Networked resources and health in South Africa: Towards a Materialist Social Network Epidemiology

Abstract

People share resources with each other across meaningful social relationships. It has been demonstrated that increased resources cause better health for individuals, yet there is no epidemiologic research investigating the impact on population health of material resources embedded in social networks. This knowledge gap is related to the use in epidemiologic studies of theories which do not account explicitly for material resources, the assumption of independence in causal inference, and the dearth of socio-centric network data linked to health outcomes and exposures. Drawing on newly developed methods in causal inference, the project examines whether: a) a financial incentive for HIV testing has effects across family relationships; b) the wealth held by one household affects the physical function of family members living in another; and c) the wealth held by one household affects mortality in the households of family members. To answer these questions, routinely collected data from two South African health and demographic surveillance systems were used to construct socio-centric family networks. Overlaid on these networks were data from HITS — a cluster randomized controlled trial studying the effect of a financial incentive on HIV testing, and data from HAALSI — a cohort study of older South Africans investigating physical function among other health outcomes. Related to a), financial incentives for HIV testing are shown to not only influence the person who is offered the incentive, but their family members as well. The offer of a financial incentive affects only the behavior of individuals who were themselves offered the incentive, however. With respect to b), wealth held by family members do not appear to affect the physical function of older individuals. Regarding c), the wealth held by a household is shown to lighten the mortality of its members as well as that of non-household family members. This effect was pronounced among working age (16-59) adults. Overall, this dissertation finds evidence that the resources held in family networks shape health across the networks. It demonstrates that it is feasible to conduct similar analyses using available theory and data and newly developed analytic methods.