Login / Signup

Excess death estimates from multiverse analysis in 2009-2021.

Michael LevittFrancesco ZontaJohn P A Ioannidis
Published in: medRxiv : the preprint server for health sciences (2022)
Excess death estimates are the ultimate assessment of the impact of multiple diseases and forces on the mortality of a population. However, their calculation can be notoriously unstable because it depends on a multitude of analytical choices. Other scientific fields have started using multiverse analysis approaches where all possible analytical choices are considered. We developed a multiverse analysis approach for excess death estimation. The approach is demonstrated with data from 33 countries for the period 2009-2021. Multiverse analysis offers a standardized way to demonstrate the sensitivity of estimates to analytic assumptions, to understand the presence of time patterns (rather than arbitrarily prespecify them), and to reveal consistent patterns that characterize excess deaths in a comparative fashion between different countries.
Keyphrases
  • type diabetes
  • gene expression
  • coronary artery disease
  • single cell
  • mass spectrometry
  • genome wide
  • deep learning
  • data analysis