Global age-sex-specific all-cause mortality and life expectancy estimates for 204 countries and territories and 660 subnational locations, 1950–2023: a demographic analysis for the Global Burden of Disease Study 2023

Comprehensive, comparable, and timely estimates of demographic metrics-including life expectancy and age-specific mortality-are essential for evaluating, understanding, and addressing trends in population health. The COVID-19 pandemic highlighted the importance of timely and all-cause mortality estimates for being able to respond to changing trends in health outcomes, showing a strong need for demographic analysis tools that can produce all-cause mortality estimates more rapidly with more readily available all-age vital registration (VR) data. The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) is an ongoing research effort that quantifies human health by estimating a range of epidemiological quantities of interest across time, age, sex, location, cause, and risk. This study-part of the latest GBD release, GBD 2023-aims to provide new and updated estimates of all-cause mortality and life expectancy for 1950 to 2023 using a novel statistical model that accounts for complex correlation structures in demographic data across age and time.

We used 24 025 data sources from VR, sample registration, surveys, censuses, and other sources to estimate all-cause mortality for males, females, and all sexes combined across 25 age groups in 204 countries and territories as well as 660 subnational units in 20 countries and territories, for the years 1950-2023. For the first time, we used complete birth history data for ages 5-14 years, age-specific sibling history data for ages 15-49 years, and age-specific mortality data from Health and Demographic Surveillance Systems. We developed a single statistical model that incorporates both parametric and non-parametric methods, referred to as OneMod, to produce estimates of all-cause mortality for each age-sex-location group. OneMod includes two main steps: a detailed regression analysis with a generalised linear modelling tool that accounts for age-specific covariate effects such as the Socio-demographic Index (SDI) and a population attributable fraction (PAF) for all risk factors combined; and a non-parametric analysis of residuals using a multivariate kernel regression model that smooths across age and time to adaptably follow trends in the data without overfitting. We calibrated asymptotic uncertainty estimates using Pearson residuals to produce 95% uncertainty intervals (UIs) and corresponding 1000 draws. Life expectancy was calculated from age-specific mortality rates with standard demographic methods. For each measure, 95% UIs were calculated with the 25th and 975th ordered values from a 1000-draw posterior distribution.