WHORiskCalculator 1.1.0

Breaking changes

• Corrected recalibration. Results change considerably from 1.0.0, and 1.1.0 results should be used instead. Version 1.0.0 replaced the derivation baseline survival with a 10-year risk built from regional age-band incidence rates. Risks are now calculated as in the published method (Kaptoge 2019) and the authors’ Stata program whocvdrisk:
  1. uncalibrated risk from the derivation model, 1 - S0^exp(LP);
  2. recalibration on the complementary log-log scale, 1 - exp(-exp(cons + slope * log(-log(1 - risk)))), using the authors’ parameters for each region, sex, endpoint, model and year;
  3. combination of CHD/MI and stroke risks assuming independence.

New features

• New recalibration argument in calculate_who_risk() and calculate_who_risk_nonlab():
  • "region" (default): WHO/GBD region-level recalibration, reproducing the published WHO risk charts and the default output of whocvdrisk.
  • "country": country-level recalibration (requires country). Countries without country-specific estimates fall back to region level with a warning.
• New year argument (1990-2017, default 2017) to choose the calendar year of the recalibration target population.
• Guam (GUM) and the Northern Mariana Islands (MNP) added to Oceania.
• New “Recalibration” section in the help pages and README.

Internal

• Recalibration parameters are stored in R/sysdata.rda, built from the authors’ Stata data file by data-raw/calibration_parameters.R. The incidence-rate tables used by 1.0.0 have been removed.
• Risk calculations are fully vectorized (no per-row mapply()).
• Added a testthat suite checking results against an independent re-implementation of the authors’ Stata code (both models, region- and country-level recalibration, all years).
• Package now depends on R (>= 2.10) because of the internal data.

WHORiskCalculator 1.0.0

• Initial CRAN submission.