Modelling Social Interaction: Can Incorporating Empirical Social Network Data in Public Health Economic Models Better Inform Public Health Interventions?

Public health systems are increasingly modelled as complex systems requiring complex interventions. Evidence shows health behaviours cluster within social networks and both influence and are influenced by social interaction. However, large-scale public health economic models often omit or poorly represent social interaction. Non-communicable health behaviours including smoking, alcohol consumption, and physical inactivity cluster within social networks. Models that fail to account for this may misrepresent health systems and produce flawed policy recommendations. Agent-based models attempt to incorporate social effects through theoretical or spatial networks, but few use empirically valid networks. Only 3% of social simulation studies in a review between 1998 and 2015 employ empirical network representations. This research aims to demonstrate the value empirical social network data into public health models. Challenges include the expense of collecting social network data, and the computational limitations of current network analysis methods. Scaling to population health models often involving populations of millions, requires efficient methods that can be feasibly used in models with limited computational budget. The key goals of this research will be to review current social interaction modelling practices, develop improved methods, and demonstrate their value by comparing empirical versus theoretical network structures in health policy models.