Describir: Avoiding lead-time bias by estimating stage-specific proportions of cancer and non-cancer deaths.