Biomedical Engineering Reference
In-Depth Information
population of hosts already connected by a dynamic exposure network. More
realistic variants might include a host-specific proclivity to utilize the interven-
tion, which may depend upon a third "media" network distributing perceived
vulnerability. Multilevel networks provide a platform for analyzing a variety of
sociocultural dynamics that may impact physical health, including socially se-
lective stressors, differential access to medical care, culturally motivated avoid-
ance of diagnosis or treatment (e.g., failure to be tested for HIV due to stigma),
social transmission of health risk behaviors (e.g., smoking), and the globaliza-
tion of personal behavior (e.g., homogenization of health beliefs, social values,
lifestyle, and behavior, as described by Garrett (12)). Multilevel networks also
provide a context for analyzing interactions between host behavior and the biol-
ogy of developing disease, such as gene
x
diet interactions in atherosclerosis or
the evolution of pathogens and immune responses within behaviorally structured
niches. Evolving variants of ActiveHost, for example, mimic observed data in
developing more powerful immune systems for sexually promiscuous individu-
als (11). In the context of disease-reactive social behavior, evolutionary analyses
also show a strong selective pressure for the development of social norms that
isolate individuals during times of illness. These norms need to be transmissible
from parent to child for population-level selection, but they need not be geneti-
cally encoded. In fact, dissemination of such norms via superimposed interven-
tion networks enjoys considerable advantage over genetic transmission due to
the enhanced speed of norm dispersal.
3.3. Synergistic Complexity
Disease-reactive social behavior creates a temporal sparseness to social
networks that combines with structural sparseness to create transient social fire-
walls at the interface between infected and uninfected subgroups. This has the
net effect of discretizing continuous disease dynamics. A pathogen that kills all
members of a subpopulation before they can convey infection to the superpopu-
lation does not suffer a quantitative reduction in penetrance; it becomes extinct.
Sparse dynamics can cut the other way, of course, with a few random links car-
rying the potential to connect an isolated outbreak to a system-wide giant com-
ponent (the "patient zero" problem (12), as illustrated in Figures 2I-2K). These
quantal dynamics constitute the primary reason that linear algebraic models per-
form poorly in predicting the course of emerging epidemics. Linear statistical
models forecast the future range of an epidemic based on its past variation, but
reactive host networks show increasingly jumpy dynamics as the size of an epi-
demic grows. Figure 7 illustrates the complex kinetics that emerge when highly
structured networks are combined with disease-reactive linkage. Figure 7A
shows results from a highly infectious epidemic spreading through a population
of 300 agents organized into interconnected blocks of 3 (families), with one
