
Projects
The Effect of Network Accuracy and Risk Categorization on Dynamic Network Models

Completed

Project Summary
The aim of the proposed project is to develop a disease transmission model that will combine agent-based modelling, temporal SNA, risk categorization and socio-economic prediction for change of the network structure. An algorithm will be developed to impute the structure of the “true” dynamic network to be included into the model. The pig industry will be used as an example, because of their importance for the Swiss agriculture and still uninvestigated network structure in Switzerland. In addition to official (legally obligated) pig movement records, we will use “grey data” on pig movements from pig health services, supplemented by semi-structured interviews with farmers. Using these data, we will develop an ovelstochastic approach for reconstructing a closer-to-complete “true” network of pig contacts from the “apparent” network based on official records. The temporal dynamics of the network will be captured by applying temporal SNA to movement data from the past six years.
Network metrics such as degree, betweenness, closeness and eigenvector centrality will be used to classify premises according to their risk for contacts to otherpremises. Econometric risk factor analysis will be applied to identify the socio-economic variables (e.g. education of farmers, urbanization, farmers' income opportunities, or changes in market structure) influencing the contact risk among premises. These socio-economic predictors will be used to simulate future changes in the network structure. To quantify the increase in model output accuracy gained by including “grey data”, the transmission model will be applied to both the “true” and “apparent” networks. Model outputs will be compared and validated using data from recent pig disease outbreaks in Switzerland, including outbreaks of porcine reproductive and respiratory syndrome, enzootic pneumonia, and progressive atrophic rhinitis. The disease transmission model will also be used to evaluate the premises-specific risk for disease propagation, spread and effectiveness of control strategies of endemic and exotic porcine infectious diseases in Switzerland.
Objectives
Develop a novel algorithm that allows us to estimate a “true” network based on officially recorded contact data.
Quantify and validate the predictive benefit of using the “true” network relative to the network based on official data using an infectious disease simulation model.
Evaluate socio-economic, farm-specific and market- driven risk factors for disease propagation.
Funding Agency
Swiss National Science Foundation
Project Members

Salome Dürr (PI)
Veterinary Public Health Institute

Hartmut Lentz Friedrich Löffler
Institute of Epidemiology

Sandro Steinbach
North Dakota State University

Christina Nathues
University of Bern

Vitaly Belik
University of Berlin