Effective Network Size Predicted From Simulations of Pathogen Outbreaks Through Social Networks Provides a Novel Measure of Structure-Standardized Group Size

Abstract

The transmission of infectious disease through a population is often modeled assuming that interactions occur randomly in groups, with all individuals potentially interacting with all other individuals at an equal rate. However, it is well known that pairs of individuals vary in their degree of contact. Here, we propose a measure to account for such heterogeneity: effective network size (ENS), which refers to the size of a maximally complete network (i.e., unstructured, where all individuals interact with all others equally) that corresponds to the outbreak characteristics of a given heterogeneous, structured network. We simulated susceptible-infected (SI) and susceptible-infected-recovered (SIR) models on maximally complete networks to produce idealized outbreak duration distributions for a disease on a network of a given size. We also simulated the transmission of these same diseases on random structured networks and then used the resulting outbreak duration distributions to predict the ENS for the group or population. We provide the methods to reproduce these analyses in a public R package, “enss.” Outbreak durations of simulations on randomly structured networks were more variable than those on complete networks, but tended to have similar mean durations of disease spread. We then applied our novel metric to empirical primate networks taken from the literature and compared the information represented by our ENSs to that by other established social network metrics. In AICc model comparison frameworks, group size and mean distance proved to be the metrics most consistently associated with ENS for SI simulations, while group size, centralization, and modularity were most consistently associated with ENS for SIR simulations. In all cases, ENS was shown to be associated with at least two other independent metrics, supporting its use as a novel metric. Overall, our study provides a proof of concept for simulation-based approaches toward constructing metrics of ENS, while also revealing the conditions under which this approach is most promising.

COVID-19 Outbreak Prediction with Machine Learning

“…Overall, the standard epidemiological models can be effective and reliable only if (a) the social interactions are stationary through time (i.e., no changes in interventions or control measures), and (b) there exists a great deal of knowledge of class R with which to compute Equation (3). To acquire information on class R, several novel models included data from social media or call data records (CDR), which showed promising results [18][19][20][21][22][23][24][25]. However, observation of the behavior of COVID-19 in several countries demonstrates a high degree of uncertainty and complexity [26].…”

COVID-19 Pandemic Prediction for Hungary; A Hybrid Machine Learning Approach

“…Secondly, the class R can be computed precisely. To better estimate the class R, several data sources can be integrated with SIR-based models, e.g., social media and call data records (CDR), which of course a high degree of uncertainty and complexity still remains [25][26][27][28][29][30][31][32]. Considering the above uncertainties involved in the advancement of SIR-based models the generalization ability are yet to be improved to achieve scalable model with high performance [33].…”

Primate Infectious Disease Ecology: Insights and Future Directions at the Human-Macaque Interface

“…R 0 is the basic reproduction number used to quantify the transmission potential of a parasite, defined as the number of secondary infections caused by a single infected individual introduced into a population made up entirely of susceptible individuals. However, following the acknowledgment that individuals do not move or interact randomly, social networks have been integrated into these models during the last decade or so (Griffin and Nunn 2012;Huang and Li 2007;McCabe and Nunn 2018;Nunn 2009).…”

“…Expanding beyond homogenous populations that contain only susceptible individuals that interact randomly, recent studies have integrated network approaches with a classic set of individual-based models used in epidemiology that classify individuals or "agents" into moving between susceptible infected and resistant (or SIR) classes or compartments (Bansal et al 2007;Brauer 2008;Grimm and Railsback 2005;Kohler and Gumerman 2000). Such integrated network-based SIR models now form the basis of many epidemiological assessments in primate systems (Griffin and Nunn 2012;Kohler and Gumerman 2000;McCabe and Nunn, 2018;Rushmore et al 2014). For instance, the likelihood of parasite transmission from A to B may be a function of (1) whether A is already infected, (2) the likelihood of a link between A and B in their social network, and (3) the per-contact transmission probability "Beta."…”