Frailty, mortality and complexity: The osteological paradox and beyond

Abstract

The primary focus of paleoepidemiological research in the past decade has largely been to establish the association between skeletal indicators of disease and mortality risk, with a variety of epidemiological statistical methods employed to study this association. The application of epidemiological methods to skeletal assemblages are not, however, without their challenges (Klaus, 2014). Issues of small sample sizes and the limited information able to be derived from skeletal materials, especially that of hidden heterogeneity giving rise to selective mortality in populations, persist despite increased sophistication in analytical and statistical methodology (Milner and Boldsen, 2017, Klaus, 2014, Wood et al., 1992). Given these limitations inherent in the study of skeletal materials from past populations, efforts to identify and adopt techniques that facilitate synthesis of knowledge both from within and beyond the field are essential to improving our understanding of the past. This thesis endeavours to contribute to paleoepidemiology and bioarchaeology more broadly through interrogation of current research methods, and the development and application of novel methods for understanding and quantifying frailty in past (and present) populations. Firstly, the thesis examines a large skeletal assemblage for the association between skeletal indicators of disease and mortality, with emphasis on mortality before adulthood and the differences in survivorship for individuals with and without these indicators. The findings from these analyses indicate the presence of sub-groups within those without skeletal indicators, as well as the differing association between varied indicators and mortality. Building upon this practical demonstration of the interpretive challenges, despite having a large sample size to facilitate multivariable analysis, the thesis then turns to alternative methodology to explore beyond the limitations of skeletal assemblages. Using a generative model, hidden heterogeneity of frailty, conceptualised as a constant hazard which as a multiplicative effect throughout the lifetime, is assessed to interrogate the predictive value of such frailty on individual mortality when also accounting for stochasticity of stressor exposure and stressor severity. The observation of this hidden heterogeneity having low predictive value for individuals, but generating observable differences at the population level, prompts the question - at what sample sizes does differences due to sources of frailty in individuals produce observable population differences in mortality? The thesis examines this threshold through a second generative model in which both hidden heterogeneity of frailty (the constant hazard) as well as an observable source of frailty (socioeconomic status producing differences in likelihood of exposure to stressors) are modelled across varying sample sizes for the visibility of difference in mortality at the population level when the risk differs for sub-groups. This capacity to explore suitable sample sizes to test assumptions is then applied to Goodman and Armelagos (1988)'s three potential explanatory hypotheses for the observed patterns of mortality in skeletal assemblages from the Dickson Mounds. Collectively, the findings of the thesis aim to demonstrate that there is a need to leverage new methodologies to explore frailty and its impacts on mortality in order to facilitate accurate interpretation of epidemiological explorations of the past. The generative models developed and utilised in this thesis provide an accessible means through which information and theory from a variety of fields may be synthesised and used to test hypotheses regarding past populations. Show more