Infection Proximity: A novel infection-specific exposure for network epidemiology

The study of contact networks, sexual or social, has proved invaluable in the study of the spread and prevention of infectious diseases. Recent research has extended beyond the contacts of a single individual to study how entire social mixing patterns, and the nature of the contact networks themselves influence disease risk. Doing so requires the use of mathematical measurements of the shape of a contact network, and the position of particular individuals within it. Many of these were not originally created with epidemiologic studies in mind, and measure only the wide-scale structure of the network. This information is of only secondary interest – a far more salient question is the positioning of infected individuals within the network, and the links between them and as-yet unexposed individuals.

We present a novel social network measurement, “Infection Proximity” (IP), which characterizes an individual's risk of disease from other infected members of their social contact networks. The performance of IP is then examined in the context of a simulated observational study of disease spread over a real-world social network, with IP acting as an exposure measurement. Analyzing these simulated studies using Cox proportional hazards models, IP is shown to be a good predictor of disease risk (HR = 3.43 95% CI = 1.54, 7.66), and superior to other common measures of network shape. Further theoretical and practical implications of IP in network epidemiology are then discussed.