HADES development announcements

Developers
323

I’m delighted to announce that DataQualityDashboard is now available on CRAN: CRAN: Package DataQualityDashboard

Thanks a million to @MaximMoinat for collaborating at the Symposium to get the final major CRAN requirement sorted in the world’s fastest hackathon :tada: :smiley:

4 Likes
324

CohortGenerator v1.0.2 has been released. This includes a few bugfixes most notably for the negate option when using cohorts as subsets.

325

TreatmentPatterns v3.1.2 has been released this includes some bug, and run-to-run stability fixes.

326

CohortMethod 6.0.0 has been released. This is a major update with many changes, so be sure to read the release notes. The biggest changes are:

  1. Unifying the settings objects, converting them to R6, similar to what was done for SelfControlledCaseSeries. This should make it easier to develop Strategus specifications, for example in a GUI. This does mean your old study code is likely incompatible with this new version.
  2. Adding the ability to specify a nesting cohort. The target and comparator will be nesting in this cohort, usually the indication. This allows the same cohorts to be used for both SCCS and CohortMethod. Attrition is precisely tracked.
  3. Added the modification to the balance diagnostic as described in George’s paper. This will allow us to pass the balance diagnostic in smaller cohorts. Note that this is currently turned off by default. See the alpha and sdmAlpha arguments in the createComputeCovariateBalanceArgs and createCmDiagnosticThresholds functions, respectively.

In addition a shout out to Kelly Li, who revised the PS trimming function, providing the various options recommend in literature. (See createTrimByPsArgs).

327

Package BigKnn has been deprecated. Previously deprecated packages Hydra and ShinyAppBuilder have been removed from HADES.

328

CohortGenerator v1.1 has been released. This release adds new functions for computing cohort attrition using inclusion rule statistics (for Circe-based cohorts) and when using cohort subsetting.

329

PatientLevelPrediction v6.6 has been released. It’s a big change. Added flexible hyperparameter tuning and custom metrics. Added ridge regression and missing indicator’s. We now default to using logit’s for rank based metric’s wherever possible to prevent numerical issues.

330

DeepPatientLevelPrediction v2.3.0 has been released. Mostly a new model RealMLP and some bug fixes.

331

CohortMethod 6.0.1 has been released. This includes some very minor changes, plus CohortMethod is now in CRAN!

332

DataQualityDashboard 2.8.7 has been released. This release contains some bugfixes.

333

SqlRender 1.19.5 has been released. This introduces some minor bugfixes.

334

SelfControlledCaseSeries 6.1.4 has been released. This contains 1 bugfix.

335

Keeper v2.0.0 has been released. This is a major release, with many changes. The most important ones are:

  1. Added an LLM-powered workflow for generating the Keeper input concept sets.
  2. Added a Shiny app to facilitate human review of Keeper profiles.
  3. Standardized LLM connectivity using the ellmer package, allowing pretty much any LLM from any provider to be used.
  4. Added functions for automatically creating and using a highly-sensitive cohort to compute both sensitivity and PPV of cohorts.
336

Characterization v3.0.0 has been released. This is a major release, with many changes. The most important ones are:

  1. Splitting the aggregateCovariates into: riskFactor, targetBaseline and caseSeries to make the inputs clearer.
  2. riskFactor analysis now has three modes: efficient (target without case vs case), CohortIncidence (target without case who have some TAR vs case) and PatientLevelPrediction (target without case and without people with outcome during washout period before TAR vs case).
  3. riskFactor analysis calculates SMD during execution and min SMD values can be specified to reduce result size.
337

OhdsiReportGenerator v2.1.0 has been released. Most changes are behind the scenes although we are now supporting different version of result tables so the latest OhdsiReportGenerator should work with results generated with older HADES packages as well as the latest HADES packages.

338

OhdsiShinyModules v3.5.0 has been released. The updates pulls in updates to the HADES packages.

339

HADES-wide release 2026Q1 has been released. This is a snapshot of the HADES packages and their dependencies, thoroughly tested and confirmed to be mutually compatible. It is intended to be a stable environment for studies and execution engines. Currently, the release is only available as an renv lock file .

340

Strategus v1.5.0 has been released. This update provides support for CohortMethod v6, Characterization v3 and updates to EvidenceSynthesis for covariate balance meta-analysis and prediction interval calculation.

341

Keeper v2.1.0 has been released. This contains some minor improvements and bug fixes.

342

SelfControlledCaseSeries v6.1.5 has been released. This contains 2 minor bug fixes.