LoginSign Up
← Back to all sparks
T

teal.modules.clinical

ANALYTICS
Velocity0.0

Provides teal modules for the standard clinical trials outputs

insightsengineering.github.io/teal.modules.clinical ↗

Roche's clinical teal modules have settled into slow, decorator-driven maintenance.

clinical-trialsshiny-modulestealdecoratorstable-formatting
◆Current state
The module library took its architectural break in 0.9.0, when every module moved to teal_data objects and dropped data_extract_spec inputs. Work since then is narrower: formatting arguments, decorator plumbing, and per-module bug fixes. The 0.13.0 candidates cut in July 2026 are a two-tag same-day release where the first tag carries all the notes.
◆Where it's heading
Development has shifted from reshaping the module API to making outputs configurable — numeric formatting in tm_t_summary, several decorators applied to one output object, categorical statistics in summary_by. Changes arrive from a wide bench of Roche engineers with occasional outside contributors, and releases go through candidates rather than straight to final.
◆Prediction
A 0.13.0 final should follow the candidates with little beyond a dependency bump; the decorator work ported from teal.modules.general is the most likely source of the next batch of changes.

◆Recent moves

  1. 1mo ago

    Numeric formatting control and multi-decorator outputs

    The substantive half of the 0.13.0 candidate pair: a numeric_formats argument for tm_t_summary, categorical statistics in summary_by, and support for applying several decorators to one output object. It continues the post-0.9.0 pattern of tuning outputs rather than reshaping module inputs.

    View source ↗
  2. 1mo ago

    teal.transform dependency bump

    The second tag of the same day's release candidate, carrying only a teal.transform version bump. The 0.13.0 notes live in the rc1 entry.

    View source ↗
  3. 2y ago

    Forest and Kaplan-Meier plots move to ggplot objects

    Forest and Kaplan-Meier modules were rebuilt on refactored tern functions and now emit ggplot rather than grob objects, with font size and relative-width controls exposed. It is the cleanup pass that follows the teal_data migration.

    View source ↗
  4. 2y ago

    Every clinical module migrates to teal_data objects

    ⚡ SPARK

    The breaking release that defines the current architecture: every module adapted to teal_data, and data_extract_spec inputs dropped in favour of choices_selected. Everything since has been refinement on top of this surface.

    View source ↗