
                                            cdmDatabaseSchema = 'cdm',
                                            resultsDatabaseSchema = 'scratch') |> 
                        execute()

A zipfile will now be created at executionSettings$exportZipFile, this contains raw csv files including cosine similarity scores and aggregated summary statistics for the full set of exposures in your cdm.

Results exploration

You must first create a database schema in a database. In principle this can be any database engine, but for large results sets postgreql is recommended as we have implemented platform specific opitmizations. If you only have a single CDM using an sqlite database should be fine.

For example:

resultsConnectionDetails <- DatabaseConnector::createConnectionDetails(dbms = "sqlite",
                                                                         server = "test.sqlite")
createResultsDataModel(resultsConnectionDetails, "main")

uploadResults(resultsConnectionDetails,
              "main",
              executionSettings$exportZipFile,
              tablePrefix = "")

TODO

• Github repo and actions
• Shiny app changes
• Import results from a zip file
• Properly implement custom cohorts and allow them in shiny app