Our health is closely related to our social interactions, such as our friendship networks. For example, it has been shown that not only infectious disease, but also health-related traits, such as obesity, depression, and smoking can spread through social networks. Understanding the relationship between social networks and health could help people devise network-based strategies to reduce the incidence of unhealthy behaviors and increase the prevalence of healthy ones.
In our most recent work we explore how a person’s position in their social network (called “centrality” by network scientists) is associated with their health. We exploit data from the NetHealth study, which has monitored smartphone usage (e.g., SMS interactions) and health-related behaviors (e.g., physical activities, heart rates, or sleep habits) of around 700 undergraduates from the University of Notre Dame over more than two years. With such data, we are able to study the relationship between individuals’ dynamic (evolving) social networks and their dynamic health-related behaviors, with a focus on physical activity.
Specifically, we look at the co-evolution of a person’s social network position (centrality) and their physical activity, with the goal of identifying groups of people who have different evolving social network profiles or different evolving physical activity profiles (or both). We are then able to explore whether and how the different groups of people differ in terms of personality, depression, and anxiety traits (measured through surveys).
NetHealth study participants whose network centralities significantly change with time show no trait differences compared to time-stable participants. However, out of these, participants whose physical activities also significantly change with time are more likely to be introverted than time-stable participants. Additionally, out of these, participants whose network centralities and physical activities also co-evolve with each other (e.g., when network centralities go up, physical activities also go up) are more likely to be anxious as well.
Such a predictive model could enable physicians and patients to not just diagnose and cure diseases early, but also act in a proactive manner before a disease even occurs.
In summary, our network analysis framework reveals several links between individuals’ social network structure, health-related behaviors, and other traits (e.g., personality or anxiety). In the future, our study could lead to the development of a predictive model that generates personalized health suggestions by exploiting rich longitudinal information (e.g., that can estimate the risk of an individual suffering from mental illness by using their social interaction, physical activity tracking, or other behavioral data from a recent time period). In the long run, such a predictive model could be translated into a health advice system that would enable physicians and patients to not just diagnose and cure diseases early, but perhaps also act in a proactive manner before a disease even occurs.
