OHDSI Home | Forums | Wiki | Github

HADES development announcements

DeepPatientLevelPrediction 1.1 has been released. This adds adds a number of bug fixes and some new features. The new features are:

  • An automatic learning rate finder has been implemented, used by setting learningRate="auto"
  • Training loop parameters are now handled by a new settings object and as a result more of the training loop parameters are customizable by the user
  • Custom metrics can be created by the user to use for early stopping and learning rate schedule.
  • Seeds have been added in various places, making the whole process more reproducible

Thanks to @lhjohn for his contributions to this release.

1 Like

DeepPatientLevelPrediction 1.1.1 has been released. One small bugfix when uploading to database.

Andromeda 0.6.3 has been release on Github and CRAN. This release add isAndromedaTable function that works with both the current andromeda tables (SQLite references) and the Andromeda v1.0 tables (Apache Arrow FileSystemDatasets). Reminder: If you depend on Andromeda please run R check with the development branch (v1.0) and report any issues on the issue tracker.

CohortGenerator v0.8.0 was released a few weeks ago (sorry for the delayed announcement). The release notes are here.

This release of CohortGenerator adds the ability to generate subsets of cohorts. So, for example, if you have a target cohort of new users of a drug, you can use CohortGenerator to define a subset definition to further subset that target based on a combination of demographic criteria, “limit” criteria (i.e. X days of prior observation) and even based on intersection with another cohort definition. Here is the Creating Cohort Subset Definitions vignette if you are interested in the details.

A big thanks to @jpegilbert who lead the development of this important new feature.

Capr v2.0.1 has been released. This is a patch release so no major changes. Updates include:

  • Additional vignettes and updated documentation
  • compile generic function to coerce Capr class to json structure
  • more informative cs print statement
  • other minor bug fixes

Please keep trying out Capr and posting issues! Feedback always welcome!

CohortDiagnostics version 3.2.0. There are no major api changes in this release so all existing scripts should maintain compatibility with this release.

The main changes are as follows:

  • Improved support for Subset operators by fixing some issues around concept sets
  • Added a database migration for existing results databases to support cohort definition sets with subsets
  • Almost all shiny code has been moved to the OhdsiShinyModules package - this is a single package to install and keep up to date
  • Some alterations to plots and tables in the DiagnosticsExplorer shiny app were changed as the ggiraph package was removed due to compatibility and installation issues on Linux environments. All plots now use ggplot2 and/or plotly
  • Support for easier publishing to Posit Connect/Rconnect and shinyapps.io (testing of shinyapps.io was limited as it was a paid service)

Please check the release notes for for details of changes.

1 Like

ResultModelManager v0.4.0 has been released.
This release adds some vignettes on usage as well as a QueryNamespace class that adds a wrapper around SqlRender functionality that allows execution of queries with fixed parameters from result model specifications.

See this guide on basic usage. The intention is that this will be useful for writing shiny apps and Quatro reports.

In addition, the function createResultsDataModel has been removed as it was added in error when migrating code.

1 Like

EvidenceSynthesis 0.4.1 has been released. The only change is the addition of a video vignette, which was also featured in the ESMARConf 2023 conference.

CohortDiagnostics version 3.2.1 . Has been released.
This is a hotfix that resolves an issue related to the cohort subsetting functionality.

Please check the release notes for for details of changes.

SqlRender 1.14.0 has been released. Contains some fixes for Spark, Oracle, Netezza, and PostgreSQL.

DatabaseConnector 6.2.0 has been released. Most important change is a large set of fixes for DataBricks.

SelfControlledCaseSeries 4.2.0 has been released. Most important change is probably fixing a bug that caused regularization to not be applied to calendar-time splines.

PheValuator 2.2.6 has been released. It creates an export folder with a set of results from the analysis as well as diagnostics for the run. This csv files may be used in shiny applications.

BrokenAdaptiveRidge v1.0.0 has been released and is now part of HADES. This package enables approximate L0 (best subset) regression at scale using Cyclops.

Cyclops v3.3.1 has been released and is now on CRAN. Features the introduction of Jeffreys prior and several fixes for likelihood profiling.

CohortMethod 5.0.0 has been released. This a major update with many changes, including a major overhaul of the framework for running multiple analyses. There are no guarantees for backwards compatibility, so please read the changelog carefully. This version should provide greater flexibility and stability than before.

DeepPatientLevelPrediction 1.1.3 has been released. This was a minor fix for the CI. Our main dependancy, torch in R, wasn’t installing correctly since a new release of it was released on CRAN.

DeepPatientLevelPrediction 1.1.4 has been released. This was a small change needed for running the package with Strategus. Now the device to develop the models on (GPU or CPU) can be controlled with an environment variable in the study package. Before it was hard coded in the study package. But users might want to select a certain GPU to use or even CPU for testing the package.

DeepPatientLevelPrediction 1.1.5 has been released. This is a bug fix for the feature introduced in 1.1.4. Things are moving a bit fast now in preparation for our first network study using the package with Strategus.

PheValuator 2.2.7 has been released. It allows for previously created models to be used in evaluations. It also fixes a bug in getPopnPrev.sql for sql server.

1 Like
t