The Effect of Disaggregation on Infection Spreading in a social network: 'More' may not be 'Merrier'

 

VARSHA KULKARNI

INDIANA UNIVERSITY

 

In the past, researchers have applied epidemiological models in the framework of complex social networks to explain dynamics of contagious diseases therein. This paper analyzes the same in the framework of a disaggregated network viz. the world as a ‘network’ disaggregated into diverse countries as ‘clusters’. First, the random matrix analysis carried out on WHO data on ‘Hepatitis/other viral’ infections of 13 countries for period 1985-2008, rules out systematic correlations in prevalence between countries or a ‘common influence’. Next the effect of disaggregation is discussed in terms of how the connectivity and population diversities affect epidemic threshold (?c). Connectivity, k is interpreted as a cultural determinant and estimated from ‘ethnic fractionalization’ data. Statistical analysis is used to estimate the model parameters- spreading rate, curing rate -and to examine the effect of population on prevalence. Agent-based modeling (abm) is employed to investigate the roles of k and population N. The analyses reveal that (i) the patterns of infection spreading from the model match well with those observed (ii) the differences in disease patterns of countries revealed by the disaggregated framework are actually a reflection of the heterogeneity across the countries. A less perfect power law fit is obtained for distribution of k. (iii) ?c depends on not just k (as established in previous studies that higher k adversely affects epidemics) but actually on its interplay with population of the region/country. It turns out that isolation may not be as effective in lowering infection spread in crowded communities as in sparsely populated regions (iv) a small but positive significant effect of population on prevalence got from statistical analysis confirms the above (v) the result of abm’s for 3 configurations – [N, k], [N, k/4] and [3N, k/4] is consistent with above findings. Thus disaggregated modeling framework explains disease dynamics in spatially separated regions in terms of their cultural and demographic aspects and the variations revealed are interesting.