Social Networks and Prevalence

NIH reports on planned research by Peter Bearman:

THE SPREAD OF AUTISM DIAGNOSIS THROUGH SPATIALLY EMBEDDED SOCIAL NETWORKS Awardee Organization: COLUMBIA UNIV NEW YORK MORNINGSIDE
Abstract Text:
Project summary/Abstract This project will study the impact of diffusion of knowledge on the increasing prevalence of autism by building a large-scale, empirically calibrated simulation model of the social and interaction networks of parents. Despite hundreds of studies, existing explanations cannot account for the bulk of the increase in autism prevalence over the past three decades. Rising awareness and knowledge about autism has not been the focus of many empirical studies even though it has been widely acknowledged as a potential salient factor in the rise of autism. We have previously demonstrated that the diffusion of knowledge about autism through spatially proximate social relations has played an important role in autism's increase. Amplified by network interactions, the diffusion of knowledge about autism may be the key driver of the temporal and spatial patterns of rising autism incidence. It may also help explain the socio-economic disparity found in the probability and timing of receiving an autism diagnosis. A systems science approach is well positioned to model such a non-linear, endogenous diffusion process in tandem with other social, institutional and environmental causes. This project will use simulation methods to model the diffusion of autism diagnoses in California. We will reconstruct the state's entire population of 3 to 9 year old children from 1992 through 2010 (~3 million per year, ~57 million children) based on block level data from the three Federal censuses and all California birth records from 1989 to 2007. We will then empirically calibrate the parents' social networks by utilizing location data on focal points (e.g., schools, malls, childcare centers, and other points where parents interact). "What-if" scenarios, including distal environmental disasters and the initial distribution of incidence, will be incorporated in the model, as will all conventional risk factors known to operate at the individual level, community level factors known to be salient, and larger institutional processes that shape diagnostic regimes over time. The simulated results will be subjected to stringent validations using the spatial and temporal data of observed autism incidence from 1992 to 2010. Our project will demonstrate that social network analysis, agent-based modeling and increasingly available geospatial and organizational data can be effectively combined to inform the epidemiology of non- contagious diseases. Specifically, we anticipate that the modeling approach developed in this project will provide answers to the most important question confronting those interested in explaining the striking increase of autism prevalence over the past three decades: what accounts for the temporal and spatial patterns we observe?