Lifespan variation during mortality transition in Bangladesh: evidence from a local surveillance and global estimates

Variation in life expectancy has become a critical dimension in understanding inequality in human populations, complementing average longevity by capturing how unequally years of life are distributed across individuals. Reducing such variation, especially by preventing premature mortality, has emerged as a key priority in global health and demographic research. For low- and middle-income countries, monitoring lifespan variation alongside life expectancy enables the detection of persistent mortality inequalities and uneven survival gains that average indicators alone may fail to reveal. Given its ongoing demographic and epidemiological changes, Bangladesh offers a relevant setting for studying these patterns. This paper examines lifespan variation in Bangladesh during its mortality transition.

We used two different sources of mortality data from 1974 to 2019: the Matlab Health and Demographic Surveillance System (HDSS), representing a long-standing rural demographic site, and the World Population Prospects 2024 (WPP), representing national-level trends. Comparing HDSS and WPP estimates allows us to assess how data resolution and national-level modeling affect the measurement of lifespan variation, which is critical for population health surveillance. We used multiple indicators of lifespan variation, including lifespan disparity, life table entropy, the Gini coefficient, standard deviation and interquartile range (IQR) of age-at-death distribution, to examine trends over time. We calculated 95% bootstrap confidence intervals to account for sampling variability and smoothing effects.

The study finds a consistent long-term decline in lifespan variation for both sexes across both datasets. Lifespan disparity declined from 19.6 to 13.8 years for Matlab males and from 20.9 to 12.4 years for Matlab females; in WPP, it dropped from 24.4 to 13.1 years for males and from 24.5 to 12.1 years for females. Similar downward trends are observed for life table entropy and Gini coefficient, with identifiable peaks during mortality crises, such as the 1974-75 famine and the 1991 cyclone. Since the early 2000s, the highest demographic variation has been observed among Matlab males, while WPP data show higher values for statistical measures in recent years- partly due to smoothing and national-level heterogeneity. Sex-specific differences and regional disparities persist, while bootstrapped confidence intervals indicate that these trends are robust despite smoothing and national-level aggregation in WPP data.

This study provides empirical evidence on lifespan variation trends during Bangladesh's mortality transition and illustrates the utility of high-resolution mortality data for monitoring inequality in LMIC populations.