Heterogeneity in Health and Mortality Studies of Population

In the field of demography, summary measures of population health aim to capture the current health and mortality conditions, facilitating comparisons across countries, other macro units, and over time. However, traditional population statistics can be affected by a compositional bias resulting from past conditions, which means they may not fully reflect the current health and mortality conditions. This research project seeks to address this compositional bias by proposing innovative methods that correct for the influence of observed and unobserved characteristics between life-table cohorts and the population implied by the period-specific conditions. These new measures are applied to key topics that are heavily debated in population studies. The first part of the study focuses on mortality statistics, including life expectancy (LE) and various lifespan inequality measures. Comparing the standard and newly proposed measures, we study the lagged compositional bias in the standard statistics. Next, by examining LE under the current population composition, we reassess the developments of best-practice LE and the phenomena of the narrowing sex gap in LE in high-income countries over recent decades. Similar to standard measures, the proposed lifespan inequality statistics can be decomposed, into three parts to understand differences between populations in mortality, differences within populations in mortality, and compositional differences between populations that reflect past conditions. In this project, we take a pioneering step by extending the concept of mortality under current conditions to also include morbidity. Applying multistate models that consider heterogeneity, we introduce innovative methods to estimate summary measures of population health, taking into account the population composition resulting from current conditions. The proposed statistics are then applied to re-examine two ongoing debates in population health: the compression vs. expansion of morbidity accompanying an increase in LE, and the gender morbidity-mortality paradox. Furthermore, we propose methods to quantify the variation in healthy and unhealthy lifespans under current conditions, along with their decomposition, similar to the approach used for mortality statistics. All of the innovative methods proposed in the study are incorporated into an open R package, making them easily accessible for researchers. By addressing the compositional bias and extending our understanding of morbidity under current conditions, our research contributes valuable insights to the field of population health. Our findings have the potential to inform public health policies and interventions, ultimately contributing to the improvement of population health worldwide.