Andromeda 1.1.0 has been released. This includes some minor performance and stability improvements.
PatientLevelPrediction 6.5.0 has been released. This was necessary since some examples were failing on CRAN and they were going to remove the package. This release includes mostly bugfixes and one new feature, the ability to skip model diagnostics when needed.
PheValuator 2.2.13 has been released. It has a bug fix for a PLP function.
CohortGenerator v0.12 has been released. This release includes enhancements to the cohort subset functionality, bug fixes and improvements to the documentation.
PheValuator 2.2.14 has been released. It improves the performance of the evaluation cohort creation SQL script.
SqlRender 1.19.3 has been released. This contains a minor fix for DDLs in InterSystems IRIS.
Andromeda 1.1.1 has been released. This contains a fix for a bug that had been haunting large-scale HADES users every since the switch to the DuckDB backend: ever so often R would just freeze indefinitely or crash.
A huge thanks to @egillax for identifying the root cause, which has also been communicated back to the DuckDB community.
PheValuator 2.2.15 has been released. It improves the performance of the model creation SQL script and the performance of PLP in the evaluation cohort step.
TreatmentPatterns 3.1.1 has been released. It contains some bug fixes and minor refactoring of internal code
EvidenceSynthesis 1.0.0 has been released. The most important changes are:
- Adding non-normal likelihood approximation using Hermite interpolation. This is much faster to compute for complex models such as SCCS.
- An experiment hierarchical Bayesian model
(Note: In OHDSI we use non-normal likelihood approximations to avoid bias in meta-analysis across sites when (outcome) counts are small and normality assumptions are violated)
ParallelLogger 3.5.0 has been released. This has 1 change, instructing Andromeda (DuckDB) to only use its fair share of threads when running in a multi-threaded setting.
DeepPatientLevelPrediction 2.2.0 has been released. This is a big release, adding support for temporal transformer models and a big refactoring because of that for the data processing and the transformer. As well added support for training from custom embeddings, including Poincare embeddings in hyperbolic space. Finally added a bunch of positional encodings and/or methods to bias the attention mechanism according to time.
DataQualityDashboard 2.7.0 has been released. This is a large release, containing:
- A brand new check which detects overlapping and back-to-back observation periods
- Several refinements to existing check logic and status assignment which will make DQD results more interpretable and actionable
- Many bugfixes
- Documentation updates
Please see the Changelog for a detailed overview of the release.
Thanks to @MaximMoinat for review and testing 
3 Likes
The Hydra and ShinyAppBuilder packages are being deprecated.
Hydrais superseded by Strategus.ShinyAppBuilderis superseded by OhdsiShinyAppBuilder
Since R doesn’t have a built-in mechanism for deprecating packages we interpreted deprecation as follows:
- The packages are moved from the Imports to the Suggests section of the Hades package DESCRIPTION, and will therefore no longer automatically be installed when the
Hadespackage is installed. - On the HADES website, the packages are marked as ‘Deprecated’, and listed in the ‘Deprecated packages’ section.
At the next major release (2026Q1) these packages will be completely removed from HADES and the HADES-wide release file.
CohortMethod 5.5.0 has been released. This offers various optimizations, and now allows the creation of simulation profiles while deleting small cell counts to protect privacy.
FeatureExtraction v3.11.0 has been released. It contains a performance improvement for tidyCovariates and Andromeda >=1.0.0. Next to this getDbCovariateData is adjusted: it has the option to specify a custom table an schema for the cohorts that have been used to generate covariates (needed for PLP)
CohortDiagnostics 3.4.2 has been released. This includes hotfixes for the results database schema.
SelfControlledCaseSeries 6.1.0 has just been released. This brings a bugfix for database platforms requiring temp table emulation, and adds support for grid-with-gradient likelihood profiles, which will be much faster to compute than the current adaptive grid. The default (for now) is still set to adaptive grid for backwards compatibility.
Strategus 1.4.0 has been released. This new version provides support for the latest releases of TreatmentPatterns (>= v3.1), SelfControlledCaseSeries (>= v6.0) and PatientLevelPrediction (v= v6.5) and EvidenceSynthesis (>= v1.0). It also fixes the spelling of the createEmptyAnalysisSpecifications function for those using GenAI to write code
KEEPER 0.2.1 has been released. It contains a bugfix for databases requiring temp emulation schema.