Abstract:
The projection of age-stratified cancer incidence and mortality
rates (and cases) is of great interest due to demographic changes,
but also therapeutical and diagnostic developments. Although (Bayesian)
age period cohort (APC) models are generally accepted, they are
not used in routine practice of epidemiologists. Reasons may include
on side that APC models have been criticised to produce wide
prediction intervals but on the other side the lack of good
stand-alone software. Here, we present a novel R-package, called
APC, which which uses integrated nested Laplace approximations (INLA)
to forecast future cancer rates and expected cases. BAPC provides
age-standardised quantities to facilitate the comparison of trends
between countries or gender, and also age-specific quantities, so
that patterns masked in the age-adjusted projection, can be inspected.
We apply BAPC to lung cancer data for females and assess the
calibration of one-step ahead predictive distributions.
This is joint work with Leonhard Held from the University of Zurich.