|dc.description.abstract||Scholars have long found compelling evidence of associations between social relationships and health, largely drawing from summary survey measures of social engagement and support. In recent decades, a growing literature not only reexamines these associations using more-specific and detailed survey-based measures of networks but also offers new understandings of how the structure and composition of networks protect or even harm physical and mental health. The three empirical chapters of this dissertation draw from four propositions to contribute to this growing literature and shed new light on how networks matter: First, the structure and composition of networks matter for health in ways not captured by summary indices alone. Second, relationships are associated with health through multiple mechanisms beyond material aid or direct disease exposure. Third, the mechanisms by which relationships matter for health are not always captured by surveys and the hypotheses that motivated their design. And, fourth, these mechanisms are specific to different outcomes and disease profiles. To evaluate these propositions, the three empirical chapters use data from two unique, multidisciplinary surveys of core discussion networks, which are derived from a classic measure of networks defined as the people (alters) with whom an individual (ego) discusses “important matters.”
In the first empirical chapter, I examine the likelihood of receiving help with household tasks and personal care activities among older adults. Difficulty with household tasks and personal care activities are common measures of functional disability and well-being in later life, and unmet needs for help with these tasks and activities are associated with a range of negative consequences, including loss of independence. Previous research on help emphasizes the importance of access to particular family and household relationships. This chapter looks to the composition and structure of the wider interpersonal environment in which older adults are embedded as captured by the core discussion network. Using logistic regression models with cross-sectional data from the second wave of National Social Life, Health, and Aging Project (NSHAP), I find that network composition and structure is associated with the odds of help when controlling for family and household relationships. I also find that the particular network measures correlated with help differ given the nature of particular tasks or activities. This chapter emphasizes that core discussion networks are not only indicators of access to helpers—including but not limited to family or household members. Rather, core discussion networks are also suggestive of different enabling processes, including different cognitive frames that condition the likelihood of activating and receiving help when coping with different functional limitations.
In the second empirical chapter, I investigate the longitudinal association of network composition and structure with cognitive function. Social relationships are frequently posited as key protective factors of cognitive function, which is critical for well-being in older age. This research is based on two hypotheses: the cognitive enrichment hypothesis, which argues social engagement directly affects cognitive function through mental stimulation; and the stress hypothesis, which argues social support and cohesion indirectly affect cognitive function by promoting positive emotions and protecting against negative emotions, which in turn affect cardiovascular and neuroendocrine functions associated with neuropathology. Using lagged dependent variable regression models with data from the second and third waves of NSHAP, I find that network size is positively associated with cognitive function while frequency of communication with alters is negatively associated with cognitive function, seemingly both supporting and contradicting the cognitive engagement hypothesis. I also find that density and average emotional closeness to alters are respectively negatively and positively associated with cognitive function as expected, but are not significant when controlling for sociodemographic and health measures. This chapter emphasizes the need not only for future research to explore these puzzling findings, but also for additional research on the effects of mild cognitive impairment and dementia on measurement error in survey-based network instruments.
In the third empirical chapter, my coauthor and I examine post-traumatic stress after Hurricane Katrina using longitudinal survey and qualitative interview data from the RISK Project. This chapter uses an innovative mixed method design to not only identify which characteristics of the core discussion network are associated with long-term psychological health following a natural disaster but also the multiple mechanisms by which those characteristics affect psychological health. Using logistic regression models, we find that geographic distance to alters is associated with a greater likelihood of post-traumatic stress five years after Hurricane Katrina. Qualitative interviews then reveal this association can be explained not by differences in access to instrumental support but rather by decreased feelings of belonging to a local community and decreased feelings of mattering to others among these with geographically dispersed networks. This chapter draws attention to the value of mixed method research, to geographic distance as a key aspect of networks, and to the salubrity of not only receiving but giving support to others.||