Methodological Aspects of Studying Human Aging, Health, and Mortality

Abstract

Age trajectories of mortality rates in human populations characterize individuals’ inequality in the duration of life. Various models of mortality rates are used in the analyses of survival data in demographic and epidemiological applications aiming to identify sources of individual differences in life span. Despite existing practice to use estimated model parameters in explanations of differences in mortality rates among different populations, in most cases differences in such estimates are difficult to interpret. The reason for this difficulty is that parameters of many demographic mortality models do not characterize either biological processes developing in aging human organisms or expose to environmental and living conditions. At the same time many processes affecting survival chances are measured in human longitudinal studies of aging, health and longevity, which suggest an opportunity for developing mortality models with parameters that can be interpreted in terms of processes and measured in such studies. The purpose of this paper is to develop an approach to mortality modeling that allows for describing the mortality rate in terms of parameters of physiological processes and declining health status that develops in aging human organisms. In contrast to traditional demographic models, which are difficult to use in the analyses of longitudinal data, our model allows for taking all these data into account. We use diffusion-type continuous time stochastic process for describing evolution of physiological state over the life course and finite-state continuous process for describing changes in health status during this period. We derive equations for respective mortality models, and approximate changes in physiological state by conditional Gaussian process, given health state. We applied this model to the analyses of longitudinal data collected in the Framingham Heart Study. The results of these analyses show that model parameters can be evaluated from longitudinal data and properly interpreted. The analyses indicate important differences in physiological dynamics among healthy and sick individuals.