Contributions to network science in public health

This thesis develops and exploits methodology within network science, driven by important applications. By use of large-scale simulations and high-quality data, we demonstrate how network science can contribute to understanding and predicting various phenomena in public health. We study peer effects of pain tolerance in a friendship network, and we use mobility networks between locations to develop spatio-temporal mathematical models of influenza. Pain tolerance is measured by how long you can hold your hand in cold water, and is censored. By extending social network methodology to handle censoring, pain tolerance among friends is found to be positively correlated. When the network is stratified on sex, we find that the peer effect is only present in males, and only through their male-male friendships. Transmission of airborne pathogens in humans on large scales is driven by the human mobility network. The mobility network is closely linked to the population density. As part of the urbanisation process, individuals cluster in geographical areas. We propose a method for generating spatial fields with controllable levels of population clustering, and simulate disease spread when population clustering is varied. Population clustering is found to be an important determinant for the effect of travel restrictions on infectious disease spread. We thus contribute to understanding how the effect of interventions can vary between countries. The spatio-temporal spread of influenza in Bangladesh is modelled by using a dynamic mobility network informed by daily mobile phone mobility data. Such data are particularly useful in low-income settings, due to scarce census data. When the model is informed by time-averaged mobility data, the results are very similar to the results with daily mobility. This is important for future studies and outbreak control. We apply the model to predict spatial spread and estimate transmissibility for influenza in Bangladesh