The age-related physiological and behavioral changes underlying the association of aging and mortality with sex and race will be assessed in five data sets (Framingham; Kaunas; Duke Longitudinal I; Duke Longitudinal II; Evans County). These data sets represent a wide range of population characteristics needed to determine whether the associations noted in one data set can be replicated in a similarly constituted analysis in the other data sets. In these analyses, we will apply methodologies appropriate to the estimation of the parameters of aging processes. Two such methodologies, which have been extensively applied to the study of such processes, will be used. The first, Grade of Membership (GOM) analysis, is applicable when there are many discrete variables used to describe biomedical and behavioral states and it is desired to determine if there exist subgroups with different age trajectories within the study population. The second, a stochastic process model of human aging and mortality, describes changes in a longitudinally followed population as a function of two inter-related processes: a) a continuous state-continuous time auto-regressive model of change on appropriate measurements, and b) a discrete state process, conditional upon the """"""""current"""""""" risk factor values, which models systematic loss from the study population. This model is well suited to describe physiological aging changes on a moderate number of well identified risk factors. In these investigations we will accomplish certain specific and substantive tasks. First, there will be continued development of the two analytic procedures. For example, the algorithm of the stochastic process model will be modified to allow much higher transition rates (e.g., 30 percent or more) than those typically found in morbidity and mortality processes. This will facilitate the analysis of certain social transitions (e.g., retirement) which occur at a high rate over a relatively short period. Second, using the GOM model we will identify subgroups with very different aging trajectories and survival chances, both within individual populations and within multiple, differently constituted populations. Third, using both methods, we will assess sex differences in the rates of change of certain physiological and cognitive parameters, the age-related correlation of physiological, cognitive and behavioral changes, and the relation of such processes to the risk of death. Of interest, is a study of the temporal lag in CVD morbidity and mortality between the sexes.
