Skip to content

Releases: stroke-modelling/stroke-outcome

Changed mRS and util imports

Choose a tag to compare

@aselaws aselaws released this 07 Aug 10:09

Functions for importing mRS and utility distributions no longer return the associated notes by default.

New mRS distributions

Choose a tag to compare

@aselaws aselaws released this 19 Jul 13:09

Recalculated mRS distributions when documentation and summaries were written. These distributions are rounded to 3 decimal places and have various small differences from the previous version. The differences are mostly from changes to the excess death rates.

v0.1.5

Choose a tag to compare

@aselaws aselaws released this 09 Nov 13:14

Minor changes from v0.1.4:

  • 🐛 Bug fix for the full cohort outcomes. Previously the three separate results dictionaries for LVO+IVT, LVO+MT, nLVO+IVT were being combined incorrectly. The case for whether the IVT results were preferable to MT results was being considered for all patients instead of just those who received both treatments.
  • New output from the main calculate_outcomes() function. The function now also returns a dictionary of the patient data used to generate the outcome results. This makes it easier to see whether stroke types were changed by the models.

v0.1.4

Choose a tag to compare

@aselaws aselaws released this 08 Nov 14:54

Most important changes:

  • Added methods for inputting patient data using a DataFrame. This removed the need for the "number of patients" argument when creating the model objects.
  • Removed the requirement to explicitly pass mRS distributions and utility weightings as input. If inputs are not given, default values are used.

Minor changes:

  • 🐛 Small bug fixes for data input sanity checks. For example floats can now quietly change to integers if they're equivalent.
  • "x" value has been renamed to "fixed probability" in outcome_utilities but not yet in the notebooks in docs/.

v0.1.3

Choose a tag to compare

@aselaws aselaws released this 18 Aug 15:43

New: discrete outcome model.