MIT Media Lab: Reality Mining

Home Visualizations About Publications Contact Us Download

Reality Mining: Complex Social Systems

Epidemiology and Information Dissemination

Computational epidemiology is the study of modeling disease propagation. In order to understand and control the spread of pathogens, it is essential to establish some of the key parameters associated with disease transmission. Determining the basic reproductive ratio (Ro) of a disease, for example, is the primary objective for many epidemiological studies. Ro defines the number of secondary cases produced by an infected individual in an entirely susceptible population. Ideally, health policies can attempt to change the parameters involved in its formulation in order to control a pathogen's spread. Unfortunately, is notoriously difficult to measure, and must be derived indirectly. Mathematical models have played an important role in assessing and understanding the dynamics of disease transmission in human populations. For example, the proportion of individuals requiring vaccination for the eradication of a disease may be formulated using Ro.

The majority of epidemiological models are based on a compartmental, SIR framework; the host population is partitioned into those that are susceptible, infected, or immune to a particular pathogen. These deterministic models assume that the rate at which new infections are acquired is proportional to the number of encounters between susceptible and infected individuals, and leads to an effective reproductive ratio that is dependent on a threshold density of susceptibles. Thus, is dependent not only on parameters intrinsic to the disease such as latent and infectious periods, but also on contacts between infectious and susceptible hosts. Compartmental models of this kind implicitly assume that the host population is well mixed, such that the probability of infection is equal for all.

Social network structures are clearly not always well mixed, however, and the complexities of host interactions may have profound implications for the interpretation of epidemiological models and clinical data. Standard mean-field models do not account for heterogeneities of risk between individuals due to the finite number, variability, and clustering of social contacts. Studies have shown that network structure can significantly affect the processes occurring on social networks, including the dynamics and evolution of infectious diseases. Some have investigated the effect of network structure on the evolution of disease traits such as infectious period and transmission rates, as well as invasion thresholds for epidemics. Others have explored the role of spatial contact structure in the evolution of virulence.

These models have used hypothetical, extreme network structures as caricatures of real host contact networks. During the recent outbreak of SARS in Singapore, contact tracing of infectious individuals showed a 'superspreader' pattern of disease transmission, although unlike for STDs, the mechanisms behind this are not understood. Furthermore, many directly transmitted diseases are distributed globally, in different types of society, with different characteristic interactions between people.

The accurate quantification of the host contacts, and therefore the associated variability in the probability of infection, is clearly of great importance. Hypothetical models are valuable for understanding the kind of effect different social network structures would have on disease spread, however we suggest that the data captured by applications such as BlueAware would give a much more realistic interpretation of human social network dynamics. With detailed data on mixing parameters within a social network, epidemiologists will be armed with more information to make predictions about our vulnerability to the next SARS, as well as greater insight into preventing future epidemics.