Many diseases can be prevented or mitigated through behaviour change, but we lack a quantitative model that can accurately predict these changes and inform policies designed to promote them. Here we introduce a quantitative model of health behaviour that takes into account individual-level barriers, the health system, and spread between individuals. We investigate limits of the model where each of these determining factors is dominant, and use them to predict behaviour from data. We apply the model to individual-level geographic barriers to mothers giving birth in a health facility, and find evidence that ease-of-access is a major determinant of delivery location. The geographic barriers allow us to explain the observed spatial distribution of this behaviour, and to accurately predict low prevalence regions. We then apply the model to the role of the health system in determining health facility usage by mothers of sick children. We show that local health facility quality does predict usage, but that this predictive power is significantly less than that gained by including unaccounted-for spatial correlation such as social influence. We also show evidence that results-based funding, rather than traditional input-based funding, increases usage. We develop a psychologically-motivated ‘complex contagion’ model for social influence and incorporate it into a general model of behaviour spread. We apply this model to short-lived behavioural fads, and show that ‘nudges’ can be very effective in systems with social influence. We successfully fit the model to data for the online spread of real-world behaviour, and use it to predict the peak time and duration of a fad before the peak occurred. Finally, we discuss ways to incorporate disease state into the model, and to relax the limits used in the rest of the thesis. We consider a model which links health behaviour to disease, and show that complex contagion leads to a feature that is not present in traditional models of disease: the survival of an epidemic depends non-trivially on the initial fraction of the population that is infected. We then introduce two possible models that include both social influence and an inhomogeneous population, and discuss the type of data that might be required to use them predictively. The model introduced here can be used to understand and predict health behaviours, and we therefore believe that it provides a valuable tool for informing policies to combat disease.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:682896 |
Date | January 2015 |
Creators | Sprague, Daniel Alexander |
Publisher | University of Warwick |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | http://wrap.warwick.ac.uk/77458/ |
Page generated in 0.0021 seconds