Excess death estimates from multiverse analysis in 2009-2021.

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• Excess death estimates are vital for public health surveillance but highly sensitive to analytical methodologies.
• Variability in analytical choices can significantly impact the magnitude and interpretation of excess death calculations.

Approach:

• A multiverse analysis approach was employed, exploring all possible time periods for reference baselines and 1-4 year projected periods.
• Utilized annual, age-stratified death data from the Human Mortality Database for 33 countries spanning 2009-2021.

Key Points:

• Different reference baseline periods caused substantial variability in absolute excess death estimates.
• Relative country rankings for specific years and year rankings for specific countries remained largely stable across baseline choices.
• Distinct mortality time patterns emerged, with pre-pandemic declines varying in steepness and significant differences in COVID-19-related excess deaths.

Conclusions:

• Multiverse analysis provides a more unbiased method for comparing international mortality trends and understanding uncertainty.
• The approach helps elucidate the nature of observed mortality peaks and provides a robust framework for long-term comparative analysis.